UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Restoring grasslands in southern Ontario sandpits : plant and soil food web responses to arbuscular mycorrhizal… Ohsowski, Brian Matthew 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2015_September_Ohsowski_Brian.pdf [ 7.5MB ]
Metadata
JSON: 24-1.0074440.json
JSON-LD: 24-1.0074440-ld.json
RDF/XML (Pretty): 24-1.0074440-rdf.xml
RDF/JSON: 24-1.0074440-rdf.json
Turtle: 24-1.0074440-turtle.txt
N-Triples: 24-1.0074440-rdf-ntriples.txt
Original Record: 24-1.0074440-source.json
Full Text
24-1.0074440-fulltext.txt
Citation
24-1.0074440.ris

Full Text

Restoring Grasslands in SouthernOntario Sandpits: Plant and Soil FoodWeb Responses to ArbuscularMycorrhizal Fungal Inoculum,Biochar, and Municipal CompostbyBrian Matthew OhsowskiB.Sc., Eastern Michigan University, 2003M.Sc., Eastern Michigan University, 2008A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE COLLEGE OF GRADUATE STUDIES(Biology)THE UNIVERSITY OF BRITISH COLUMBIA(Okanagan)May 2015c© Brian Matthew Ohsowski, 2015AbstractHabitat destruction and land use change are among the anthropogenic influences affect-ing many ecosystems. Sandpit mining often restricts grassland plant restoration efforts dueto the abioticially stressed mine substrate, a lack of viable plant symbionts, and disruptedmulti–trophic interactions in soil food webs. Recently excavated sandpits can be amelio-rated with soil amendments and arbuscular mycorrhizal fungal inoculum to address thesedegraded substrate conditions, potentially improving plant performance and acceleratingsoil food web development. This dissertation describes the results of a multi–year grasslandrestoration project established in southern Ontario that optimized industrial–scale grass-land restoration protocol in post–extraction sandpits. This research tested the effect of soilamendment rate (municipal compost, biochar) and arbuscular mycorrhizal (AM) fungalinoculum (Rhizophagus irregularis) in a grassland plant plug trial and a seed applicationtrial. In the plant plug trial, the multi–year effects of the experimental treatments onplant growth, AM fungal colonization of roots, soil microbial biomass (i.e. bacteria andfungi) and soil animal abundance (i.e. nematodes, Collembola, and mites) were exploredover two growing seasons. In the plant plug trial, 20 T ha−1 (tons hectare−1) of compostmixed with a low rate of biochar (10 T ha−1) yielded the largest positive effect on totalplant biomass, microbial community biomass, and soil animal abundance after two growingseasons. AM fungal inoculation did not influence total plant biomass or soil food web devel-opment during this trial. In the seed application trial, the multi–year effects of increasingrates of compost and biochar (0 T ha−1 to 40 T ha−1) were explored for total plant coverover three growing seasons. AM fungal inoculation combined with high rates of compost(20 T ha−1 and 40 T ha−1) and biochar (20 T ha−1 and 40 T ha−1) resulted in the highestplant cover over three growing seasons compared to controls. Our results indicate that: (1)co–amending mine substrates with compost, biochar, and AM fungal inoculum are practicalland management tools that improve grassland plant growth while increasing soil food webdevelopment, (2) AM fungal inoculum increases plant cover when applying seed with highrates of compost + biochar, and (3) amending post–mine substrates with biochar as a soli-tary amendment may increase biotic stress in the sandpit environment during restoration.I suggest that restoration practitioners emphasize soil community development in tandemwith plant community growth when restoring sandpits to maximize restoration success.iiPrefaceThe introduction (Chapter 1) is an adaptation of a published review in the Journalof Applied Soil Ecology (Ohsowski et al. 2012). The published work, The potential ofsoil amendments for restoring severely disturbed grasslands, was wholly drafted by BrianM. Ohsowski with editorial comments by Drs. Miranda Hart, John Klironomos and KariDunfield. The research for this study was carried out in a post–extraction sandpit locatednear Port Rowan, Ontario, Canada (Chapter 3–Chapter 4 ). The Ontario Aggregate Re-search Corporation (TOARC) graded the site and installed a nine–wire fence before plotinstallation in 2010. I designed and implemented both experiments (i.e. plant plug trialand seed application trial). My field assistant, Andre´ Aude´t, and I manually installed theplots and plant plugs in the summer of 2010. In May of 2011, I applied the seed and AMfungal inoculum to establish the seed application trial. I collected plant biomass data withthe help of field assistants in the summer of 2011 and 2012. In Chapter 2, I designed andimplemented the partial least squares regression experiment. Plant data was collected withmy field assistant, Sarah Kruis, in September 2012. I conducted the statistical analysis andcomposed the manuscript. In Chapter 3, I collected plant biomass data, soil cores, and per-formed root washing/preservation for AM fungal percent colonization with help from Andre´Aude´t. AM fungal colonization was conducted by the Soil Analysis Laboratory, Universityof California, Riverside. In Chapter 4, I collected soils for the soil food web analysis in theplant plug trial. Soil organisms were measured by the Soil Analysis Laboratory, Univer-sity of California, Riverside. I performed all linear model analyses for soil organisms. Dr.Anita Antoninka at Northern Arizona University assisted me with the proper implemen-tation and execution of the structural equation models in the statistical program AMOS.I created the a priori hypotheses, co–ran the structural equation models, interpreted thedata, and composed the manuscript. All thesis chapters were written with the guidance ofDrs. Miranda Hart and John Klironomos. Thesis chapters were reviewed by the membersof my supervisory committee: Drs. Melanie Jones and David Scott from the University ofBritish Columbia Okanagan Campus and Dr. Kari Dunfield from the University of Guelphin Guelph, Ontario.iiiTABLE OF CONTENTSTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxDedication . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiChapter 1: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Landscape restoration and successional theory . . . . . . . . . . . . . . . . . 11.1.1 Historical context . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1.2 Post–mine areas as primary succession models . . . . . . . . . . . . 21.2 An ecological context of degraded system restoration . . . . . . . . . . . . . 31.3 Grassland vegetation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.4 Soil food webs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51.4.1 Soil microbial communities as indicators for post–mine substrate re-covery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.4.2 Soil animals as indicators for post–mine substrate recovery . . . . . 71.5 Techniques for improving disturbed soils in grassland restoration projects . 81.5.1 Vegetation–derived biochar . . . . . . . . . . . . . . . . . . . . . . . 81.5.2 Leaf and yard waste compost . . . . . . . . . . . . . . . . . . . . . . 101.5.3 Arbuscular mycorrhizal fungal inoculation of grassland plants . . . . 111.6 Review conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121.7 Research objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13Chapter 2: Improving Plant Biomass Estimation . . . . . . . . . . . . . . . . 15ivTABLE OF CONTENTS2.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 152.1.1 Techniques to predict plant biomass . . . . . . . . . . . . . . . . . . 162.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182.2.1 Species selection and data collection . . . . . . . . . . . . . . . . . . 182.2.2 Measured plant traits . . . . . . . . . . . . . . . . . . . . . . . . . . 182.2.3 Model creation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 192.2.4 Data transformation, auto–scaling, and polynomial terms . . . . . . 212.2.5 Variable reduction and model averaging . . . . . . . . . . . . . . . . 212.2.6 Partial least squares regression and linear regression models . . . . . 222.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 232.3.1 Variable selection in partial least squares regression models . . . . . 232.3.2 Comparing models for predicting plant biomass in the training dataset 232.3.3 Comparing models for predicting plant biomass in the test dataset . 252.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262.4.1 Variable selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262.4.2 The statistical advantage of using partial least squares regressionwhen prediction plant biomass . . . . . . . . . . . . . . . . . . . . . 282.4.3 Practical applications of partial least squares regression . . . . . . . 292.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31Chapter 3: The Restoration of Grassland Vegetation in Post–ExtractionSandpits . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 323.1.1 Biochar and compost as sandpit amendments . . . . . . . . . . . . . 323.1.2 Arbuscular mycorrhizal fungi as inoculum . . . . . . . . . . . . . . . 333.1.3 Synergisms among biochar, compost, and arbuscular mycorrhizas . . 343.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353.2.1 Research site establishment . . . . . . . . . . . . . . . . . . . . . . . 353.2.2 Experimental design . . . . . . . . . . . . . . . . . . . . . . . . . . . 363.2.3 Plants used in restoration . . . . . . . . . . . . . . . . . . . . . . . . 373.2.4 Plant plug trial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 393.2.5 Seed application trial . . . . . . . . . . . . . . . . . . . . . . . . . . 453.2.6 Statistical analyses . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473.3.1 Plant plug trial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473.3.2 Seed application trial . . . . . . . . . . . . . . . . . . . . . . . . . . 533.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62vTABLE OF CONTENTS3.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69Chapter 4: Soil Food Webs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 704.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 704.1.1 Arbuscular mycorrhizal fungal inoculum . . . . . . . . . . . . . . . . 704.1.2 Biochar . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 714.1.3 Compost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 714.1.4 Synergisms among biochar, compost, and arbuscular mycorrhizas . . 714.1.5 Hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724.2 Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 724.2.1 Soil collection and organismal analyses . . . . . . . . . . . . . . . . . 724.2.2 Statistical analyses for soil biota . . . . . . . . . . . . . . . . . . . . 744.2.3 Soil food web analysis with structural equation modeling . . . . . . 744.3 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 774.3.1 Soil food web structural equation model selection . . . . . . . . . . . 774.3.2 Microbial community biomass and soil animal abundance . . . . . . 774.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 964.4.1 Soil food web response to AM fungal inoculation . . . . . . . . . . . 964.4.2 Soil food web response to biochar . . . . . . . . . . . . . . . . . . . . 974.4.3 Soil food web response to compost . . . . . . . . . . . . . . . . . . . 984.4.4 Soil food web response to compost and biochar . . . . . . . . . . . . 1004.4.5 Interactions among soil microbial biomass and soil animal abundance 1014.4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103Chapter 5: Management Recommendations for Grassland Restoration inPost–Extraction Sandpits . . . . . . . . . . . . . . . . . . . . . . . 1055.1 Plant species selection and sourcing . . . . . . . . . . . . . . . . . . . . . . 1065.1.1 Soil amendments and commercial AMF inoculum . . . . . . . . . . . 1085.2 Purchasing soil amendments and inoculum for a restoration project . . . . . 1095.3 Site preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1105.4 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110Chapter 6: Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1136.1 Strengths and limitations of the dissertation research . . . . . . . . . . . . . 1166.2 Future directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118viTABLE OF CONTENTSAppendicesAppendix A: R Code for AM Fungal Plant Plug Root Colonization in thePlant Plug Trial . . . . . . . . . . . . . . . . . . . . . . . . . . . 147Appendix B: R Code for AM Fungal Root Colonization in the Plant PlugTrial Field Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . 149Appendix C: R Code for Plant Plug Trial Biomass Predictions and PlotMass Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . 154Appendix D: R Code for Seed Trial Plant Cover . . . . . . . . . . . . . . . . 239Appendix E: R Code for Soil Food Web Organisms . . . . . . . . . . . . . . 245viiList of TablesTable 2.1 Experimental or observational situations to employ non–destructivebiomass estimation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16Table 2.2 Measured plant traits included in the LR and optimized partial leastsquares regression models. Partial least squares regression componentselection based on lowest root mean squared error from cross–validation(RMSECV) using 10–fold cross–validation. Plant measurement abbre-viations: 30c = circumference at height of 30cm; bc = basal circum-ference; cd = maximum canopy diameter; fc = frond count; fl = frondlength (blade length + stipe length); hc = circumference at half plantheight; ln = number of leaves; sl = stipe length; th = total plantheight; wph = resting height of falling plate meter. . . . . . . . . . . 21Table 2.3 Summary statistics for PLS and LR model training datasets. R-squared(R2) and root mean squared error (RMSE) values are based on Pmassversus Rmass estimates where slope = 1 and intercept = 0. . . . . . . 25Table 2.4 Summary statistics for PLS and LR model externally predicted data.R-squared (R2) and root mean squared error (RMSE) values are basedon Pmass verses Rmass estimates where slope = 1 and intercept = 0. 26Table 3.1 Experimental treatments for the seed application experiment. All treat-ment levels are fully factorial. Each treatment combination was appliedto one plot only. Total number of plots was 72. . . . . . . . . . . . . . 37Table 3.2 The eight grassland plant species used in the plant plug trial and seedapplication trial. The abbreviation column indicates the plant codescheme associated with Figure 3.1. The final two columns indicate theabundance (i.e. number of plant plugs) of all species in each plot andthe core sampling areas for the plant plug trial. . . . . . . . . . . . . . 38Table 3.3 Morphological characters measured in the field for the six plant speciesin September 2011 and September 2012. The most parsimonious com-bination of variables was selected via Bayesian Information Criterionmodel selection to create the predictive models. . . . . . . . . . . . . . 42viiiLIST OF TABLESTable 3.4 Morphological characters selected for the six plant species measuredin September 2011 and September 2012. The variables given in thetable were selected via Bayesian Information Criterion model selectionto create the predictive models using partial least square regression. . 43Table 3.5 Partial least squares regression diagnostics for the six plant species mea-sured in September 2011 and September 2012. All prediction data isbased on variables selected via Bayesian Information Criterion modelselection (Table 3.4). Mass data is given in grams (g) dry weight basedon weighed plants used to create the standard curve. For each species,predicted plant mass from the partial least squares regression modelwas subtracted from the reference plant mass (Pmass – Rmass )± 1 stan-dard deviation (SD) to calculate within–model estimates. When Pmass= Rmass, predicted mass is equal to reference mass, thus representsa perfect prediction. R-squared, root mean squared error (RMSE),and p–values were calculated for Pmass – Rmass using linear regressionfor each plant species to indicate prediction accuracy. All regressiondiagnostics are based on a slope = 1 and intercept = 0. . . . . . . . 44Table 3.6 Seeding rate in grams (g) for the eight grassland plant species usedin the seed application trial. Plot size was 10.2 m2. Seeds were cold–moist stratified at 4 ◦C for one month until the time of sowing in thefield (May 2011). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45Table 4.1 Direct, indirect, and total standardized regression estimates of soilamendments on the soil community and N–fixing forbs generated bythe structural equation model. Significant direct pathway estimates aregiven in bold text (p < 0.05). . . . . . . . . . . . . . . . . . . . . . . . 82Table 4.2 Direct, indirect, and total standardized regression estimates of soil mi-crobes and N–fixing forbs on the soil community generated by the struc-tural equation model. Significant direct pathway estimates are givenin bold text (p < 0.05). . . . . . . . . . . . . . . . . . . . . . . . . . . 94Table 4.3 Direct, indirect, and total standardized regression estimates among thesoil animals generated by the structural equation model. Significantdirect pathway estimates are given in bold text (p < 0.05). . . . . . . 95Table 5.1 The projected materials cost of land rehabilitation in abandoned sand-pits in southern Ontario. Two viable options are available for prairiesystem rehabilitation: seed addition or plug addition. Note that thecost per ha decreases as the rehabilitation area increases. . . . . . . . 111ixList of FiguresFigure 2.1 Workflow for predicting plant biomass with partial least squares re-gression. Plant measurement abbreviations: 30c = circumference atheight of 30 cm; bc = basal circumference; bl = fern blade length; cd= maximum canopy diameter; fl = frond length (blade length + stipelength); hc = circumference at half plant height; ln = leaf number;lp = longest pinna per blade; pi = pinnae number per blade; shc =seed head count; sl = stipe length; th = total plant height; wph =resting height of falling plate meter. . . . . . . . . . . . . . . . . . . . 20Figure 2.2 Graphs of predicted (Pmass) vs. reference (Rmass) plant biomass us-ing the optimized PLS models and the LR model for the three plantspecies. The blue(PLS) and red(LR) points represent internally pre-dicted data used to train each model (n = 35). Black points representexternal data predictions from the test dataset using only predictorvariables (n = 6). Each dashed line indicates a perfect prediction(Pmass = Rmass) with a slope = 1 and intercept = 0. . . . . . . . . . 24Figure 3.1 Diagram of the plant plug layout with plant positioning. Each hexag-onal cell signifies the location and identity of one plant taxa added tothe plot as a plant plug. All plots have the same plug configurationto minimize spatial variability. Plug spacing = 33 cm. Plants sampledin the core are indicated in beige. See Table 3.2 for plug abbreviations. 41Figure 3.2 Collecting photographic data to analyze percent plant cover. A right–angled monopod was designed to take over–head photographs used toestimate plant cover in the seed application trial. The monopod wasraised and leveled with the camera on a delayed setting to capturea picture for cover estimation in the SamplePoint software. (PhotoTaken: September 2012) . . . . . . . . . . . . . . . . . . . . . . . . . 46xLIST OF FIGURESFigure 3.3 Percent AM fungal colonization of greenhouse grown plant plug roots.Plant plugs were randomly selected just prior to sowing plant plugs inthe field (June 2010). Ten plant plugs from each treatment level (± R.irregularis) of all eight species were analyzed for AM colonization ofroots using t-tests comparing inoculated and non–inoculated plants.Raw data ± 1 SD is presented in the graph. Each asterisk represents ap–value (***) < 0.001 for comparisons between inoculation treatmentlevels. Replication = 10. . . . . . . . . . . . . . . . . . . . . . . . . . 48Figure 3.4 AM fungal colonization of the mixed community of field roots in theplant plug trial. Panel (a) represents the graph of raw data with errorbars (± 1 SD) based on the most parsimonious linear mixed effectsmodel. Experimental treatment replication = 9. The left graph panelrepresents data after one growing season. Labels on the x–axis: None= no soil amendment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1biochar, 20CP = 20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar+ 20 T ha−1 compost, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1compost. Statistical output (b) shows significant main effect termsand interactions. Main effects included in the model were: Amend-ment, AM inoculation, Plot Height, and Year. % explained devianceis abbreviated as % Expl. Dev. in the output. Model terms withnegative regression slopes are indicated in parentheses. . . . . . . . . 49Figure 3.5 Predicted total plant biomass in the plant plug trial. Panel (a) rep-resents the graph of raw data with error bars (± 1 SD) based on themost parsimonious linear mixed effects model. Experimental treat-ment replication = 9. The left graph panel represents data afterone growing season. Labels on the x–axis: None = no soil amend-ment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP =20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar + 20 T ha−1 com-post, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1 compost. Statis-tical output (b) shows significant main effect terms and interactions.Main effects included in the model were: Amendment, AM inocula-tion, Plot Height, and Year. % explained deviance is abbreviated as% Expl. Dev. in the output. Note that model terms with negativeregression slopes are indicated in parentheses around the significancelevels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51xiLIST OF FIGURESFigure 3.6 Predicted Andropogon gerardii biomass in the plant plug trial. Panel(a) represents the graph of raw data with error bars (± 1 SD) basedon the most parsimonious linear mixed effects model. Experimentaltreatment replication = 9. The left graph panel represents data afterone growing season. Labels on the x–axis: None = no soil amend-ment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP =20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar + 20 T ha−1 com-post, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1 compost. Statis-tical output (b) shows significant main effect terms and interactions.Main effects included in the model were: Amendment, AM inocula-tion, Plot Height, and Year. % explained deviance is abbreviated as% Expl. Dev. in the output. Note that model terms with negativeregression slopes are indicated in parentheses around the significancelevels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52Figure 3.7 Predicted Lespedeza capitata biomass in the plant plug trial. Panel(a) represents the graph of raw data with error bars (± 1 SD) basedon the most parsimonious linear mixed effects model. Experimentaltreatment replication = 9. The left graph panel represents data afterone growing season. Labels on the x–axis: None = no soil amend-ment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP =20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar + 20 T ha−1 com-post, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1 compost. Statis-tical output (b) shows significant main effect terms and interactions.Main effects included in the model were: Amendment, AM inocula-tion, Plot Height, and Year. % explained deviance is abbreviated as% Expl. Dev. in the output. Note that model terms with negativeregression slopes are indicated in parentheses around the significancelevels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54xiiLIST OF FIGURESFigure 3.8 Predicted Desmodium canadense biomass in the plant plug trial. Panel(a) represents the graph of raw data with error bars (± 1 SD) basedon the most parsimonious linear mixed effects model. Experimentaltreatment replication = 9. The left graph panel represents data afterone growing season. Labels on the x–axis: None = no soil amend-ment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP =20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar + 20 T ha−1 com-post, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1 compost. Statis-tical output (b) shows significant main effect terms and interactions.Main effects included in the model were: Amendment, AM inocula-tion, Plot Height, and Year. % explained deviance is abbreviated as% Expl. Dev. in the output. Note that model terms with negativeregression slopes are indicated in parentheses. . . . . . . . . . . . . . 55Figure 3.9 Predicted Panicum virgatum biomass in the plant plug trial. Panel(a) represents the graph of raw data with error bars (± 1 SD) basedon the most parsimonious linear mixed effects model. Experimentaltreatment replication = 9. The left graph panel represents data afterone growing season. Statistical output (b) shows significant main ef-fect terms and interactions. Main effects included in the model were:Amendment, AM inoculation, Plot Height, and Year. % explaineddeviance is abbreviated as % Expl. Dev. in the output. Note thatmodel terms with negative regression slopes are indicated in paren-theses around the significance levels. . . . . . . . . . . . . . . . . . . 56Figure 3.10 Predicted Symphyotrichum laeve biomass in the plant plug trial. Panel(a) represents the graph of raw data with error bars (± 1 SD) basedon the most parsimonious linear mixed effects model. Experimentaltreatment replication = 9. The left graph panel represents data afterone growing season. Labels on the x–axis: None = no soil amend-ment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP =20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar + 20 T ha−1 com-post, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1 compost. Statis-tical output (b) shows significant main effect terms and interactions.Main effects included in the model were: Amendment, AM inocula-tion, Plot Height, and Year. % explained deviance is abbreviated as% Expl. Dev. in the output. Note that model terms with negativeregression slopes are indicated in parentheses around the significancelevels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57xiiiLIST OF FIGURESFigure 3.11 Predicted Liatris cylindracea biomass in the plant plug trial. Panel(a) represents the graph of raw data with error bars (± 1 SD) basedon the most parsimonious linear mixed effects model. Experimentaltreatment replication = 9. The left graph panel represents data afterone growing season. Labels on the x–axis: None = no soil amend-ment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP =20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar + 20 T ha−1 com-post, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1 compost. Statis-tical output (b) shows significant main effect terms and interactions.Main effects included in the model were: Amendment, AM inocula-tion, Plot Height, and Year. % explained deviance is abbreviated as% Expl. Dev. in the output. Note that model terms with negativeregression slopes are indicated in parentheses around the significancelevels. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58Figure 3.12 Raw data wireframe graph (a) of total native plant cover in the seedapplication trial based on the most parsimonious linear mixed effectsmodel. Panels represent the three analyzed growing seasons (Fall2011–Fall 2013). The gradient bar on the left indicates increasing% cover from magenta → cyan. Results are based on the most par-simonious statistical model. Significance levels and interactions forthe model terms are given in the statistical output table(b). y–axis= % total plant cover; x–axis = biochar rate, z–axis = compost rate.AM fungal inoculation and relative plot height are not included in thegraph due to visual complexity. % explained deviance accounts for theproportion of variation explained by each model term. Replication = 1. 60Figure 3.13 Diagnostic boxplots and a scatterplot for each main model term an-alyzing total native plant cover when all other factors were held con-stant in the seed application trial. Panels A–D are boxplots represent-ing the raw data distribution for each categorical model term includedin the linear mixed effect model. Panel E is a scatterplot of the rela-tive plot height in meters compared to total native plant cover. Thesurveyed plots with relative plot height values closer to zero are higheron the landscape. At the field site, surface soils of plots higher on thelandscape were observed to dry more rapidly than plots lower on thelandscape. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61xivLIST OF FIGURESFigure 4.1 A priori hypotheses used to construct the most parsimonious struc-tural equation soil food web model (Model 3). Exogenous variables aredisplayed in shaded gray boxes. Endogenous variables are displayedin white boxes. The residual error associated with each endogenousvariable is displayed as (ε). Single headed arrows indicate direct path-ways. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76Figure 4.2 Bacterial biomass collected during the second growing season of theprairie restoration (September 2012). Data were analyzed with linearmodels to test treatment–level effects. Panel (a) represents raw data± 1 SD; n = 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1biochar, 10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC+20CP = 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP =10 T ha−1 biochar + 20 T ha−1 compost. Significant main effect termsand interactions shown in (b). Model term estimates represent theexpected change from the model intercept (i.e. control plots). . . . . 78Figure 4.3 Fungivorous nematode abundance collected during the second season(September 2012). Generalized linear models with a negative bino-mial distribution link function were used to test the treatment ef-fects. Panel (a) represents raw data ± 1 SD; n = 9. x–axis: None= no soil amendment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1biochar, 20CP = 20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar+ 20 T ha−1 compost, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1compost. Significant main effect terms and interactions shown in (b).Model estimates represent the expected change from the intercept. . 79Figure 4.4 Collembola abundance collected during the second growing season ofthe prairie restoration (September 2012). Generalized linear mod-els with a negative binomial distribution link function were used totest the treatment effects. Panel (a) represents raw data ± 1 SD; n= 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1 biochar,10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP= 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1biochar + 20 T ha−1 compost. Significant main effect terms and in-teractions shown in (b). Model term estimates represent the expectedchange from the model intercept (i.e. control plots). . . . . . . . . . 80xvLIST OF FIGURESFigure 4.5 Structural equation soil food web model for the plant plug experi-ment. Exogenous variables are displayed in shaded gray boxes. En-dogenous variables are displayed in white boxes. The residual errorassociated with each endogenous variable is displayed as (ε). Struc-tural equation model line weights are scaled to the direct pathwaystandardized regression estimates given in each boxes. Blue (positive)and red (negative) arrows indicate significant standardized regressionestimates (p < 0.05). Yellow (positive) and orange (negative) arrowsindicate trends in standardized regression estimates (0.05 < p < 0.1).Dashed lines are non–significant paths with standardized regressionestimates > 0.1. Regression estimates < 0.1 are not included to sim-plify the data presentation. A full description of direct, indirect, andtotal model estimates are given in Table 4.1 – 4.3. Squared multiplecorrelations are reported within endogenous variable boxes. Squaredmultiple correlations were calculated for each endogenous variable todetermine explained variance. . . . . . . . . . . . . . . . . . . . . . . 81Figure 4.6 Fungal biomass collected during the second growing season of theprairie restoration (September 2012). Data were analyzed with linearmodels to test treatment–level effects. Panel (a) represents raw data± 1 SD; n = 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1biochar, 10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC+20CP = 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP =10 T ha−1 biochar + 20 T ha−1 compost. Significant main effect termsand interactions shown in (b). Model term estimates represent theexpected change from the model intercept (i.e. control plots). . . . . 84Figure 4.7 Fungal:bacterial biomass ratios collected during the second growingseason of the prairie restoration (September 2012). Generalized linearmodels with a negative binomial distribution link function were usedto test the treatment effects. Panel (a) represents raw data ± 1 SD;n = 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1 biochar,10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP= 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1biochar + 20 T ha−1 compost. Significant main effect terms and in-teractions shown in (b). Model term estimates represent the expectedchange from the model intercept (i.e. control plots). . . . . . . . . . 85xviLIST OF FIGURESFigure 4.8 Bacteriovorus nematode abundance collected during the second grow-ing season of the prairie restoration (September 2012). Generalizedlinear models with a negative binomial distribution link function wereused to test the treatment effects. Panel (a) represents raw data ±1 SD; n = 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1biochar, 10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC+20CP = 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP =10 T ha−1 biochar + 20 T ha−1 compost. Significant main effect termsand interactions shown in (b). Model term estimates represent theexpected change from the model intercept (i.e. control plots). . . . . 86Figure 4.9 Predatory nematode abundance collected during the second growingseason of the prairie restoration (September 2012). Generalized linearmodels with a negative binomial distribution link function were usedto test the treatment effects. Panel (a) represents raw data ± 1 SD;n = 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1 biochar,10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP= 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1biochar + 20 T ha−1 compost. Significant main effect terms and in-teractions shown in (b). Model term estimates represent the expectedchange from the model intercept (i.e. control plots). . . . . . . . . . 87Figure 4.10 Oribatid mite abundance collected during the second growing seasonof the prairie restoration (September 2012). Generalized linear mod-els with a negative binomial distribution link function were used totest the treatment effects. Panel (a) represents raw data ± 1 SD; n= 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1 biochar,10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP= 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1biochar + 20 T ha−1 compost. Significant main effect terms and in-teractions shown in (b). Model term estimates represent the expectedchange from the model intercept (i.e. control plots). . . . . . . . . . 88xviiLIST OF FIGURESFigure 4.11 Predatory mite abundance collected during the second growing seasonof the prairie restoration (September 2012). Generalized linear mod-els with a negative binomial distribution link function were used totest the treatment effects. Panel (a) represents raw data ± 1 SD; n= 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1 biochar,10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP= 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1biochar + 20 T ha−1 compost. Significant main effect terms and in-teractions shown in (b). Model term estimates represent the expectedchange from the model intercept (i.e. control plots). . . . . . . . . . 89Figure 4.12 Negative standardized regression estimates in the soil food web modelfor the grassland restoration plant plug experiment Exogenous vari-ables are displayed in shaded gray boxes. Endogenous variables aredisplayed in white boxes. Structural equation model line weights arescaled to the direct pathway standardized regression estimates givenin each boxes. Red arrows indicate significant standardized regressionestimates (p < 0.05) and orange arrows indicate trends in standardizedregression estimates (0.05 < p < 0.1). Dashed lines are non–significantpaths with standardized regression estimates > 0.1. Regression esti-mates < 0.1 are not included to simplify the data presentation. Afull description of direct, indirect, and total model estimates are givenin Table 4.1 – 4.3. Squared multiple correlations are reported withinendogenous variable boxes. Squared multiple correlations were calcu-lated for each endogenous variable to determine explained variance. . 90xviiiLIST OF FIGURESFigure 4.13 Positive standardized regression estimates in the soil food web modelfor the grassland restoration plant plug experiment. Exogenous vari-ables are displayed in shaded gray boxes. Endogenous variables aredisplayed in white boxes. The residual error associated with eachendogenous variable is displayed as (ε). Structural equation modelline weights are scaled to the direct pathway standardized regressionestimates given in each boxes. Blue arrows indicate significant stan-dardized regression estimates (p < 0.05) and yellow arrows indicatetrends in standardized regression estimates (0.05 < p < 0.1). Dashedlines are non–significant paths with standardized regression estimates> 0.1. Regression estimates < 0.1 are not included to simplify the datapresentation. A full description of direct, indirect, and total model es-timates are given in Table 4.1 – 4.3. Squared multiple correlationsare reported within endogenous variable boxes. Squared multiple cor-relations were calculated for each endogenous variable to determineexplained variance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92xixAcknowledgmentsI would like to express my gratitude to the faculty, staff, and students at the Universityof British Columbia and the University of Guelph who helped make this degree possible.In particular, the following people have led to the culmination of this degree:− I would like thank my supervisors, Drs. John Klironomos and Miranda Hart, for theirencouragement, advice, constructive criticism, and friendship throughout my degree.I would not have achieved this milestone and become the person I am today withoutyou.− To my committee members, Drs. Melanie Jones, David Scott, and Kari Dunfield, yourcomments, constructive criticisms, and support have been essential to the developmentand completion of this enormous undertaking.− To my field assistants, Andre Audet, Jeremy Booth, Nicola Day, Sarah Kruis, yourtireless efforts and enduring friendship during long days in the field have been essentialto my project’s completion. You are as much a part of this project as I am.− To Josh Stumpf, thank you for your support during the R learning curve. My gratitudefor your friendship and patience knows no bounds.− To the Mary Gartshore and Peter Carson, your knowledge of the restoration projectsand local flora has been invaluable when developing my research.− To the Nature Conservancy of Canada, especially Tom Bradstreet and Wendy Crid-land, thank you for donating the land to conduct my research and develop my project.− To TOARC, especially Danielle Solondz and David Sterrett, thank you for your fi-nancial support and editorial comments throughout this project.− I would like to thank Dr. Kari Dunfield, Mike Mucci, and Tannis Slimmon at theUniversity of Guelph and Dr. Scott Petrie at Bird Studies Canada. You provided mewith a scientific home base to conduct my research. I am eternally grateful.− Thank you to my parents, Christine Sullivan and Larry Ohsowski, for your continuedsupport as I have pursed the life of a perpetual student.xxDedicationTo my family, friends, and colleagues...xxiChapter 1IntroductionHuman–induced disturbance is pervasive among all ecosystems as the result of wasteaccumulation, industrial pollution, resource extraction, and urban sprawl (Hannah et al.1995). Previous land use, ranging from industrial spoils (e.g. mine tailings, contaminatedbrownfields) to road construction, dictates the approach of a restoration project (Jacksonand Hobbs 2009). For my purposes, I define a severely disturbed landscape as an areamanipulated in such a way that the pre–existing habitat can no longer be maintained. Iwill focus on the restoration of grassland vegetation and ecosystem processes after post–minesand extraction.The definition of restoration success is largely dependent upon the goals of the restora-tion practitioner. Goals can range from achieving diversity indices (e.g. organism richnessand abundance), vegetative structure (e.g. percent cover, biomass, vegetative profiles), orecosystem process reestablishment (e.g. nutrient cycling and soil stabilization) (Ruiz-Jaenand Aide 2005). The current paradigm in restoration tends to be phytocentric while under–emphasizing belowground food webs (van der Heijden et al. 2008, Kardol and Wardle 2010)and soil ecological knowledge (Callaham Jr. et al. 2008). Furthermore, restoration projectstend to evaluate short–term outcomes for vegetative and microbial production, as well assoil processes.Practitioners need viable techniques that influence the recovery of the entire ecosystem.After severe disturbance (i.e. post–mine areas), edaphic conditions and soil communitiesmay not support diverse plant communities. The addition of inoculum and soil conditionerscan address some components of the soil environment. Soil amendments should create moresuitable conditions for diverse and productive plant communities. With an ecosystem–level approach to restoration, native plant production is the consequence of the restorationpractice, not the focus.1.1 Landscape restoration and successional theory1.1.1 Historical contextHistorically, ecological succession has been viewed in terms of stable, climax communityendpoints (Clements 1916, Odum 1969). Current thought recognizes that community diver-11.1. Landscape restoration and successional theorysity is shaped by environmental fluctuations at large spatial, temporal, and organizationalscales (Pimm 1991). Successional pathways can be multi-directional, driven by stochasticprocesses and disturbance, thus long-term community stability will never be maintained(Glenn-Lewin and van der Maarel 1992). This implies that an ecosystem has multiple, al-ternative stable states separated by unstable transitions (Scheffer et al. 2001). Alternativestable states depend upon the surrounding biotic community, order of organism arrival,and inherent system randomness. In terms of restoration, degraded systems are often ina persistent stable state (Suding et al. 2004). Plant establishment and soil building in de-graded habitats may be slow to recover by natural successional processes without humanintervention.Restoration ecology and successional theory often address similar questions, albeit fromdifferent perspectives. Successional pathways are comprised of temporal changes in com-munity assembly, biodiversity, and biogeochemical cycles (Walker et al. 2007). Habitatrestoration manipulates these processes to accelerate target community establishment (Har-ris, 2009). Successional research, generally confined to one ecosystem, addresses time scalesrelated to vascular plant life history (10–200 years). In contrast, landscape restoration oper-ates on broad spatial scales (e.g. altitude gradients, moisture gradients, catchment basins)(del Moral et al. 2007), focusing on the duration of human involvement (120 years).A practical application of successional mechanisms in restoration has not been broadlydeveloped for practitioners (Walker et al. 2007). Restoration ecologists must acknowledgethe potentially persistent stable state of degraded systems. Feedback mechanisms betweenbiotic and abiotic factors in degraded systems may suppress plant establishment and com-munity sustainability. For example, Ash et al. (1994) described abandoned waste areas innorthwest England that had reduced plant cover and diversity after a century followingdisturbance.1.1.2 Post–mine areas as primary succession modelsAbandoned mine lands (e.g. ore extraction, gravel pits) are analogous to natural pri-mary succession events such as volcanic activity or glacial retreat. The extraction processcompletely removes flora, fauna, and soils of the previous system. Following resource ex-haustion, post–mine areas are typically characterized by low soil organic matter (SOM)content, low fertility, and poor physio-chemical and biological properties (Bradshaw 2000).The resulting raw substrate (i.e. subsoils and rock material) is a stark contrast to the abioticand biotic soil complexity of the original habitat. As a consequence, natural reestablish-ment of above- and belowground communities in abandoned mine areas is typically slow(Bradshaw 1997).21.2. An ecological context of degraded system restorationSuppressed regeneration of biotic communities may be due to reduced biological com-plexity in post–mine substrates. The deposition and subsequent heterotrophic turnover oforganic matter is a critical link for facilitating plant establishment. Restoration projects indegraded soils must include attempts to rehabilitate, at least in part, biological complexity.Biological colonization requires a source of energy and nutrients, which may be initiallylacking in post–mine substrates (Frenot et al. 1998). One solution is to add organic detrituscontaining natural microbial assemblages (e.g. bacteria (Tscherko et al. 2003, Bardgettet al. 2007), cyanobacteria (Nemergut et al. 2007), and fungi (Hodkinson et al. 2002)).These microbes actively turn over organic substrates and prime biogeochemical cycles.1.2 An ecological context of degraded system restorationClearly, soil health is paramount to restoration success in devastated landscapes. Soilmicrobial communities play a major role in the development and sustainability of soil health(Anderson 2003). Soil health is defined as the capacity of soil to function as a living system,sustaining biotic productivity, and maintaining ecosystem services (Doran and Zeiss 2000).Soil microbial communities are well correlated with plant primary production (Bardgett andWardle 2003, van der Heijden et al. 2008, Heneghan et al. 2008, Benayas et al. 2009) andintegral in the recycling of organic matter and nutrients (Wardle et al. 2002). Decomposers(Harte and Kinzig 1993, Reynolds et al. 2003), mycorrhizal fungi (Klironomos 2002) andnitrogen-fixing bacteria (van der Heijden et al. 2008) are key soil functional groups in therhizosphere (i.e. soil area directly influenced by plant tissues and secretions). Soil microbialcommunities are also important for soil stabilization via stable aggregate formation (Rillig2004, Six et al. 2004). These factors can ultimately mediate successional dynamics and plantcommunity composition (Wardle et al. 2004), thus contributing to the reestablishment ofnatural systems in severely disturbed landscapes.Edaphic characteristics, resource availability, and soil microorganisms mediate above-ground biotic responses to include primary productivity (Baer et al. 2004), organic matterdecomposition rates (Smith and Bradford 2003), and plant community structure (Baer et al.2003, Heneghan et al. 2008). CENTURY (Parton et al. 1993), an established ecosystem–level model of plant–soil biogeochemical cycles, models the links among plant productivity,decomposition, climate and land management options. Among its many functions, CEN-TURY emphasizes the role of carbon management decisions under natural and agriculturalscenarios. Restoration projects that appropriately manage soil organic matter dynamicsand soil microbial feedbacks may increase production and carbon storage in disturbed habi-tats (Ojima et al. 1993). Practitioners should emphasize soil carbon cycles and microbialprocesses in tandem with plant establishment in damaged ecosystems (Cairns 2000).31.3. Grassland vegetation1.3 Grassland vegetationGrassland productivity varies with habitat classification, ranging from shortgrass steppe(least productive) to tallgrass prairies (most productive) (Knapp and Smith 2001). Grass-land productivity is ultimately dictated by the availability of three limiting resources: light,water, and nitrogen (Baer et al. 2003). Resource availability is determined by patternsin precipitation (Sala et al. 1988), soil characteristics (Briggs and Knapp 1995), herbivory(Knapp et al. 1999), and periodic fires (Knapp and Seastedt 1986). Plant production ingrasslands will ultimately depend upon adaptations to spatial and temporal availabilitiesof these limiting resources.Grassland restoration in severely degraded habitats must recognize the factors thatshape and maintain these communities. Grassland plants are evolutionarily adapted to thementioned environmental context. Restoration projects incorporating locally adapted plantpopulations are more likely to improve rates of establishment and persistence (Pywell et al.2002). Resulting plant communities are expected to more closely resemble natural grasslandremnants and encourage the conservation of rare flora and fauna.Four functional groups composed of herbaceous perennials dominate grassland commu-nities: perennial C4 grasses, C3 graminoids (grasses and sedges), nitrogen-fixing species(primarily Fabaceae), and late summer flowering, drought-hardy composites (Asteraceae)(Kindscher and Wells, 1995). Cool season C3 grasses have traits that provide early seasonplant cover, nutrient-rich plant tissues beneficial to herbivores, and have decreased light re-quirements ideal for shady refugia. Compared to cool season grasses, warm season C4 grassesexhibit higher water-use efficiency, higher plant biomass potential, late season growth, andtolerance of full sun exposure (Tiessen et al. 1993). Composite forbs are integral in rapidlycolonizing open soil (especially after grazing or fire disturbances), supporting pollinatorpopulations, and driving overall plant community diversity indices (Pokorny et al. 2004).Forbs in the legume family (Fabaceae) form a symbiotic relationship with nitrogen-fixingbacteria. Nitrogen-fixing bacteria are found within legume root nodules, and convert biolog-ically unavailable atmospheric N2 gas into forms of nitrogen usable by plants. In exchangefor usable nitrogen, the plant delivers a supply of nutrition in the form of carbohydrates.Nitrogen-rich legumes within grasslands can contribute to the total nitrogen pool of soilsduring growth and after senescence (Oelmann et al. 2007). Soil nitrate and ammoniumlevels are usually limited within grasslands due to rapid utilization and immobilization byprimary producers and microbial decomposers (Risser and Parton 1982). The introductionof N-fixing plants may affect the structure and function of grassland systems.Restoration projects that incorporate multiple functional groups and high numbers ofspecies are more likely to achieve community sustainability (Piper and Pimm 2002). Long–41.4. Soil food websterm ecosystem stability depends on communities containing species or functional groupsthat are capable of differential response to disturbance (McCann 2000, Hooper et al. 2005).Studies of grassland ecosystems indicate that increased diversity can be expected, on av-erage, to give rise to resistance and resilience (Tilman et al. 1997, Tilman and Downing1994). Higher species diversity may also lead to increased plant production due to speciescomplementarity (Cardinale et al. 2007). Restoration projects that maintain high speciesdiversity with varied functional traits could increase the likelihood of achieving long-termcommunity stability.1.4 Soil food websMicrobial communities (i.e. bacteria and fungi) play a fundamental role in drivingbiogeochemical cycles in terrestrial ecosystems. Carbon cycling and plant nutrient avail-ability are dictated by bacterial and fungal communities, subsequently mediating plantproductivity and soil development in habitat restoration (Harris 2009). Fungivorous andbacteriovorus soil animals (i.e. grazing nematodes, Oribatid mites and Collembola) directlyor indirectly consume microorganisms embedded within organic matter, thus contributingto litter breakdown and soil mixing (Lavelle et al. 2006). Microorganisms associated withlitter have high nutritional value compared to detritus and become a critical food sourcethat links fungal and bacterial communities to soil animal abundance (Bardgett and Cook1998). In conjunction with abiotic conditions, food resource availability determines thepopulation size of soil animals within a soil food web (Ingham et al. 1985).Restoring a grassland ecosystem can be challenging in post–mine sandpits. Land man-agers often emphasize aboveground plant biodiversity when rehabilitating sandpits whilelargely ignoring the contribution of soil biota to plant community productivity. Adopting aholistic restoration strategy that focuses on soil recovery can positively influence the restora-tion trajectory of a degraded area (Heneghan et al. 2008). Thus, re–establishing a functionaldetrital food web is an essential component of recovering soils in severely degraded systems.Mine activities, such as sandpit excavation, disrupts and diminishes the multi–trophicinteractions among soil biota, consequently reducing the beneficial ecosystem services asso-ciated with soil food webs (de Vries et al. 2012, Arau´jo et al. 2013, Zhao and Neher 2013). Inthe case of aggregate extraction sites, substrate conditions are stressful as these systems lackhigh concentrations of essential nutrients, soil organic matter, and a large water–holdingcapacity. Thus, the growth of all organisms is often restricted and the ecological connectionsbetween soil-plants-microbes are usually severed (Maiti 2013). Successfully restoring soilfood webs in conjunction with plant communities depends on alleviating abiotic stress inmine substrate (McKinley et al. 2005).51.4. Soil food websTo address the depauperate conditions of post–mine areas, reclamation tools such assoil amendments (i.e. compost and biochar) and microbial inoculants (i.e arbuscular my-corrhizas) are often required to increase plant production (Refer to Chapters 1 and 3).Incorporating organic soil amendments and arbuscular mycorrhizal inoculants strengthensthe feedback links among plants-soils-microbes to ultimately re–establish decomposition cy-cles and accelerate soil development (Elkins et al. 1984, Ros et al. 2003). To accomplishthis, researchers must recognize the links among restoration protocol (i.e. amendment ap-plication), soil microorganisms, soil animals, and ecosystem functioning when revegetatingseverely disturbed areas (Coleman and Whitman 2005). To date, an explicit protocol usefulto land managers that will increase soil microbial biomass and soil animal abundance inpost–mine grassland restoration does not exist.1.4.1 Soil microbial communities as indicators for post–mine substraterecoveryFungi and bacteria are key decomposers in soils that are responsible for nutrient avail-ability, nutrient transformations, and litter breakdown. Therefore, estimating the biomassof these microbial constituents in recovering soils are a proxy to soil function (Visser andParkinson 1992, Karlen et al. 1997) and can be used to evaluate ecological restoration soils(Harris 2009). Assessing the biomass of microbial assemblages elucidates the stage of soildevelopment and food resource availability for grazing soil animals in the soil food web.The ratio of fungal and bacterial biomass can be a useful tool to assess soil develop-ment status. Severe anthropogenic disturbance, such as mining, often shifts the dominanceof soil microbial communities from fungal–dominated to bacterial–dominated assemblagesdue to poor physiochemical conditions in recently exposed substrates (Frey et al. 1999,Bailey et al. 2002, Mummey et al. 2002). Bacterial–based soil food webs are common inecosystems with poorly developed, low organic matter soils such as mined or conventionalagriculture landscapes (Kardol and Wardle 2010). In contrast to mine systems, naturalgrassland soils are dominated by fungal decomposers due to the higher volume of com-plex organic matter from nutrient–rich litter inputs (Bardgett and McAlister 1999, Harris2009). Reduced disturbance facilitates an extensive hyphal fungal network, allowing fungito access spatially separated limiting nutrients in the soils via fungal translocation (Beareet al. 1992). Restoration ecologists should target fungal–dominated systems to indicate asuccessful grassland restoration (Bardgett and McAlister 1999, Smith et al. 2003).61.4. Soil food webs1.4.2 Soil animals as indicators for post–mine substrate recoverySoil animals (i.e. nematodes, Collembola, mites) are ecosystem engineers that enhancebiological, chemical, and physical soil properties that benefits the growth of plants. Lavelleet al. (2006) suggested that soil animals enhance nutrient release in the plant rhizosphere,stimulate mutualistic associations, and positively affect soil physical structure. After asevere disturbance, soil animal communities are removed or severely reduced (Curry andGood 1992), thus the ecosystem services provided by these soil organisms are non-existent.Restoration practitioners should create soil recovery plans that promote high densities of di-verse soil animals to facilitate multi–trophic interactions among soil microbial communitiesand soil animals.Multi–trophic interactions in soil food webs are based on soil animal feeding preferences.Grazing soil animals (i.e. bacteriovorus and fungivorous nematodes, Collembola, and Orib-atid mites) depend upon soil microorganisms as a food resource. Grazing nematodes altersoil nutrient cycles and influence organic matter decomposition by ingesting large quantitiesof fungal and bacterial communities residing in plant litter (Yeates et al. 1993). The domi-nant microbial community in soils has been shown to determine the abundance of nematodefeeding groups. Greater bacterial production supports mainly bacteriovorus fauna (Hendrixet al. 1986) while fungivorous feeding soil animals are expected to thrive in fungal–rich soil(Beare et al. 1992). Comparatively, Collembola and Oribatid mites feed on soil microbialcommunities but also ingest litter, influencing microbial populations and litter turnoverrates in soil systems (Ha¨ttenschwiler et al. 2005, Frouz et al. 2006). Collembola, often con-sidered consumers of fungi and bacteria, are also known to be predatory on rotifers withsome species consuming nematodes when available (Wallwork 1976, Lee and Widden 1996).Thus, the activities of grazing soil animals can ultimately influence nutrient cycles and litterretention in soil by regulating microbial decomposition rates in a restoration project.Predators in the soil food web can have a top–down trophic cascade effect on the micro-bial production at the base of the soil food web (Lenoir et al. 2007). Predatory nematodesfeed upon lower trophic levels such as rotifers, protists, and other soil nematodes. Grazingnematode abundance has been shown to be reduced by the feeding activities of predatorynematodes, ultimately influencing microbial productivity (Laakso and Seta¨la¨ 1999). Fur-thermore, fungal activity has also been shown to increase when predatory mites consumeCollembola in a tri–trophic interaction study (Hedlund and O¨hrn 2000). Overall, the com-plex multi–trophic interactions in soil food webs can indirectly influence soil function andlitter decomposition rates in restoration project via consumption of grazing soil animals.Soil food webs are important indicators for determining restoration success and soildevelopment trajectories. These organisms are invaluable to a restoration project becauseof the ecosystem services provided by the activities of soil animals. Therefore, soil animal71.5. Techniques for improving disturbed soils in grassland restoration projectsabundance has been used as indications of soil quality as they are sensitive to disturbanceand land management practices (Wardle et al. 1995, Roub´ıc˘kova´ et al. 2013). Nematodesare easily collected, respond rapidly to environmental change, and can be easily sortedinto functional feeding groups based on morphology. Nematodes have been successfullyused to evaluate mine area recovery in a variety of disturbance scenarios (Biederman et al.2008, Ha´neˇl 2008, Courtney et al. 2011). Soil microarthropods are also useful indicators ofsuccessional stage and soil system recovery as population growth is reliant on food resourcesin their immediate environment (Ferris et al. 2001, Parisi et al. 2005). Thus, soil animalabundance should be incorporated into the restoration of degraded systems to assess thedevelopment trajectory of soils.1.5 Techniques for improving disturbed soils in grasslandrestoration projectsThe positive effects of soil amendments on plant and microbial production within agri-cultural systems, restoration projects, and greenhouse experiments have been extensivelyrecognized. In the following section, I review three typical amendments that are widely ac-cessible to restoration practitioners. These amendments (i.e. biochar, compost, arbuscularmycorrhizas) have had promising results both in greenhouse and field settings.1.5.1 Vegetation–derived biocharApplication of black carbon to soils is expected to build soil organic matter, enhancenutrient biogeochemical cycles, lower bulk soil density, increase bio–available water, andreduce nutrient leaching (Shrestha et al. 2010). Black carbon consists of all C rich residues,ranging from partly charred material to graphite and soot particles, resulting from theincomplete combustion of organic materials (Schmidt and Noack 2000). Research has shownthat prairie soils contain substantial amounts of black carbon resulting from a 10,000 yearlegacy of prairie fires (Skjemstad et al. 2002, Brodowski et al. 2005). Laird (2008) estimatesthat between 5% and 15% of the total organic carbon in natural Midwestern prairie soils iscomposed of black carbon. Within boreal forests, short–term soil fertility effects have beenattributed to increased charcoal fractions in the soil after naturally occurring fires (Wardleet al. 1998).Historically, human agricultural practices (i.e. terra preta soils in the central Amazon)have long recognized plant growth benefits of black carbon soil supplements (Glaser et al.2002) . Terra preta literally translates to black earth in Portuguese. These ancient soils(500–7,000 YBP) have been anthropogenically amended with black carbon, bones, and ma-81.5. Techniques for improving disturbed soils in grassland restoration projectsnure. Compared to adjacent infertile soils (terra comum or common soils), the concentrationof black carbon in terra preta soils is seventy times greater. Furthermore, these soils stillexhibit three times more soil organic matter, nitrogen, and phosphorus in comparison toneighboring terra comum soils (Glaser 2007).One form of refined black carbon being used in environmental management is biochar, orcarbon–rich charcoal (Lehmann et al. 2009). To create biochar, organic materials (i.e. feed-stocks) are heated to temperatures between 300 oC and 800 oC in a low oxygen environment.Anoxic conditions during heating leads to the incomplete combustion of the organic matter,thus producing biochar. Feedstocks may include agricultural wastes, forestry wastes, woodpellets, or manures. The high temperatures used in pyrolysis induce molecule polymer-ization within feedstocks to produce aromatic and aliphatic compounds (Sohi et al. 2009).This creates a stable product demonstrated to be a long–term carbon storage pool for atmo-spheric CO2 in addition to being a beneficial soil amendment (Lehmann et al. 2006). Whenincorporated into soils, initial degradation of biochar by chemical oxidation and microbialprocesses has been noted (Bruun et al. 2008, Nguyen et al. 2008). The recalcitrant proper-ties of black carbon stocks eventually stabilize and resist microbial degradation within soilsfor 100–1000+ years (Glaser et al. 2002).Amended soils benefit from biochar’s large, oxidized surface area and porous structure.Soils amended with biochar have an increased soil charge density (potential cation exchangecapacity [CEC] per unit surface area) in comparison to non–amended soils (Liang et al.2006). Biochar improves: (1) soil nutrient availability and retention (i.e. major cations,phosphorus, total nitrogen) (Lehmann et al. 2003), (2) acidic soil conditions, (3) organicmatter adsorption (Shrestha et al. 2010), and (4) soil aeration (Shrestha et al. 2010).Biochar as a soil amendment has generated promising results within agricultural systemsand greenhouse experiments. Recent research has demonstrated that biochar amended soilshave greater crop biomass (Rondon et al. 2007, Major et al. 2010) and enhanced biologicalN–fixation in leguminous crops (Rondon et al. 2007). A meta–analysis by Biederman andHarpole (2012) shows that biochar increases aboveground productivity, crop yield, soil mi-crobial biomass, rhizobia nodulation, and soil nutrients compared to controls. The fertilizereffect induced in plants may be explained by the retention of beneficial nutrients and pHneutralization.Indirectly, plant growth may be stimulated by increased mycorrhizal associations (Nishioand Okano 1991, Ishii and Kadoya 1994) and soil microorganism activity (Thies and Rillig2009). Warnock et al. (2007) proposed four mechanisms that may benefit arbuscular myc-orrhizal fungi in soils with biochar: (1) positively changing physio–chemical soil properties(i.e. CEC, bioavailability of phosphate [PO4−] in low P soils), (2) promoting beneficialsoil organisms (i.e. phosphate solubilizing bacteria, mycorrhization helper bacteria), (3)91.5. Techniques for improving disturbed soils in grassland restoration projectsadsorbing plant secretions that may alter mycorrhizal root colonization, and (4) providinga grazing refuge in biochar’s porous structure. In general, increased soil microbial activityin biochar amended soils may also be attributed these hypothesized mechanisms for AMfungi.Biochar soil amelioration in severely degraded landscapes has the potential to increasegrassland plant production, enrich soil microbial populations, and stimulate arbuscularmycorrhizal persistence. Biochar is hypothesized to reduce nutrient leaching in well–drainedsoils. Nutrient retention in impoverished post–mine substrates should increase productivityby stimulating biotic–abiotic feedbacks.1.5.2 Leaf and yard waste compostAgricultural societies have historically recognized that ameliorating fields with compostresults in improved soil conditions. Soil disturbance (e.g. mining or tillage) generally de-creases SOM pools due to erosion and disruption of the biogeochemical mechanisms and mi-crobial communities associated with SOM pools (McLauchlan 2006). When added to soils,composted material increases soil fertility by increasing: (1) soil organic carbon (Crecchioet al. 2004, Walter et al. 2006), (2) available soil nitrogen (Eriksen et al. 1999, Gabrielleet al. 2005), phosphorus (Wortmann and Walters 2007), and micronutrients (i.e. iron, cop-per, zinc)(Hargreaves et al. 2008), (3) water holding capacity (Movahedi-Naeini and Cook2000), (4) cation exchange capacity (McConnell et al. 1994), (5) soil aggregation (Bressonet al. 2001, Annabi et al. 2007, Abiven et al. 2009), and (6) neutralization of acid soils(Mkhabela and Warman 2005).Leaf and yard waste (LYW), largely composed of community organic waste, is typicallycomposted at large scales. During aerobic LYW composting, theromphilic microbes assim-ilate and mineralize complex organic compounds while releasing heat, water vapor, CO2,and ammonia waste products. The remaining non–mineralized organic material is humifiedto form the stable end product, compost. Civic and environmental benefits of compostingLYW include waste volume reduction, microbial pathogen and weed sterilization (due tohigh temperatures), and odor suppression (Jakobsen 1995). LYW compost derived frommunicipal processing facilities is utilized in gardens, organic agriculture, land reclamation,and slope stabilization projects.Research demonstrates direct increases to crop biomass (Montemurro et al. 2006) andnutritional quality (Allievi et al. 1993) in compost amended soils. Compost addition stronglyinfluences soil microbial communities by increasing microbial biomass, respiration rates,and soil enzyme activity (Albiach et al. 2001). As bacterial and fungal decomposers utilizeand sequester carbon in amended soils, concentrations of total nitrogen and phosphorusincrease over time (Iglesias-Jime´nez 2001, Wolkowski 2003). Long–term ramifications of101.5. Techniques for improving disturbed soils in grassland restoration projectsmicrobial community activity (Ros et al. 2006) and soil biochemical characteristics (Garc´ıa-Gil et al. 2004) due to compost ameliorations have been noted. Pascual et al. (1999)found that microbial biomass, soil basal respiration, and dehydrogenase activity recovered tolevels similar to adjacent Mediterranean soils eight years after a single compost amendment.Within restoration projects specifically, compost bolstered arbuscular mycorrhizas inoculumpersistence, thus benefiting native plant cover (Noyd et al. 1996, Celik et al. 2004). In semi–arid soil restorations, extensive work from Caravaca et al. suggest short–term (Caravacaet al. 2002b;a; 2003b) and medium-term (Caravaca et al. 2003a) influences of mycorrhizalinoculations and compost ameliorations. Sharp increases in plant primary production wereattributed to the abiotic–biotic link between bioavailable phosphorus supplied by compostresidues and AM fungal phosphorus uptake.In sandy soils with low SOM, compost improves soil structure (Wahba and Darwish2008), bioavailable nutrients (P, K, Mg) (Weber et al. 2007), total inorganic N (Busby et al.2007), plant production (Mkhabela and Warman 2005), and soil microbial activity(Ros et al.2006). Low SOM and poor physio–chemical properties in post–mine substrates are expectedto have restricted microbial community activity and depleted nutrients. LYW compostamendments should increase microbial activity (Ros et al. 2003), mycorrhizal persistence(Gaur and Adholeya 2005), and increase plant biomass. To date, few studies have researchedthe effect of compost application to native plants and mycorrhizal communities in severelydegraded post–mine substrates (Busby et al. 2007).1.5.3 Arbuscular mycorrhizal fungal inoculation of grassland plantsArbuscular mycorrhizal (AM) fungi are globally distributed soil microorganisms thatform symbiotic associations with more than 80% of terrestrial plants (Smith and Read 2008).These obligate biotrophs constitute a major fraction of the plant–associated soil microbialcommunity. In exchange for host plant–derived photosynthate, arbuscular mycorrhizasbenefit plants by: (1) increasing soil nutrient acquisition and subsequent assimilation intoplant tissues (especially phosphorus), (2) protecting target plant roots from pathogens, (3)enhancing seedling performance, (4) improving plant water relations, and (5) improving soilstabilization. In addition to improved target plant performance, AM fungal communitiesdirectly relate to the biodiversity of plant communities (van der Heijden et al. 1998).Positive plant growth responses to mycorrhizas have stimulated the emergence of biotechcompanies promoting the use of commercially–produced AM fungal inoculum as a soil en-hancement agent. In horticultural systems (Azco´n-Aguilar and Barea 1997) and landscaperestoration (Miller and Jastrow 1992), mycorrhizal inoculum has been recommended to in-crease plant growth performance. The intentional movement of mycorrhizal fungal speciesis growing, but the potential negative ecological ramifications of non–native arbuscular my-111.6. Review conclusionscorrhizal invasion are poorly understood (Schwartz et al. 2006). Evidence indicates thatsymbiotic associations between plants and fungus range from parasitic to beneficial depend-ing on host plant/AM fungal pairings (Klironomos 2003). Depending on the biogeochemicalcontext and AM fungal–plant associations within an ecosystem, AM fungal inoculation mayyield positive, neutral, or negative plant growth effects in the field. Furthermore, a recentgreenhouse study by Mummey et al. (2009) indicated that plant pre–inoculation with AMfungi may have unintended implications for resident AM fungal communities. AM fungalinoculum may restrict assembly potentials in resident soil AM fungal communities withdivergent phylogenies, thus suppressing plant growth and foliar nutrients. As research in-dicates AM fungal phylogeny diversity in host plant roots directly correlates to increasedplant growth responses (Maherali and Klironomos 2007), restricting native soil inoculumpotentials could have ramifications to plant production and soil feedback mechanisms in arestoration project.The ramifications of pre–inoculating native plants with AM fungal inoculum in severelydegraded habitats have not been thoroughly addressed. Evidence indicates that after majorsoil disturbances such as agricultural tilling, native AM fungal associations are fracturedand strongly diminished (Jansa et al. 2002; 2003) . The resident AM fungal communityin post–mine substrates is expected to be strongly reduced compared to natural grasslandsoils. To date, the resident AM fungal community soil inoculum potential within post–minesubstrates has not yet been identified. The AM fungal inoculum potential in severely dis-turbed sites should be determined by spore immigration rates, soil nutrient availability (i.e.phosphorus availability), plant identity (i.e. obligate mycorrhizal plants vs. facultative my-corrhizal plants vs. non–mycorrhizal plants) in the degraded area, and time since landscapedisturbance (Allen and Allen 1980).To date, some AM fungal inoculation research has been conducted in non-toxic post–mine reclamation areas. These field studies indicate that AM fungi benefit native plantproduction and establishment in severely degraded areas (Johnson 1998, Matias et al. 2009).Mycorrhizal inoculum is anticipated to benefit plant production in post–mine substrates dueto a lack of an existing AM fungal community.1.6 Review conclusionsIt is imperative that restoration practitioners integrate soil ecological knowledge intothe reclamation of degraded habitats. Emphasizing an ecosystem–level approach to grass-land restoration in degraded areas should reduce landscape recovery time and reduce plantfailure. When used in combination, the addition of mycorrhizal fungi, biochar, and compostapproaches the goal of a viable soil environment for sustainable plant growth.121.7. Research objectivesIt is clear that soil amendments are necessary to restore severely disturbed landscapes ina reasonable time–frame. A checklist or key could be developed to facilitate identification offactors that are important for determining the most appropriate amendments and practices.Application rates of biochar and compost could be determined experimentally to establisha feasible restoration protocol under a variety of restoration scenarios. Since universalapplication of soil microorganisms may not always be beneficial, more studies testing the useof locally bolstered inoculum sources should be conducted to eliminate the environmentalimpact of foreign inoculum.The list of amendments discussed is by no means complete. Other amendments mayinclude inoculation (e.g. nitrogen–fixing bacteria, earthworms), and organic materials (e.g.biosolids, hydrogels, paper mill sludge). Further research into the integration of theseamendments into severely degraded landscapes during restoration projects needs to be con-ducted. As we make advancements in biotechnology and soil conditioners, we can reducelong–term maintenance costs and create a foundation for sustainable above– and below-ground communities.1.7 Research objectivesThe research conducted for my Ph.D. dissertation tests the efficacy of industrially feasi-ble rates of soil amendments (i.e. compost, biochar) and a commercial AM fungal inoculum(R. irregularis) on grassland plant restoration and soil food web development in a post–extraction sandpit. To accomplish this, I installed a large–scale restoration research sitenear Port Rowan, Ontario, Canada in the summer of 2010. Two planting strategies wereimplemented: greenhouse grown plant plugs and direct seeding. In September of each year,plant response data was collected for two growing seasons in the plant plug trial (2011–2012) and three growing seasons in the seed application trial (2011–2013). A soil food webanalysis was conducted for data collected from the plant plug experiment in September2012. The results of this dissertation will directly inform land management protocol whenrestoring grassland plants in post–mine sandpits. The three main research objectives of thisPh.D. dissertation are as follows:Objective #1 Develop a minimally destructive statistical method to increase mea-surement accuracy and reduce data collection time when estimating aboveground plantbiomass. I hypothesized that plant biomass predictive models using multiple mor-phological plant traits would be more accurate and robust compared to single planttrait model estimates. The rationale for this hypothesis that increased informationwould be acquired on plant morphology in the field, thus leading to higher predictionaccuracy for each plant species.131.7. Research objectivesObjective #2 Determine the multi–year plant response of both planting strategiesto soil amendments and the commercial AM fungal isolate in a post-mine sandpit.In both experiments, I hypothesized that: (a) AM fungal inoculation, compost, andbiochar addition would increase total plant community mass in the plug trial andvegetative cover in the direct seeding trial compared to non–amended controls, and(b) plots with the highest rates of compost + biochar and inoculated with the AMfungal isolate will yield the highest plant responses. The rationale for these hypotheseswas that water and nutrient stress of the plants would be ameliorated when the minesubstrate is ameliorated. When all amendments are combined, maximal plant responsewas anticipated due to higher nutrient inputs and retention from the soil amendmentsand increased nutrient acquisition due to the plant–fungal symbiosis.Objective #3 Determine the soil food web response to the addition of soil amend-ments and a AM fungal isolate in sandpit substrate. I hypothesized that: (a) addingcompost, biochar, and AM fungal inoculum singly would increase soil microbialbiomass (i.e. bacteria and fungi) and soil animal abundance (i.e. nematodes, Collem-bola, and mites), and (b) co–amending sandpits with compost, biochar, and AMfungal inoculum will show the greatest response in soil biota. The rationale for thesehypotheses was that nutrient stress would be alleviated due to increased belowgroundproduction and rhizosphere activity, and organic matter in sandpit substrate.14Chapter 2Improving Plant BiomassEstimation2.1 BackgroundAboveground plant biomass is an important measurement relevant across multiple dis-ciplines. Plant biomass is often considered a good approximation of productivity, especiallyin grassland communities (Hector et al. 1999). Directly measuring plant biomass, however,requires destructive sampling, thus severely disrupting the plant community of interest andburdens the researcher with a large labor cost. Harvesting, drying, and weighing a largevolume of plants restricts a researcher’s ability to minimize plant community destruction,rapidly collect a high volume of data, and track plant production of an individual at multipletime points (See Table 2.1 for several detailed scenarios). Therefore, minimally destructiveestimation methods are useful when data collection is time sensitive, labor force constraintsexist, large–scale plant harvesting is impractical, and sampling design requires repeatedmeasures (Catchpole and Wheeler 1992).Minimally destructive measures have been suggested as alternatives to harvesting plantbiomass, with varying success. Techniques include measuring community attributes per unitarea (e.g. percent cover, point intercept transects, photographic image analysis (Byrne et al.2011)) and individual estimates of plant mass (e.g. simple linear regression estimates andallometric equations). However, such approaches tend to have high variability and inherentsubjectivity, reducing the predictive power of the models. For example, several authors havefound good correlations between biomass and point–intercept methods but high variabilitystill exists depending upon sampling intensity (Jonasson 1988, Glatzle et al. 1993, Vittoz andGuisan 2007). This experimental error due to the sampling method represents potentiallyimportant biotic variation that is unaccounted for in a study.152.1. BackgroundTable 2.1: Experimental or observational situations to employ non–destructivebiomass estimation.Annual Net Primary Production (ANPP):Plant species respond differently to seasonal and inter-annual environmental varia-tion resulting in different biomass maxima throughout the growing season (Briggsand Knapp 1995, Fay et al. 2003, Moya-Larao and Corcobado 2008). MeasuringANPP from one or two time points, such as mid/ late season community biomass,underestimates ANPP due to fluctuations in biomass by different species throughoutthe season. With minimally destructive methods, researchers track plant speciesbiomass throughout the year to accurately determine ANPP of individual plantspecies without severely disrupting the plant community.Repeated Measures / Time Series Experiments:Researchers are interested in spatial and temporal aspects of plant growth in a time–series. Phenological responses of individual plants can be tracked over the durationof an experiment with minimally destructive prediction methods. The establishmentof a standard curve to predict biomass at a plant’s life stage would allow for high–throughput sample replication while being minimally invasive.Sampling Accuracy:Increased accuracy and precision in biomass estimates leads to higher confidence inresults. Increased measurement accuracy reduces variability in dataset predictions,thus reducing statistical noise due to prediction error. Error reduction enhances theability to detect discrete differences among experimental treatments and controls.2.1.1 Techniques to predict plant biomassTwo main methods have been used to predict biomass of a plant individual using mini-mally destructive techniques. The first approach uses published allometric equations devel-oped mainly to estimate tree biomass. The second approach creates a predictive standardcurve from the relationship between a measured plant trait such as plant height and plantmass. These predictive methods have been shown to be unreliable in their ability to con-sistently predict plant biomass. Thus, I introduce a new multivariate statistical approach,partial least squares (PLS) regression, to increase predictive accuracy of estimating plantmass in the field.Historically, allometric equations have been used to estimate plant biomass in the field.Allometric equations are mathematical functions published in the literature that are oftenlimited to the estimation of woody plant biomass. These equations are constructed usingeasily measured predictor variables such as diameter at breast height and total plant heightto estimate biomass (Picard et al. 2012). The advantage of allometric equations is that a162.1. Backgroundresearcher can apply these equations to estimate tree biomass without creating a standardcurve. However, allometric equations developed to measure herbaceous plants are relativelyscarce and can be highly variable in prediction accuracy when available (e.g. (Elliott andClinton 1993)).When using published allometric equations, a researcher must be aware of variabilityin plant morphology due to experimental treatments and local environmental conditionsthat will reduce prediction performance. In addition, allometric equations may not mea-sure predictor variables that optimally estimate plant biomass (Chave et al. 2004). Thus,published allometric equations are not readily usable when measuring plants under uniqueexperimental conditions.Alternatives to allometric equations have been used by a measuring predictor variableregressed against plant mass (Catchpole and Wheeler 1992). Plant traits that describe astrong linear relationship between plant biomass and a predictor variable are measured tocreate a standard curve. Once a standard curve is established, rapid measurement of similarvegetation is straightforward. When using this method, the researcher must balance theprecision of the mathematical relationship, conform to statistical assumptions, and weighthe costs associated with direct versus indirect measurements.When establishing a standard curve for predicting plant biomass, the strength of therelationship between a plant trait and harvested biomass determines the predictive perfor-mance of the model. To achieve the most accurate prediction, the standard curve must meetthe following assumptions: a linear relationship, equal residual variance (i.e. homoscedas-ticity), normal distribution of residuals, no highly influential outliers, and no strong multi-collinearity among predictor variables (Zuur et al. 2007). Multiple regression models wouldincrease plant estimation accuracy by taking into account several plant traits but must beapproached with caution. When predicting plant biomass, morphological plant traits are of-ten strongly correlated, violating the assumption of predictor independence and potentiallyreducing predictive power of the external samples (Graham 2003).Partial least squares regression (PLS) is a statistical method commonly used in com-putational chemistry that predicts a response variable from multiple, collinear predictorvariables (Wold et al. 2001). PLS is a robust generalization of multiple linear regressionand principle component regression that extracts orthogonal factors (i.e. latent variables)from predictors while taking into account the response variable (Abdi 2010). PLS is becom-ing increasingly popular in ecological data analysis (Carrascal et al. 2009). PLS is valuablewhen two conditions exist: (1) the dataset has a high number of predictor variables relativeto the number of samples and/or (2) high collinearity amongst predictor variables existssuch as the case for most plant biomass estimations. Continuous and categorical data can172.2. Methodsbe used simultaneously in PLS, an essential feature when measuring morphological aspectsof plants.PLS is a powerful statistical method that will maximize the predictive accuracy andprecision of plant mass estimation in the field. PLS should be used in combination withvariable reduction techniques such as Bayesian Information Criterion (BIC) model selectionto optimize the predictive model and reduce field measurements (Mehmood et al. 2011).Compared to destructive plant harvesting, minimally destructive PLS plant biomass esti-mation will increase sampling volume, reduce data collection time, and minimize labor. Thestatistical assumptions associated with PLS make it well–suited to estimating abovegroundplant biomass. Thus, using a multivariate plant biomass prediction approach with PLS ul-timately increases measurement accuracy and precision and will outperform other methodsin the field.In this chapter, I propose a highly accurate, customizable approach to estimating plantbiomass with minimal plant destruction in the field using PLS. My method achieves this bycollecting a set of simple measures from the plant population under study. I propose thatPLS is an accessible and powerful technique for the estimation of plant biomass compared toother plant estimation alternatives. Thus, I hypothesize that partial least squares regressionwill increase plant biomass prediction performance in three distinct plant growth formscompared to simple linear regression models using a single predictor variable, plant height.2.2 Methods2.2.1 Species selection and data collectionI selected three plant species representing extreme differences in morphology/growthhabit in order to determine the robustness of my approach. I tested a small shrub (Cor-nus racemosa Lam., grey dogwood), a tussock grass (Sporobolus cryptandrus Torr., sanddropseed), and a fern with radial rhizomes (Osmunda claytoniana L., interrupted fern). InSeptember 2012, 41 individuals from each plant species were selected from southern On-tario’s hardwood forest near Simcoe, Ontario, Canada along 50 m transects. Plants alongeach transect were selected to capture the range of sizes and shapes present in the populationto establish a standard curve.2.2.2 Measured plant traitsEasily measured plant traits that had a potential to estimate plant biomass were cus-tomized for the growth form of each species. These selected traits were based on mea-surement variations stemming from plant height and circumference, structural counts (i.e.182.2. Methodsnumber of sand dropseed seed heads), and a weighted plate estimator to approximate plantdensity (Rayburn and Rayburn 1998). A weighted plate apparatus was constructed using40.0 cm length x 40.0 cm width x 3.2 mm depth acrylic plexiglass sheet and a 122.0 cm lengthx 1.9 cm width wooden dowel. A large hole was drilled into the center of the plexiglass plateto insert the wooden dowel. In addition, four small holes were drilled into each corner ofthe acrylic sheet to attach strings to raise and lower the plate. When taking measurements,the wooden dowel was set–up near the center of the each plant and the weighted platewas lowered until four leaves touched the plate. Plant height was recorded by measuringheight of the plate from the soil surface. After measuring plant height, the weighted platewas then lowered to rest upon the plant. The resting plate height gives an estimate ofplant density and plate height at rest was recorded. Circumference measures were madeby gathering and compressing the plant material and measuring the circumference of thevegetation using a cloth measuring tape. Circumference measurements were also collectedin a similar manner at half of the plant height and at 30 cm from the soil surface. Thismeasurement was considered as the plant’s basal circumference may be highly variable. Alist of measurements collected for each species is given in Figure 2.1. After measuring eachindividual, aboveground biomass was clipped at the soil surface, dried in a forced air ovenat 60 ◦C for approximately 3–6 days, and weighed. Plants were considered to be dry whenthe biomass weights stabilized within ± 0.3 g.In the case of the interrupted fern, plant morphology was distinctly different from thesand dropseed and grey dogwood growth forms. Interrupted ferns have radial rhizomes withmultiple fronds growing in a circular cluster. Plant measurements were taken per frond andsubsequently averaged and summed per individual. The average and sum measurementswere the variables used to create the statistical models.2.2.3 Model creationTwo datasets, a training dataset and a test dataset, were created from the 41 individualsmeasured and weighed for each plant species. The training dataset was built by randomlyselecting thirty–five plants from each plant population and subsequently used to calculatethe standard curve. The remaining six plants were used as external data points to establisha test dataset. Samples from the test dataset were not included in the creation of thepredictive model to remove any potential influence when generating the standard curve.The same training and test datasets were used when creating the partial least squaresregression (PLS) and linear regression (LR) prediction models.192.2. MethodsData CollectionPredictive ModelSample PredictionPartial LeastSquares Regression Biomass PredictionField DataOven-Dry PlantsGrassShrub Fernthcdlnbchcpiblsl30cthshcwphbcPlant MeasurmentsWeighHarvestlpBayesian InformationCriterion ModelsFigure 2.1: Workflow for predicting plant biomass with partial least squares regression.Plant measurement abbreviations: 30c = circumference at height of 30 cm; bc = basalcircumference; bl = fern blade length; cd = maximum canopy diameter; fl = frond length(blade length + stipe length); hc = circumference at half plant height; ln = leaf number;lp = longest pinna per blade; pi = pinnae number per blade; shc = seed head count; sl =stipe length; th = total plant height; wph = resting height of falling plate meter.202.2. Methods2.2.4 Data transformation, auto–scaling, and polynomial termsEach variable was transformed to approximate normality to maximize the statisticalperformance of each predictive model (Table 2.2). Data were normalized (mean centeredand auto–scaling) as PLS is sensitive to fluctuations in scale and variance among predictorvariables. Diagnostic plots indicated potential curvilinear relationships after auto–scalingbetween predictor and response variables. Polynomial terms (2nd and 3rd orders) werecalculated for each response variable after auto–scaling and included in the variable selectioncalculations (Schielzeth 2010).Table 2.2: Measured plant traits included in the LR and optimized partial least squaresregression models. Partial least squares regression component selection based on lowestroot mean squared error from cross–validation (RMSECV) using 10–fold cross–validation.Plant measurement abbreviations: 30c = circumference at height of 30cm; bc = basalcircumference; cd = maximum canopy diameter; fc = frond count; fl = frond length (bladelength + stipe length); hc = circumference at half plant height; ln = number of leaves; sl= stipe length; th = total plant height; wph = resting height of falling plate meter.Species Model Comp RMSECV Predictors(mass ~ x1+x2+...xn)grey dogwood PLS 3 37.4gsqrt(mass)~ bc + bc2 + cd + cd2+ hc + sqrt(ln) + sqrt(ln)2 + thLR NA 55.7g sqrt(mass)~ thinterrupted fern PLS 3 13.8glog(mass) ~ sqrt−1(∑(sl)) + log(fc)+ sqrt−1(x¯(bl)) +log(∑(fl))LR NA 16.8g log(mass) ~ log(x¯(fl))sand dropseed PLS 4 30.9gsqrt(mass)~ bc + bc2 +square(30c) + cube(th) +cube(th)2 + sqrt(wph)LR NA 42.3g sqrt(mass)~ cube(th)2.2.5 Variable reduction and model averagingUsing the training dataset (n = 35), all possible combinations of transformed variablesand associated (2nd and 3rd order polynomials were scored using Bayesian InformationCriterion (BIC) model selection in the MuMin package (Barton´ 2013) in R (R-Core-Team2013). MuMin’s dredge function was used to force the condition that polynomial regressioncoefficients must be evaluated in conjunction with 1st order regression coefficients to ensureproper fitting of the model (Schielzeth 2010, Symonds and Pither 2012). Models with BIC212.2. Methods≤ 2 are considered to be equivalent. Therefore, BIC models ≤ 2 were averaged and variableimportance values for the predictor variables were extracted. The plant measurementsselected by BIC model selection represent the optimized variables that will best predictplant mass in the dataset (Johnson and Omland 2004, Grueber et al. 2011).2.2.6 Partial least squares regression and linear regression modelsPLS regression models were created with the pls package (Mevik and Wehrens 2007)in R. The optimized predictor variables and transformations used to calculate each plant’sPLS model are given in Table 2.2. Simple linear regression (LR) models were calculatedwith the plant height predictor variable to establish a standard curve in R’s lm function.Predicted biomass from allometric equations were not compared to PLS and LR modelsbecause published equations did not exist for the three measured plant species in thisstudy.The number of orthogonal components (i.e. latent variables) extracted from the PLSmodels was determined to evaluate each component’s contribution to overall predictivefit. The number of components to retain in each model was determined using the rootmean squared error of cross–validation (RMSECV) calculated from 10–fold cross–validation.RMSECV is a diagnostic metric used to test each component’s contribution to the overallpredictive fit of the model. The latent variable with the lowest average RMSECV indicatesthe number of components to retain in the PLS model, thus maximizing each model’spredictive performance.After generating standard curves from the PLS and LR models, plant mass was predictedand back transformed (Pmass) for the training and test datasets. Pmass was subtracted fromthe corresponding reference plant mass (Rmass) weighed in the laboratory to determine howwell the model predicted each data point. A perfect model prediction for a sample is equal tozero (i.e. Pmass - Rmass = 0). Root mean squared error (RMSE) and R–squared estimatesfor the linear relationship between Pmass versus Rmass were calculated to determine thetraining and test dataset’s actual predictive performance. The regression slope used tocalculate RMSE and R–squared for the Pmass versus Rmass training dataset is equal to aslope = 1 with an intercept = 0. Mean and standard deviations of Pmass - Rmass werecalculated for the training and test datasets.222.3. Results2.3 Results2.3.1 Variable selection in partial least squares regression modelsBIC model selection reduced the number of variables from the full model in the inter-rupted fern (5 → 4) and sand dropseed (5 → 4) datasets. Comparatively, grey dogwoodmodels included all five field measured variables in the optimized PLS models (Table 2.2).Field measured variables retained in the PLS models for all three species accounted for as-pects of plant diameter, height, and structural counts. The most descriptive variables, scaledfrom 0.00 to 1.00, were determined by relative variable importance measures. The followingare variables with high influence in each plant’s PLS model: grey dogwood (basal circumfer-ence (1.00), canopy diameter (1.00), plant height (1.00), leaf number (1.00), circumference athalf height (0.95)), interrupted fern (Σ frond length (0.72), Σ stipe length (0.59), Σ bladelength (0.46), Σ frond count (0.43)), and sand dropseed (circumference at 30 cm (1.00),weighted plate height (1.00), basal circumference (0.98), plant height (0.90)). Weightedplate measurements were only relevant when predicting the biomass of sand dropseed. PLSmodels used for grey dogwood and sand dropseed corrected for curvilinear relationships be-tween plant biomass and several field measured variables (i.e. 2nd order polynomial terms)(Table 2.2).2.3.2 Comparing models for predicting plant biomass in the trainingdatasetLR models using plant height predicted plant biomass well in the field but optimized PLSmodels consistently performed better in prediction diagnostics for all three plant species(Tables 2.2 & 2.3). PLS RMSECV calculated in each training dataset was 37.4 g (greydogwood), 13.8 g (interrupted fern), and 30.9 g (sand dropseed). Comparatively, trainingdataset RMSECV predication accuracy in LR models was reduced for all three plant speciesusing plant height as the predictor variable (55.7 g (grey dogwood), 16.8 g (interrupted fern),and 42.3 g (sand dropseed)).The RMSECV model performance indicators translated to higher prediction accuracywhen evaluating Pmass versus Rmass in each training dataset. All optimized PLS modelshad R-squared values ranging between 0.985 – 0.995. Predicted Pmass versus Rmass regres-sion diagnostics were more variable in LR models with R-squared values ranging from 0.784– 0.945. In all LR models, the lower R-squared values are a result of reduced model per-formance when predicting heavier plants in the population (Figure 2.2). Therefore, linearregression models introduced higher variability when predicting heavier plants. In compar-ison, PLS models accurately predicted training dataset plant mass across all plant weightsresulting in higher R-squared values (Figure 2.2).232.3. Results●●trainingtestPredictedDry Mass (g)1001505020000 50 100 150 200●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●trainingtest0 50 100 150 200100150502000●●●●●●●●●●●●●●●● ●●●●●●●●●●●●●●●●●●●●0 20 40 60 8040602008040602000 20 40 60 808050752510000 25 50 75 100●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●0 25 50 75 1005075251000●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●●PredictedDry Mass (g)PredictedDry Mass (g)ReferenceDry Mass (g)ReferenceDry Mass (g)PLS LRGrey DogwoodInterrupted FernInterrupted FernSand Dropseed Sand DropseedGrey Dogwood●●●●●●●●●●●●●●●●●●●●●●●●●●●● ● ●●●●●●●●●●●●●●●●●●●●●●●●●● ●●●●●●●●●●●trainingtest●●trainingtest●●trainingtest●●trainingtestFigure 2.2: Graphs of predicted (Pmass) vs. reference (Rmass) plant biomass using theoptimized PLS models and the LR model for the three plant species. The blue(PLS) andred(LR) points represent internally predicted data used to train each model (n = 35).Black points represent external data predictions from the test dataset using only predictorvariables (n = 6). Each dashed line indicates a perfect prediction (Pmass = Rmass) with aslope = 1 and intercept = 0.242.3. ResultsTable 2.3: Summary statistics for PLS and LR model training datasets. R-squared (R2)and root mean squared error (RMSE) values are based on Pmass versus Rmass estimateswhere slope = 1 and intercept = 0.Species mass ± 1 SD height ± 1 SD Model RMSE R2grey dogwood 42.9g ± 41.3g 77.7cm ± 32.9cm PLS 4.6g 0.995LR 25.2g 0.872interrupted fern 17.1g ± 12.9g 81.0cm ± 18.1cm PLS 2.8g 0.986LR 5.1g 0.945sand dropseed 32.9g ± 23.6g 88.3cm ± 19.0cm PLS 3.4g 0.994LR 14.2g 0.794Model prediction performance in the training dataset, best determined by RMSE values,indicated that all PLS models consistently outperformed LR models when comparing Pmass- Rmass data. LR model RMSE increased as average plant mass increased (Table 2.3). Thisindicates that LR prediction accuracy is reduced when measuring plants with higher biomassin the field thus introducing higher prediction variability (RMSE: 25.2 g grey dogwood, 5.1 ginterrupted fern, 14.2 g sand dropseed). In comparison, PLS models prediction accuracywas consistent across all plant growth forms thus reducing variability when predicting plantbiomass (RMSE: 4.6 g grey dogwood, 2.8 g interrupted fern, 3.4 g sand dropseed).2.3.3 Comparing models for predicting plant biomass in the test datasetSimilar to models using the training set data, external data points predicted by PLSmodels consistently outperformed LR models. Higher variability in LR models was shown inall diagnostic tests compared to the PLS models. R-squared of Pmass versus Rmass PLS testdata points ranged between (0.995 – 0.995) while LR R-squared data ranged between (0.755– 0.943). RMSE of PLS model Pmass versus Rmass was consistently lower when using LRmodels to predict species biomass. Average Pmass - Rmass ± 1 SD for externally predicteddata using PLS was highest in grey dogwood (4.4 g ± 5.8 g) and lowest in interrupted fern(0.4 g ± 2.2 g) (Table 2.4). In comparison, average Pmass - Rmass ± 1 SD using LR hadreduced predictive model performance with highest prediction variability in grey dogwood(−11.1 g ± 21.6 g) and lowest variability in interrupted fern (−1.5 g ± 9.6 g). Using theexternal dataset, PLS models had superior prediction performance with lower variability inall diagnostic tests.252.4. DiscussionTable 2.4: Summary statistics for PLS and LR model externally predicted data. R-squared(R2) and root mean squared error (RMSE) values are based on Pmass verses Rmass estimateswhere slope = 1 and intercept = 0.Species Model x¯(Pmass-Rmass) ± 1 SD RMSE R2grey dogwood PLS 4.4g ± 5.8g 5.7g 0.995LR -11.1g ± 21.6g 15.5g 0.943interrupted fern PLS 0.4g ± 2.2g 2.8g 0.995LR -1.5g ± 9.6g 12.5g 0.926sand dropseed PLS -2.5g ± 2.3g 3.3g 0.996LR -2.5g ± 19.0g 17.4g 0.7552.4 DiscussionPartial least squares regression was a superior predictive methodology compared tosimple linear regression in the three plant species selected in this study. PLS predictedplant biomass had high accuracy and precision in datasets across distinct plant growthforms. This indicates a distinct advantage of using a multivariate approach to predict plantbiomass in the field since growth form did not strongly influence the predictive performanceof PLS.2.4.1 Variable selectionResponse variable selection is the most crucial step when creating a predictive standardcurve to estimate plant mass. A statistical model is only as good as the response vari-ables included in the analyses. For example, if all input variables are weakly correlated toplant mass, the best models chosen by model selection techniques will still result in poorlyperforming models in the field. Therefore, model selection techniques are used to identifythe best model selected from a complete set of variable combinations based on statisticalsupport (Johnson and Omland 2004). BIC model selection optimizes the most descriptivecombination of variables that fit the data and improves predictive performance to build arepresentative model. BIC model selection is more statistically conservative than AkaikeInformation Criterion (AIC) calculations (Burnham and Anderson 2004). In the case of thisstudy, this statistical property is advantageous as BIC model selection will identify fewerplant response variables to input into PLS analyzes, ultimately reducing field measurement262.4. Discussiontime. In my methodology, I employed BIC model selection due to its more conservativeapproach when selecting PLS variables.In my study, the variables selected in grey dogwood and sand dropseed were stronglycorrelated to plant mass as shown by variable importance indicators. On the other hand,variables to predict interrupted fern biomass were not as strong. This may be an artifact ofusing the sum and average frond measurements in the field. This approach may have maskedinformation in the individual frond measurements and reduced the predictive performanceof each variable. Despite the weak correlation of the individual variables to plant mass inthe interrupted fern population, the multivariate PLS approach still resulted in an accuratemodel with good predictive performance and highlights the usefulness of the technique inthe field.Plant height was reasonably correlated with plant biomass in the LR models. Thus, ithas been used as an easily measured surrogate for plant biomass in the field (Singh et al.1975, Catchpole and Wheeler 1992). In the field, simple linear regression models usingone measurement variable have been shown to have low to moderate statistical accuracydepending on ecosystem type and vegetative structure (Catchpole and Wheeler 1992). Toaddress reduced sampling accuracy, Bonham (1982) noted the need for the development ofa multivariate sampling method to predict plant biomass. Models were developed to esti-mate sampling cost for increased variable collection but no predictive model was proposed.Thus, variable selection methods in this paper did not address the statistical assumptionviolations associated with collinear predictors in multiple regression. To account for this,the multivariate PLS approach in my study shows the importance of incorporating sev-eral predictor variables to estimate plant mass for these three species while accounting forvariable collinearity. Incorporating variables that estimated plant density, circumference,and structural counts highlighted the morphological variation in the field as it pertains tocorrelating plant biomass.An approach that employs only one predictor variable largely ignores the fact that twoplants with identical heights may have distinctly different plant volumes in the field. Thisultimately leads to higher variance in prediction accuracy and less reliable results. Thiseffect is shown in my study when LR models of Pmass versus Rmass become increasinglyunreliable as Rmass increases (Figure 2.2). Predicting the biomass of larger individuals withLR reduces measurement accuracy and introduces uncertainty into the model. Increasedvariability in prediction performance is indicated by lower R–squared and higher RMSEestimates in the training (Table 2.3) and external (Table 2.4) datasets compared to PLS.The best way to avoid selecting weakly correlated response variables is to measure alarge number of estimators in the field and iteratively reduce the number of predictor vari-ables post–hoc using model selection techniques (Johnson and Omland 2004). Creative272.4. Discussionmeasurements that are non–traditional in ecology (i.e. weighted plate measurements, plantcircumference at a height of 30 cm) may yield surprisingly strong correlations when esti-mating the plant biomass of a target plant (Rayburn and Rayburn 1998, Rayburn andLozier 2003). Once a suite of measurements have been collected, the number of variablesin full models should be reduced to obtain a more parsimonious prediction model, removeirrelevant variables, and maximize measurement efficiency (Andersen and Bro 2010). Thisultimately leads to highly accurate prediction models balanced against the cost of labor inthe field.2.4.2 The statistical advantage of using partial least squares regressionwhen prediction plant biomassIn comparison to LR, PLS improves the precision and accuracy of estimating biomassacross the herbaceous and shrub phenotypes by incorporating a multivariate estimationapproach (Figure 2.2). PLS models have no statistical restrictions when variables exhibitmulticollinearity (Wold et al. 2001), thus allowing for the incorporation of all variablesthat adequately describe aboveground plant architecture and morphological variation. Thisfeature of PLS results in exemplary predictive performance in the field compared to alltested LR models. For example, the optimized PLS model for grey dogwood includesfive collinear measurements (basal circumference, canopy diameter, circumference at halfplant height, leaf number, and plant height). These response variables would violate theassumptions of traditional multiple regression methods even though the several estimatorswould be advantageous when predicting plant mass (Graham 2003). As shown in all PLSmodels, the multivariate approach led to more robust and accurate statistical models usingboth the training and test datasets compared to LR.Variance in externally and internally predicted data tended to increase with higheraverage plant mass in the field for the three plant species (Figure 2.2). Larger predictionerror can reasonably be expected due to the higher variability associated with plant growthrate response during competition for water, light and nutrients (Poorter and Nagel 2000).Hence, I show that samples predicted in the grey dogwood model (mass: 42.9 g ± 41.3 g)has the highest RMSE values while interrupted fern models (17.1 g ± 12.9 g) had the lowestRMSE in the regression diagnostics for PLS and LR. Thus, higher variation would beexpected in plant populations with a larger range of plant mass in the field. But in allcases, LR was more sensitive to plants with higher average biomass and standard deviationsresulting in less robust biomass prediction models using plant height as the sole predictorvariable. PLS models, on the other hand, exhibited high predictive performance across allplants, regardless of plant biomass ranges in the field. The robustness of the multivariate282.4. Discussionapproach largely accounts for more morphological variability thus increasing the reliabilityof the models.LR models were also less reliable when measuring larger plants within each plant pop-ulation. LR models exhibited higher variation in biomass predictive performance near theupper end of plant mass in all species. This effect was most pronounced when evaluatingpredicted mass versus reference mass in the sand dropseed population, but was present inall evaluated plant species (Figure 2.2). This means that the plant height response variablein the LR models is less descriptive when predicting plants with larger biomass in the field,most likely due to increased morphological variation at larger sizes in the plant populationdue to variability in field response (Poorter and Nagel 2000). The use of a single responsevariable in all species highlights the need for multivariate measurements since predictionaccuracy is not uniformly reliable across the entire plant population (Gholz et al. 1979).Comparatively, all PLS models performed equally well and had relatively uniform pre-diction accuracy regardless of mass in all species (Figure 2.2). In all PLS models, predictedmass linearly increases near the perfect prediction slope indicating excellent predictive per-formance in the training and test datasets. RMSE diagnostics show that 66% of the pre-dicted data in the externally predicted test dataset will fall within 2.8 g (interrupted fern),3.3 g (sand dropseed), 5.7 g (grey dogwood) of the reference mass compared to the largerRMSE prediction errors associated with LR models. Therefore, I show that more uncer-tainty is introduced when predicting plant biomass using only one response variable. Thus,PLS is shown to be a more robust statistical technique that increases prediction confidencein experimental scenarios.2.4.3 Practical applications of partial least squares regressionThis paper follows a workflow (Figure 2.1) that integrates common statistical techniques(i.e. BIC model selection, data–transformations) with PLS. The intensity of labor cost andtime necessary to create a PLS model in an experimental setting was similar to collectinga single response variable using LR. As shown by (Bonham 1982), optimizing the alloca-tion of response variables for model input will maximize labor efficiencies and reduce datacollection costs. Thus, implementing PLS into data collection schemes will increase plantprediction accuracy without introducing significantly higher opportunity costs such as in-creased sampling time. This approach is easily adapted to a variety of field and greenhousesituations, thus increasing sample replication, work efficiency, and prediction accuracy.When creating a PLS model, choose a suite of morphological traits that are measuredquickly and accurately under greenhouse or field conditions. As PLS evaluates categoricaland continuous variables, morphological measurements should be tailored to the plant(s)of interest. This approach incorporates the flexibility to choose the number of variables292.4. Discussionto be used in the predictive model. BIC model averaging extracts predictor variables withthe highest parsimony. If a large suite of variables was measured, a complex model with5 or more measurement variables could potentially be extracted. In this case, samplingvolume during data collection may be a higher priority when weighed against final predictionaccuracy. As BIC models ≤ 2 are considered equivalent, each equivalent model should becalculated using the PLS algorithm. RMSECV results using 10–fold cross–validation canbe subsequently compared to determine the predictive capabilities within each model. PLSmodels should be optimized to create the best model for external predicting data while alsoconsidering sampling efficiency.Compared to published allometric equations, PLS is customizable to a researcher’s studysystem or greenhouse experiment. Allometric equations for predicting biomass have beenshown to differ as a function of morphologic features and environmental conditions (Niklasand Enquist 2002). Thus, reliance upon published equations is not necessary for herbaceousplants and shrubs when utilizing PLS. Unlike published allometric equations, a drawbackof this technique is the creation of the biomass standard curve. The destruction of a smallsubset of plants is inevitably required for all non–destructive biomass prediction analyses.Destructive harvesting can be accounted for when designing an experiment. A researchercan adjust the experimental design of a project by increasing sample replication with theintent of destructive harvesting, creating a preliminary experiment under the same envi-ronmental conditions and harvesting its biomass, or choosing a representative populationin the field similar to the population of interest.Several statistical methodologies have been proposed to predict multivariate, collineardatasets. The main alternatives to PLS regression are principal component regression(PCR), ridge regression (RR), and artificial neural networks (ANN) (Hastie et al. 2001).Compared to PLS, PCR does not account for variance associated with response variablesand resulting models tend to be less parsimonious with higher variability. Studies evaluatingPLS performance compared to RR show similar (Frank and Friedman 1993) or marginallybetter (Yeniay and Goktas 2002) predictive performance using external datasets while bothoutperform PCR. In general, PLS models are more parsimonious, easier to interpret, andmore user friendly than RR. Alternatively, advances in computational statistics and machinelearning suggest that ANN will create better predictive models than all of the precedinglinear regression techniques. Currently, ANN methodologies are not widely used in ecol-ogy. ANN computations have a steep learning curve as the underlying statistics do not usecommon statistical methods. Thus, ANN methods are less accessible and more difficult toimplement compared to PLS regression techniques.302.5. Summary2.5 SummaryIn this chapter, I describe a double sampling method for accurately estimating individualherbaceous plant and small shrub biomass in the field. Partial least squares regression isa robust statistical technique that should be employed to accurately predict plant biomassin ecological experiments. In comparison to liner regression using a sole predictor variable,partial least squares regression increases prediction confidence and reliability in ecologicalexperiments.This chapter is intended to be a simple, customizable guide for ecologists and landmanagers. My approach maximizes aboveground biomass prediction accuracy with highmeasurement efficiency using simple statistical methods and inexpensive tools in the field.The customizable nature of this technique makes PLS a powerful statistical tool for re-searchers in ecological and environmental science.31Chapter 3The Restoration of GrasslandVegetation in Post–ExtractionSandpits3.1 BackgroundOntario’s sand plain prairies support a high biodiversity of regionally unique plants,insects, and animals (Gartshore et al. 1987). Surveys indicate that approximately 22%of Ontario’s rare plant species are found in these prairie ecosystems (Ontario-Biodiversity-Council 2010). Many of these species have been elevated to endangered status due to habitatloss from land–use change, invasive species colonization, and fire suppression. It is estimatedthat Ontario’s prairies occupy less than three percent of their original coverage (Rodger1998). Increasing patch size on marginal lands through prairie restoration will facilitate thesurvival of sensitive habitat in addition to supporting species at risk. Excavated sandpits arecandidate areas to restore prairie plant species but edaphic conditions limit the spontaneousdevelopment of high diversity plant communities (Wali 1999, Prach and Hobbs 2008). Ifpost–mine substrate is left unassisted, plant communities can take decades to recover, ifever (Bradshaw 1997).3.1.1 Biochar and compost as sandpit amendmentsWhen added to soils, researchers suggest that biochar alters physiochemical soil proper-ties by directly releasing nutrients or indirectly altering plant available nutrient concentra-tions (Chan and Xu 2009). Several studies, limited to agricultural systems, indicate thatplant nutrient bioavailability of macro- (P,K) and micro–nutrients (Ca,Mg) have increasedin response to charcoal application (Lehmann et al. 2003, Major et al. 2010, Rondon et al.2007). Meta–analysis shows that biochar significantly translated to increased agriculturalcrop biomass and plant tissue macro–nutrients across all soil types and climates (Biedermanand Harpole 2012). Biochar’s largest positive influence on agricultural plant production hasbeen shown in acidic, nutrient poor soils (Jeffery et al. 2011).323.1. BackgroundTo date, there is little information on the effect of biochar on native plant growth in arestoration setting. Biochar amendments influenced grassland plant biomass inconsistently(Adams et al. 2013) and has the potential to cause shifts in species composition withinmanaged grasslands (Schimmelpfennig et al. 2014). To date, no field study has investigatedbiochar as a soil conditioner when restoring grassland plants in degraded landscapes. Thelarge–scale implication of biochar as a land management tool to grow native grassland plantsstill remains unexplored.As a solitary soil amendment, compost has demonstrated ameliorative effects on soilsin agricultural and mine restoration settings (Shiralipour et al. 1992, Giusquiani et al.1995, Oue´draogo et al. 2001). Compost amendment increases organic matter content, wa-ter holding capacity, and soil nutrients, thus improving soil quality in degraded systems(Termorshuizen et al. 2004). Soil organic matter is a major component of soil quality be-cause it directly or indirectly contributes to physical, chemical, and biological propertiesof functioning soils (Lal 2009). Compost strongly influences soil microbial communities byincreasing in microbial biomass, respiration rates, and soil enzyme activity (Allievi et al.1993). Microbial activity and soil fertility are generally related as compost is mineralizedby microorganisms, thus releasing important elements (C, N, P and S) to the soil solution(Frankenberger and Dick 1983). Thus, increasing soil organic matter in soil is essential torestoring degraded landscapes by alleviating infertile conditions through the reestablish-ment decomposition cycles. As a land management tool, compost application to severelydegraded landscapes increases grassland plant survivorship and primary production thusinfluencing restoration success (Hortenstine and Rothwell 1972, Norland and Veith 1995,Noyd et al. 1996).3.1.2 Arbuscular mycorrhizal fungi as inoculumIn comparison to natural systems, post–mine areas have reduced arbuscular mycorrhizaldiversity and abundance in addition to low nutrients and organic material (Stahl et al. 1988,Ganesan et al. 1991, Diaz and Honrubia 1994). This compounds the nutrient stress of theseenvironments because some plants may be unable to establish and persist simply becausethey lack important microbial symbionts. Even if pre–mine area topsoil is stockpiled andretained, mining activities have been shown to degrade the efficacy of pre–mine populationsof arbuscular mycorrhizas (Stahl et al. 1988). Thus, target plants can be inoculated withfungal propagules to facilitate plant production in disturbed mine areas (Bi et al. 2003,Taheri and Bever 2010).Arbuscular mycorrhizal inoculum has been used in the restoration of mine areas for morethan thirty years because of the ability to enhance plant establishment and survival. Plantsinoculated with AM fungi in post–mine areas show positive growth responses (Khan 1981,333.1. BackgroundJohnson 1998, Enkhtuya et al. 2005, Rydlova´ et al. 2008). Several studies highlight the needto screen AM fungal isolates in order to determine their efficacy in the abiotically stressededaphic conditions (Taheri and Bever 2010, Pu¨schel et al. 2011). Not all commercial AMfungal isolates will be adapted to the harsh abiotic conditions present in post–mine areas.3.1.3 Synergisms among biochar, compost, and arbuscular mycorrhizasIn degraded mine systems, the use of compost to restore plant communities is effectivebut its land management potential may be underestimated. Co–amending soils with biocharand compost may be synergistic as biochar’s high cation exchange capacity and large surfacearea has the potential retain nutrients released from mineralized compost (Fischer andGlaser 2012). Initial studies show mixed results when co–amending soils with biochar andcompost in terms of plant growth. Research on cultivar production ranged from a neutral(Vitis vinifera L., Schmidt et al. 2014) to positive (Avena sativa L., Schulz and Glaser 2012;Samanea saman F.Muell. and Suregada multiflora (A.Juss.) Baill., Ghosh et al. 2014)impact on the growth and quality of plants in soils co–amended with biochar compared tocompost only treatments. To date, no field studies have investigated the impact of biocharco–amended with compost on native grassland plants.Combining AM fungal inoculum with biochar and compost is anticipated to promotelarger gains in plant community biomass compared to adding soil amendments alone. Theapplication of organic amendments have a positive effect on the proliferation of natural AMfungi in agricultural systems (Harinikumar et al. 1990). AM fungi are able to exploit nutri-ents released by mineralization of organic matter due to the activities of soil microorganisms(Hodge et al. 2010). The combination of organic amendments in degraded systems and AMfungal inoculum has been shown to produce larger plants compared to either treatmentalone when reclaiming desertified areas with shrubs (Caravaca et al. 2003b) and coal-minespoil banks with biofuel cultivars (Pu¨schel et al. 2011). Studies that have investigatedthe effect of biochar and AM fungal inoculation on plant biomass show inconclusive re-sults. Both positive (Warnock et al. 2007) and negative (Birk et al. 2009, Warnock et al.2010) effects on plant biomass are dependent upon pyrolysis temperature and quality of thebiochar produced. Therefore, determining the optimized combination of compost, biochar,and AM fungal inoculum to increase plant response is essential when restoring grasslandsin post–mine systems.Considering the goal of grassland plant community restoration, the effect of AM fungalinoculation, municipal compost, and biochar has never been tested in degraded systems.In this multi–year study, two large–scale experiments using common grassland plants weresown in a recently excavated sandpit in southern Ontario, Canada. In the first experiment,a fully factorial combination of compost [CP], biochar [BC], and AM fungal inoculum were343.2. Methodsapplied to greenhouse grown plant plugs in a post–mine sandpit for two growing seasons.In a second experiment, a direct seeding experiment measured total plant cover along gra-dients of industrially feasible rates of compost and biochar with and without AM fungalinoculum. I hypothesized that AM fungal inoculation, compost, and biochar addition wouldindividually increase plant dry mass and cover compared to non–amended controls. Therationale for this hypothesis was that additions were expected to alleviate water and nu-trient stress in post–mine sandpit substrates. I further hypothesized that plots with thehighest rates of compost + biochar and AM fungal inoculum would yield the highest plantdry mass and total cover. The rationale for this hypothesis was that the largest nutrientinput, retention, and acquisition was expected through the fungal symbiosis. This will beevident by increased plant growth and total cover when compared to non–amended con-trol plots. The overall aim of this study was to prescribe industrially feasible abiotic andbiotic soil amendments to facilitate the long–term growth of a grassland plant communityin post–mine sandpits while understanding the role of biochar, compost, and AM fungi invegetative restoration.3.2 Methods3.2.1 Research site establishmentMy research site was established on a recently active sand extraction area (0.5hectares(ha)) near Port Rowan, Ontario, Canada (42 ◦40’17”N, 80 ◦28’46”W, elevation211 m). The mine area, surrounded by Carolinian forest on three sides, is dominated byblack oak (Quercus velutina Lam.), sassafras (Sassafras albidum (Nutt.) Nees.), and tuliptree (Liriodendron tulipifera L.) with interspersed exotic Scots pine (Pinus sylvestris L.).In 2010, the north side of the research site had a cover crop of soybean (Glycine max (L.)Merr.) followed by seeding with endemic grassland vegetation in 2011.In the summer of 2010, the mine area was graded flat by an earthmover and a nine–wire fence was installed on the research site perimeter before the experimental plots wereestablished to minimize deer browsing. The mine area substrate was poorly developedand composed of unconsolidated mineral substrate with no evidence of coarse soil organicmaterial. The exposed sand substrate was easily eroded by wind which created a slightberm at the field site after one growing season. In 2011, relative height of each plot wasmeasured where difference between the highest to lowest plots in the plant plug trial andseed application trial was 0.76 meters(m) and 1.02 m respectively.353.2. Methods3.2.2 Experimental designI tested the effects of soil amendments (biochar, compost and AM fungal inoculation) onthe establishment and growth of endemic grassland plants in a post-mining sandpit usingtwo planting approaches: plant plugs, whereby plants were established in greenhouse andplanted as plugs, and direct seeding at the site. Two different plant response measurementmethods were used in the field: predicted plant biomass estimation in the plant plug trial,and vegetative cover estimation for the seed application trial. In the plant plug trial, aminimally destructive biomass estimation methodology was used because each plant pluglocation was known and could be tracked over multiple years. This methodology allowsfor the precise estimation of individual plant response to the experimental treatments overtime. In the seed application trial, vegetative cover was estimated because tracking thegrowth of plant individuals was less feasible in the field. Therefore, vegetative cover wasused as a non–destructive proxy for plant response, thus estimating plant germination andestablishment rates in the field.BiocharI used biochar created from wood pellet feed stock that was pyrolyzed at 500 ◦C in anindustrial scale non–oxygenated vacuum reactor. My biochar was supplied by the large-scalebiochar producing facility, New Earth Renewable Energy Inc., based in Quebec, Canada.As biochar is a relatively unknown commodity as a soil amendment, I tested two industriallyfeasible biochar rates (5 T ha−1 and 10 T ha−1) in the plant plug experiment. These ratesbalance amendment cost against the potential plant growth benefit of biochar relevant tothe industrial–scale restoration of sandpit areas. In the direct seeding experiment, I testedsix rates of biochar ranging from no biochar to rates at the upper end of cost feasibility insandpit restoration (0 T ha−1 to 40 T ha−1) (Table 3.1).CompostI used compost derived from municipal lawn and leaf urban waste streams distributed byTry Recycling in London, Ontario, Canada. I tested one industrially feasible compost rate(20 T ha−1) in the plant plug experiment. As compost is relatively less expensive comparedto biochar, compost can be applied at a higher rate when budgeting for an industrial–scalegrassland restoration project. In the direct seeding experiment, I tested six rates of compostranging from 0 T ha−1 to 40 T ha−1(Table 3.1).363.2. MethodsTable 3.1: Experimental treatments for the seed application experiment. All treatmentlevels are fully factorial. Each treatment combination was applied to one plot only. Totalnumber of plots was 72.Biochar Level Compost Level AM Level0.0 T ha−1 0.0 T ha−1 No inoculum2.5 T ha−1 2.5 T ha−1 Rhizophagus irregularis5.0 T ha−1 5.0 T ha−110.0 T ha−1 10.0 T ha−120.0 T ha−1 20.0 T ha−140.0 T ha−1 40.0 T ha−1Factorial = biochar level × compost level × AM fungal inoculum levelArbuscular mycorrhizal fungal inoculumI used a commercial arbuscular mycorrhizal fungal inoculant, Rhizophagus irregularis(Blaszk., Wubet, Renker & Buscot) C. Walker & A. Schu¨ßler (2010), supplied by Mikro–Teklocated in Timmins, Ontario, Canada. In the plant plug experiment, each plug container wasinoculated with 20 spores contained in a proprietary powdered medium. The spore mediumwas added just below the soil surface of each plant plug container during seed sowing inApril 2010. In the seed application experiment, each plot received 2 liters of water withsuspended with spores. To mix the solution, a proprietary water–soluble powdered mediumcontaining the spores was added to a watering can and applied evenly over the plot followingseed compaction. Spores were applied at Mikro–Tek’s recommended rate of 1000 spores/m2.3.2.3 Plants used in restorationThe eight grassland plant species selected for this project met the following criteria:plant species that are common in Ontario prairies, tolerant of sandy soils and dry conditions,endemic to the study area, and known to associate with arbusuclar mycorrhizal fungi.Details about these plants are given in Table 3.2.373.2.MethodsTable 3.2: The eight grassland plant species used in the plant plug trial and seed application trial. The abbreviation columnindicates the plant code scheme associated with Figure 3.1. The final two columns indicate the abundance (i.e. number of plantplugs) of all species in each plot and the core sampling areas for the plant plug trial.Species Common Name Abbreviation plantsplotplantscoreC4 GrassesAndropogon gerardii Vitman Big Bluestem AG 11 5Panicum virgatum L. Switchgrass PV 11 4C3 GrassesElymus canadensis L. Canada Wild Rye EC 8 3Bromus kalmii A. Gray Prairie Brome BK 8 4N–Fixing ForbsDesmodium canadense L. Showy Tick–trefoil DC 11 5Lespedeza capitata Michx. Roundhead Bushclover LZCA 3 3Composite ForbsLiatris cylindracea Michx. Ontario Blazing Star LC 10 4Symphyotrichum laeve(L.) A´. Lo¨ve & D. Lo¨ve var. laeve Smooth Blue Aster SL 10 5TOTAL 72 33383.2. Methods3.2.4 Plant plug trialIn the plant plug trial, plants were grown as plugs for 16 weeks in a commercial green-house by Pterophylla / St. Williams Nursery & Ecology Centre in St. Williams, Ontario,Canada from March 1st to June 24th. No supplemental light or heating was used in thegreenhouse. Plant plugs were grown in 72 cell Landmark plug trays each filled with 57 cu-bic centimeters of a proprietary growing medium containing pine bark, sphagnum peat, leafand yard waste compost and perlite. At the time of plug sowing, half of the plug containerswere inoculated with AM fungal spores in the greenhouse. The growing medium used inthe plugs was not sterilized to mimic industrial conditions. Background AM fungal com-munities were anticipated in non–inoculated plant plugs due to potential growing mediumcontamination in the industrial–scale greenhouse setting. The plant source material wascollected from local plant populations by Pterophylla / St. Williams Nursery & EcologyCentre in the vicinity of the restoration project.Plots (size: 10.2 m2) were established by June 22nd, 2010 using a fully–crossed, random-ized factorial design, and monitored for 3 growing seasons (2010–2012). The two factorswere: soil amendments (no amendment, 5 T ha−1 biochar, 10 T ha−1 biochar, 20 T ha−1compost, 5 T ha−1 biochar + 20 T ha−1 compost, 10 T ha−1 biochar + 20 T ha−1 compost)and Rhizophagus irregularis inoculation (±). Each of the 12 factorial combinations wasreplicated ten times for a total of 120 plots. Compost and biochar were raked into theupper 6 cm of substrate in May 2010. Control plots were not amended and were plantedwith non–inoculated plant plugs. A one meter buffer zone separated each plot to minimizeplant interactions.Native plant plugs were transplanted to the field between June 24th, 2010 – July 1st,2010. Seventy–two plant plugs per plot were pre–mapped to have identical positions acrossall field plots (plug spacing = 33 cm)(Figure 3.1). A hexagonal plug arrangement waschosen to minimize spatial variability. Of the 8,640 plant plug positions, only two plugswere incorrectly planted as noted during vegetative censuses.AM fungal quantificationAM fungal colonization of roots was quantified for greenhouse grown plant plugs (June2010) and field plots (September 2011 / 2012) in the plant plug trial. For plugs, ten non–inoculated and ten inoculated plugs from each of the eight plant species were randomlyselected in the greenhouse to assess root colonization before adding plugs in the field. Ineach field plots, sixteen soil cores per plot were collected and pooled near designated pluglocations in September 2011 / September 2012 to minimize spatial variability. Plugs andthe pooled field soil cores were washed free of soil in a 1 mm sieve to extract the roots.393.2. MethodsRoots were removed, cut into 1 cm pieces, and preserved in 50% ethanol until microscopicanalysis. Roots were stained with Chlorazol Black E (Brundrett et al. 1984) and thencounted systematically under a microscope using the gridline intersect method (McGonigleet al. 1990).Plant biomass estimationI used partial least squares (PLS) regression to predict plant biomass using the samemodel creation methodology described in Chapter 2. A subset of randomly selected plots,one from each factor combination, was destructively harvested to create a PLS standardcurve for six of eight plant species between September 14th, 2011 and September 16th, 2011.Similarly, a second set of plots was also destructively harvested between August 28th, 2012and August 31th, 2012 as plants grew larger and morphological predictor characteristicswere anticipated to change from first to second growing season. In both years, the C3grasses (see above) were not estimated; living plant tissue was not available in Septemberdue to early season senescence and poor plant performance, resulting in unreliable partialleast squares regression estimates in the field. Since aboveground biomass harvesting mayhave introduced a plant growth bias in subsequent growing seasons, plots harvested inthe first year were excluded from the final analyses. Plots destructively harvested in thesecond growing season were included in final analyses as they were not disturbed prior toharvesting.Biomass was estimated for plants in the center of each plot (i.e. the ”core area”(Figure3.1)) in September 2011 and 2012. Core area plants were measured to reduce any con-founding edge effects present in the field plots. Thirty–three (33) plant plugs in the corearea were measured for each plot for a total of 3,960 plug locations measured per growingseason.I measured morphological plant characters related to height, diameter, and stem countswhen appropriate for each plant species for the Fall 2011 and Fall 2012 growing seasons(Table 3.3). The predictor variables were selected via BIC model selection then used tomeasure the remainder of the 3,960 plant plugs in the field each season (Table 3.4). Partialleast squares predicted mass was subtracted from the corresponding reference plant mass(Pmass–Rmass) ± 1 standard deviation (SD) to estimate prediction error. A value of zeroindicates Pmass = Rmass, hence a perfect prediction. Statistical details of measurementaccuracy for each species are given in Table 3.5.403.2. MethodsLZCA12PV01LC02BK03PV04SL05AG06LC07BK08BK17BK11BK67BK37BK21BK31PV16PV36PV58PV59PV60PV62PV52PV39PV65AG18AG27AG19AG13AG38 AG54AG15 AG48AG72AG66LC09LC34LC41LC43LC45LC63LC14LC22DC10DC20DC28DC35DC42DC61DC69DC64DC40DC24 DC57SL25SL44SL29SL30SL55SL53 SL68SL32SL56EC26EC33 EC49EC50EC71EC70EC47EC23LZCA51LZCA46Figure 3.1: Diagram of the plant plug layout with plant positioning. Each hexagonal cellsignifies the location and identity of one plant taxa added to the plot as a plant plug. Allplots have the same plug configuration to minimize spatial variability. Plug spacing = 33 cm.Plants sampled in the core are indicated in beige. See Table 3.2 for plug abbreviations.413.2. MethodsTable 3.3: Morphological characters measured in the field for the six plant species inSeptember 2011 and September 2012. The most parsimonious combination of variables wasselected via Bayesian Information Criterion model selection to create the predictive models.Species Year Measured Variables)C4 GrassesAndropogon gerardii 2011height; weigh plate; # of tillers; # of seed heads;basal circumference2012height; weigh plate; # of tillers; # of seed heads;basal circumference; circumference @ 30cmPanicum virgatum 2011height; weigh plate; # of tillers; # of seed heads;basal circumference2012height; weigh plate; # of tillers; # of seed heads;basal circumference; circumference @ 30cmN–Fixing ForbsDesmodium canadense 2011stem length; # of stems; length of steminflorescence; basal circumference2012stem length; # of stems; length of steminflorescence; basal circumference; circumference@ 30cmLespedeza capitata 2011stem length; # of stems; # of stems withinflorescence; inflorescence length; basalcircumference2012stem length; # of stems; # of stems withinflorescence; inflorescence length; basalcircumferenceComposite ForbsLiatris cylindracea 2011# of leaves; height; # of inflorescence; length ofstems with inflorescence2012# of leaves; height; # of inflorescence; length ofstems with inflorescenceSymphyotrichum laeve 2011stem length; # of stems; # of stems withinflorescence; inflorescence length2012stem length; # of stems; # of stems withinflorescence; inflorescence length423.2. MethodsTable 3.4: Morphological characters selected for the six plant species measured in Septem-ber 2011 and September 2012. The variables given in the table were selected via BayesianInformation Criterion model selection to create the predictive models using partial leastsquare regression.Species Year Comp Predictor Vars(mass ~ x1 + x2 + ... xn)C4 GrassesAndropogon gerardii 2011 2 (height) + (weigh plate)2012 3(weigh plate) + (circumference @ 30cm) +(circumference @ 30cm)2 + (# of seedheads) + (# of seed heads)2Panicum virgatum 2011 2 (basal circumference) + (weigh plate)2012 2(circumference @ 30cm) + (weigh plate) +(weigh plate)2 + (# of seed heads)N–Fixing ForbsDesmodium canadense 2011 2(mean stem length) + (sum stem length) +(sum stem length)22012 1(basal circumference) + (circumference @30cm)Lespedeza capitata 2011 2 (sum stem length)+ (mean stem length)2012 2 (sum stem length) + (sum stem length)2Composite ForbsLiatris cylindracea 2011 2(# of leaves) + (# of leaves)2 + (# ofleaves)3 + (height) + (height)22012 1 (# of leaves) + (height)Symphyotrichum laeve 2011 3(sum stem length) + (mean stem length) +(mean stem length)22012 4(mean inflorescence length) + (meaninflorescence length)2 + (sum stem length)+ (sum stem length)2433.2.MethodsTable 3.5: Partial least squares regression diagnostics for the six plant species measured in September 2011 and September 2012.All prediction data is based on variables selected via Bayesian Information Criterion model selection (Table 3.4). Mass data isgiven in grams (g) dry weight based on weighed plants used to create the standard curve. For each species, predicted plant massfrom the partial least squares regression model was subtracted from the reference plant mass (Pmass – Rmass ) ± 1 standarddeviation (SD) to calculate within–model estimates. When Pmass = Rmass, predicted mass is equal to reference mass, thusrepresents a perfect prediction. R-squared, root mean squared error (RMSE), and p–values were calculated for Pmass – Rmassusing linear regression for each plant species to indicate prediction accuracy. All regression diagnostics are based on a slope = 1and intercept = 0.Species Year mass ± 1 SD Rep (Pmass–Rmass) ± 1 SD RMSE R2 p–valueC4 GrassesAndropogon gerardii 2011 8.5g ± 5.1g 36 0.2g ± 2.4g 9.9g 0.944 <0.0012012 30.7g ± 22.9g 41 0.9g ± 5.1g 28.6g 0.983 <0.001Panicum virgatum 2011 15.7g ± 11.1g 34 -0.7g ± 6.2g 17.6g 0.892 <0.0012012 94.1g ± 84.9g 41 -1.2g ± 21.5g 100.0g 0.970 <0.001N–Fixing ForbsDesmodium canadense 2011 33.5g ± 20.1g 36 -0.4g ± 7.6g 36.1g 0.962 <0.0012012 29.6g ± 21.2g 41 -1.0g ± 12.9g 38.1g 0.873 <0.001Lespedeza capitata 2011 2.7g ± 2.2g 25 -0.1g ± 1.0g 2.9g 0.923 <0.0012012 2.2g ± 3.3g 25 -0.1g ± 1.1g 1.9g 0.924 <0.001Composite ForbsLiatris cylindracea 2011 2.5g ± 1.3g 35 0.0g ± 0.6g 3.1g 0.957 <0.0012012 0.9g ± 0.8g 41 0.9g ± 0.8g 1.0g 0.917 <0.001Symphyotrichum laeve 2011 7.9g ± 4.1g 36 -0.3g ± 2.8g 10.2g 0.901 <0.0012012 7.9g ± 9.7g 41 -0.2g ± 1.8g 6.9g 0.981 <0.001443.2. Methods3.2.5 Seed application trialThe seed application experimental plots were set–up adjacent to the plug experimentusing the same plot dimensions and soil amendment incorporation protocol. Plots wereestablished in August of 2010 using a fully–crossed, randomized factorial design and moni-tored for 3 growing seasons (2011–2013). Factors were: six rates of each amendment givenin Table 3.1 and Rhizophagus irregularis inoculation (±). Amendment and inoculationcombinations were not replicated, for a total of seventy–two plots. To minimize overwinterseed mortality and undesired seed movement via wind scour, seeding and inoculation werenot done until May 2011.In April 2011, seeds were pre–weighed into bags, mixed with moist vermiculite, andstored at 4 ◦C in the refrigerator for one month (Table 3.6). This process of cold–moiststratification promotes rapid spring germination of dormant plant seeds. In May 2011,cold–moist stratified seeds were hand sown and lightly mixed with a steel rake into thesandpit substrate. A seed roller was used to press the seed into the sandpit floor to ensuresoil–seed contact. Mycorrhizal inoculum was added to one set of the amendments via aliquid medium containing spores at the recommended rate. Seeds were applied at doublethe standard rate for recommended for tallgrass prairie restoration projects to ensure plantestablishment in the experiment.Table 3.6: Seeding rate in grams (g) for the eight grassland plant species used in the seedapplication trial. Plot size was 10.2 m2. Seeds were cold–moist stratified at 4 ◦C for onemonth until the time of sowing in the field (May 2011).Species Seeding Rate (g)Andropogon gerardii 7.0 gPanicum virgatum 1.5 gElymus canadensis 5.5 gBromus kalmii 2.5 gDesmodium canadense 1.5 gLespedeza capitata 1.5 gLiatris cylindracea 4.0 gSymphyotrichum laeve 6.0 g453.2. MethodsPercent cover estimationA photographic time–series technique was used to estimate the percent cover of plantgrowth in the seed application trial. This approach tracked the germination and establish-ment rates of the seeded grassland species across the treatment levels. An angle cameramonopod was constructed to take overhead pictures in each plot (Figure 3.2). Photos werecropped to analyze a 2.6 m2 area. Total plant cover was measured using the SamplePointsoftware (Booth and Cox 2008). In SamplePoint, a 100 point overlaying grid was used toclassify pixels as grass, composite forb, N–fixing forb, soil, or plant litter. Native plant coverwas estimated from the summation of grasses, forbs, and N–fixing forbs and subsequentlydivided by total pixels estimated. Photographs of each plot were taken in September forthree growing seasons (2011–2013).Figure 3.2: Collecting photographic data to analyze percent plant cover. A right–angledmonopod was designed to take over–head photographs used to estimate plant cover in theseed application trial. The monopod was raised and leveled with the camera on a delayedsetting to capture a picture for cover estimation in the SamplePoint software. (Photo Taken:September 2012)463.3. Results3.2.6 Statistical analysesLinear mixed effects models were used to test treatment significance for the plant re-sponse estimates and AM colonization of field roots. Linear mixed effects models use mul-tivariate statistical procedures that account for random variability associated with plotsat the field site. AM colonization of plant plug roots in inoculated versus non–inoculatedtreatments were evaluated using a t–test. Data transformations were employed when nec-essary to approximate a normal distribution of model residuals. Relative plot height wasincluded as a covariate in all linear mixed effects models. Linear mixed effect model se-lection procedures iteratively removed non–significant variables using Chi–squared tests.This resulted in the most parsimonious models to analyze statistical significance for eachresponse variable. Linear mixed effects models were analyzed using the lme4 package in R(R-Core-Team 2013, Bates et al. 2014). Significance levels (p–values) derived from MarkovChain Monte Carlo methods, % explained deviance (an R–squared proxy, abbreviated: %Expl. Dev.), and main level post–hoc comparisons were calculated using the R packageLMERConvenienceFunctions by Tremblay and Ransijn (2013).3.3 Results3.3.1 Plant plug trialAM fungal establishment in greenhouse plug rootsAll eight plant species were colonized by AM fungi in the greenhouse (Figure 3.3). Theapplication of R. irregularis inoculum resulted in significant increases in percent colonizationin all species compared to non–inoculated plants (p<0.001) (mean ranges of AM fungi ininoculated roots: 16.9% (E. canadensis) – 30.1% (Andropogon gerardii)). As expectedin the unsterile greenhouse environment, low levels of AM fungal colonization of rootswere detected in non–inoculated plant plugs across all plant species (<5.0% mean AMcolonization of roots). These results indicate that the AM fungal inoculum treatment wasestablished at the onset of the plant plug trial.AM fungal establishment in field rootsR. irregularis inoculum persisted in the field after two growing seasons. Significantincreases in AM colonization rates of field roots were detected in inoculated vs. non–inoculated plots (p<0.001)(Figure 3.4a). AM fungal inoculated treatments nearly doubledin the rate of root colonization between September 2011 and September 2012 (mean AMfungal colonization of roots: 22.4% (2011) to 45.8% (2012)). Mean AM fungal colonization473.3. Resultsrates of roots in the non–inoculated plots tripled from 5.5% (2011) to 15.8% (2012). Rootsin inoculated plots experienced a larger relative increase in AM fungal colonization betweenSeptember 2011 and September 2012 compared to non–inoculated treatments (p<0.001).Plant biomass responses in the plant plug trialIncreases in total plant biomass were influenced by soil amendments (p=0.056), growingseason (p<0.001), and the relative plot height covariate (p=0.068) in the plant plug trial. Nosignificant difference in total plant biomass was detected in plots receiving soil amendmentscompared to non–amended controls (p>0.05). No significant interactions among the modelterms were detected.Plant SpeciesAM Colonization of Roots(% Arbuscules + % Vesicles)0%10%20%A. gerardiiP. virgatumC 4  Grass30%B. kalmiiE. canadensisC 3  GrassD. canadenseL. capitataN-FixerS. laeveL. cylindraceaComposite****** ************ ******InoculatedNot InoculatedFigure 3.3: Percent AM fungal colonization of greenhouse grown plant plug roots. Plantplugs were randomly selected just prior to sowing plant plugs in the field (June 2010). Tenplant plugs from each treatment level (± R. irregularis) of all eight species were analyzedfor AM colonization of roots using t-tests comparing inoculated and non–inoculated plants.Raw data ± 1 SD is presented in the graph. Each asterisk represents a p–value (***) <0.001 for comparisons between inoculation treatment levels. Replication = 10.483.3. Results(a)InoculatedNot Inoculated0%20%40%September 2011 September 201260%AM Colonization of Roots(% Arbuscules + % Vesicles)None 5BC10BC20CP20CP + 5BC20CP + 10BC None 5BC10BC20CP20CP + 5BC20CP + 10BC(b)Model Terms p–value sig. level % Expl. Dev.AM Inoculation <0.001 *** 33.64%Year <0.001 *** 24.19%Plot Height (Dry → Wet) 0.003 (**) 0.94%InteractionsYear × Plot Height 0.056 . 0.39%Year × AM Inoculation <0.001 *** 3.35%Amendment × AM Inoculation × Year 0.035 * 1.27%Significance: *** ≤ <0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Relationships with negative regression slopes are indicated by parentheses.Figure 3.4: AM fungal colonization of the mixed community of field roots in the plant plugtrial. Panel (a) represents the graph of raw data with error bars (± 1 SD) based on the mostparsimonious linear mixed effects model. Experimental treatment replication = 9. The leftgraph panel represents data after one growing season. Labels on the x–axis: None = nosoil amendment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1compost, 5BC +20CP = 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1biochar + 20 T ha−1 compost. Statistical output (b) shows significant main effect termsand interactions. Main effects included in the model were: Amendment, AM inoculation,Plot Height, and Year. % explained deviance is abbreviated as % Expl. Dev. in the output.Model terms with negative regression slopes are indicated in parentheses.493.3. ResultsBiochar amendments compared to non–amended control plots, direct effects of biocharwere not detected on total plant biomass (p>0.05)(Figure 3.5a). Unexpectedly, biochar onlyamendments significantly reduced total plant biomass compared to the majority of compostand compost + biochar amended plots (Figure 3.5a). Only Andropogon gerardii respondednegatively to the addition of biochar compared to non–amended control plots (Figure 3.6a).All other measured plant species exhibited no direct response to either biochar rate.Compost amendment The total plant biomass response to the compost only amend-ment was positive compared to non-amended controls although not statistically significant(p=0.125). Andropogon gerardii biomass was reduced in the presence of compost comparedto non–amended control plots (Figure 3.6a). Desmodium canadense experienced significantincreases in plant biomass in the presence of compost amendments. No other direct compostonly effect were detected in plant response for the four other plant species in this trial.AM fungal inoculation AM fungal inoculation did not significantly influence total plantbiomass in the plant plug trial (Figure 3.5a) although each species varied in plant biomasswhen inoculated with R. irregularis. The biomass of Panicum virgatum (p<0.001) and Les-pedeza capitata (p=0.021) responded positively to R. irregularis inoculation. In contrast,Andropogon gerardii and Liatris cylindracea biomass was significantly reduced in AM in-oculated plots (p<0.05). No inoculation response was detected for Symphyotrichum laeveand Desmodium canadense. Altogether, interspecies variation in plant response to AMinoculation resulted in a neutral effect on the total biomass response in the community.Synergistic effects of biochar, compost, and AM fungal inoculation Direct totalplant biomass effects of 10 T ha−1 of biochar + 20 T ha−1 of compost were positive comparedto non–amended control plots although not statistically significant (p=0.117).(Figure 3.5a).Co–amended plots with 10 T ha−1 of biochar + 20 T ha−1 of compost significantly increasedtotal plant biomass compared to plots with 5 T ha−1 of biochar (p=0.037) and 10 T ha−1of biochar (p=0.008) amended plots. The interaction of AM fungal inoculum and soilamendments did not significantly influence total plant biomass. No significant differenceswere detected when comparing total plant biomass in compost and compost + biochar.503.3. Results(a)0200400September 2011 September 2012600Total Plant Biomass(grams)None 5BC 10BC20CP20CP + 5BC20CP + 10BC None 5BC10BC20CP20CP + 5BC20CP + 10BC(b)Model Terms p–value sig. level % Expl. Dev.Amendment 0.056 . 2.95%Year <0.001 *** 3.99%Plot Height (Dry → Wet) 0.068 . 0.90%Sig. Post–Hoc Comparisons p–value sig. level5BC → 20CP 0.039 *5BC → 20CP + 10BC 0.037 *10BC → 20CP 0.008 **10BC → 20CP + 5BC 0.052 .10BC → 20CP + 10BC 0.008 **Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Relationships with negative regression slopes are indicated by parentheses.Figure 3.5: Predicted total plant biomass in the plant plug trial. Panel (a) represents thegraph of raw data with error bars (± 1 SD) based on the most parsimonious linear mixedeffects model. Experimental treatment replication = 9. The left graph panel representsdata after one growing season. Labels on the x–axis: None = no soil amendment, 5BC= 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP= 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1compost. Statistical output (b) shows significant main effect terms and interactions. Maineffects included in the model were: Amendment, AM inoculation, Plot Height, and Year. %explained deviance is abbreviated as % Expl. Dev. in the output. Note that model termswith negative regression slopes are indicated in parentheses around the significance levels.513.3. Results(a)InoculatedNot Inoculated050150September 2011 September 2012A. gerardii Biomass(grams)None 5BC 10BC 20CP20CP + 5BC20CP + 10BCNone 5BC 10BC 20CP20CP + 5BC20CP + 10BC100(b)Model Terms p–value sig. level % Expl. Dev.Amendment 0.048 (*) 1.45%Year <0.001 *** 17.61%Plot Height (Dry → Wet) 0.061 . 0.45%AM Inoculation 0.022 (*) 0.67%Sig. Post–Hoc Comparisons p–value sig. levelNone → 5BC 0.094 (.)None → 10BC 0.012 (*)None → 20CP + 5BC 0.003 (**)None → 20CP + 10BC 0.027 (*)10BC → 20CP 0.069 .20CP → 20CP + 5BC 0.018 (*)Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Relationships with negative regression slopes are indicated by parentheses.Figure 3.6: Predicted Andropogon gerardii biomass in the plant plug trial. Panel (a) repre-sents the graph of raw data with error bars (± 1 SD) based on the most parsimonious linearmixed effects model. Experimental treatment replication = 9. The left graph panel repre-sents data after one growing season. Labels on the x–axis: None = no soil amendment, 5BC= 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP= 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1compost. Statistical output (b) shows significant main effect terms and interactions. Maineffects included in the model were: Amendment, AM inoculation, Plot Height, and Year. %explained deviance is abbreviated as % Expl. Dev. in the output. Note that model termswith negative regression slopes are indicated in parentheses around the significance levels.523.3. ResultsLespedeza capitata was positively influenced by the interaction among the soil amend-ments and AM fungal inoculum compared to non–amended control (Figure 3.7a). Desmod-ium canadense experienced biomass gains in the presence of all compost + biochar amendedtreatments compared to biochar only and non–amended plots (Figure 3.8a). Desmodiumcanadense’s large response to compost and compost + biochar amended plots stronglyinfluenced total biomass results in September 2011 (Figure 3.5a). Comparatively, only An-dropogon gerardii responded negatively to the compost + biochar treatments compared tonon–amended control plots (Figure 3.6a). Overall, these results indicate that compost +biochar addition with AM fungal inoculation enhances the plant community response atthe species level in post–mine sandpits.Growing season Interspecies growth response was highly variable after two full growingseasons. In all models, growing season explained the highest amount of variation in thespecies biomass datasets. The biomass of N-fixing forbs (Figure 3.7a & Figure 3.8a) andcomposite forbs (Figure 3.10a & Figure 3.11a) was significantly reduced between September2011 and September 2012. Comparatively, the C4 grasses (Andropogon gerardii and Pan-icum virgatum) experienced biomass gains between September 2011 and September 2012.The C4 grasses were amongst the largest contributors to total biomass, accounting for totalbiomass gains from September 2011 to September 2012.Plot height covariate A trend was detected in the response of total biomass to the plotheight covariate (p=0.068)(Figure 3.5a). This indicates that plant biomass increased inplots lower on the landscape regardless of treatment. Similarly, the plot height covariatewas significant and positive for all measured plant species except for Desmodium canadense(p>0.05) and Panicum virgatum (p>0.05). The influence of the plot height covariate in thetotal biomass analysis may have been suppressed since Desmodium canadense and Panicumvirgatum are among the largest contributors to total plant biomass.3.3.2 Seed application trialNote: Two plots in the southeast corner of the seed application trial were removed fromthe analysis due to close proximity to the research site’s water table. These outlier plots,5 T ha−1 biochar + AM fungi (% cover in 2013: 53%) and 5 T ha−1 biochar - AM fungi (%cover in 2013: 61%), were nearest to a former wet depression at the field site and exhibitedhigh vegetative density compared to all other plots. Mean total % cover ± 1 SD excludingoutlier plots was 21% ± 9% in 2013.The eight species sown at the mine site accounted for the vast majority of the vegeta-tion in seed application trial. Non–seeded volunteer plant cover was negligible throughout533.3. Results(a)InoculatedNot Inoculated0515September 2011 September 201220L. capitata Biomass(grams)None 5BC10BC20CP20CP + 5BC20CP + 10BC None 5BC10BC20CP20CP + 5BC20CP + 10BC10(b)Model Terms p–value sig. level % Expl. Dev.Year <0.001 (***) 4.00%Plot Height (Dry → Wet) 0.035 * 0.50%AM Inoculation 0.021 * 0.60%InteractionsAmendment × AM Inoculation × Year 0.014 * 1.61%Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Relationships with negative regression slopes are indicated by parentheses.Figure 3.7: Predicted Lespedeza capitata biomass in the plant plug trial. Panel (a) representsthe graph of raw data with error bars (± 1 SD) based on the most parsimonious linear mixedeffects model. Experimental treatment replication = 9. The left graph panel representsdata after one growing season. Labels on the x–axis: None = no soil amendment, 5BC= 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP= 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1compost. Statistical output (b) shows significant main effect terms and interactions. Maineffects included in the model were: Amendment, AM inoculation, Plot Height, and Year. %explained deviance is abbreviated as % Expl. Dev. in the output. Note that model termswith negative regression slopes are indicated in parentheses around the significance levels.543.3. Results(a)0100200September 2011 September 2012D. canadense Biomass(grams)None 5BC 10BC 20CP20CP + 5BC20CP + 10BCNone 5BC 10BC 20CP20CP + 5BC20CP + 10BC(b)Model Terms p–value sig. level % Expl. Dev.Amendment <0.001 *** 9.27%Year <0.001 (***) 14.05%Sig. Post–Hoc Comparisons p–value sig. levelNone → 20CP <0.001 ***None → 20CP + 5BC <0.001 ***None → 20CP + 10BC <0.001 ***5BC → 20CP <0.001 ***5BC → 20CP + 5BC <0.001 ***5BC → 20CP + 10BC <0.001 ***10BC → 20CP <0.001 ***10BC → 20CP + 5BC <0.001 ***10BC → 20CP + 10BC <0.001 ***Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Relationships with negative regression slopes are indicated by parentheses.Figure 3.8: Predicted Desmodium canadense biomass in the plant plug trial. Panel (a)represents the graph of raw data with error bars (± 1 SD) based on the most parsimoniouslinear mixed effects model. Experimental treatment replication = 9. The left graph panelrepresents data after one growing season. Labels on the x–axis: None = no soil amend-ment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost,5BC +20CP = 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1 biochar+ 20 T ha−1 compost. Statistical output (b) shows significant main effect terms and in-teractions. Main effects included in the model were: Amendment, AM inoculation, PlotHeight, and Year. % explained deviance is abbreviated as % Expl. Dev. in the output.Note that model terms with negative regression slopes are indicated in parentheses.553.3. Results(a)InoculatedNot Inoculated0100200September 2011 September 2012300P. virgatum Biomass(grams)(b)Model Terms p–value sig. level % Expl. Dev.Year <0.001 *** 43.34%AM Inoculation <0.001 *** 0.88%Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Relationships with negative regression slopes are indicated by parentheses.Figure 3.9: Predicted Panicum virgatum biomass in the plant plug trial. Panel (a) representsthe graph of raw data with error bars (± 1 SD) based on the most parsimonious linear mixedeffects model. Experimental treatment replication = 9. The left graph panel representsdata after one growing season. Statistical output (b) shows significant main effect termsand interactions. Main effects included in the model were: Amendment, AM inoculation,Plot Height, and Year. % explained deviance is abbreviated as % Expl. Dev. in the output.Note that model terms with negative regression slopes are indicated in parentheses aroundthe significance levels.563.3. Results(a)02040September 2011 September 2012S. laeve Biomass(grams)None 5BC10BC20CP20CP + 5BC20CP + 10BC None 5BC10BC20CP20CP + 5BC20CP + 10BC60(b)Model Terms p–value sig. level % Expl. Dev.Year <0.001 (***) 34.74%Plot Height (Dry → Wet) 0.001 ** 1.86%InteractionsAmendment × Year 0.094 . 1.65%Year × Plot Height <0.001 *** 2.21%Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Relationships with negative regression slopes are indicated by parentheses.Figure 3.10: Predicted Symphyotrichum laeve biomass in the plant plug trial. Panel (a)represents the graph of raw data with error bars (± 1 SD) based on the most parsimoniouslinear mixed effects model. Experimental treatment replication = 9. The left graph panelrepresents data after one growing season. Labels on the x–axis: None = no soil amend-ment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost,5BC +20CP = 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1 biochar+ 20 T ha−1 compost. Statistical output (b) shows significant main effect terms and in-teractions. Main effects included in the model were: Amendment, AM inoculation, PlotHeight, and Year. % explained deviance is abbreviated as % Expl. Dev. in the output.Note that model terms with negative regression slopes are indicated in parentheses aroundthe significance levels.573.3. Results(a)InoculatedNot Inoculated0510September 2011 September 201215L. cylindracea Biomass(grams)None 5BC10BC20CP20CP + 5BC20CP + 10BC None 5BC10BC20CP20CP + 5BC20CP + 10BC20(b)Model Terms p–value sig. level % Expl. Dev.Year <0.001 (***) 70.96%Plot Height (Dry → Wet) <0.001 *** 1.09%InteractionsPlot Height × Year 0.045 * 0.38%Amendment × Inoculation 0.081 (.) 0.93%Year × Inoculation 0.039 (*) 0.40%Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Relationships with negative regression slopes are indicated by parentheses.Figure 3.11: Predicted Liatris cylindracea biomass in the plant plug trial. Panel (a) repre-sents the graph of raw data with error bars (± 1 SD) based on the most parsimonious linearmixed effects model. Experimental treatment replication = 9. The left graph panel repre-sents data after one growing season. Labels on the x–axis: None = no soil amendment, 5BC= 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP= 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1compost. Statistical output (b) shows significant main effect terms and interactions. Maineffects included in the model were: Amendment, AM inoculation, Plot Height, and Year. %explained deviance is abbreviated as % Expl. Dev. in the output. Note that model termswith negative regression slopes are indicated in parentheses around the significance levels.583.3. Resultsthe study (mean % cover: 1.1%, range: from 0.0% - 16.0%). When present, non–seededvegetation was dominated by patches of the perennial herb Artemisia campestris (commonname: field wormwood). Pooled C3 and C4 grasses largely dominated vegetative cover afterthree growing seasons (2013 mean % cover: 17.4%, range: 2.0% – 42.0%). The cover ofN–fixing forbs was second most abundant by the third growing season (2013 mean % cover:3.0%, range: 0.0% – 19.0%). The establishment and survival of composite forbs was sparseafter three growing seasons (2013 mean % cover: 0.4%, range: 0.0% – 7.0%), leading to anegligible contribution to total plant cover.Plant cover in the seed application trialAs main effects, compost rate (p=0.025) and growing season (p<0.001) were the mostinfluential drivers of total plant cover. Model term interactions show variable positive andnegative plant cover responses depending upon factor combination. Significant increasesin total plant cover were largely driven by plots with three-way and four-way interactionsamong biochar, compost, AM fungal inoculation, and growing season (Figure 3.12a). Theplot height covariate significantly influenced total cover where plots higher on the landscapehad more plant cover regardless of treatment when accounting for growing season (p=0.002).Biochar amendments When all other factors were held constant, increasing ratesof biochar did not significantly influence total plant cover in the seed application trial(p>0.05)(Figure 3.13). In addition, the plot height covariate and growing season did notalter the influence of biochar in the field (p>0.05). The addition of biochar, regardless ofapplication rate, resulted in no direct influence on total plant cover in this experiment.Compost amendments With all other factors held constant, increasing rates of compostwere a main driver of increasing total plant cover in the direct seeding study (p=0.025).The compost x year interaction (p=0.001) revealed a significant negative total plant coverresponse mainly driven by large variation in plots adding 40 T ha−1 compost (Figure 3.13).AM fungal inoculation No direct influence of AM fungal inoculation was detected in theseed application trial (p>0.05)(Figure 3.13). A trend of decreasing plant cover was detectedin plots adding AM fungal inoculum when accounting for growing season (p=0.065).593.3. Results(a)0%10%20%30%40%50%60%FALL 2013FALL 2011 FALL 2012Cover (%)BiocharCompostCover (%)BiocharCompostCover (%)BiocharCompostNative Plant Cover (%)(b)Model Terms p–value sig. level % Expl. Dev.Compost 0.025 * 0.92%Year <0.001 *** 8.23%InteractionsCompost × Biochar 0.062 (.) 0.64%Compost × Year 0.001 (**) 1.91%Biochar × Inoculation 0.093 (.) 0.52%Inoculation × Year 0.067 (.) 0.62%Inoculation × Plot Height (Dry → Wet) 0.075 (.) 0.58%Year × Plot Height (Dry → Wet) 0.002 (**) 1.79%Compost × Biochar × Year 0.040 * 0.77%Compost × Biochar × Inoculation 0.093 . 0.52%Compost × Year × Inoculation 0.018 * 1.03%Compost × Biochar × Inoculation × Year × Plot Height 0.012 * 1.17%Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Relationships with negative regression slopes are indicated by parentheses.Figure 3.12: Raw data wireframe graph (a) of total native plant cover in the seed applicationtrial based on the most parsimonious linear mixed effects model. Panels represent the threeanalyzed growing seasons (Fall 2011–Fall 2013). The gradient bar on the left indicatesincreasing % cover from magenta → cyan. Results are based on the most parsimoniousstatistical model. Significance levels and interactions for the model terms are given in thestatistical output table(b). y–axis = % total plant cover; x–axis = biochar rate, z–axis =compost rate. AM fungal inoculation and relative plot height are not included in the graphdue to visual complexity. % explained deviance accounts for the proportion of variationexplained by each model term. Replication = 1.603.3. Results60%Compost Rate (T ha -1 )Native Plant Cover (%)50%40%30%20%10%0%0 2.5 5 10 20 40A)60%50%40%30%20%10%0%0 2.5 5 10 20 40Biochar Rate (T ha -1 )B)60%AM InoculationNative Plant Cover (%)50%40%30%20%10%0%non-inoculatedC)inoculated60%50%40%30%20%10%0%2010 2011 2012YearD)60%Relative Plot Height (m) Native Plant Cover (%)50%40%30%20%10%0%E)0.2 0.4 0.6 0.8 1.0Figure 3.13: Diagnostic boxplots and a scatterplot for each main model term analyzing totalnative plant cover when all other factors were held constant in the seed application trial.Panels A–D are boxplots representing the raw data distribution for each categorical modelterm included in the linear mixed effect model. Panel E is a scatterplot of the relative plotheight in meters compared to total native plant cover. The surveyed plots with relativeplot height values closer to zero are higher on the landscape. At the field site, surface soilsof plots higher on the landscape were observed to dry more rapidly than plots lower on thelandscape.613.4. DiscussionSynergistic effects of biochar, compost, and AM fungal inoculation Plots withincreasing rates of biochar and compost resulted in significant increases in total plant coverwhen accounting for growing season (p=0.040). When adding R. irregularis inoculum, totalplant cover was significantly increased only when combined with increasing rates of compost+ growing season (p=0.018) or compost + biochar (p=0.093). Overall, plot inoculation withAM fungal inoculum was most effective for increasing total plant cover when combinedwith high rates of biochar and compost while accounting for plot height and growing seasoninfluences (p=0.012).Growing season Growing season explained the largest amount of variation in the model(Expl. Dev: 8.23%) (Figure 3.12a). Total % cover across all treatments was sparse afterone growing season, gradually increasing by the second and third growing seasons (mean %cover: 14.9% (2011), 17.7% (2012), and 20.9% (2013)).Plot height covariate The plot height covariate did not directly influence total plantcover in the model when all other factors were held constant (Figure 3.13). Plot heightwas a significant covariate when determining increasing plant cover in the interaction termscombining biochar, compost, AM fungal inoculation, and growing season (p=0.012). Plotswith a higher position at the field site exhibited increased plant biomass when accountingfor growing season (p=0.002).3.4 DiscussionThese field trials show that compost is the most influential driver to improve plantresponse in post-mine sandpit areas regardless of planting method. Supplementing compostwith biochar and a commercial AM fungal inoculant largely accentuates its effectiveness byfurther increasing plant response in the seed application trial. The combination of compost,biochar, and AM inoculum are effective land management tools to restore grassland plantsin severely disturbed post-mine sandpits.Plant response to AM fungal inoculationThe effectiveness of AM fungal inoculation depended upon the restoration plantingmethod in the field. Total plant biomass in the plant plug experiment was not significantlyinfluenced by R. irregularis inoculum. Although AM colonization was significantly greaterin inoculated plots in the plant plug trial, this did not translate to a consistent plantresponse. Comparatively, total cover in the seed application trial responded positivelyto the AM fungal inoculum addition only in the presence of high rates of compost and623.4. Discussionbiochar. In both trials, my hypothesis that AM fungal inoculation would directly accountfor increased total plant response in post–mine sandpits was not supported.Studies indicate that the application of AM fungi in a mine restoration setting generatespositive plant responses from seed compared to non–inoculated controls (Noyd et al. 1996,Johnson 1998, de Souza et al. 2010). But, these studies also suggest that the applicationof AM fungi does not consistently increase plant response for all plant species or restora-tion scenarios. In the seed application trial, the model suggests that AM fungal inoculumincreased total plant cover in the presence of increasing compost and compost + biocharrates. Under similar mine area conditions, Pu¨schel et al. (2011) shows that plant responsewas maximized in mine spoil banks co–amended with AM fungal isolates and organic mat-ter treatments due to increased resource availability. Contrary to this study, plant responseto AM fungal inoculum in mine areas can vary in the presence of soil conditioners such asorganic matter or fertilizer (Gryndler et al. 2008). Gryndler et al. (2008) revealed that theaddition of compost increased plant performance in reclaimed clay substrate to the detri-ment of a plant biomass effect from AM fungal inoculation. Therefore, the addition of AMfungal inoculum does not universially benefit plant growth in severely degraded mine areas.In the plant plug trial, inoculated and non–inoculated plants were colonized by R. irreg-ularis and the inoculation effect persisted in the field for two growing seasons. Yet, planthost–fungal pairings in this study yielded mixed results in terms of plant response to thecommercial inoculum. P. virgatum and L. capitata responded positively to the commer-cial inoculum while A. gerardii and L. cylindracea biomass decreased. Plant–mycorrhizalassociations range from mutualistic to parasitic depending upon the environmental contextand host species (Johnson et al. 1997). Klironomos (2003) shows that AM fungal–plantpairings are known to be unpredictable in biomass responses along a mutualistic–parasiticcontinuum. In my mixed community of grassland plants, the effectiveness of commercialinoculum was shown to be variable depending upon plant species. Thus, the universalapplication of a single AM fungal isolate may not benefit all target plants when restoringgrassland habitat.In the plant plug trial, AM fungal colonization was initially present in the non–inoculatedcontrol plugs due to non–sterile greenhouse conditions. A mycorrhizal effect induced byR. irregularis on plant response may have been significant if inoculated plant plugs werecompared to sterile, non–colonized plant plug roots. In a transplant study of Sporoboluswrightii into agricultural fields, greenhouse grown plants inoculated with AM fungi hadgreater survival, larger basal diameters, and increased tiller production after two growingseasons (Richter and Stutz 2002). That study compared plants growing in sterilized soilversus plants inoculated with AM fungi. Therefore, it is possible that the backgroundmycorrhizal community present in non–inoculated plant plugs grown in the greenhouse may633.4. Discussionhave also benefited plant growth in the field. Thus, a mycorrhizal effect from R. irregularismay be minimized due to the presence of a background AM fungal community in my study.Compared to the plant plug trial, a background community of AM fungi was not in-troduced into the seed application experiment in the non–inoculated controls. The sandpitdid not have a legacy of plant growth prior to planting. Therefore, inoculum potential inthe mine substrate was anticipated to be very low (Allen and Allen 1980, Stahl et al. 1988).Furthermore, any established hyphal networks present before the disturbance would havereduced infectivity in the severely disturbed mine system (Jasper et al. 1989). In my study,the addition of AM fungi was potentially more effective in promoting a plant response inthe seed application trial due to a lack of an infective background AM fungal communityat the site.Over time, the immigration of AM fungal propagules adapted to post–mine conditionsis expected to develop and benefit new plant recruits in degraded areas (Ganesan et al.1991). Thorne et al. (2013) found no difference in dominant prairie plant growth whencomparing AM fungi collected from a 30-year reclaimed mine spoil and a tallgrass prairiesoil. The commercial fungal isolate used in that experiment, R. irregularis, may not bewell–adapted to mine land conditions, thus exhibiting an inconsistent plant response in theplant plug trial. Taheri and Bever (2010) found that plants growing in mines are particularlydependent on strong AM fungal partners under harsh edaphic condition. Larger plantbiomass responses were induced when locally adapted AM fungi from recovering mine areaswere applied as inoculum. Inoculating plants with AM fungi isolated from sandy, post–minehabitats may increase target plant biomass due to local adaptation of the inoculum.Plant response to biocharBiochar as a solitary amendment was not an effective land management tool to promoteplant growth in either trial. Applying biochar alone reduced plant growth compared tomost compost and compost + biochar amended plots in the plant plug experiment whileexhibiting no significant difference from non–amended controls. In the seed applicationexperiment, biochar rate did not directly influence total plant cover but was an effectivetool to promote plant response when used in combination with compost and AM fungalinoculum. In both trials, my hypothesis that biochar would directly account for increasedtotal plant response in post–mine sandpits was not supported.The high cation exchange capacity of biochar may have bound nutrients in the soilsolution, thus introducing more abiotic stress into the plant plug trial (Liang et al. 2006).Positive vegetative response to biochar application in disturbed mine substrates has beenattributed to increased water holding capacity, nutrient retention, and reduced soil bulkdensity (Fellet et al. 2011, Kelly et al. 2014). These nutrient retention properties may have643.4. Discussionbeen to be detrimental to rapidly growing plants in my study. Xu et al. (2013) found thatbiochar application to sandy soils resulted in pH changes and mineral sorption to alternutrient bioavailability and reduce predicted total phosphorous. In metal contaminatedmine sites, Beesley et al. (2014) found high metal adsorption rates when co–amending soilswith biochar and compost. The potential nutrient retention (Ding et al. 2010) and release(Mukherjee and Zimmerman 2013) of biochar alone did not result in increased total plantbiomass in the plant plug trial. Further research should be conducted to determine theefficacy of applying biochar as a solitary amendment in unconsolidated mine substrates.The properties of biochar as a solitary amendment may only be beneficial dependingupon the environmental context. Experiments on biochar amendments have been largelyrestricted to agricultural soil systems and test plants have exhibited mixed growth results(Major et al. 2010, Jones et al. 2012, Filiberto and Gaunt 2013). A solitary study investi-gating biochar’s effect on a native and invasive grassland plant in a greenhouse experimentindicates increased biomass of the native plant, Andropogon gerardii, while reducing thegrowth of the invasive plant, Lespedeza cuneata (Adams et al. 2013). The results of mystudy contradict Adams et al. (2013) as biochar had no significant influence on the biomassof Andropogon gerardii. A key difference between the studies was that the greenhouseexperiment used natural soils collected from a 2–yr–old prairie restoration site on formeragricultural land. Biochar’s effectiveness as a soil amendment may be more promising whenrestoring former agricultural landscapes as compared to low quality post–mine soils.The negative effects of biochar were not evident in the seed application trial. With allother factors held constant, biochar rate had no direct effect on total native plant cover. Onereason may be that plant plug individuals grew much larger and faster than seeded plantsover the same time period. Plant competition for soil nutrient resources may have outpaceda fertilization benefit gained by low rates of biochar at the site. The increased root stockassociated with gains in aboveground plant tissues would have a high nutrient and waterdemand in the mine substrate (Craine and Dybzinski 2013). Comparatively, the growthof the grassland plants in the seed application trial was stunted over the same growingperiod. The nutrient requirements of plants in the seed application trial may have beenreduced, thus minimizing the negative effects attributed to biochar. Therefore, restorationpractitioners should approach the application of biochar to abioticially stressed mine areaswith caution. Unintended reductions in plant community response may cascade through thesystem by creating more stressful plant growth conditions after the application of biocharin the field.653.4. DiscussionPlant response to compostCompost as a solitary amendment was the most effective soil amendment to promoteplant growth in both trials. Applying compost alone promoted positive plant growth inthe plant plug trial with compared non–amended and biochar amended plots. In the seedapplication experiment, increasing the rate of compost directly contributed to total plantcover and its effectiveness increased when used in combination with biochar and AM fungalinoculum. In both trials, my hypothesis that compost would directly account for increasedtotal plant response in post–mine sandpits was supported.It is not surprising that compost resulted in improved plant growth, likely due to itswell–known fertilizer effect, its ability to increase water retention, and create higher cationexchange capacity in soils (Shiralipour et al. 1992). Similarly, compost has been shown topromote plant production from seed in other severely disturbed mine restoration scenarios(Hortenstine and Rothwell 1972, Norland and Veith 1995, Noyd et al. 1996). A one–timeapplication of compost is shown to be an effective, readily available technical reclamationtool that accelerates prairie plant growth from seed in my study. The long–term residualeffects of compost are anticipated to continue to benefit plant growth in upcoming growingseasons (Diacono and Montemurro 2010).In the seed experiment, high rates of compost were the most influential technical recla-mation tool to alleviate abiotic plant stress as evident by increased total plant cover. Areduction in plant cover was detected in plots adding compost when accounting for growingseason. Investigating the patterns in the raw data, high variability was detected in plotsadding compost at 40 T ha−1 with increasing rates of biochar. This created large variancein the final predictive model for this treatment level with all other factors held constant.As the model interaction terms increased in complexity, total plant cover values showeda positive response as biochar rates and AM fungal inoculum were incorporated into thelinear mixed effects model. Increased plot replication in the seed application trial wouldhave accounted for the inevitable natural environmental stochasticity at the field site.Synergisms among biochar, compost, and arbuscular mycorrhizasCombining biochar, compost, and AM fungal inoculum had minimal effect in the plantplug trial in terms of total plant biomass compared to non–amended controls. The highestrates of compost + biochar, 20 T ha−1 of compost + 10 T ha−1 of biochar, had the largestpositive influence on total plant biomass compared to control although not significant. Inthe seed application trial, the synergistic effect of increasing rates of compost and biocharcombined with AM fungal inoculation was effective in promoting total plant cover in thefield. My hypothesis that the synergistic effect of all amendments would account for the663.4. Discussionlargest plant response in each trial was weakly supported in the plant plug trial and stronglysupported in the seed application trial.Compost rates in the plant plug experiment may have been too low to elicit a strongplant response in the plant plug trial. Noyd et al. (1996) shows a significant increase inplant grassland plant cover in taconite mine spoils when increasing compost rates from22.4 T ha−1 to 44.8 T ha−1 after three growing seasons. In clay spoils, Pu¨schel et al. (2008a)indicates that three high compost amendment rates (100 T ha−1, 200 T ha−1, 500 T ha−1)significantly increased the flax biomass compared to controls. However, only negligible in-creases in plant biomass were detected among these compost treatments. As such, Pu¨schelet al. (2008a) suggested that lowering compost rates would be more cost effective for indus-trial applications when restoring mine spoil areas. When growing plant plugs in Ontario’ssandpits, the optimal compost application rate should be determined to maximize plantresponse at an industrial scale.In the seed application trial, the most effective amendment rates occurred when bothbiochar and compost were applied at 20 T ha−1 or greater. Studies investigating co–amendedsoils indicate mixed plant growth results when adding compost and biochar to soils (Ghoshet al. 2014, Schmidt et al. 2014). In my study, higher rates of compost and biochar inextremely degraded systems may be required to produce a larger plant response. Moreresearch needs to be conducted to determine optimal rates of biochar and compost whenrestoring plant communities in Ontario’s post-mine sandpits.The effect of biochar may have been enhanced by the concurrent addition of compost bycharging biochar surfaces and promoting a plant response in both trials (Fischer and Glaser2012). It has been shown that biochar and compost amended soils stimulates microbialgrowth and respiration rates thus enhancing decomposition rates and nutrient availabilityduring the composting process (Steinbeiss et al. 2009). Fischer and Glaser (2012) showthat plant growth generally increased with increasing amendment of biochar and compostamendments, especially in nutrient poor, sandy soil. I similarly suggest that mixing highbiochar rates in conjunction with high compost rates compliments the nutrient retentionproperties of biochar in soils when restoring post–mine sandpits.The addition of AM fungal inoculum in the seed application trial was most effectiveas compost and biochar rates increased. Therefore, the alleviation of stressful edaphicconditions by the soil amendments may have facilitated the effectiveness of the plant host–fungal pairings leading to a positive plant response in the field. In the plant plug trial,the addition of 20 T ha−1 of compost and/or 5 T ha−1 and 10 T ha−1 of biochar may havebeen too low favor a positive plant response due to AM fungal inoculation in all plantspecies. Using locally adapted inoculum may be more effective in promoting a plant responsein restoration scenarios when applying lower rates of soil amendments (Gryndler et al.673.4. Discussion2008). Further research needs to be conducted regarding the most appropriate mycorrhizalinoculum to include in the restoration of tallgrass prairie species in abandoned sandpits.In my study, plug plants were germinated under stress–free conditions in the greenhouse.In contrast, germinating plant seeds under field conditions had to overcome the stressfulenvironment of post-mine substrate. When restoring from seeds in field, plants benefitedmore from the AM fungal symbiotic association due to the alleviation of stress, especiallyin the presence of high compost and biochar rates. This was most likely due to increasedaccess to soil nutrients and more favorable water balance in the plant provided by theAM fungal symbiosis. The difference in AM fungal inoculum efficacy between plugs andseedling trials may be due to early soil conditions experienced by germinating seedlings.Spontaneous plant succession of post–mine soils is often restricted by poor soils, a lack ofbiotic symbionts, and restricted local seed immigration (Prach and Hobbs 2008). Whenspontaneous succession of plants in quarries is observed, weedy annual plant species oftenpersist (Khater et al. 2003). Thus, even with the introduction of grassland plant seedsto mine areas, plant germination would be restricted if soil amendments and AM fungalinoculum are not incorporated.In the presence of increasing amendments in the seed application trial, AM fungi hada greater access to a pool of nutrients and greater water availability that was provided bythe compost and biochar. Research by Hodge and Fitter (2010) shows that the AM fungalsymbiosis with plants improves fungal nitrogen acquisition from decomposing organic matterwhich is especially beneficial in nitrogen limited systems. The transfer of nitrogen to theplant can therefore benefit plant establishment and production when restoring of degradedmine areas (Govindarajulu et al. 2005). These results indicate that AM fungal inoculum willbe most effective after the stressful abiotic conditions are improved with soil amendments.Plant growth dynamicsIn the plant plug experiment, four of the six measured plants, (Desmodium canadense,Lespedeza capitata, Symphyotrichum laeve, Liatris cylindracea, had significantly reducedplant biomass between Fall 2011 and Fall 2012 despite the addition of soil amendments.The selection of these plant species was not optimal for long–term growth in post–minesandpits. Only the C4 grasses, Andropogon gerardii and Panicum virgatum, had significantbiomass increases during the same time period. The large increase in C4 grass biomassaccounted for the majority of the detected increase in total plant biomass between Fall2011 and Fall 2012. Long–term monitoring of these plots needs to be conducted in order totrack the potential biodiversity loss in this trial.In the seed application experiment, relative plot height had a significant influence ontotal native plant cover. Conflicting results were identified in the model depending upon683.5. Summaryinteractions among model terms. The interaction between growing season and relative plotheight led to a significant decline total native plant cover. High plant cover variation inplots adding no compost exists within this two–way interaction, thus driving this signifi-cant decline in the model. Conversely, plot height was a significant term in the five–wayinteraction among the experimental variables and growing season. The resolution of thefive–way interaction model most accurately represents the true experimental design in thistrial. I speculate that a decrease in the relative plot height would increase water availabil-ity at the site. In this trial, higher water retention would be expected as compost ratesand biochar rates increase (Aggelides and Londra 2000, Movahedi-Naeini and Cook 2000,Abel et al. 2013). As abiotic soil measurements were not collected during the study, a de-tailed soil moisture analysis needs to be determine to understand the driving abiotic factorsdetermining variation in the total plant cover measurements.3.5 SummaryThe restoration of tallgrass prairie plants in post–mine aggregate sites is a viable man-agement option in southern Ontario. Increasing grassland community diversity on marginal,anthropogenically influenced lands through prairie restoration will ensure the survival ofsensitive habitat in addition to supporting species at risk. But, the harsh edaphic charac-teristics present in the post–mine sandpits restricts plant community development in thefield. Thus, restoring self–sustaining and diverse grassland communities is not possiblewithout acknowledging the impoverished conditions of post–mine substrates.Incorporating land management tools to mitigate the harsh abiotic conditions of post–extraction substrate is therefore necessary to increase plant production in target grasslandcommunities. The technical reclamation tools investigated in this study led to higher seedestablishment rates and total plant biomass when used in combination. A single applica-tion of high rates (20 T ha−1) of biochar and compost at the onset of an industrial–scalerestoration project can lessen site maintenance costs, increase plant community recoverytime, and promote vegetative biodiversity.69Chapter 4Soil Food Webs4.1 BackgroundEcosystem productivity and fertility are characterized by organic matter inputs, primaryproduction, and microbial energy pathways (Wardle et al. 2004). The high density and di-versity of biota in soils influence these ecosystem services by altering soil water storage,litter decomposition rates, and nutrient cycles (Doran and Zeiss 2000). Ultimately, ecosys-tem services provided by microbial communities and soil animals in belowground food webscan influence plant community growth in mine areas restoration (Wardle 1999). Thus, thisstudy explored the bottom–up effects of multi–trophic group interactions in belowgroundfood webs during a grassland restoration in a post–extraction sand pit.Grassland restoration projects focused on recovering soil food webs in post–mine sand-pits must overcome harsh edaphic conditions and the lack of soil organic matter. Overtime, soil invertebrates and microorganisms recolonize restored areas without assistance,but management intervention can accelerate the establishment of desirable species (Curryand Good 1992). Incorporating soil amendments into post–mine areas can increase soilorganic matter, fertility, and water-holding capacity, ultimately influencing soil food webdevelopment. Increasing plant production with amendments and arbuscular mycorrhizascan further contribute to the development of an active plant rhizosphere, thus further stim-ulating the soil food web (Kuzyakov 2002).4.1.1 Arbuscular mycorrhizal fungal inoculumThe plant–AM fungal association extends the biologically active zone around plant roots(i.e. rhizosphere) to incorporate the influence of the plant symbiont (i.e mycorrhizosphere).Soil areas under the influence of the mycorrhizosphere have rapid water and nutrient uptake,high concentrations of exudates, high root turnover, and increased respiration (Garbaye1991). The plant-AM fungal symbiosis has been shown to enhance rhizobial N–fixationby legumes (Amora-Lazcano et al. 1998) and increase bacterial populations (Johanssonet al. 2004) compared to non–mycorrhizal plants. Therefore, soil animals may benefitfrom increased microbial activity associated with the mycorrhizosphere due to higher foodavailability. Thus, increased plant root biomass and mycorrhizosphere activity due to AM704.1. Backgroundfungal inoculation can translate into greater soil food web production in these zones of highroot exudates and plant root turnover.4.1.2 BiocharBiochar as a soil amendment is anticipated to prime fungal and bacterial biomass byproviding carbon substrates, retaining soil macro–nutrients, and/or providing suitable mi-crobial refugia (Lehmann et al. 2011, Lou et al. 2014). Lower temperature biochar (250 ◦Cto 400 ◦C) is anticipated to most improve soil fertility and stimulate microbial communities(Novak et al. 2009).The mechanisms of biochar’s influence on microbial biomass and soil animal abundanceare understudied (McCormack et al. 2013). Thus, it is difficult to predict the effects ofbiochar in post–mine restoration on soil biotic communities due to the dearth of research andthe complexity of biological and physical interactions in soil. But, I anticipate that biochar’sprojected physiochemical benefits in soils will increase plant growth and soil microbialcommunities, thus increasing food resources for higher trophic levels in the soil food web.4.1.3 CompostCompost increases microbial community biomass, soil respiration rates, and soil enzymeactivity by providing bacteria and fungi with decomposable substrate (Allievi et al. 1993).Compost has been shown to favor the development of fungal–dominated systems by addingcomplex organic matter to degraded systems (Biederman and Whisenant 2009, Biederman2013). Microbial growth and soil fertility are closely related as compost is decomposed bythe microbial community. Thus, the decomposition of compost releases important elements(N, P) into the soil solution to be taken up by organisms (Frankenberger and Dick 1983).Compost can be considered a bio–inoculant with an associated community of fungi andbacteria, nematodes, and microarthropods (Cernova 1970, Streit et al. 1985, Steel et al.2013a;b). As soil animal communities are severely diminished in post–mine habitats, theaddition of compost is anticipated to give compost amended areas a soil food web headstart. Furthermore, increased microbial biomass associated with compost should providesoil animals with an increased food supply. Thus, compost is expected to be an essentialsoil amendment crucial to the development of soil food webs.4.1.4 Synergisms among biochar, compost, and arbuscular mycorrhizasThis study is the first to test the concurrent application of biochar, compost, and com-mercial inoculum on soil food webs. Although conceptually recommended to incorporatebiochar and compost simultaneously (Fischer and Glaser 2012), no studies have researched714.2. Methodsthe soil food web response to the application of compost, biochar, and AM fungal inoculumin a restoration setting. As compost is expected to positively ameliorate soils, co–amendedsoils with biochar can further increase water and nutrient retention. Increased mycorrhizo-sphere activity and plant root biomass associated with AM fungal inoculation is anticipatedto further accelerate the development of soil food webs in post–mine sandpits. Thus, co–amended soils with biochar, compost, and AM fungal inoculum are expected to exhibit thegreatest increase in soil food web response.4.1.5 HypothesesTo my knowledge, the relationship among compost, biochar, arbuscular mycorrhizal(AM) fungal inoculation, and the soil food web has never been tested when restoring grass-land vegetation in post–mine sandpits. I hypothesized that compost, biochar, and AMfungal inoculation would individually increase soil microbial biomass compared to non–amended controls. In addition, I hypothesized that the concurrent addition of compost,biochar and AM fungal inoculum would yield the largest response in plant functional groupbiomass and soil microbial community biomass. The rationale for these hypotheses is thatthe amendments and AM fungal inoculation would alleviate of water and nutrient stress andincrease mycorhizosphere activity in post–mine sandpit substrates. A structural equationmodel was constructed to describe the direct, indirect, and total effects driving soil foodweb response among plant functional biomass, soil organisms abundance, and soil amend-ments (i.e. compost, biochar, and AM fungal inoculum). As a guide to create my structuralequation model, trophic interactions from Hunt et al. (1987) form the basis in determiningrelationships within soil food webs. The goal of this study was to prescribe industriallyfeasible abiotic and biotic soil amendments to facilitate soil development and determine thetrajectory of soils in recovering post–mine sandpits when restoring grassland communities.4.2 MethodsI tested the fully factorial effects of soil amendments (biochar, compost) and AM fungalinoculation on soil microbial biomass and soil animal abundance in the post-mine sandpit.Soils were collected from plots in the plant plug trial after two growing seasons. Fullexperimental design details of the plant plug trial are described in Chapter 3: Methods.4.2.1 Soil collection and organismal analysesIn September 2012, sixteen soil cores (2.54 cm diameter) were collected to a depth of12 cm from each plot in the plant plug experiment. Soil cores were sampled directly adjacent724.2. Methodsto aboveground plant tissue for each of the six C4 grasses, five N–fixing forbs, and fivecomposite forbs within the core sampling area. Soil cores were sampled directly adjacent toeach plant at each plug location to minimize plant destruction. The soil corer was cleaned ofsubstrate with a clean cloth and water between plot sampling to minimize contamination.Collected soils were pooled at the plot level and homogenized. Soil samples were storedin a cooler on ice in the field until final storage at 4◦C. Soil microbial biomass and soilanimal abundance were analyzed at the Soil Analysis Laboratory, University of California,Riverside.Bacterial and fungal biomassBacterial and fungal biomass was estimated by differential fluorescent staining (DFS)following an adapted protocol by Klironomos et al. (1996). The DFS was composed of amixture of europium(III)thenoyl–trifluoroacetonate and a fluorescent intensifier (Andersonand Westmoreland 1971). For fungal biomass estimation, 200 ml of soil was suspended with1 ml of DFS stain for 1 hour. Once stained, the suspension was filtered through nitrocellulosefilter paper using a 50% ethanol wash. Filters were then mounted on microscope slides forvisual inspection under UV light at 620 nm. Active cellular material was visually highlightedwith red fluorescence under UV light. Fungal biomass was calculated from images takenby computer imaging software. Hyphal length was measured to estimate milligrams (mg)of fungal biomass kilogram (kg)−1 soil using conversion factors [hyphal diameter: 1.65micrometers (µm)(Kjøller and Struwe 1982), density: 0.33 g cm−3 (van Veen and Paul 1979),C content: 45% (Swift et al. 1979)].To estimate bacterial biomass, soil dilution aliquots were stained with DFS for 1 hour.After staining, filters were rinsed with a 50% ethanol wash and slides mounted for visualinspection using UV microscopy at 620 nm. Active cellular material was visualized by redfluorescence and images taken with computer imaging software. Bacterial biomass wasestimated using a conversion factor of 6.4 × 10−14 g carbon cell−1 calculated by Hunt andFogel (1983). Results are given in mg of bacterial biomass kg−1 soil.Nematode enumerationNematodes were extracted by the same wet sieve sucrose centrifuge approach for extract-ing arbuscular mycorrhizal spores as described in Klironomos et al. (1993). Soil sampleswere suspended in water and passed through a series of mesh sieves decreasing in pore size(1.0 mm – 45µm). After rinsing with water, the material retained in the 45µm sieve wassuspended on top of a 60% sucrose solution and centrifuged for 20 minutes. Nematodes werecollected via a pipette at the sucrose–water interface. Nematode individuals were sorted734.2. Methodsand counted under a microscope. Individuals were classified into one of three functionalfeeding groups: bacterial feeding, fungal feeding, and predatory. Categorization was basedupon nematode morphological characteristics. Abundance is reported as # of individualsg−1 soil.Soil arthropod enumerationA high efficiency canister–type soil arthropod extractor (Lussenhop 1971) was used toextract mites and Collembola onto dishes containing picric acid as described in Klironomoset al. (1996). Soil arthropods were counted and classified into Collembola, microbial feeding(Oribatid) mites, and predatory mites based on morphological characteristics. Abundanceis reported as # of individuals g−1 soil.4.2.2 Statistical analyses for soil biotaLinear models were used to test treatment–level effects on bacterial and fungal biomass.Fungal and bacterial biomass data approximated a normal data distribution after a logtransformation. Generalized linear models with a negative binomial distribution link func-tion were used to test the treatment effects on soil animal abundance (i.e. nematodes,Collembola, mites). As with most count data, all soil animal abundance data displayedcharacteristics of over–dispersion (the variance of the response variable exceeded the mean)and is best analyzed using a generalized linear models (Bolker et al. 2009).Linear and generalized linear models for each soil functional group were reduced usingmodel selection procedures. Full models including treatment factors, covariates, and inter-actions were iteratively reduced to remove non–significant variables using χ2 tests. Thisresulted in the most parsimonious model to test differences in the response variable. Lin-ear models were analyzed using the base package in R (R-Core-Team 2013). Generalizedlinear models were analyzed with the glm.nb function from the MASS package to calculateestimates from a negative binomial distribution.4.2.3 Soil food web analysis with structural equation modelingStructural equation modeling was used to test multivariate hypotheses and their in-terdependencies among soil functional groups, experimental treatments, and plant func-tional group biomass. A priori soil food web hypotheses, developed from prior literatureknowledge, were determined among exogenous variables (i.e. soil amendments, AM fungalinoculation) and endogenous variables (i.e. biota in the soil food web, plant functionalgroup biomass)(Figure 4.1). Before running the analysis, a covariance matrix of relation-744.2. Methodsships among all variables suggested that log transformations were appropriate to betterapproximate normality of residuals.Plant biomass was estimated by partial least squares regression (PLS) for C4 grasses, C3grasses, and N–fixing forbs (See Chapter 3 for full details). Plant functional group biomasswas pooled at the plot–level in September 2012 and used as a variable in the structuralequation model.Structural equation models were conducted in IBM’s SPSS program extension AMOS.Overall fit of the a priori hypotheses was tested by evaluating χ2 tests and the comparativefit index (CFI). An acceptable fit of a model is indicated by non–significant p–values inχ2 tests and CFI values over 0.93 (Byrne 2013). Generalized least squares (GLS) with abootstrap correction was used to calculate squared multiple correlations and standardizedpath coefficients. Squared multiple correlations were calculated for each endogenous vari-able to determine explained variance by incoming predictor variables. Standardized pathcoefficients were calculated from maximum likelihood distributions based on each variable’sstandard deviations (Grace 2006). Standardized path coefficients are interpreted as theexpected change in response variable for each unit increase of the explanatory variable.Three structural equation models were compared by selecting the model with the lowestAIC value to remove non–significant explanatory variables in the soil food web. Removingnon–significant variables increases statistical power and improves structural equation modelparsimony. Note that predatory and Oribatid mites were pooled in all models due to lowcorrelations and large variances in this data.Model 1: Exogenous variables: compost rate, biochar rate, AM fungal inoculation;Endogenous plant biomass variables: composite forbs, N-fixing forbs, C4 grassesModel 2: Exogenous variables: compost rate, biochar rate; Endogenous plantbiomass variables: composite forbs, N-fixing forbs, C4 grassesModel 3: Exogenous variables: compost rate, biochar rate; Endogenous plantbiomass variables: N-fixing forbsOnly biologically relevant standardized direct pathways with estimates greater than 0.1or less than -0.1 were included in Figure 4.5 – Figure 4.12. Pathway significance (p<0.05)and trends (0.05<p<0.1) are indicated by colored arrows. Non–significant pathways (p>0.1)are indicated by dashed lines. All direct, indirect, and total standardized regression coeffi-cients are given in Tables 4.1 – Table 4.3.754.2. MethodsFigure 4.1: A priori hypotheses used to construct the most parsimonious structural equationsoil food web model (Model 3). Exogenous variables are displayed in shaded gray boxes.Endogenous variables are displayed in white boxes. The residual error associated with eachendogenous variable is displayed as (ε). Single headed arrows indicate direct pathways.764.3. Results4.3 Results4.3.1 Soil food web structural equation model selectionModel #1 (χ2=17.381, df=13, p=0.182; CFI=0.955; AIC=201.4), Model #2(χ2=17.276, df=11, p=0.100; CFI=0.940; AIC=177.3), and Model #3 (χ2=7.9, df=5,p=0.162; CFI=0.970; AIC=129.9) had an acceptable fit for testing a priori hypothesesas shown by non–significant χ2 comparisons and CFI values > 0.93. Nevertheless, Model#3 (Exogenous variables: soil amendment rate; Endogenous plant biomass variables: N-fixing forbs) was chosen as the final model because it had the highest statistical power andparsimony as shown by lowest AIC values. Therefore, AM fungal inoculation, compositeforb biomass, and C4 grass variables were removed from the final model as they did nothave sufficient explanatory power. Inoculated and non–inoculated plots were pooled by soilamendment rate (replication = 18).4.3.2 Microbial community biomass and soil animal abundanceThe influence of AM fungal inoculum on soil biotaThe response of the soil microbial community to the AM fungal inoculation of plantplugs was not significant in this study. No significant effect of AM fungal inoculation wasdetected for fungal biomass or the fungal:bacterial ratio as the AM fungal inoculation termwas dropped from both linear models. Plots inoculated with AM fungi exhibited a positivetrend in bacterial biomass at the study site (p=0.085)(Figure 4.2a).Fungal feeding nematode (p=0.035) and Collembola (p=0.043) abundance was signif-icantly reduced in plots inoculated with AM fungi (Figure 4.3a & 4.4a respectively). Allother grazing and predatory soil animals were not significantly influenced by R. irregularisinoculation.The influence of biochar on soil biotaBiochar’s influence on soil microbial community biomass and soil animal abundancewas consistently negative compared to non–amended control plots. Although not alwayssignificant, both biochar rates reduced microbial biomass and soil animal abundance in allcases as evident by negative coefficient estimates in the statistical output of the models.Biochar as a solitary amendment had no significant direct impact on the soil microbialbiomass, soil animal abundance, or N–fixing forb biomass in the structural equation model(Figure 4.5 & Table 4.1).774.3. Results(a)Bacterial Biomass(mg kg -1 soil)InoculatedNot Inoculated0.01.02.03.0None 5BC10BC20CP20CP + 5BC20C P + 10BC(b)Model Terms Estimate SE p–value sig. levelIntercept 0.312 0.033 <0.001 ***5BC -0.116 0.044 0.010 (**)10BC -0.109 0.044 0.015 (*)20CP 0.064 0.044 0.149 n.s.20CP + 5BC 0.190 0.044 <0.001 ***20CP + 10BC 0.194 0.047 <0.001 ***AM Fungi 0.044 0.025 0.085 .Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Significantly different intercepts with negative values in parenthesesFigure 4.2: Bacterial biomass collected during the second growing season of the prairierestoration (September 2012). Data were analyzed with linear models to test treatment–level effects. Panel (a) represents raw data ± 1 SD; n = 9. x–axis: None = no soil amend-ment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost,5BC +20CP = 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1 biochar+ 20 T ha−1 compost. Significant main effect terms and interactions shown in (b). Modelterm estimates represent the expected change from the model intercept (i.e. control plots).784.3. Results(a)Fungal Feeding Nematodes(abundance)InoculatedNot Inoculated0204050None 5BC 10BC 20CP20CP + 5BC20CP  + 10BC1030(b)Model Terms Estimate SE p–value sig. levelIntercept 0.400 0.837 0.633 n.s.5BC -0.844 1.205 0.484 n.s.10BC -10.130 10.204 0.321 n.s.20CP 1.301 1.114 0.243 n.s.20CP + 5BC 2.390 1.034 0.021 *20CP + 10BC 1.620 1.067 0.129 n.s.AM Inoculation -5.700 2.703 0.035 (*)Interactions5BC × AM Inoculation 5.741 3.069 0.061 .10BC × AM Inoculation 17.930 10.559 0.090 .20CP × AM Inoculation 5.938 2.917 0.042 *20CP + 5BC × AM Inoculation 5.960 2.931 0.042 *20CP + 10BC × AM Inoculation 7.006 2.860 0.014 *AM Inoculation × Plot Height 0.134 0.062 0.030 *5BC × AM Inoculation × Plot Height -0.163 0.070 0.021 (*)10BC × AM Inoculation × Plot Height -0.331 0.192 0.086 (.)20CP × AM Inoculation × Plot Height -0.134 0.071 0.059 (.)20CP + 5BC × AM Inoculation × Plot Height -0.165 0.067 0.014 (*)20CP + 10BC × AM Inoculation × Plot Height -0.184 0.068 0.007 (**)Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Significantly different intercepts with negative values in parenthesesFigure 4.3: Fungivorous nematode abundance collected during the second season (Septem-ber 2012). Generalized linear models with a negative binomial distribution link functionwere used to test the treatment effects. Panel (a) represents raw data ± 1 SD; n = 9. x–axis:None = no soil amendment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP =20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP= 10 T ha−1 biochar + 20 T ha−1 compost. Significant main effect terms and interactionsshown in (b). Model estimates represent the expected change from the intercept.794.3. Results(a)Collembola(abundance)InoculatedNot Inoculated02550None 5BC10BC20CP20CP + 5BC20C P + 10BC(b)Model Terms Estimate SE p–value sig. levelIntercept 1.591 0.456 <0.001 ***5BC 0.071 0.478 0.882 n.s.10BC -0.560 0.492 0.256 n.s.20CP -0.105 0.477 0.825 n.s.20CP + 5BC 1.562 0.469 <0.001 ***20CP + 10BC 2.151 0.461 <0.001 ***AM Inoculation -1.071 0.530 0.043 (*)InteractionsAM Inoculation × Plot Height 0.029 0.014 0.032 *Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Significantly different intercepts with negative values in parenthesesFigure 4.4: Collembola abundance collected during the second growing season of the prairierestoration (September 2012). Generalized linear models with a negative binomial distribu-tion link function were used to test the treatment effects. Panel (a) represents raw data ± 1SD; n = 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1 biochar, 10BC = 10 T ha−1biochar, 20CP = 20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar + 20 T ha−1 compost,10BC +20CP = 10 T ha−1 biochar + 20 T ha−1 compost. Significant main effect terms andinteractions shown in (b). Model term estimates represent the expected change from themodel intercept (i.e. control plots).804.3. Results0.13-0.14-0.12-0.13-0.130.160.150.350.230.250.710.94 0.620.61 0.830.870.320.64 0.63-0.18-0.44-0.16-0.32 -0.18-0.23-0.39-0.18R 2  = 0.477R 2  = 0.312R 2  = 0.482R 2  = 0.413R 2  = 0.446R 2  = 0.332R 2  = 0.393 R 2  = 0.347Figure 4.5: Structural equation soil food web model for the plant plug experiment. Exoge-nous variables are displayed in shaded gray boxes. Endogenous variables are displayed inwhite boxes. The residual error associated with each endogenous variable is displayed as(ε). Structural equation model line weights are scaled to the direct pathway standardizedregression estimates given in each boxes. Blue (positive) and red (negative) arrows indicatesignificant standardized regression estimates (p < 0.05). Yellow (positive) and orange (neg-ative) arrows indicate trends in standardized regression estimates (0.05 < p < 0.1). Dashedlines are non–significant paths with standardized regression estimates > 0.1. Regressionestimates < 0.1 are not included to simplify the data presentation. A full description ofdirect, indirect, and total model estimates are given in Table 4.1 – 4.3. Squared multiplecorrelations are reported within endogenous variable boxes. Squared multiple correlationswere calculated for each endogenous variable to determine explained variance.814.3. ResultsTable 4.1: Direct, indirect, and total standardized regression estimates of soil amendmentson the soil community and N–fixing forbs generated by the structural equation model.Significant direct pathway estimates are given in bold text (p < 0.05).Observed Effects(λ)Predictor → Response Direct Indirect TotalAmendments → MicrobesBC → bacterial biomass -0.18 -0.01 -0.18BC → fungal biomass -0.13 0.00 -0.12CP → bacterial biomass 0.15 -0.03 0.12CP → fungal biomass 0.15 0.01 0.17BC + CP → bacterial biomass 0.64 -0.04 0.60BC + CP → fungal biomass 0.63 0.02 0.65Amendments → N–fixing forbsBC → N–fixing forbs 0.07 0.00 0.07CP → N–fixing forbs 0.35 0.00 0.35BC + CP → N–fixing forbs 0.61 0.00 0.61Amendments → NematodesBC → bact. feeding nematodes -0.07 0.02 -0.05BC → fungal feeding nematodes -0.05 0.06 0.01BC → predatory nematodes -0.02 0.01 -0.02CP → bact. feeding nematodes 0.13 -0.01 0.12CP → fungal feeding nematodes 0.23 -0.07 0.16CP → predatory nematodes 0.32 -0.11 0.21BC + CP → bact. feeding nematodes 0.71 -0.05 0.66BC + CP → fungal feeding nematodes 0.87 -0.28 0.59BC + CP → predatory nematodes 0.83 -0.30 0.54Amendments → MicroarthropodsBC → Collembola -0.04 0.02 -0.02BC → mites -0.13 0.04 -0.10CP → Collembola 0.16 -0.14 0.02CP → mites 0.07 -0.08 -0.01BC + CP → Collembola 0.94 -0.45 0.49BC + CP → mites 0.62 -0.19 0.43824.3. ResultsWhen applied as a solitary amendment, both biochar rates negatively influenced bacte-rial biomass compared to non–amended control plots (5 T ha−1 biochar, p=0.010; 10 T ha−1biochar, p=0.015)(Figure 4.2a). Fungal biomass was not significantly influenced by eitherbiochar rate in this study (5 T ha−1 biochar, p=0.441; 10 T ha−1 biochar, p=0.119)(Figure4.6a). Across all soil amendment treatments, the only significant increase in fungal:bacterialratios compared to non–amended controls occurred at the 5 T ha−1 biochar amendment rate(p=0.002)(Figure 4.7a).Biochar’s influence on soil nematode abundance was variable, ranging from a neutralto significantly negative responses. Bacterial feeding nematode abundance was significantlyreduced in the 10 T ha−1 biochar application rate (p=0.005) but unaffected in 5 T ha−1biochar treatments when compared to non–amended controls (Figure 4.8a). Fungal feedingnematode abundance was not directly influenced by biochar (5 T ha−1 biochar, p=0.484;10 T ha−1 biochar, p=0.321)(Figure 4.3a). No predatory nematodes were detected in eitherbiochar amendment rates (Figure 4.9a).Direct effects of biochar on the soil arthropods was not predictable after generalizedlinear model analysis. Oribatid mite abundance was significantly reduced in plots withboth biochar rates (5 T ha−1 biochar, p=0.014; 10 T ha−1 biochar, p=0.047) while Collem-bola abundance (5 T ha−1 biochar, p=0.882; 10 T ha−1 biochar, p=0.256) and predatorymite abundance (5 T ha−1 biochar, p=0.146; 10 T ha−1 biochar, p=0.363) was unaffectedby biochar application rate(Figures 4.4a, 4.10a, & 4.11a).Bacterial biomass tended to be reduced in the presence of biochar in the structural equa-tion model (direct pathway coefficient = -0.18, p=0.059) (Figure 4.12). Soil fungal biomass(p=0.171) and soil animal abundance (bacterial feeding nematodes (p=0.497); fungal feed-ing nematodes (p=0.648); predatory nematodes (p=0.827); Collembola (p=0.723); mites(p=0.219)) remained largely unaffected by the introduction of biochar in the post–minesandpit. Although not significant, all direct pathway coefficients were negative for soil mi-crobial biomass and soil animal abundance relationships suggesting a negative impact onthe soil food web compared to control plots (See Table 4.1 for direct pathway values).The influence of compost on soil biotaThe effect of 20 T ha−1 compost on the soil food web was largely neutral in this post–minerestoration trial compared to non–amended plots. Only fungal biomass was significantlyincreased by compost addition in the study (p=0.010) (Figure 4.6a). Bacterial biomass(p=0.149) and fungal:bacterial ratios (p=0.822) were not influenced by the addition ofcompost. Among soil animals, the abundance of predatory nematodes was significantlyincreased by 20 T ha−1 of compost (p<0.001). All other soil animals had no significantresponse to the compost amendment compared to non–amended controls (bacterial feeding834.3. Results(a)Fungal Biomass(mg kg -1 soil)0.01.02.03.04.0None 5BC10BC20CP20CP + 5BC20C P + 10BC(b)Model Terms Estimate SE p–value sig. levelIntercept 0.333 0.033 <0.001 ***5BC -0.036 0.046 0.441 n.s.10BC -0.073 0.046 0.119 n.s.20CP 0.122 0.046 0.010 **20CP + 5BC 0.252 0.046 <0.001 ***20CP + 10BC 0.261 0.046 <0.001 ***Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Figure 4.6: Fungal biomass collected during the second growing season of the prairie restora-tion (September 2012). Data were analyzed with linear models to test treatment–level ef-fects. Panel (a) represents raw data ± 1 SD; n = 9. x–axis: None = no soil amendment, 5BC= 5 T ha−1 biochar, 10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP= 5 T ha−1 biochar + 20 T ha−1 compost, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1compost. Significant main effect terms and interactions shown in (b). Model term estimatesrepresent the expected change from the model intercept (i.e. control plots).844.3. Results(a)Fungal:Bacterial RatioNone 5BC10BC20CP20CP + 5BC20CP + 10BC0246(b)Model Terms Estimate SE p–value sig. levelIntercept 0.267 0.226 0.238 n.s.5BC 0.847 0.274 0.002 **10BC 0.343 0.299 0.251 n.s.20CP 0.069 0.311 0.822 n.s.20CP + 5BC 0.094 0.309 0.760 n.s.20CP + 10BC 0.055 0.311 0.860 n.s.Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Figure 4.7: Fungal:bacterial biomass ratios collected during the second growing season of theprairie restoration (September 2012). Generalized linear models with a negative binomialdistribution link function were used to test the treatment effects. Panel (a) represents rawdata ± 1 SD; n = 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1 biochar, 10BC =10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar + 20 T ha−1compost, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1 compost. Significant main effectterms and interactions shown in (b). Model term estimates represent the expected changefrom the model intercept (i.e. control plots).854.3. Results(a)Bacterial Feeding Nematodes(abundance)InoculatedNot Inoculated0204060None 5BC 10BC 20CP20CP + 5BC20CP  + 10BC(b)Model Terms Estimate SE p–value sig. levelIntercept 2.448 0.712 0.001 ***5BC -1.380 0.952 0.147 n.s.10BC -3.412 1.210 0.005 (**)20CP -0.316 0.960 0.742 n.s.20CP + 5BC 1.814 0.893 0.042 *20CP + 10BC 0.684 0.915 0.046 *Plot Height (Dry → Wet) -0.052 0.025 0.040 (*)Interactions10BC × AM Inoculation 4.642 1.668 0.005 **5BC × Plot Height 0.058 0.029 0.046 *10BC × Plot Height 0.100 0.034 0.003 **20CP × Plot Height 0.060 0.036 0.096 .20CP + 10BC × Plot Height 0.060 0.0305 0.049 *AM Inoculation × Plot Height 0.071 0.033 0.030 *10BC × AM Inoculation × Plot Height -0.169 0.047 <0.001 (***)20CP × AM Inoculation × Plot Height -0.076 0.045 0.090 (.)20CP + 10BC × AM Inoculation × Plot Height -0.093 0.041 0.025 (*)Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Significantly different intercepts with negative values in parenthesesFigure 4.8: Bacteriovorus nematode abundance collected during the second growing seasonof the prairie restoration (September 2012). Generalized linear models with a negative bino-mial distribution link function were used to test the treatment effects. Panel (a) representsraw data ± 1 SD; n = 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1 biochar,10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar+ 20 T ha−1 compost, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1 compost. Signifi-cant main effect terms and interactions shown in (b). Model term estimates represent theexpected change from the model intercept (i.e. control plots).864.3. Results(a)Predatory Nematodes(abundance)051015None 5BC10BC20CP20CP + 5BC20C P + 10BC(b)Model Terms Estimate SE p–value sig. levelIntercept -2.140 0.749 0.004 **5BC - - - -10BC - - - -20CP 3.401 0.796 <0.001 ***20CP + 5BC 3.959 0.792 <0.001 ***20CP + 10BC 4.184 0.791 <0.001 ***Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Figure 4.9: Predatory nematode abundance collected during the second growing season ofthe prairie restoration (September 2012). Generalized linear models with a negative bino-mial distribution link function were used to test the treatment effects. Panel (a) representsraw data ± 1 SD; n = 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1 biochar,10BC = 10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar+ 20 T ha−1 compost, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1 compost. Signifi-cant main effect terms and interactions shown in (b). Model term estimates represent theexpected change from the model intercept (i.e. control plots).874.3. Results(a)Microbial Feeding Mites(abundance)0204060None 5BC10BC20CP20CP + 5BC20C P + 10BC(b)Model Terms Estimate SE p–value sig. levelIntercept 1.386 0.273 <0.001 ***5BC -1.041 0.424 0.014 (*)10BC -0.811 0.407 0.047 (*)20CP -0.028 0.386 0.942 n.s20CP + 5BC 1.843 0.370 <0.001 ***20CP + 10BC 0.905 0.375 0.016 *Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Note: Significantly different intercepts with negative values in parenthesesFigure 4.10: Oribatid mite abundance collected during the second growing season of theprairie restoration (September 2012). Generalized linear models with a negative binomialdistribution link function were used to test the treatment effects. Panel (a) represents rawdata ± 1 SD; n = 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1 biochar, 10BC =10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar + 20 T ha−1compost, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1 compost. Significant main effectterms and interactions shown in (b). Model term estimates represent the expected changefrom the model intercept (i.e. control plots).884.3. Results(a)Predatory Mites(abundance)051015None 5BC10BC20CP20CP + 5BC20C P + 10BC(b)Model Terms Estimate SE p–value sig. levelIntercept -0.118 0.419 0.778 n.s.5BC -0.981 0.674 0.146 n.s.10BC -0.575 0.632 0.363 n.s.20CP -0.724 0.564 0.893 n.s.20CP + 5BC 1.545 0.549 0.005 **20CP + 10BC 1.658 0.548 0.002 **Significance: *** ≤ 0.001 | ** ≤ 0.010 | * ≤ 0.050 | . ≤ 0.100Figure 4.11: Predatory mite abundance collected during the second growing season of theprairie restoration (September 2012). Generalized linear models with a negative binomialdistribution link function were used to test the treatment effects. Panel (a) represents rawdata ± 1 SD; n = 9. x–axis: None = no soil amendment, 5BC = 5 T ha−1 biochar, 10BC =10 T ha−1 biochar, 20CP = 20 T ha−1 compost, 5BC +20CP = 5 T ha−1 biochar + 20 T ha−1compost, 10BC +20CP = 10 T ha−1 biochar + 20 T ha−1 compost. Significant main effectterms and interactions shown in (b). Model term estimates represent the expected changefrom the model intercept (i.e. control plots).894.3. Results-0.14-0.12-0.13-0.13-0.18-0.44-0.16-0.32 -0.18-0.23-0.39-0.18R 2  = 0.477R 2  = 0.312R 2  = 0.482R 2  = 0.413R 2  = 0.446R 2  = 0.332R 2  = 0.393 R 2  = 0.347Figure 4.12: Negative standardized regression estimates in the soil food web model for thegrassland restoration plant plug experiment Exogenous variables are displayed in shadedgray boxes. Endogenous variables are displayed in white boxes. Structural equation modelline weights are scaled to the direct pathway standardized regression estimates given ineach boxes. Red arrows indicate significant standardized regression estimates (p < 0.05) andorange arrows indicate trends in standardized regression estimates (0.05 < p < 0.1). Dashedlines are non–significant paths with standardized regression estimates > 0.1. Regressionestimates < 0.1 are not included to simplify the data presentation. A full description ofdirect, indirect, and total model estimates are given in Table 4.1 – 4.3. Squared multiplecorrelations are reported within endogenous variable boxes. Squared multiple correlationswere calculated for each endogenous variable to determine explained variance.904.3. Resultsnematodes, p=0.742; fungal feeding nematodes, p=0.243; Collembola, p=0.825; Oribatidmites, p=0.942; predatory mites, p=0.893).As a solitary amendment, compost had a significant positive influence on the soil foodweb for several biotic variables in the structural equation model (direct pathway coefficients:N–fixing forb biomass = 0.35 (p<0.001); predatory nematode abundance = 0.32 (p=0.004);fungal feeding nematode abundance = 0.23 (p=0.026))(Figure 4.13 & Table 4.1). A positivetrend in increasing fungal biomass was detected in plots with compost only addition (directpathway coefficients = 0.15 (p=0.089)).Synergistic effects of biochar, compost, and AM fungal inoculationThe synergistic interaction of compost and biochar had a large positive effect on the soilmicrobial community, but no such interaction occurred with AM fungal inoculation and soilamendments. Regardless of biochar rate, compost + biochar significantly increased bacterialand fungal biomass (p<0.001) compared to non–amended controls (Figures 4.2a & 4.6a).Biochar + compost significantly increased all biotic variables (i.e. soil microbial communitybiomass, soil animal abundance, and N–fixing forb biomass) in the soil food web structuralequation model (all p–values<0.001)(Figure 4.5 & Table 4.1). No significant change infungal:bacterial ratios was detected in the compost + biochar treatments compared to non–amended controls.Soil animal abundance responded positively in plots with added compost + biochar.Bacterial feeding nematode abundance increased significantly in both compost + biochartreatments (20 T ha−1 compost + 5 T ha−1 biochar, p=0.042; 20 T ha−1 compost +10 T ha−1 biochar, p=0.046) while fungal feeding nematode abundance was only signifi-cantly increased in the 20 T ha−1 compost + 5 T ha−1 biochar treatment (p=0.021) andno significant response in the 20 T ha−1 compost + 10 T ha−1 biochar treatment (p=0.129)(Figures 4.3a & 4.8a). Predatory nematode abundance was significantly increased in bothcompost + biochar treatments compared to controls (p<0.001)(Figure 4.9a). A consistentpositive response in fungal feeding nematode abundance was also detected when adding AMfungal inoculum in conjunction with all amendment rates (Figure 4.3a).Collembola abundance (20 T ha−1 compost + 5 T ha−1 biochar, p<0.001; 20 T ha−1compost + 10 T ha−1 biochar, p<0.001), Oribatid mite abundance (20 T ha−1 compost +5 T ha−1 biochar, p<0.001; 20 T ha−1 compost + 10 T ha−1 biochar, p=0.016), and preda-tory mite abundance (20 T ha−1 compost + 5 T ha−1 biochar, p<0.005; 20 T ha−1 compost+ 10 T ha−1 biochar, p=0.002) responded positively to both compost + biochar treatments(Figures 4.4a, 4.10a, & 4.11a).The standardized direct pathway coefficients in biochar + compost treatments weredemonstrably larger (range of direct pathway coefficients: N–fixing forb biomass (0.61) –914.3. Results0.130.160.150.350.230.250.710.94 0.620.61 0.830.870.320.64 0.63R 2  = 0.477R 2  = 0.312R 2  = 0.482R 2  = 0.413R 2  = 0.446R 2  = 0.332R 2  = 0.393 R 2  = 0.347Figure 4.13: Positive standardized regression estimates in the soil food web model for thegrassland restoration plant plug experiment. Exogenous variables are displayed in shadedgray boxes. Endogenous variables are displayed in white boxes. The residual error asso-ciated with each endogenous variable is displayed as (ε). Structural equation model lineweights are scaled to the direct pathway standardized regression estimates given in eachboxes. Blue arrows indicate significant standardized regression estimates (p < 0.05) andyellow arrows indicate trends in standardized regression estimates (0.05 < p < 0.1). Dashedlines are non–significant paths with standardized regression estimates > 0.1. Regression es-timates < 0.1 are not included to simplify the data presentation. A full description ofdirect, indirect, and total model estimates are given in Table 4.1 – 4.3. Squared multiplecorrelations are reported within endogenous variable boxes. Squared multiple correlationswere calculated for each endogenous variable to determine explained variance.924.3. ResultsCollembola abundance (0.94)) compared to biochar only and compost only plots (Table4.1). In the biochar + compost treatments, large indirect effects were detected due tonegative interactions among the soil animals (Tables 4.1 & 4.3). Thus, the biochar +compost treatments resulted in reduced total effects in the soil food web model comparedto direct effects (range of total pathway coefficients: mite abundance (0.43) – bacterialfeeding nematode abundance (0.66)). Comparatively, the total effects of the biochar +compost treatments are consistently more influential on all soil biota than compost alone(range of total pathway coefficients: mite abundance (-0.01) – N–fixing forb biomass (0.35))and biochar alone (range of total pathway coefficients: bacterial biomass (-0.18) – N–fixingforb biomass (0.07))(Table 4.1).The influence of plot height on soil biotaRelative plot height had no significant direct effect on the majority of soil food webbiota. Only bacterial feeding nematode abundance had a direct negative response to de-creasing height of plots on the landscape. The interaction of soil amendments, AM fungalinoculation, and decreasing plot height indicated a consistent reduction in fungal feedingnematode abundance (Figure 4.3a). Collembola abundance increased significantly with plotheight and the inoculated plots (p=0.043).The influence of plant functional group biomass on the soil food webModel #2 containing the C4 grass and composite forb biomass variables had an ac-ceptable fit to the proposed a priori hypotheses (χ2=17.276, df=11, p=0.100; CFI=0.940;AIC=177.3). Compared to Model #3 (AIC=129.9), dropping the C4 grass and compos-ite forb biomass variables greatly improved model performance Therefore, these variableswere dropped from the final structural equation model. N–fixing biomass was significantlyincreased as a result of compost addition (direct pathway coefficient: 0.35, p<0.001) andbiochar + compost addition (direct pathway coefficient: 0.61, p<0.001). Although soilamendment rate influenced N–fixing biomass, no significant direct effect of N–fixing biomasswas observed on soil microbial biomass or soil animal abundance (Table 4.2).The influence of bacterial and fungal biomass on soil animal abundanceNo significant covariance effect was detected between bacterial biomass and fungalbiomass (direct pathway coefficient: 0.00, p=0.950)(Table 4.2). Bacterial biomass didnot significantly influence the abundance of soil animals in the study (bacterial feed-ing nematodes (p=0.393), predatory nematodes (p=0.292); Collembola (p=0.197); mites(p=0.476))(Table 4.2). Comparatively, fungal biomass significantly reduced fungal feed-934.3. ResultsTable 4.2: Direct, indirect, and total standardized regression estimates of soil microbesand N–fixing forbs on the soil community generated by the structural equation model.Significant direct pathway estimates are given in bold text (p < 0.05).Observed Effects(λ)Predictor → Response Direct Indirect TotalN–fixing forbs → Soil CommunityN–fixing forbs → bacterial biomass -0.07 0.00 -0.07N–fixing forbs → fungal biomass 0.04 0.00 0.04N–fixing forbs → bact. feeding nematodes 0.00 0.01 0.01N–fixing forbs → fungal feeding nematodes 0.01 -0.02 0.00N–fixing forbs → predatory nematodes -0.16 0.01 -0.15N–fixing forbs → Collembola 0.00 0.06 0.06N–fixing forbs → mites 0.00 0.04 0.04Bacteria → Soil Communitybacterial biomass → fungal biomass 0.00 0.00 0.00bacterial biomass → bact. feeding nematodes -0.09 0.00 -0.09bacterial biomass → predatory nematodes -0.12 0.00 -0.12bacterial biomass → Collembola -0.14 0.04 -0.10bacterial biomass → mites -0.08 0.02 -0.06Fungi → Soil Communityfungal biomass → bacterial biomass 0.00 0.00 0.00fungal biomass → fungal feeding nematodes -0.44 0.00 -0.44fungal biomass → predatory nematodes -0.01 0.10 0.10fungal biomass → Collembola 0.04 0.14 0.18fungal biomass → mites -0.08 0.01 -0.07944.3. ResultsTable 4.3: Direct, indirect, and total standardized regression estimates among the soilanimals generated by the structural equation model. Significant direct pathway estimatesare given in bold text (p < 0.05).Observed Effects(λ)Predictor → Response Direct Indirect TotalGrazing nematodes → Predatory nematodesbacterial feeding nematodes → predatory nematodes 0.02 0.00 0.02fungal feeding nematodes → predatory nematodes -0.23 0.00 -0.23Grazing nematodes → Microarthropodsbacterial feeding nematodes → Collembola 0.02 -0.01 0.01bacterial feeding nematodes → mites -0.21 0.00 -0.21fungal feeding nematodes → Collembola -0.39 0.07 -0.32fungal feeding nematodes → mites 0.04 -0.04 0.00Predatory nematodes → Microarthropodspredatory nematodes → Collembola -0.32 0.00 -0.32predatory nematodes → mites -0.18 -0.08 -0.26Collembola → MitesCollembola → mites 0.25 0.00 0.25ing nematode abundance (direct pathway coefficient: -0.44, p<0.001), but did not signifi-cantly influence any other soil animal group (predatory nematodes (p=0.962); Collembola(p=0.724); mites (p=0.507)) (Table 4.2).Soil animal functional feeding group interactionsSignificant negative direct effects were observed in the proposed feeding hierarchy basedon my a priori hypotheses (Figure 4.12 & Table 4.3). Fungal feeding nematode abundancewas negatively correlated with predatory nematode abundance (direct pathway coefficient:-0.23, p=0.033) and Collembola abundance (direct pathway coefficient: -0.39, p<0.001). Inaddition, predatory nematode abundance had a significant negative effect on Collembolaabundance (direct pathway coefficient: -0.32, p<0.001). Negative trends were detected be-tween bacterial feeding nematodes→ mites (direct pathway coefficient: -0.18, p=0.055) andpredatory nematodes → mites (direct pathway coefficient: -0.18, p=0.076). The only sig-nificant, positive direct path coefficient detected in soil animal interactions was Collembola→ mites (direct pathway coefficient: 0.25, p=0.015).954.4. Discussion4.4 DiscussionAfter two growing seasons, this study showed that soil amendments can significantly af-fect soil microbial biomass and soil animal abundance in post–mine sandpits. I found thatAM fungal inoculation of plant plug roots had little influence on soil food web structurecompared to non–inoculated controls. As a solitary amendment, biochar had a largely neg-ative effect on the soil food web although this effect was not always significant. Conversely,biochar mixed with compost promoted large significant increases in the soil microbial com-munity and soil animal abundance. Thus, I clearly show that soil food web developmentis highly dependent upon amendment choice during grassland restoration in the degradedsandpit.4.4.1 Soil food web response to AM fungal inoculationN–fixing plant biomass, soil microbial biomass, and soil animal abundance were unaf-fected by AM fungal inoculation. The structural equation model clearly showed that AMfungal inoculation did not have sufficient explanatory power to describe any direct effectinfluences on soil microbial biomass and soil animal abundance. Thus, my hypothesis thatAM fungal inoculation of plant plugs would increase the response of microbial communitybiomass and soil animal abundance in post–mine sandpits was not supported.Arbuscular mycorrhizal fungi are key components of the soil microbiota and interactwith other microorganisms in the rhizosphere (Bowen and Rovira 1999). AM fungi caninfluence belowground soil food webs by increasing plant root biomass through nutrientacquisition, therefore increasing litter inputs (Langley and Hungate 2003). Jastrow et al.(1998) determined strong positive direct and indirect effects by AM fungi on fine roots,microorganisms, and soil aggregation using path analysis. Increased microbial biomass andrhizosphere activity can subsequently support soil nematodes and microarthropods throughbottom–up cascading mechanisms (Scherber et al. 2010).One reason for no effect may be due to the presence of a background AM fungal commu-nity in non–inoculated plant plug roots at the time of planting due to unsterilzed greenhouseconditions in the nursery. Rowe et al. (2007) has suggested that locally collected field inocu-lum is more effective than commercial inoculum for establishing late-successional species.This background community of local AM fungi may have formed a strong partnership withthe plants used in this restoration, thus contributing to a positive plant biomass responsein the field. In desertified semi–arid systems, the establishment of the shrub with localvs. non–native AM fungal inoculum showed increases in soil enzyme activity compared tocontrols but no difference between inoculum source (Alguacil et al. 2005). Therefore, theinfluence of the AM fungal inoculum on the development of the soil food web may have964.4. Discussiongone undetected because of a beneficial belowground plant response by the background AMfungal community in the non–inoculated plots. Thus, the commercial AM fungal inoculum,R. irregularis, was not an effective land management tool to increase total plant biomass,plant functional group biomass, or soil food web biomass and abundance in this study.4.4.2 Soil food web response to biocharBiochar had neutral to negative influence soil microbial biomass and soil animal abun-dance in this study. My hypothesis that both biochar amendments would positively increasesoil microbial biomass, soil animal abundance, and N–fixing plant abundance due to ame-liorative effects in sandpit substrate was not supported.Biochar application as a land management tool has been proposed to assist soil recov-ery in severely degraded systems (Blackwell et al. 2009). To date, most mine reclamationstudies using biochar have investigated soil chemical properties under laboratory conditions(i.e. pH, cation exchange capacity, heavy metal sequestration) (Fellet et al. 2011; 2014,Kelly et al. 2014). When investigated, the response of soil microbes to biochar in the lit-erature are mixed. Kelly et al. (2014) found that microbial biomass was not altered bybiochar amendments in mine tailings while a meta–analysis of plant and microbial biomassby Biederman and Harpole (2012) found that biochar addition increased aboveground pro-ductivity, crop yield, soil microbial biomass, and favorable tissue macro–nutrients across allsoil types and climates. Graber et al. (2010) suggested that shifts in soil microbial activitywere indirect and arose from biochar stimulating plant growth, thus inducing a plant exu-date effect in the rhizosphere. My study suggests that this mechanism is unlikely as plantbiomass decreases were not detected by the structural equation model yet decreases in thesoil microbial community were detected.In my study, biochar may have introduced nutrient stress associated with post–minesandpits, reducing microbial biomass and soil animal abundance. One potential mechanismfor reduced biotic response in reduced soil microbial abundance is biochar’s high cation ex-change capacity strongly adhering limited nutrients in post–mine soils (Steiner et al. 2007,Xu et al. 2013). Ultimately, feedstock source and pyrolysis time determines nutrient leach-ing rates, chemical properties, and hydrophobicity of biochar in soils, thus dictating a soilbiotic response (Singh et al. 2010, Kinney et al. 2012). Furthermore, biochar’s hydrophobicnature may have repelled soil moisture, causing negative trends in soil microbial biomass(Kinney et al. 2012). Complex biogeochemical interactions will ultimately determine re-source availability for biotic communities and may need to be optimized for soil conditions.The long–term ecological effect of biochar application needs to be investigated in terms ofsoil community development and plant response under field conditions.974.4. DiscussionReductions in soil animal abundance were also detected when applying biochar althoughnot significant. This is most likely due to reduced fungal and bacterial biomass creatinga limiting microbial food resource for grazing soil animals. Mikola and Seta¨la¨ (1998) es-tablished a trophic dynamic microcosm experiment testing the soil interactions among mi-crobes, microbivorous nematodes, and predatory nematodes. This study suggested thatincreased microbial productivity leads to the increased biomass of microbes followed by alagging response time in the microbivorous nematode trophic level. As nematode recoveryafter severe disturbance is slow (Bongers and Ferris 1999), the reduced abundance of graz-ing soil animals in biochar only plots most likely contributed to the minimized abundanceof predatory soil animals. To date, biochar’s influence on soil food web structure is rela-tively unknown (Lehmann et al. 2011). My study results are contrary to the hypothesesproposed by McCormack et al. (2013) where the addition of biochar was anticipated to in-crease microbial and soil animal resource availability. The direct influence of biochar on themulti–trophic interactions warrants further study in restoration and agriculture. Biochar’suse as a land management tool to assist soil food web development is questionable as shownby my results.4.4.3 Soil food web response to compostCompost application did not affect most of the organisms in my study. While it in-creased the abundance of N–fixing plant biomass, fungal biomass, and predatory nematodeabundance, I could not detect an influence of compost in other groups. This is contrary tomy hypothesis which predicted that compost would increase soil microbial biomass and soilanimal abundance due to ameliorated soil conditions.Compost was expected to promote increased soil microbial community growth due to im-proved nutrient and water retention profiles in compost amended soils (Bastida et al. 2008,Larney and Angers 2012). Long–term and short–term studies indicate that urban compostprimes microbial community decomposition and increases plant–available macro–nutrientsin agricultural soils (Weber et al. 2007, Hadas and Portnoy 1997). Jones et al. (2010)concluded that compost additions in bauxite–processing residue sand positively influencedwater retention and nutrient profiles, thus increasing soil microbial activity. The influenceof organic amendments has been shown to favor the growth of fungal community comparedto bacteria communities (Jastrow et al. 2007). Fungi are more efficient decomposers ofcompost amendments compared to bacteria due to large hyphal networks and efficient nu-trient acquisition and translocation mechanisms (Lucas et al. 2014). My study confirmsthat fungal biomass was significantly increased by the addition of compost compared to apositive, but non–significant influence on bacterial biomass.984.4. DiscussionAn important indicator in re–establishing a soil microbial community is the relativeproportions of bacterial and fungal biomass (Bardgett and McAlister 1999) with naturalgrassland systems being dominated by fungal communities (Harris 2009). Mummey et al.(2002) suggested that fungal:bacterial ratios in mine spoils can approach ratios in naturalsoils after 20 years following restoration although total biomass is comparatively reduced.Compost addition in my study did not significantly influence fungal:bacterial biomass ra-tios compared to non–amended controls. This is surprising as the compost amendment re-sulted in significant increases in fungal biomass but did not significantly influence bacterialbiomass. This result suggests that soil conditions improved due to the compost amendmentbut gains in fungal biomass were not pronounced compared to gains in bacterial biomass. Asfungal:bacterial biomass ratios were only measured after two growing seasons, these ratiosare expected to increase as above– and belowground litter inputs accumulate, ultimatelyfavoring fungal dominance in the soil food web (Holtkamp et al. 2008).I had anticipated that compost would have a larger influence on soil food web struc-ture due to the alleviation abiotic stress in impoverished soils, increased microbial foodresources, and more feeding substrate. The addition of farm composts containing cropresidue and manure showed increased fungal feeding nematodes in a soil incubation studydue to increased food resources (Steel et al. 2012). Jørgensen and Hedlund (2013) showedthat Collembola and predatory mites had increased fecundity when adding a fungal inocu-lated clover amendment to soils, highlighting the importance of fungal biomass for grazinganimal fecundity and prey attraction.After two growing seasons, my study showed that no direct influence of compost wasdetected on soil animal abundance. This is surprising as compost significantly increasedfungal biomass, a food source for fungal feeding nematodes, Collembola, and Oribatidmites. As Collembola and Oribatid mites also consume litter, the addition of compostedplant material had little influence on population densities. Therefore, predatory nematodesand mites subsequently had low population densities most likely attributed to low preyabundances. Several studies have linked rates of organic matter mineralization to microbialproduction and the biomass of soil microbivores and predators (Seastedt 1984, Bardgettet al. 1998, Laakso et al. 2000).As increases in resource availability drives microbial production (Baer et al. 2003), thechosen compost rate may have been too low to overcome the harsh abiotic sandpit condi-tions, resulting in a subdued response in the soil food web. Thus, I speculate that largestobstacle for soil food web development is the harsh conditions in post–mine areas. Increas-ing the amount of organic matter may have a stronger influence on the development ofthe soil food web by further ameliorating soil conditions and increasing food resources forgrazing soil animals.994.4. Discussion4.4.4 Soil food web response to compost and biocharSoil food web responses were much more pronounced when compost and biochar wereapplied together. My hypothesis that co–amending soils with compost + biochar and AMfungal inoculation would positively influence soil microbial biomass and soil animal abun-dance was partially supported in my study. The compost + biochar treatments had a largepositive influence on soil food web development but an effect of AM fungal inoculation wasnot detected.This is the first study to investigate the influence of compost and biochar on soil food webstructure. As suggested by Fischer and Glaser (2012), a synergistic interaction of compost+ biochar can positively influence soil conditions, leading to a large positive impact on soilfood web structure. Potential mechanisms may be increased water and nutrient retention,buffered pH, or creation of microbial refugia in biochar’s highly porous structure with a largenutrient pulse supplied by compost. Fischer and Glaser (2012) also indicate that compostmay charge biochar’s surfaces to slowly release nutrients to soils, increase soil aeration,and reduce leaching losses. These soil amelioration mechanisms may have overcome theabiotic conditions associated with post–mine sandpits, significantly increasing soil microbialbiomass followed by soil animal abundance compared to solitary amendments or controls.Food resources are key to the development of multi–trophic belowground food webs(Hunt et al. 1987). Increased knowledge regarding the linkages within decomposer food websrequires an understanding of the importance of resource availability upon the growth andabundance microbial communities and the associated consumer trophic levels (Wardle 2006).Adding compost with biochar clearly influenced bacterial and fungal biomass by creatinga bottom–up trophic cascade effect within the soil food web. By co–mixing amendments,large increases in soil microbial communities translated to increased abundance of grazingnematodes, Collembola, Oribatid mites. Increased abundance of grazing soil animals createda prey resource for predatory nematodes and mites leading to a more complete trophichierarchy in the soil food web. Thus, the soil environment with compost + biochar wasmore tolerable for the growth and development of a belowground soil microbial communities,inducing a positive response in soil nematode and microarthropod abundance.Compost + biochar treatments did not influence fungal:bacterial biomass ratios whencompared to control plots even though substantial increases in fungal and bacterial biomasswere detected. The addition of organic amendments to severely degraded areas drivespositive changes in microbial activity as estimated by soil microbial biomass carbon (Roset al. 2003). This suggests that an influence of increased resource availability may haveequally benefited the growth of both fungal and bacterial functional groups. As biocharpersists in soils for over 100+ years (Lehmann et al. 2009), the growth benefits gained by1004.4. Discussionsoil microbial communities and belowground plant response from biochar’s physiochemicalproperties are anticipated to be long–term (Glaser et al. 2002).My study results clearly confirm that the development of soil microbial communitiesand soil animals is significantly enhanced when biochar and compost are used in tandem.When used as a land management tool, co–amending soils with compost + biochar canaccelerate the development of soil food webs, indicating that post–mine substrate recoveryis substantial when compared to non–amended controls. As the soil food web is developed,functioning decomposition and nutrient cycles can translate into greater plant response inthe field due to the ecosystem services provided by multi–trophic interactions of soil biota(de Vries et al. 2013). Thus, when a researcher approaches restoration with a holistic ecosys-tem perspective, the incorporation of compost + biochar is an essential to soil regenerationin degraded mine areas.4.4.5 Interactions among soil microbial biomass and soil animalabundanceThis study showed that increasing fungal biomass significantly reduced fungal feedingnematode abundance while increasing Collembola abundance. Bacterial biomass had nosignificant effect on bacterial feeding nematode and Collembola grazers. These relationshipsin the fungal and bacterial energy pathways channels did not coincide with my hypothesisthat increased food resources in the soil microbial community would correlate to highergrazing soil animal densities.As described previously, amendment choice had a direct influence on microbial biomassand soil animal density with compost + biochar amendments significantly increasing allsoil animals in the belowground food web. These results indicate that the soil edaphicconditions improved the growth of food resources for grazing and predatory animals. Con-versely, when investigating direct relationships in the fungal and bacterial energy channels,unexpected patterns emerged. Increasing fungal biomass resulted in a significant nega-tive correlation with fungal feeding nematodes while no significant correlation was detectedbetween bacterial biomass and bacterial feeding nematodes.Bacterial and fungal nematode response to soil amendments is not equivocal due todifferent life history strategies in these groups (Ferris and Bongers 2006). Generally, bac-terial feeding nematodes have short life cycles and high reproductive potential to quicklyrespond to bacterial blooms in soils (Bongers and Ferris 1999). Fungal feeding nematodes,on the other hand, are longer–lived and reproduce more slowly compared to bacterial feed-ing nematodes, thus are less likely to respond to changing conditions (Ferris et al. 2001).Disturbance severity will ultimately dictate the composition of bacterial vs. fungal feedingnematodes with severe disturbance shifting soil systems towards bacterial feeding energy1014.4. Discussionchannels (Bongers 1990). Thus, in the fungal pathway, the negative correlation betweenfungal biomass and fungal feeding nematodes may be due to a lag in rapid nematode re-sponse to the available fungal food resource. Bacterial colonizers present in the sandpitmay have a more ephemeral response to bacterial biomass leading to no direct correlationbetween these trophic levels. Long–term monitoring of the grazing nematodes populationsis needed to determine the recovery trajectory of this system.Furthermore, Yeates et al. (1993) points out that nematode feeding group identificationmay not be well delineated in practice. Bacterial feeding nematodes are generally classi-fied by having a wide mouth, but bacterial feeders have been known to feed upon fungalfood resources (Gupta et al. 1979). Fungal feeding nematodes are classified as possessinga stylet but some genera with this feature are known to feed on plant roots or be verte-brate predators (Bongers and Bongers 1998). The large nematode family, Tylenchidae, iscommonly considered root feeding nematodes but have been shown to feed on fungal foodresources (Okada et al. 2005). Distinguishing between plant feeding nematodes and fungalfeeding nematodes is important to understand interactions between food resources. In mystudy, nematodes classified as fungal feeding nematodes may feed upon plant roots in therhizosphere, leading to unexpected correlations between fungal biomass and fungal feedingnematodes by inaccurately attributing food resource being consumed.Fungal biomass had a positive total effect on predatory nematode abundance. Con-versely, bacterial biomass had a negative total effect on predatory nematode abundance.Wardle et al. (1995) showed that top predatory nematodes were regulated by microbialbiomass while fungal and bacterial feeding nematode responses were more variable in thebelowground interactions. Li et al. (2014) found that organic enrichment in an agriculturalsetting shifted grazing fungal dominance to the fungal energy channel while increasingpredatory nematodes. In my study, the ephemeral life cycle response of bacterial feedingnematodes may not be a stable food source for predatory nematodes. Thus, the longer–lived fungal feeding nematodes in the fungal energy channel may be a more nutritious foodsource with a more stable population to support nematode predators.Compost + biochar soil amendments showed large increases in soil microarthropodsabundance compared to controls. This indicates that a food resource is available to supportthese soil animals in the food web. When investigating the relationship in the structuralequation model, Collembola responded positively to increases in fungal biomass while anegative response was detected for bacterial biomass although no strong relationship ex-ists. Soil mites had no strong response to the microbial community. Results from fieldstudies in microarthropod populations are often ambiguous as invertebrate population sizecan increase or decrease with amendment type (Bardgett and Cook 1998, Jørgensen andHedlund 2013). Complex trophic interactions present in soil food webs occur as a number1024.4. Discussionof direct and indirect interactions occur between species is difficult to predict (Bengtssonet al. 1996). More research needs to be conducted on the reliability of fungal and bacterialbiomass measurements when determining resource availability for grazing soil animals andpredators.Sandpit resource extraction strongly diminishes populations of soil animals in the sys-tem. Slow recovery may be expected as the soil environment develops over time and responseof microarthropods can be highly variable due to environmental heterogeneity (Curry andGood 1992, Menta 2012). As evident in control plots, soil animal abundances were ex-tremely low after two growing seasons. In former agricultural lands, colonization of newareas is unpredictable and responds differently to successional changes in plant communities(Scheu and Schulz 1996, Korthals et al. 2001). Most soil organisms are considered to havelimited abilities of overcoming soil heterogeneity and have restricted movement (Ojala andHuhta 2001). In my study system, compost + biochar amended plots ultimately improvedsoil conditions to support a higher abundance of soil microbes and soil animals comparedto control plots. Thus, as soil organisms disperse to the field site, these amended plotsare anticipated to better support the survival of newly arriving immigrants compared tounamended plots.To accelerate recovery of soil food webs, post–mine sandpits may benefit from an in-oculation of soil food web biota. Assuming soil animals survive the composting process,it is reasonable to conclude that compost would act as soil food web inoculant that con-tained a high abundance of bacteria, fungi, nematodes, and soil microarthropods (Cernova1970, Streit et al. 1985, Steel et al. 2013b). After two growing seasons, compost additionsalone unexpectedly did not alter soil animal abundance significantly as shown by marginallyimproved soil animal densities compared to non–amended plots. Conversely, improved bio-geochemical conditions of compost + biochar amended soils may have allowed for soil animalsurvival associated with compost. Further research on the potential of using compost as asoil food web inoculum should be conducted under various restoration scenarios.4.4.6 SummaryBased on my results, the recovery of soil food webs when restoring grasslands in post–mine aggregate sites is a viable management option in southern Ontario. As ecosystemproductivity and soil fertility are closely tied to soil biota, land managers should targetthe development of soil food webs in tandem with phyto–centric goals to maximize plantproduction in a restoration project. As shown in this study, mining sand strongly reducessoil microbial communities and soil animal abundance even after two years of habitat re-covery with grassland plant plugs. The harsh substrate conditions in non–amended controlplots suggests that the recovery time of a soil food web would be slow if no management1034.4. Discussionaction was implemented. Therefore, land management tools are necessary to accelerate thedevelopment of functioning soil food webs in severely disturbed habitats.In my study, the technical reclamation tools (i.e. compost, biochar, and AM fungal in-oculation) induced a variable response within the soil food web. The application of biocharalone added stress to the post–mine substrate and further restricted soil food web devel-opment in the field while compost had a negligible effect compared to control plots. Incontrast, co–amending soils with compost + biochar led to large increases in soil food webdevelopment in the field. The application of compost + biochar in an industrial–scalerestoration project should promote increases in soil microbial biomass and soil animal pro-duction leading by improving soil conditions at the site. Increasing the function of soil foodwebs can ultimately drive aboveground plant community production due to the ecosystemservices provided. These ecosystem services can lead to reduced site maintenance costs,increase plant community recovery time, and promote vegetative biodiversity.104Chapter 5Management Recommendations forGrassland Restoration inPost–Extraction SandpitsRestoring a grassland plant community is challenging when attempting to recreate nat-ural habitat in post–mine sandpits. Native plant growth in sandpits is hampered by stress-ful abiotic conditions and disrupted connections among plants–microbes–soil animals at-tributed to severe disturbance and low organic matter. As shown in this study, only C4grasses and N–fixing forbs responded positively in sandpit substrate during the plant plugand seed application trials. Composite forbs and C3 grasses exhibited poor plant responsein the study, regardless of treatment. Thus, recreating highly diverse prairie ecosystemsremains a challenge, even after addressing the harsh conditions of sandpit substrate usingsoil amendments and AM fungal inoculum.My results show that soil amelioration can benefit plant response when restoring grass-land vegetation as plugs or seeds. When directly seeding in sandpits, significant increasesin plant response were achieved by concurrently amending soils with high rates of biocharand compost (20 T ha−1 and 40 T ha−1) and the recommended rate of the commercial AMfungal inoculum, Rhizophagus irregularis. In the plant plug trial, no significant differencesin total plant biomass were detected in plots adding 20 T ha−1 of compost + 10 T ha−1 ofbiochar although a positive trend was indicated. In this case, the rate of 20 T ha−1 of com-post + 10 T ha−1 of biochar may have been too low to create a strong positive plant growthresponse in the plug experiment. But, in terms of soil food web development, the largebiotic response of fungi, bacteria, and soil animals to 20 T ha−1 of compost + 10 T ha−1of biochar in the soil food web indicates more favorable conditions for soil microbial andanimal growth after two growing seasons. Therefore, the addition of high rates of biocharand compost improved soil conditions to accelerate soil food web development and increaseplant response in the field compared to non–amended and non–inoculated control plots.1055.1. Plant species selection and sourcing5.1 Plant species selection and sourcingIncorporating plant material as seed or plugs is essential when recreating prairie–likehabitat in post–mine sandpits to reach plant community targets. When left to naturalplant recolonization, control areas (i.e. no plant plugs added) were sporadically colonizedby weedy, ephemeral plants with low biomass. Native seed recruitment was minimal inthese control areas even with the incorporation of soil amendments in non–vegetated plots.Therefore, incorporating native plant material as seed or plant plugs is essential whenrestoring grassland habitat in post–mine aggregate sites.The decision to rehabilitate prairies with native plant seeds or plugs will be determinedby desired speed of recovery and future maintenance considerations. Seeding the landscapeincorporates drawback such as:− slower and less successful plant establishment− possible increased time to achieve rehabilitation certification− increased site maintenance requirements (i.e. reseeding applications)− increased influence of weedy, invasive plant species (i.e. herbicide applications maybe necessary)The upfront cost of sowing native plant plugs with soil amendments is initially morecost prohibitive than direct seeding (Table 5.1). The advantage of restoring with plantplugs is high plant biomass production by the C4 grass and N–fixing forbs compared toseed growth over a similar growing period. Accelerated growth rates in the plant plug trialcan increase soil stabilization by binding substrate with native plant roots and reducingwind scour. From personal observation, plant plug addition reduced surface erosion bywind energy immediately at the time of plug installation. Compared to the plant plug trial,plant growth in the seed application trial was stunted after three growing seasons comparedto plants starting as plugs. Thus, integrating plug installation with native seeded may bea cost effective hybrid technique to minimize seed loss and stabilize the mine substrate ata restoration site.This restoration project used locally–collected seed mixtures which were adapted toregional growing conditions. Locally–sourced plant material has been suggested to positivelyinfluence plant response in restoration projects with greatest success in soils experiencinglower disturbance (Lesica and Allendorf 1999, Buisson et al. 2006). In southern Ontario,high diversity seed mixes can range from 10 - 30 plant species to include a mixture of warmseason (C4) grasses, cool season (C3) grasses, legumes (i.e. nitrogen–fixing forbs), and1065.1. Plant species selection and sourcingcomposite wild flowers (Delaney et al. 2000). As shown by this study, C4 grasses and N–fixing forbs will likely dominate the vegetative community in a sandpit restoration projectafter several years of plant growth. Further research on seed mixture proportions needs tobe investigated to achieve the best results for creating a high diversity plant community inpost–extraction sandpits.This study shows that plant response to soil amendments and AM fungal inoculationvaries among plant species. Therefore, plant selection must be considered on a case by casebasis. The environmental tolerance of each plant species to amended post–mine substrateconditions favored some species, while being a detriment to another. All plant species,except for Andropogon gerardii, had a neutral to positive response to the compost or compost+ biochar amendment addition. All plant species growing in biochar only treatments had aneutral to negative biomass response compared to control. Panicum virgatum and Lespedezacapitata biomass was significantly greater in AM inoculated compared to non–inoculatedplants while the opposite was true for Andropogon gerardii. In general, the addition ofcompost and biochar amendments benefited the growth of most species in the post–minesubstrate by alleviating abiotic stress. Conversely, plant response to the commercial AMinoculum was dependent upon species and planting method thus its use in tallgrass prairierestoration is context–dependent.The composite forbs, Symphyotrichum laeve and Liatris cylindracea, did not performwell in either restoration trial. In the seed application trial, composite forb cover wasnegligible after three growing seasons. In the plug trial, significant reductions in compositeforb biomass were exhibited in the field after two growing seasons. The chosen compositeforb species in this trial were not ideal candidates for this post–mine restoration. In thefollowing years, I expect that these species will be non–existent in the two trials.In contrast, the C4 grasses and N–fixing forbs tolerated the post–mine sandpit environ-ment and were responsive to the applied soil amendments. These species were the largestcontributor to total plant biomass in the field. Thus, sand pit restoration should include amixture of these plant functional groups to increase plant community biomass.Inconsistent plant species responses highlights the need to have clearly stated goals inrestoration management plans when recreating prairies in severely disturbed areas. If totalplant response is the key component to determine restoration success, then the applicationof biochar, compost, and AM inoculum would be an effective tool to assist the restorationof grassland plants in post–mine sandpits. If a practitioner is targeting a specific suite ofplant species and managing for species at risk, higher caution must be used when choosingamendments and AM fungal inoculum. Target plants may be adversely affected by theaddition of soil amendments in the field and/or choice of mycorrhizal inoculum. Context–dependent abiotic and biotic scenarios ultimately determine the success of each restoration.1075.1. Plant species selection and sourcing5.1.1 Soil amendments and commercial AMF inoculumCommercial AM fungal inoculumThe arbuscular mycorrhizal inoculum, Rhizophagus irregularis, was most effective duringseed application when co–amended with high rates of compost and biochar. No significanteffects on the total plant biomass and soil food web development were detected in the plantplug experiment. This may be a result of background AM fungal present in unsterilizedplant plug soils in the commercial greenhouse.As shown by other mine land reclamation studies, AM fungal inoculum benefits thegrowth of plants in severely degraded mine areas (Rao and Tak 2002, Rydlova´ et al. 2008, Wuet al. 2009) but plant response due to AM fungal inoculation can be enhanced by the additionof soil amendments (Gryndler et al. 2008, Pu¨schel et al. 2008a). My research supports thefinding that adding AM fungal inoculum enhances plant response when amended withcompost in the seed trial. Co-amending soils with increasing rates of compost and biocharfurther facilitates the plant growth in this system. In this sandpit restoration, adding acommercial AM inoculum is appropriate when establishing grassland plants from seed whenapplying soil amendments. When growing plants from plugs in an unsterilzed greenhousesetting, my results show that land managers do not need to apply AM fungal inoculum ifbiochar and compost are not added as no significant plant response was detected in thistrial after two growing seasons.Biochar as a soil amendmentBiochar has been shown to be a beneficial land management tool to enhance plant pro-duction in tropical agriculture (Major et al. 2010). To date, no research has been conductedon the role of biochar in the restoration of grasslands in severely degraded post–mine areas.My results show that the incorporation of biochar as a solitary amendment for grasslandrestoration in sandpits should not be used. The negative responses detected in plant pluggrowth and soil food web development indicate that biochar addition further stresses thesubstrate and restricts the development of biota in the recovering system. Comparatively,high rates of biochar had no effect on total plant cover after three growing seasons in thedirect seeding trial. Therefore, solitary biochar addition should be approached with cautionas plant response can be hampered in a restoration project. More research needs to beconducted for the most appropriate restoration scenarios to add biochar with the goal ofincreasing plant response.1085.2. Purchasing soil amendments and inoculum for a restoration projectCompost as a soil amendmentCompost has been shown to increase plant response in degraded mine areas by alleviatingstressful abiotic conditions in many studies (Noyd et al. 1996, Pu¨schel et al. 2011). My studyindicates that compost positively influences plant growth in the plant plug trial although notsignificant compared to non–amended controls. In the seed application trial, high levels ofcompost addition is the main driver of total plant cover. If a land manager is presented withan amendment choice, compost as a solitary amendment outweighs the use of biochar asa soil amendment. Incorporating 20 T ha−1 to 40 T ha−1 of compost into sandpit substratehas the largest potential to positively influence soil conditions as shown by increased plantresponse in grassland vegetation.Synergism among biochar, compost, and AM fungal inoculumThe results of this study indicate that the concurrent addition of municipal compost,biochar and mycorrhizal inoculum are simple land management tools that improve plantperformance and soil food web development in post–extraction aggregate sites. In the plantplug experiment, 20 T ha−1 of compost mixed with 10 T ha−1 of biochar had the highestpositive effect on plant biomass, soil microbial biomass, and soil animal abundance. Thus,the amelioration of stressful abiotic conditions in the sandpit was achieved when compostand biochar were used together in the plant plug trial. AMF inoculation combined with highrates of compost (20 T ha−1 to 40 T ha−1) and low rates of biochar (20 T ha−1 to 40 T ha−1)resulted in the highest plant cover in the seed experiment.The rates of biochar and compost need to be optimized to achieve the highest plant re-sponse at industrially feasible costs. My results suggest that low rates of biochar (5 T ha−1and 10 T ha−1) combined with higher rates of compost (20 T ha−1 and 40 T ha−1) mayachieve significant responses in the plant community while being cost effective. As thecost of biochar is substantially higher than that of municipal compost, adding 20 T ha−1and 40 T ha−1 of biochar is not a cost effective amendment at this time (Table 5.1). Fundsto restore grassland vegetation in a sandpit could be more effectively used by incorporatingplant plugs with a high diversity seed mixture.5.2 Purchasing soil amendments and inoculum for arestoration projectRhizophagus irregularis can be purchased as a seed coat powder from Myke Pro R©(www.usemykepro.com) and applied at the rate suggested by the manufacturer. The in-1095.3. Site preparationoculum recommended for agricultural crops, Myke Pro R© PS3, would be the most effectiveAMF inoculum for grassland restoration in sandpits.Compost can be purchased locally at most landscape supply locations across Ontario.The approximate cost of compost is $40–$50 per metric ton plus delivery. Compost istypically generated from municipal waste collection streams and is readily available forpurchase.In comparison, the U.S. Biochar Initiative reports the cost of biochar as $500 per ton ex-cluding shipping (http://biochar-us.org/, 2014). Currently, production facilities of biocharare not widespread, making large quantities of biochar less readily available to the landmanager.Although the cost and availability of biochar may be prohibitive in 2014, the soil con-ditioning effect of this amendment when co–amended the compost may become a viableoption in the future. As carbon taxes are on the horizon, landholders may soon be ableto generate offset carbon credits from activities that reduce emissions or sequester carbon,including biochar application. These offset carbon credits may defer the cost associatedwith biochar.5.3 Site preparationWhen preparing the pit floor substrate for a grassland restoration project, the areashould be roughly graded flat to allow for ease of planting. Once graded, compost andbiochar can be tilled into the upper 10 cm of sandpit substrate before planting occurs. Irecommend minimizing the time between compost incorporation and planting to reducethe colonization of unwanted weedy plants. Seeds and/or plant plugs can be sown byhand or with machinery depending upon the scale of the project. Ideally, seeds should becompacted with a seed roller to ensure solid contact with the pit floor. I do not recommendreincorporating long–term storage stock piles into the site. A high density of weedy plantswill have developed on the stock–piled topsoil and would potentially out compete the growthof seeded native vegetation.5.4 SummaryMy goal was to optimize cost and effectively establish a tallgrass prairie ecosystem. Isuggest that integrating both planting approaches (i.e. plant plugs and seed) will be themost effective strategy for ecosystem establishment. I recommend incorporating 20 T ha−1to 40 T ha−1 of compost into the substrate before planting and /or seeding the site. Ifavailable, co–amend sandpit substrate with low rates of biochar (5 T ha−1 to 15 T ha−1).1105.4. SummaryTable 5.1: The projected materials cost of land rehabilitation in abandoned sandpits insouthern Ontario. Two viable options are available for prairie system rehabilitation: seedaddition or plug addition. Note that the cost per ha decreases as the rehabilitation areaincreases.Approx. Cost to Establish One Hectare of Prairie GrassesPrairie Rehabilitation w/ SeedSeed Application / ha (no grading required) $3,000Miscellaneous Costs (Transportation, etc.) $500Total $3,500Prairie Rehabilitation w/ PlugsPlug Cost ($1.00 × 20,000 plants / ha [1 plant / 0.5 m2] $20,000Miscellaneous Costs (Transportation, labour, etc.) $2,750Total $22,750AmendmentsAMF Inoculum (4 kg inoculum = 5.3 ha coverage) $400Compost [$45 / metric ton × 20 T ha−1] $900Biochar [$500 / metric ton × 10 T ha−1] $5,0001115.4. SummaryIncorporate plant plugs composed of legumes and warm season grasses at a rate of oneplug per square meter. These plants have a high survivorship and growth success at thesite, which will maximize the cost effectiveness of plant plugs. Sow a high diversity plantseed mixture containing warm season grasses, cool season grasses, legumes, and wildflowersamong the plant plugs. When planting, incorporating AM fungal inoculum can furtherpromote vegetative establishment and growth. Incorporating all of the investigated amend-ments is an effective restoration strategy that compliments the desired outcome of grasslandplant establishment and soil development.112Chapter 6ConclusionSevere land disturbance is pervasive among all ecosystems as a result of anthropogenicactivities. As a society, we have a responsibility to repair the destruction that accompaniesresource extraction and land–use change. Therefore, it is imperative that we restore viableecosystems that support plant and animal communities in severely impacted sites to accountfor regional habitat loss. In this study, I show that the restoration of grassland plantsin post–mine aggregate sites is a viable management option in southern Ontario. Afterresource depletion in mine areas, the substrate conditions that are a legacy of aggregateextraction are a stark contrast to functional soils in natural habitats. The edaphic conditionsof sand extraction restricts plant growth and soil food web development as shown by thecontrol plots at my research site. My goal was to recreate functional grassland habitatwith ecological characteristics that resemble reference sites. Land management tools (i.ecompost, biochar, and AM fungal inoculation) were anticipated to accelerate plant growthand soil recovery, translating into aboveground and belowground biota recovery on marginallands.A summary of the support garnered for the three main thesis objectives is addressedbelow:Objective #1 Develop a minimally destructive statistical method to increase mea-surement accuracy and reduce data collection time when estimating aboveground plantbiomass. The sampling method developed to estimate individual herbaceous plantand small shrub biomass in the field via partial least squares regression was superiorto linear regression statistical techniques. Partial least squares regression was shownto be a robust statistical technique that should be used to accurately predict plantbiomass in ecological experiments. In comparison to liner regression using a sole pre-dictor variable, partial least squares regression increases prediction confidence andreliability in ecological experiments.Objective #2 Determine the multi–year plant response of both planting strategiesto soil amendments and the commercial AM fungal isolate in a post-mine sandpit.Incorporating land management tools to mitigate the harsh abiotic conditions ofpost–extraction substrate is necessary to increase plant production in target grass-113Chapter 6. Conclusionland communities. As solitary amendments, the incorporation of the commercialarbuscular mycorrhizal fungal isolate and biochar did not significantly improve plantgrowth in either trial. In both field trials, the addition of compost was the mostinfluential driver of plant production in the post–mine sandpit. In the plant plugtrial, a trend was detected when comparing the total plant biomass in plots adding20 T ha−1 compost and 20 T ha−1 compost + 10 T ha−1 biochar compared to control.When compost was supplemented with biochar and the arbuscular mycorrhizal fun-gal inoculum (Rhizophagus irregularis) in the seed application trial, grassland plantresponse was largely accentuated. Thus, co–amending sandpit soils with biochar, com-post, and mycorrhizal inoculum increased the effectiveness of the restoration protocolin this trial. A single application of high rates 20 T ha−1 of biochar and compost atthe onset of an industrial–scale restoration project will lessen site maintenance costs,increase plant community recovery time.Objective #3 Determine the soil food web response to the addition of soil amend-ments and a AM fungal isolate in sandpit substrate. As shown in this study, miningsand strongly reduces soil microbial communities and soil animal abundance even af-ter two years of habitat recovery. Non–amended control plots had low soil food webabundance across all trophic levels. Thus, the natural recovery time of a soil food webwould be slow if no management action is implemented. The application of biocharalone added stress to the post–mine substrate and further restricted soil food webdevelopment compared to control plots. In contrast, compost and arbuscular mycor-rhizal inoculum had a negligible effect compared to control plots. Co–amending soilswith compost + biochar led to large increases in soil food web development across alltrophic levels indicating improved soil conditions at the site. Increasing the functionof soil food webs ultimately drives aboveground plant community production due tothe ecosystem services provided. These ecosystem services can lead to reduced sitemaintenance costs, increase plant community recovery time, and promote vegetativebiodiversity.The restoration of severely disturbed mine areas necessitates interventions that addressstressful soil conditions (Se´re´ et al. 2008). Restoration projects have successfully used munic-ipal compost in mine areas with low organic matter content to promote plant growth (Noydet al. 1996, Gryndler et al. 2008, Pu¨schel et al. 2008b) and soil community development(Ros et al. 2003; 2006, Biederman et al. 2008). My research shows marginal improvementof plant growth and soil food web development due to compost in the plant plug trial. The20 T ha−1 compost rate applied in the plant plug trial may have insufficient stimulate largeproduction changes compared to controls.114Chapter 6. ConclusionBiochar has been shown to improve soil nutrient availability and retention, reduce soilacidity, and adsorb organic matter (Lehmann et al. 2003, Shrestha et al. 2010). As biocharresearch is sparse in restoration, the successful application of biochar to improve bioticresponse will ultimately depend upon soil type, source feedstock, pyrolysis conditions, andbiochar application rates (Verheijen et al. 2014). Biochar as a soil amendment to increaseplant growth has generated promising results within agricultural systems and greenhouseexperiments. Research has demonstrated that biochar amended soils have greater cropbiomass (Rondon et al. 2007, Major et al. 2010, Biederman and Harpole 2012) and enhancedbiological N–fixation in leguminous crops (Rondon et al. 2007). In contrast to these studies,a plant growth effect was not detected in my grassland restoration experiment.In the field of mine restoration, a dearth of information exists on the effect of biocharin facilitating the growth of plants, development of microbial communities, and soil faunaabundance (Lehmann et al. 2011). My research indicates that biochar as a solitary amend-ment results in no significant improvement in plant growth. Furthermore, biochar addedstress to the system as shown by the reduction of microbial and soil fauna abundance inthe plant plug trial. In the seed application trail, increasing rates of biochar were onlybeneficial for promoting plant cover when soils were co–amended with increasing compostand inoculation with Rhizophagus irregularis.Biotic symbionts such as arbuscular mycorrhizal fungi have been used as inoculum tofacilitate plant production in severely degraded habitats (Johnson 1998, Gryndler et al.2008, Rydlova´ et al. 2008). The effectiveness of mycorrhizal inoculation on plant growthcan vary by the combination of plant species, soil disturbance type, and the selection of anarbuscular mycorrhizal fungal isolate (Taheri and Bever 2010, Pu¨schel et al. 2011, Thorneet al. 2013). My research indicates that the addition of a commercial arbuscular mycorrhizalfungal isolate, Rhizophagus irregularis, successfully established in the plant plug trial andpersisted over the study period. No arbuscular mycorrhizal effect was detected on totalplant biomass in the plant plug trial, but individual plants had a varied response to theaddition of the inoculum. This indicates that plants did not respond equally to the additionof the isolate.Background arbuscular mycorrhizal colonization was also detected in the non–inoculatedcontrol plants in the plant plug trial and persisted throughout the study period. Thus,the arbuscular mycorrhizal fungi present in the non–inoculated controls may have benefitedgrowth and nutrient acquisition of plants in these plots. Furthermore, several studies suggestthat arbuscular mycorrhizas should be collected directly from similar mine sites as thesefungi will be better adapted to field conditions when restoring plants (Noyd et al. 1995,Taheri and Bever 2010). As I used a commercial inoculum in this study, the isolate used inmy experiment may not have been well adapted to post–mine conditions in the field.1156.1. Strengths and limitations of the dissertation research6.1 Strengths and limitations of the dissertation researchThis dissertation research makes several strong contributions to the field of restorationecology and ecological sampling methodology. The techniques developed in this study areapplicable to mine restoration projects across the world. Severely degraded soil conditionsare pervasive within the resource extraction industry. This research clearly shows that minesoil conditions need to be altered before the restoration of natural vegetation is attempted,especially when establishing plants from seed. Furthermore, the use of plant plugs in aresearch setting is a novel technique to restore grasslands in post–mine areas. Growingplants as plugs may have assisted the grassland plants in overcoming initial abiotic soilconditions compared to establishing plants from seed. Thus, the use of plant plugs wasshown to be a reliable restoration technique, especially for C4 grasses.Selecting the appropriate suite of soil amendments carefully is essential when restor-ing post–mine sandpits. Land managers should be aware that all soil amendments andarbuscular mycorrhizal inoculum will not ubiquitously produce positive growth responsesfor plants, soil microbial communities, and soil animals. A single application of high rates(20 T ha−1) of biochar and compost with arbuscular mycorrhizal fungal inoculum at theonset of an industrial–scale grassland restoration project can maximize plant response fromseed and improve soil food web development. Thus, this translates to a restoration projectthat more closely approximates reference site conditions when compared to no managementintervention.During this research, I developed a new statistical technique to more accurately esti-mate herbaceous plant and small shrub biomass in the field. The improved measurementaccuracy can reduce the error in plant biomass estimation and increase plant measurementexperimental replication. The increased resolution in plant biomass estimation will proveto be an invaluable tool to experimentally measure plant growth in a variety of ecologicalscenarios. A limitation during the development of this technique was the low numbers ofspecies measured in the field. Ideally, a large suite of plants would have been measured totest the robustness of this technique across many different types of vegetation in southernOntario.A strength of this research was testing biochar as a soil amendment when restoring veg-etation in severely degraded habitats. My research clearly shows that biochar as a solitaryamendment provided no positive plant growth effects and is detrimental to the developmentof soil food webs. Comparatively, co–amending soils with biochar, compost, and Rhizoph-agus irregularis significantly improved the biotic response of the soil food web and plantcover in the seed application trial, indicating improved soil conditions belowground. Whenapproaching grassland restoration from an ecosystem perspective, the combination of all1166.2. Future directionsamendments was shown to yield the most interesting prospects as a reclamation tool torestore vegetation and belowground soil food webs.A limitation of this study was the replication in the seed application trial. As naturalvariability would be anticipated in field, the single replication of each factor combinationsled to high variation in total native plant cover. Plot size and labor was a limiting factorthat restricted the replication in the seed application experiment. Ideally, a smaller scaleproject could have been used to increase replication and statistical resolution.6.2 Future directionsI would like to continue the investigation of the interactions among biochar, organicamendments, and arbuscular mycorrhizal fungi. My career as a restoration ecologist willfocus on the recovery of plants and soils in severely degraded habitats. As biochar is a toolto create sustainable biofuels, improve soils conditions, and increase carbon sequestration,I think its role as in restoration needs to be explored more thoroughly. Thus, I would liketo experiment with designing and creating ways to more cost-effectively produce biocharfrom inexpensive feed stocks and incorporate the resulting char into restoration plans.In addition, restoration projects are often limited to short–term monitoring. I would liketo continue to monitor the long–term research site established during my PhD program. Theplots I have established will be available to be monitored indefinitely. I plan to continue totrack the plant growth rates and soil food web recovery over multiple time points throughoutmy career. This long–term monitoring will prove to be an invaluable tool to investigate theinfluence of the land management tools on the recovery trajectory of the community. Asthe plant plug and seed application trial was only monitored for two years and three yearsrespectively, this may not have been enough time to show large significant effects on theplant growth dynamics among the treatments. As the effect of solitary biochar additionto mine soils is unknown, there may be a lag in the response of plants and soil food webdevelopment early in the establishment of the ecosystem.117BibliographyAbdi, H. 2010. Partial least squares regression and projection on latent structure regression(PLS Regression). Wiley Interdisciplinary Reviews: Computational Statistics 2(1):97–106.Abel, S., A. Peters, S. Trinks, H. Schonsky, M. Facklam, and G. Wessolek. 2013. Impactof biochar and hydrochar addition on water retention and water repellency of sandy soil.Geoderma 202:183–191.Abiven, S., S. Menasseri, and C. Chenu. 2009. The effects of organic inputs over time onsoil aggregate stability - A literature analysis. Soil Biology and Biochemistry 41(1):1–12.Adams, M., T. Benjamin, N. Emery, S. Brouder, and K. D. Gibson. 2013. The effect ofbiochar on native and invasive prairie plant species. Invasive Plant Science and Manage-ment 6(2):197–207.Aggelides, S. M. and P. A. Londra. 2000. Effects of compost produced from town wastesand sewage sludge on the physical properties of a loamy and a clay soil. BioresourceTechnology 71(3):253–259.Albiach, R., R. Canet, F. Pomares, and F. Ingelmo. 2001. Organic matter components andaggregate stability after the application of different amendments to a horticultural soil.Bioresource Technology 76(2):125–129.Alguacil, M. M., F. Caravaca, and A. Roldan. 2005. Changes in rhizosphere microbial ac-tivity mediated by native or allochthonous AM fungi in the reafforestation of a Mediter-ranean degraded environment. Biology and Fertility of Soils 41(1):59–68.Allen, E. B. and M. F. Allen. 1980. Natural re–establishment of vesicular–arbuscularmycorrhizae following stripmine reclamation in Wyoming. Journal of Applied Ecology17(1):139–147.Allievi, L., A. Marchesini, C. Salardi, V. Piano, and A. Ferrari. 1993. Plant quality and soilresidual fertility six years after a compost treatment. Bioresource Technology 43(1):85–89.118BibliographyAmora-Lazcano, E., M. M. Va´zquez, and R. Azco´n. 1998. Response of nitrogen–transforming microorganisms to arbuscular mycorrhizal fungi. Biology and Fertility ofSoils 27(1):65–70.Andersen, C. M. and R. Bro. 2010. Variable selection in regressiona tutorial. Journal ofChemometrics 24(11):728–737.Anderson, J. R. and D. Westmoreland. 1971. Direct counts of soil organisms using a fluo-rescent brightener and a europium chelate. Soil Biology and Biochemistry 3(1):85–87.Anderson, T. 2003. Microbial eco–physiological indicators to asses soil quality. Agriculture,Ecosystems and Environment 98(1):285–293.Annabi, M., S. Houot, C. Francou, M. Poitrenaud, and Y. L. Bissonnais. 2007. Soil aggregatestability improvement with urban composts of different maturities. Soil Science Societyof America Journal 71(2):413–423.Arau´jo, A. S. F., N. Eisenhauer, L. A. P. Nunes, L. F. C. Leite, and S. Cesarz. 2013. Soilsurface–active fauna in degraded and restored lands of northeast Brazil. Land Degradation& Development doi: 10.1002/ldr.2247.Ash, H. J., R. P. Gemmell, and A. D. Bradshaw. 1994. The introduction of native plantspecies on industrial waste heaps: A test of immigration and other factors affectingprimary succession. Journal of Applied Ecology 31(1):74–84.Azco´n-Aguilar, C. and J. M. Barea. 1997. Applying mycorrhiza biotechnology to horticul-ture: Significance and potentials. Scientia Horticulturae 68(1):1–24.Baer, S., J. Blair, S. Collins, and A. Knapp. 2004. Plant community responses to resourceavailability and heterogeneity during restoration. Oecologia 139(4):617–629.Baer, S. G., J. M. Blair, S. L. Collins, and A. K. Knapp. 2003. Soil resources regulateproductivity and diversity in newly established tallgrass prairie. Ecology 84(3):724–735.Bailey, V. L., J. L. Smith, and H. Bolton Jr. 2002. Fungal–to–bacterial ratios in soilsinvestigated for enhanced C sequestration. Soil Biology and Biochemistry 34(7):997–1007.Bardgett, R. D. and R. Cook. 1998. Functional aspects of soil animal diversity in agriculturalgrasslands. Applied Soil Ecology 10(3):263–276.Bardgett, R. D. and E. McAlister. 1999. The measurement of soil fungal:bacterial biomassratios as an indicator of ecosystem self–regulation in temperate meadow grasslands. Bi-ology and Fertility of Soils 29(3):282–290.119BibliographyBardgett, R. D. and D. A. Wardle. 2003. Herbivore–mediated linkages between abovegroundand belowground communities. Ecology 84(9):2258–2268.Bardgett, R. D., S. Keiller, R. Cook, and A. S. Gilburn. 1998. Dynamic interactions betweensoil animals and microorganisms in upland grassland soils amended with sheep dung: Amicrocosm experiment. Soil Biology and Biochemistry 30(4):531–539.Bardgett, R. D., A. Richter, R. Bol, M. H. Garnett, R. Ba¨umler, X. Xu, E. Lopez-Capel,D. A. C. Manning, P. J. Hobbs, I. R. Hartley, and W. Wanek. 2007. Heterotrophic micro-bial communities use ancient carbon following glacial retreat. Biology Letters 3(5):487–490.Barton´, K. 2013. MuMIn: Multi-model inference. R package version 1.9.5. Available fromhttp://CRAN.R–project.org/package=MuMIn.Bastida, F., E. Kandeler, J. L. Moreno, M. Ros, C. Garc´ıa, and T. Herna´ndez. 2008.Application of fresh and composted organic wastes modifies structure, size and activityof soil microbial community under semiarid climate. Applied Soil Ecology 40(2):318–329.Bates, D., M. Maechler, B. Bolker, and S. Walker. 2014. lme4: Linear mixed-effectsmodels using Eigen and S4. R package version 1.0-6 Available from http://CRAN.R–project.org/package=lme4.Beare, M. H., R. W. Parmelee, P. F. Hendrix, W. Cheng, D. C. Coleman, and D. A.Crossley Jr. 1992. Microbial and faunal interactions and effects on litter nitrogen anddecomposition in agroecosystems. Ecological Monographs 62(4):569–591.Beesley, L., O. S. Inneh, G. J. Norton, E. Moreno-Jimenez, T. Pardo, R. Clemente, andJ. J. C. Dawson. 2014. Assessing the influence of compost and biochar amendments onthe mobility and toxicity of metals and arsenic in a naturally contaminated mine soil.Environmental Pollution 186:195–202.Benayas, J. M. R., A. C. Newton, A. Diaz, and J. M. Bullock. 2009. Enhancement ofbiodiversity and ecosystem services by ecological restoration: A meta–analysis. Science325(5944):1121–1124.Bengtsson, J., H. Seta¨la¨, and D. W. Zheng. 1996. Food webs and nutrient cycling in soils:Interactions and positive feedbacks. Pages 30–38 in G. A. Polis and K. O. Winemiller,editors. Food webs: Integration of patterns and dynamics. Chapman and Hall, New York,New York, U.S.A.120BibliographyBi, Y. L., X. L. Li, P. Christie, Z. Q. Hu, and M. H. Wong. 2003. Growth and nutrientuptake of arbuscular mycorrhizal maize in different depths of soil overlying coal fly ash.Chemosphere 50(6):863–869.Biederman, L. A. 2013. Directing ecological restoration: Impact of organicamendments on above– and belowground ecosystem characteristics. PhD The-sis. Available at http://repository.tamu.edu/bitstream/handle/1969.1/ETD–TAMU–1371/BIEDERMAN–DISSERTATION.pdf?sequence=1.Biederman, L. A. and W. S. Harpole. 2012. Biochar and its effects on plant productivityand nutrient cycling: a metaanalysis. Global Change Biology Bioenergy 5(2):202–214.Biederman, L. A. and S. W. Whisenant. 2009. Organic amendments direct grass populationdynamics in a landfill prairie restoration. Ecological Engineering 35(5):678–686.Biederman, L. A., T. W. Boutton, and S. G. Whisenant. 2008. Nematode community de-velopment early in ecological restoration: The role of organic amendments. Soil Biologyand Biochemistry 40(9):2366–2374.Birk, J. J., C. Steiner, W. C. Teixiera, W. Zech, and B. Glaser. 2009. Microbial response tocharcoal amendments and fertilization of a highly weathered tropical soil. Pages 309–324in W. I. Woods, W. G. Teixeira, J. Lehmann, C. Steiner, A. WinklerPrins, and L. Re-bellato, editors. Amazonian dark earths: Wim Sombroek’s vision. Springer Netherlands.Blackwell, P., G. Reithmuller, and M. Collins. 2009. Biochar application to soil. Pages207–226 in C. Lehman and S. Joseph, editors. Biochar for environmental management:Science and technology. Earthscan, London.Bolker, B. M., M. E. Brooks, C. J. Clark, S. W. Geange, J. R. Poulsen, M. H. M. Stevens,and J.-S. S. White. 2009. Generalized linear mixed models: A practical guide for ecologyand evolution. Trends in Ecology and Evolution 24(3):127–135.Bongers, T. 1990. The maturity index: An ecological measure of environmental disturbancebased on nematode species composition. Oecologia 83(1):14–19.Bongers, T. and M. Bongers. 1998. Functional diversity of nematodes. Applied Soil Ecology10(3):239–251.Bongers, T. and H. Ferris. 1999. Nematode community structure as a bioindicator in envi-ronmental monitoring. Trends in Ecology and Evolution 14(6):224–228.Bonham, C. D. 1982. Optimum allocation in multivariate double sampling for biomassestimation. Journal of Range Management 35(6):777–779.121BibliographyBooth, D. T. and S. E. Cox. 2008. Image–based monitoring to measure ecological changein rangeland. Frontiers in Ecology and the Environment 6(4):185–190.Bowen, G. D. and A. D. Rovira. 1999. The rhizosphere and its management to improveplant growth. Advances in Agronomy 66:1–102.Bradshaw, A. 1997. Restoration of mined lands–using natural processes. Ecological Engi-neering 8(4):255–269.Bradshaw, A. 2000. The use of natural processes in reclamation–advantages and difficulties.Landscape and Urban Planning 51(2):89–100.Bresson, L. M., C. Koch, Y. Le Bissonnais, E. Barriuso, and V. Lecomte. 2001. Soil surfacestructure stabilization by municipal waste compost application. Soil Science Society ofAmerica Journal 65(6):1804–1811.Briggs, J. M. and A. K. Knapp. 1995. Interannual variability in primary production intallgrass prairie: Climate, soil moisture, topographic position, and fire as determinantsof aboveground biomass. American Journal of Botany 82(8):1024–1030.Brodowski, S., A. Rodionov, L. Haumaier, B. Glaser, and W. Amelung. 2005. Revised blackcarbon assessment using benzene polycarboxylic acids. Organic Geochemistry 36(9):1299–1310.Brundrett, M. C., Y. Piche, and R. L. Peterson. 1984. A new method for observing the mor-phology of vesicular–arbuscular mycorrhizae. Canadian Journal of Botany 62(10):2128–2134.Bruun, S., E. S. Jensen, and L. S. Jensen. 2008. Microbial mineralization and assimilationof black carbon: Dependency on degree of thermal alteration. Organic Geochemistry39(7):839–845.Buisson, E., K. D. Holl, S. Anderson, E. Corcket, G. F. Hayes, F. Torre, A. Peteers, andT. Dutoit. 2006. Effect of seed source, topsoil removal, and plant neighbor removal onrestoring California coastal prairies. Restoration Ecology 14(4):569–577.Burnham, K. P. and D. R. Anderson. 2004. Multimodel inference understanding AIC andBIC in model selection. Sociological Methods and Research 33(2):261–304.Busby, R. R., H. Allen Torbert, and D. L. Gebhart. 2007. Carbon and nitrogen mineraliza-tion of non–composted and composted municipal solid waste in sandy soils. Soil Biologyand Biochemistry 39(6):1277–1283.122BibliographyByrne, B. M. 2013. Structural equation modeling with AMOS: Basic concepts, applications,and programming. Mahwah, NJ: Lawrence Erlbaum.Byrne, K. M., W. K. Lauenroth, P. B. Adler, and C. M. Byrne. 2011. Estimating above-ground net primary production in grasslands: A comparison of nondestructive methods.Rangeland Ecology and Management 64(5):498–505.Cairns, J. 2000. Setting ecological restoration goals for technical feasibility and scientificvalidity. Ecological Engineering 15(3):171–180.Callaham Jr., M. A., C. C. Rhoades, and L. Heneghan. 2008. A striking profile: Soil ecologi-cal knowledge in restoration management and science. Restoration Ecology 16(4):604–607.Caravaca, F., J. M. Barea, and A. Rolda´n. 2002a. Synergistic influence of an arbuscularmycorrhizal fungus and organic amendment on Pistacia lentiscus L. seedlings afforestedin a degraded semiarid soil. Soil Biology and Biochemistry 34(8):1139–1145.Caravaca, F., T. Herna´ndez, C. Garc´ıa, and A. Rolda´n. 2002b. Improvement of rhizosphereaggregate stability of afforested semiarid plant species subjected to mycorrhizal inocula-tion and compost addition. Geoderma 108(1):133–144.Caravaca, F., D. Figueroa, C. Azco´n-Aguilar, J. M. Barea, and A. Rolda´n. 2003a. Medium–term effects of mycorrhizal inoculation and composted municipal waste addition on theestablishment of two Mediterranean shrub species under semiarid field conditions. Agri-culture, Ecosystems and Environment 97(1):95–105.Caravaca, F., D. Figueroa, A. Rolda´n, and C. Azco´n-Aguilar. 2003b. Alteration in rhizo-sphere soil properties of afforested Rhamnus lycioides seedlings in short–term responseto mycorrhizal inoculation with Glomus intraradices and organic amendment. Environ-mental Management 31(3):412–420.Cardinale, B. J., J. P. Wright, M. W. Cadotte, I. T. Carroll, A. Hector, D. S. Srivastava,M. Loreau, and J. J. Weis. 2007. Impacts of plant diversity on biomass production increasethrough time because of species complementarity. Proceedings of the National Academyof Sciences 104(46):18123–18128.Carrascal, L. M., I. Galva´n, and O. Gordo. 2009. Partial least squares regression as analternative to current regression methods used in ecology. Oikos 118(5):681–690.Catchpole, W. R. and C. J. Wheeler. 1992. Estimating plant biomass: A review of tech-niques. Australian Journal of Ecology 17(2):121–131.123BibliographyCelik, I., I. Ortas, and S. Kilic. 2004. Effects of compost, mycorrhiza, manure and fertilizeron some physical properties of a Chromoxerert soil. Soil and Tillage Research 78(1):59–67.Cernova, N. M. 1970. Regularities in the distribution of microarthropods in compost heaps.Pedobiologia 10:365–372.Chan, K. Y. and Z. Xu. 2009. Biochar: Nutrient properties and their enhancement. Pages67–84 in J. Lehmann and S. Joseph, editors. Biochar for environmental management:Science and technology. Earthscan, London.Chave, J., R. Condit, S. Aguilar, A. Hernandez, S. Lao, and R. Perez. 2004. Error propaga-tion and scaling for tropical forest biomass estimates. Philosophical Transactions of theRoyal Society of London. Series B: Biological Sciences 359(1443):409–420.Clements, F. E. 1916. Plant succession: An analysis of the development of vegetation.Carnegie Institution, Washington, D.C.Coleman, D. C. and W. B. Whitman. 2005. Linking species richness, biodiversity and ecosys-tem function in soil systems. Pedobiologia 49(6):479–497.Courtney, R., A. M. Keith, and T. Harrington. 2011. Nematode assemblages in bauxiteresidue with different restoration histories. Restoration Ecology 19(6):758–764.Craine, J. M. and R. Dybzinski. 2013. Mechanisms of plant competition for nutrients, waterand light. Functional Ecology 27(4):833–840.Crecchio, C., M. Curci, M. D. R. Pizzigallo, P. Ricciuti, and P. Ruggiero. 2004. Effectsof municipal solid waste compost amendments on soil enzyme activities and bacterialgenetic diversity. Soil Biology and Biochemistry 36(10):1595–1605.Curry, J. P. and J. A. Good. 1992. Soil faunal degradation and restoration. Pages 171–215in L. Rattan and B. A. Stewart, editors. Soil restoration. Springer New York.de Souza, R. G., B. T. Goto, D. K. Alves da Silva, F. S. Barbosa da Silva, E. V. S. B.Sampaio, and L. C. Maia. 2010. The role of arbuscular mycorrhizal fungi and cattlemanure in the establishment of Tocoyena selloana Schum. in mined dune areas. EuropeanJournal of Soil Biology 46(3):237–242.de Vries, F. T., M. E. Liiri, L. Bjørnlund, M. A. Bowker, S. Christensen, H. M. Seta¨la¨, andR. D. Bardgett. 2012. Land use alters the resistance and resilience of soil food webs todrought. Nature Climate Change 2(4):276–280.124Bibliographyde Vries, F. T., E. The´bault, M. Liiri, K. Birkhofer, M. A. Tsiafouli, L. Bjørnlund, H. B.Jørgensen, M. V. Brady, S. Christensen, and P. C. de Ruiter. 2013. Soil food web prop-erties explain ecosystem services across European land use systems. Proceedings of theNational Academy of Sciences 110(35):14296–14301.del Moral, R., L. Walker, and J. Bakker. 2007. Insights gained from succession for therestoration of landscape structure and function. Pages 19–44 in L. R. Walker, J. Walker,and R. J. Hobbs, editors. Linking restoration and ecological succession. Springer, NewYork, New York.Delaney, K., L. Rodger, P. A. Woodlife, G. Rhynard, and P. Morris. 2000. Planting theseed – A guide to establishing prairie and meadow communities in southern Ontario.Environment Canada, Government of Canada. 61pp.Diacono, M. and F. Montemurro. 2010. Long-term effects of organic amendments on soilfertility. A review. Agronomy for Sustainable Development 30(2):401–422.Diaz, G. and M. Honrubia. 1994. A mycorrhizal survey of plants growing on mine wastesin Southeast Spain. Arid Land Research and Management 8(1):59–68.Ding, Y., Y. Liu, W. Wu, D. Shi, M. Yang, and Z. Zhong. 2010. Evaluation of biochareffects on nitrogen retention and leaching in multi–layered soil columns. Water, Air, andSoil Pollution 213(1):47–55.Doran, J. W. and M. R. Zeiss. 2000. Soil health and sustainability: Managing the bioticcomponent of soil quality. Applied Soil Ecology 15(1):3–11.Elkins, N. Z., L. W. Parker, E. Aldon, and W. G. Whitford. 1984. Responses of soil biotato organic amendments in stripmine spoils in northwestern New Mexico. Journal of En-vironmental Quality 13(2):215–219.Elliott, K. J. and B. D. Clinton. 1993. Equations for estimating biomass of herbaceousand woody vegetation in early–successional southern Appalachian pine–hardwood forests.USDA Forest Service Research Note SE–365. 7p.Enkhtuya, B., M. Po¨schl, and M. Vosa´tka. 2005. Native grass facilitates mycorrhizal coloni-sation and P uptake of tree seedlings in two anthropogenic substrates. Water, Air, andSoil Pollution 166(1):217–236.Eriksen, G. N., F. J. Coale, and G. A. Bollero. 1999. Soil nitrogen dynamics and maizeproduction in municipal solid waste amended soil. Agronomy Journal 91(6):1009–1016.125BibliographyFay, P. A., J. D. Carlisle, A. K. Knapp, J. M. Blair, and S. L. Collins. 2003. Productivityresponses to altered rainfall patterns in a C4–dominated grassland. Oecologia 137(2):245–251.Fellet, G., L. Marchiol, G. Delle Vedove, and A. Peressotti. 2011. Application of biochar onmine tailings: Effects and perspectives for land reclamation. Chemosphere 83(9):1262–1267.Fellet, G., M. Marmiroli, and L. Marchiol. 2014. Elements uptake by metal accumulatorspecies grown on mine tailings amended with three types of biochar. Science of the TotalEnvironment 468:598–608.Ferris, H. and T. Bongers. 2006. Nematode indicators of organic enrichment. Journal ofNematology 38(1):3–12.Ferris, H., T. Bongers, and R. G. M. De Goede. 2001. A framework for soil food webdiagnostics: Extension of the nematode faunal analysis concept. Applied Soil Ecology18(1):13–29.Filiberto, D. M. and J. L. Gaunt. 2013. Practicality of biochar additions to enhance soiland crop productivity. Agriculture 3(4):715–725.Fischer, D. and B. Glaser. 2012. Synergisms between compost and biochar for sustainablesoil amelioration. Management of Organic Waste 10:167–198.Frank, L. E. and J. H. Friedman. 1993. A statistical view of some chemometrics regressiontools. Technometrics 35(2):109–135.Frankenberger, W. T. and W. A. Dick. 1983. Relationships between enzyme activities andmicrobial growth and activity indices in soil. Soil Science Society of America Journal47(5):945–951.Frenot, Y., J. C. Gloaguen, M. Cannavacciuolo, and A. Bellido. 1998. Primary successionon glacier forelands in the subantarctic Kerguelen Islands. Journal of Vegetation Science9(1):75–84.Frey, S. D., E. T. Elliott, and K. Paustian. 1999. Bacterial and fungal abundance andbiomass in conventional and no–tillage agroecosystems along two climatic gradients. SoilBiology and Biochemistry 31(4):573–585.Frouz, J., D. Elhottova´, V. Kura´zˇ, and M. Sˇourkova´. 2006. Effects of soil macrofauna onother soil biota and soil formation in reclaimed and unreclaimed post mining sites: Resultsof a field microcosm experiment. Applied Soil Ecology 33(3):308–320.126BibliographyGabrielle, B., J. Da-Silveira, S. Houot, and J. Michelin. 2005. Field–scale modelling ofcarbon and nitrogen dynamics in soils amended with urban waste composts. Agriculture,Ecosystems and Environment 110(3):289–299.Ganesan, V., S. Ragupathy, B. Parthipan, D. B. R. Rani, and A. Mahadevan. 1991. Distri-bution of vesicular–arbuscular mycorrhizal fungi in coal, lignite, and calcite mine spoilsof India. Biology and Fertility of Soils 12(2):131–136.Garbaye, J. 1991. Biological interactions in the mycorrhizosphere. Experientia 47(4):370–375.Garc´ıa-Gil, J. C., S. B. Ceppi, M. I. Velasco, A. Polo, and N. Senesi. 2004. Long–term effectsof amendment with municipal solid waste compost on the elemental and acidic functionalgroup composition and pH–buffer capacity of soil humic acids. Geoderma 121(1):135–142.Gartshore, M. E., D. A. Sutherland, and J. D. McCracken. 1987. The natural areas inventoryof the regional municipality of Haldimand–Norfolk. The Norfolk Field Naturalists. Volume1: 112pp.Gaur, A. and A. Adholeya. 2005. Diverse response of five ornamental plant species to mixedindigenous and single isolate arbuscular–mycorrhizal inocula in marginal soil amendedwith organic matter. Journal of Plant Nutrition 28(4):707–723.Gholz, H. L., C. C. Grier, A. G. Campbell, and A. T. Brown. 1979. Equations for estimatingbiomass and leaf area of plants in the Pacific Northwest. Forest Research Lab, OregonState University, Corvallis, Oregon, U.S.A. 43pp.Ghosh, S., L. F. Ow, and B. Wilson. 2014. Influence of biochar and compost on soil proper-ties and tree growth in a tropical urban environment. International Journal of Environ-mental Science and Technology doi: 10.1007/s13762–014–0508–0.Giusquiani, P. L., M. Pagliai, G. Gigliotti, D. Businelli, and A. Benetti. 1995. Urban wastecompost: Effects on physical, chemical, and biochemical soil properties. Journal of Envi-ronmental Quality 24(1):175–182.Glaser, B. 2007. Prehistorically modified soils of central Amazonia: A model for sustainableagriculture in the twenty–first century. Philosophical Transactions of the Royal Societyof London, Series B: Biological Sciences 362(1478):187–196.Glaser, B., J. Lehmann, and W. Zech. 2002. Ameliorating physical and chemical propertiesof highly weathered soils in the tropics with charcoal–a review. Biology and Fertility ofSoils 35(4):219–230.127BibliographyGlatzle, A., A. Mechel, and M. E. V. Lourenco. 1993. Botanical components of annualMediterranean grassland as determined by point–intercept and clipping methods. Journalof Range Management 46(3):271–274.Glenn-Lewin, D. C. and E. van der Maarel. 1992. Patterns and processes of vegetationdynamics. Pages 11–59 in D. C. Glenn-Lewis, R. K. Peet, and T. T. Veblen, editors.Plant succession: Theory and prediction. Chapmann and Hall, London.Govindarajulu, M., P. E. Pfeffer, H. Jin, J. Abubaker, D. D. Douds, J. W. Allen, H. Bu¨cking,P. J. Lammers, and Y. Shachar-Hill. 2005. Nitrogen transfer in the arbuscular mycorrhizalsymbiosis. Nature 435(7043):819–823.Graber, E. R., Y. M. Harel, M. Kolton, E. Cytryn, A. Silber, D. R. David, L. Tsechansky,M. Borenshtein, and Y. Elad. 2010. Biochar impact on development and productivity ofpepper and tomato grown in fertigated soilless media. Plant and Soil 337(1):481–496.Grace, J. B. 2006. Structural equation modeling and natural systems. Cambridge UniversityPress.Graham, M. H. 2003. Confronting multicollinearity in ecological multiple regression. Ecology84(11):2809–2815.Grueber, C. E., S. Nakagawa, R. J. Laws, and I. G. Jamieson. 2011. Multimodel inferencein ecology and evolution: Challenges and solutions. Journal of Evolutionary Biology24(4):699–711.Gryndler, M., R. Sudova´, D. Pu¨schel, J. Rydlova, M. Janouskova, and M. Vosatka. 2008.Cultivation of high-biomass crops on coal mine spoil banks: Can microbial inoculationcompensate for high doses of organic matter? Bioresource Technology 99(14):6391–6399.Gupta, M. C., R. S. Singh, and K. Sitaramaiah. 1979. A new endoparasitic fungus onXiphinema and cultivation of Rhabditis spp. and Aphelenchus avenae on same fungus.Nematologica 25(1):142.Hadas, A. and R. Portnoy. 1997. Rates of decomposition in soil and release of available nitro-gen from cattle manure and municipal waste composts. Compost Science and Utilization5(3):48–54.Hannah, L., J. L. Carr, and A. Lankerani. 1995. Human disturbance and natural habitat:A biome level analysis of a global data set. Biodiversity and Conservation 4(2):128–155.128BibliographyHa´neˇl, L. 2008. Nematode assemblages indicate soil restoration on colliery spoils af-forested by planting different tree species and by natural succession. Applied Soil Ecology40(1):86–99.Hargreaves, J. C., M. S. Adl, and P. R. Warman. 2008. A review of the use of compostedmunicipal solid waste in agriculture. Agriculture, Ecosystems and Environment 123(1):1–14.Harinikumar, K. M., D. J. Bagyaraj, and B. C. Mallesha. 1990. Effect of intercropping andorganic soil amendments on native VA mycorrhizal fungi in an oxisol. Arid Land Researchand Management 4(3):193–197.Harris, J. 2009. Soil microbial communities and restoration ecology: Facilitators or follow-ers? Science 325(5940):573–574.Harte, J. and A. P. Kinzig. 1993. Mutualism and competition between plants and de-composers: Implications for nutrient allocation in ecosystems. The American Naturalist141(6):829–846.Hastie, T., R. Tibshirani, and J. J. H. Friedman. 2001. The elements of statistical learning,vol. 1. Springer New York.Ha¨ttenschwiler, S., A. V. Tiunov, and S. Scheu. 2005. Biodiversity and litter decompositionin terrestrial ecosystems. Annual Review of Ecology, Evolution, and Systematics 36:191–218.Hector, A., B. Schmid, C. Beierkuhnlein, M. C. Caldeira, M. Diemer, P. G. Dimitrakopoulos,J. A. Finn, H. Freitas, P. S. Giller, J. Good, R. Harris, P. Ho¨gberg, K. Huss-Danell,J. Joshi, A. Jumpponen, C. Ko¨rner, P. W. Leadley, M. Loreau, A. Minns, C. P. H. Mulder,G. O’Donovan, S. J. Otway, J. S. Pereira, A. Prinz, D. J. Read, M. Scherer-Lorenzen,E.-D. Schulze, A.-S. D. Siamantziouras, E. M. Spehn, A. C. Terry, A. Y. Troumbis,F. I. Woodward, S. Yachi, and J. H. Lawton. 1999. Plant diversity and productivityexperiments in european grasslands. Science 286(5442):1123–1127.Hedlund, K. and M. S. O¨hrn. 2000. Tritrophic interactions in a soil community enhancedecomposition rates. Oikos 88(3):585–591.Hendrix, P. F., R. W. Parmelee, D. A. Crossley, D. C. Coleman, E. P. Odum, and P. M.Groffman. 1986. Detritus food webs in conventional and no–tillage agroecosystems. Bio-science 374–380.129BibliographyHeneghan, L., S. P. Miller, S. Baer, M. A. Callaham, J. Montgomery, M. Pavao-Zuckerman,C. C. Rhoades, and S. Richardson. 2008. Integrating soil ecological knowledge into restora-tion management. Restoration Ecology 16(4):608–617.Hodge, A. and A. H. Fitter. 2010. Substantial nitrogen acquisition by arbuscular mycorrhizalfungi from organic material has implications for N cycling. Proceedings of the NationalAcademy of Sciences 107(31):13754–13759.Hodge, A., T. Helgason, and A. H. Fitter. 2010. Nutritional ecology of arbuscular mycor-rhizal fungi. Fungal Ecology 3(4):267–273.Hodkinson, I. D., N. R. Webb, and S. J. Coulson. 2002. Primary community assembly onland–the missing stages: Why are the heterotrophic organisms always there first? Journalof Ecology 90(3):569–577.Holtkamp, R., P. Kardol, A. van der Wal, S. C. Dekker, W. H. van der Putten, and P. C.de Ruiter. 2008. Soil food web structure during ecosystem development after land aban-donment. Applied Soil Ecology 39(1):23–34.Hooper, D. U., F. S. Chapin Iii, J. J. Ewel, A. Hector, P. Inchausti, S. Lavorel, J. H. Law-ton, D. M. Lodge, M. Loreau, and S. Naeem. 2005. Effects of biodiversity on ecosystemfunctioning: A consensus of current knowledge. Ecological Monographs 75(1):3–35.Hortenstine, C. C. and D. F. Rothwell. 1972. Use of municipal compost in reclamation ofphosphate–mining sand tailings. Journal of Environmental Quality 1(4):415–418.Hunt, G. A. and R. Fogel. 1983. Fungal hyphal dynamics in a western Oregon Douglas–firstand. Soil Biology and Biochemistry 15(6):641–649.Hunt, H. W., D. C. Coleman, E. R. Ingham, R. E. Ingham, E. T. Elliott, J. C. Moore,S. L. Rose, C. P. P. Reid, and C. R. Morley. 1987. The detrital food web in a shortgrassprairie. Biology and Fertility of Soils 3(1):57–68.Iglesias-Jime´nez, E. 2001. Nitrogen availability from a mature urban compost determinedby the 15N isotope dilution method. Soil Biology and Biochemistry 33(3):409–412.Ingham, R. E., J. A. Trofymow, E. R. Ingham, and D. C. Coleman. 1985. Interactions ofbacteria, fungi, and their nematode grazers: Effects on nutrient cycling and plant growth.Ecological Monographs 55(1):119–140.Ishii, T. and K. Kadoya. 1994. Effects of charcoal as a soil conditioner on citrus growthand vesicular–arbuscular mycorrhizal development. Journal of the Japanese Society forHorticultural Science (Japan) 63(3):529–535.130BibliographyJackson, S. T. and R. J. Hobbs. 2009. Ecological restoration in the light of ecological history.Science 325(5940):567–569.Jakobsen, S. T. 1995. Aerobic decomposition of organic wastes 2. Value of compost as afertilizer. Resources, Conservation and Recycling 13(1):57–71.Jansa, J., A. Mozafar, T. Anken, R. Ruh, I. Sanders, and E. Frossard. 2002. Diversityand structure of AMF communities as affected by tillage in a temperate soil. Mycorrhiza12(5):225–234.Jansa, J., A. Mozafar, G. Kuhn, T. Anken, R. Ruh, I. R. Sanders, and E. Frossard. 2003.Soil tillage affects the community structure of mycorrhizal fungi in maize roots. EcologicalApplications 13(4):1164–1176.Jasper, D. A., L. K. Abbott, and A. D. Robson. 1989. Soil disturbance reduces the infectivityof external hyphae of vesicular–arbuscular mycorrhizal fungi. New Phytologist 112(1):93–99.Jastrow, J. D., R. M. Miller, and J. Lussenhop. 1998. Contributions of interacting bio-logical mechanisms to soil aggregate stabilization in restored prairie. Soil Biology andBiochemistry 30(7):905–916.Jastrow, J. D., J. E. Amonette, and V. L. Bailey. 2007. Mechanisms controlling soil carbonturnover and their potential application for enhancing carbon sequestration. ClimaticChange 80(1):5–23.Jeffery, S., F. G. A. Verheijen, M. Van der Velde, and A. C. Bastos. 2011. A quantitativereview of the effects of biochar application to soils on crop productivity using meta–analysis. Agriculture, Ecosystems and Environment 144(1):175–187.Johansson, J. F., L. R. Paul, and R. D. Finlay. 2004. Microbial interactions in the mycorrhi-zosphere and their significance for sustainable agriculture. FEMS Microbiology Ecology48(1):1–13.Johnson, J. B. and K. S. Omland. 2004. Model selection in ecology and evolution. Trendsin Ecology and Evolution 19(2):101–108.Johnson, N. C. 1998. Responses of Salsola kali and Panicum virgatum to mycorrhizal fungi,phosphorus and soil organic matter: Implications for reclamation. Journal of AppliedEcology 35(1):86–94.Johnson, N. C., J.-H. Graham, and F. A. Smith. 1997. Functioning of mycorrhizal associa-tions along the mutualism–parasitism continuum. New Phytologist 135(4):575–585.131BibliographyJonasson, S. 1988. Evaluation of the point intercept method for the estimation of plantbiomass. Oikos 52(1):101–106.Jones, B. E. H., R. J. Haynes, and I. R. Phillips. 2010. Effect of amendment of bauxiteprocessing sand with organic materials on its chemical, physical and microbial properties.Journal of Environmental Management 91(11):2281–2288.Jones, D. L., J. Rousk, G. Edwards-Jones, T. H. DeLuca, and D. V. Murphy. 2012. Biochar–mediated changes in soil quality and plant growth in a three year field trial. Soil Biologyand Biochemistry 45:113–124.Jørgensen, H. B. and K. Hedlund. 2013. Organic amendment and fungal species in combi-nation can alter collembolan fitness. Soil Biology and Biochemistry 65:316–321.Kardol, P. and D. A. Wardle. 2010. How understanding aboveground–belowground linkagescan assist restoration ecology. Trends in Ecology and Evolution 25(11):670–679.Karlen, D. L., M. J. Mausbach, J. W. Doran, R. G. Cline, R. F. Harris, and G. E. Schuman.1997. Soil quality: A concept, definition, and framework for evaluation (A guest editorial).Soil Science Society of America Journal 61(1):4–10.Kelly, C. N., C. D. Peltz, M. Stanton, D. W. Rutherford, and C. E. Rostad. 2014. Biocharapplication to hardrock mine tailings: Soil quality, microbial activity, and toxic elementsorption. Applied Geochemistry 43:35–48.Khan, A. G. 1981. Growth responses of endomycorrhizal onions in unsterilized coal waste.New Phytologist 87(2):363–370.Khater, C., A. Martin, and J. Maillet. 2003. Spontaneous vegetation dynamics and restora-tion prospects for limestone quarries in Lebanon. Applied Vegetation Science 6(2):199–204.Kinney, T. J., C. A. Masiello, B. Dugan, W. C. Hockaday, M. R. Dean, K. Zygourakis, andR. T. Barnes. 2012. Hydrologic properties of biochars produced at different temperatures.Biomass and Bioenergy 41:34–43.Kjøller, A. and S. Struwe. 1982. Microfungi in ecosystems: Fungal occurrence and activityin litter and soil. Oikos 391–422.Klironomos, J. N. 2002. Feedback with soil biota contributes to plant rarity and invasivenessin communities. Nature 417(6884):67–70.132BibliographyKlironomos, J. N. 2003. Variation in plant response to native and exotic arbuscular mycor-rhizal fungi. Ecology 84(9):2292–2301.Klironomos, J. N., P. Moutoglis, B. Kendrick, and P. Widden. 1993. A comparison ofspatial heterogeneity of vesicular–arbuscular mycorrhizal fungi in two maple–forest soils.Canadian Journal of Botany 71(11):1472–1480.Klironomos, J. N., M. C. Rillig, and M. F. Allen. 1996. Below–ground microbial and micro-faunal responses to Artemisia tridentata grown under elevated atmospheric CO2. Func-tional Ecology 10(4):527–534.Knapp, A. K. and T. R. Seastedt. 1986. Detritus accumulation limits productivity of tall-grass prairie. BioScience 36(10):662–668.Knapp, A. K. and M. D. Smith. 2001. Variation among biomes in temporal dynamics ofaboveground primary production. Science 291(5503):481–484.Knapp, A. K., J. M. Blair, J. M. Briggs, S. L. Collins, D. C. Hartnett, L. C. Johnson,and E. G. Towne. 1999. The keystone role of bison in North American tallgrass prairie.BioScience 49(1):39–50.Korthals, G. W., P. Smilauer, C. Van Dijk, and W. H. Van der Putten. 2001. Linkingabove– and below–ground biodiversity: Abundance and trophic complexity in soil asa response to experimental plant communities on abandoned arable land. FunctionalEcology 15(4):506–514.Kuzyakov, Y. 2002. Review: Factors affecting rhizosphere priming effects. Journal of PlantNutrition and Soil Science 165(4):382.Laakso, J. and H. Seta¨la¨. 1999. Sensitivity of primary production to changes in the archi-tecture of belowground food webs. Oikos 87:57–64.Laakso, J., H. Seta¨la¨, and A. Paloja¨rvi. 2000. Influence of decomposer food web structureand nitrogen availability on plant growth. Plant and Soil 225(1):153–165.Laird, D. A. 2008. The charcoal vision: A win–win–win scenario for simultaneously produc-ing bioenergy, permanently sequestering carbon, while improving soil and water quality.Agronomy Journal 100(1):178–181.Lal, R. 2009. Challenges and opportunities in soil organic matter research. European Journalof Soil Science 60(2):158–169.133BibliographyLangley, J. A. and B. A. Hungate. 2003. Mycorrhizal controls on belowground litter quality.Ecology 84(9):2302–2312.Larney, F. J. and D. A. Angers. 2012. The role of organic amendments in soil reclamation:A review. Canadian Journal of Soil Science 92(1):19–38.Lavelle, P., T. Decans, M. Aubert, S. Barot, M. Blouin, F. Bureau, P. Margerie, P. Mora,and J. P. Rossi. 2006. Soil invertebrates and ecosystem services. European Journal of SoilBiology 42:S3–S15.Lee, Q. and P. Widden. 1996. Folsomia candida, a ”fungivorous” collembolan, feeds prefer-entially on nematodes rather than soil fungi. Soil Biology and Biochemistry 28(4):689–690.Lehmann, J., J. Pereira da Silva, C. Steiner, T. Nehls, W. Zech, and B. Glaser. 2003.Nutrient availability and leaching in an archaeological Anthrosol and a Ferralsol of theCentral Amazon basin: Fertilizer, manure and charcoal amendments. Plant and Soil249(2):343–357.Lehmann, J., J. Gaunt, and M. Rondon. 2006. Bio–char sequestration in terrestrialecosystems–a review. Mitigation and Adaptation Strategies for Global Change 11(2):395–419.Lehmann, J., C. Czimczik, D. A. Laird, and S. Sohi. 2009. Stability of Biochar in Soil. Pages183–205 in J. Lehmann and S. Joseph, editors. Biochar for environmental management:Science and technology. Earthscan, London.Lehmann, J., M. Rillig, J. Thies, C. A. Masiello, W. C. Hockaday, and D. Crowley. 2011.Biochar effects on soil biota–a review. Soil Biology and Biochemistry 43(9):1812–1836.Lenoir, L., T. Persson, J. Bengtsson, H. Wallander, and A. Wiren. 2007. Bottom–up ortop–down control in forest soil microcosms? Effects of soil fauna on fungal biomass andC/N mineralisation. Biology and Fertility of Soils 43(3):281–294.Lesica, P. and F. W. Allendorf. 1999. Ecological genetics and the restoration of plant com-munities: Mix or match? Restoration Ecology 7(1):42–50.Li, X., Q. Liu, Z. Liu, W. Shi, D. Yang, and E. Tarasco. 2014. Effects of organic and othermanagement practices on soil nematode communities in tea plantation: A case study insouthern China. Journal of Plant Nutrition and Soil Science 177(4):604–612.Liang, B., J. Lehmann, D. Solomon, J. Kinyangi, J. Grossman, B. O’neill, J. O. Skjemstad,J. Thies, F. J. Luizao, and J. Petersen. 2006. Black carbon increases cation exchangecapacity in soils. Soil Science Society of America Journal 70(5):1719–1730.134BibliographyLou, Y., S. A. Clay, A. S. Davis, A. Dille, J. Felix, A. H. M. Ramirez, C. L. Sprague,and A. C. Yannarell. 2014. An affinity–effect relationship for microbial communities inplant–soil feedback loops. Microbial Ecology 67(4):866–876.Lucas, S. T., E. M. D’Angelo, and M. A. Williams. 2014. Improving soil structure bypromoting fungal abundance with organic soil amendments. Applied Soil Ecology 75:13–23.Lussenhop, J. 1971. A simplified canister–type soil arthropod extractor. Pedobiologia 11:40–45.Maherali, H. and J. N. Klironomos. 2007. Influence of phylogeny on fungal communityassembly and ecosystem functioning. Science 316(5832):1746–1748.Maiti, S. K. 2013. Ecology and ecosystem in mine–degraded land. Pages 21–37 in Ecorestora-tion of the coalmine degraded lands. Springer India.Major, J., M. Rondon, D. Molina, S. Riha, and J. Lehmann. 2010. Maize yield and nutritionduring 4 years after biochar application to a Colombian savanna oxisol. Plant and Soil333(1):117–128.Matias, S. R., M. C. Pagano, F. C. Muzzi, C. A. Oliveira, A. A. Carneiro, S. N. Horta,and M. R. Scotti. 2009. Effect of rhizobia, mycorrhizal fungi and phosphate-solubilizingmicroorganisms in the rhizosphere of native plants used to recover an iron ore area inBrazil. European Journal of Soil Biology 45(3):259–266.McCann, K. S. 2000. The diversity–stability debate. Nature 405(6783):228–233.McConnell, D., A. Shiralipour, and W. Smith. 1994. Compost impact on soil/plant proper-ties. Pages 89–91 in Composting source separated organics JG Press, Emmaus, PA.McCormack, S. A., N. Ostle, R. D. Bardgett, D. W. Hopkins, and A. J. Vanbergen. 2013.Biochar in bioenergy cropping systems: impacts on soil faunal communities and linkedecosystem processes. Global Change Biology Bioenergy 5(2):81–95.McGonigle, T. P., M. H. Miller, D. G. Evans, G. L. Fairchild, and J. A. Swan. 1990. A newmethod which gives an objective measure of colonization of roots by vesicular–arbuscularmycorrhizal fungi. New Phytologist 115(3):495–501.McKinley, V. L., A. D. Peacock, and D. C. White. 2005. Microbial community PLFAand PHB responses to ecosystem restoration in tallgrass prairie soils. Soil Biology andBiochemistry 37(10):1946–1958.135BibliographyMcLauchlan, K. 2006. The nature and longevity of agricultural impacts on soil carbon andnutrients: A review. Ecosystems 9(8):1364–1382.Mehmood, T., H. Martens, S. Sæbø, J. Warringer, and L. Snipen. 2011. A partial leastsquares based algorithm for parsimonious variable selection. Algorithms for MolecularBiology 6(1):27.Menta, C. 2012. Soil fauna diversity–function, soil degradation, biological indices, soilrestoration. Pages 59–94 in G. A. Lameed, editor. Biodiversity conservation in a diverseworld. InTech, Rijeka, Croatia.Mevik, B. H. and R. Wehrens. 2007. The pls package: Principal component and partialleast squares regression in R. Journal of Statistical Software 18(2):1–24.Mikola, J. and H. Seta¨la¨. 1998. Productivity and trophic-level biomasses in a microbial-based soil food web. Oikos 82(1):158–168.Miller, R. M. and J. D. Jastrow. 1992. The application of VA mycorrhizae to ecosystemrestoration and reclamation. Chapman and Hall, New York, New York.Mkhabela, M. S. and P. R. Warman. 2005. The influence of municipal solid waste compost onyield, soil phosphorus availability and uptake by two vegetable crops grown in a Pugwashsandy loam soil in Nova Scotia. Agriculture, Ecosystems and Environment 106(1):57–67.Montemurro, F., M. Maiorana, G. Convertini, and D. Ferri. 2006. Compost organic amend-ments in fodder crops: Effects on yield, nitrogen utilization and soil characteristics. Com-post Science and Utilization 14(2):114–123.Movahedi-Naeini, S. A. R. and H. F. Cook. 2000. Influence of municipal compost on tem-perature, water, nutrient status and the yield of maize in a temperate soil. Soil Use andManagement 16(3):215–221.Moya-Larao, J. and G. Corcobado. 2008. Plotting partial correlation and regression inecological studies. Web Ecology 8:35–46.Mukherjee, A. and A. R. Zimmerman. 2013. Organic carbon and nutrient release from arange of laboratory–produced biochars and biochar–soil mixtures. Geoderma 193:122–130.Mummey, D. L., P. D. Stahl, and J. S. Buyer. 2002. Microbial biomarkers as an indicator ofecosystem recovery following surface mine reclamation. Applied Soil Ecology 21(3):251–259.136BibliographyMummey, D. L., P. M. Antunes, and M. C. Rillig. 2009. Arbuscular mycorrhizal fungipre–inoculant identity determines community composition in roots. Soil Biology and Bio-chemistry 41(6):1173–1179.Nemergut, D. R., S. P. Anderson, C. C. Cleveland, A. P. Martin, A. E. Miller, A. Seimon,and S. K. Schmidt. 2007. Microbial community succession in an unvegetated, recentlydeglaciated soil. Microbial Ecology 53(1):110–122.Nguyen, B. T., J. Lehmann, J. Kinyangi, R. Smernik, S. J. Riha, and M. H. Engelhard. 2008.Long–term black carbon dynamics in cultivated soil. Biogeochemistry 89(3):295–308.Niklas, K. J. and B. J. Enquist. 2002. On the vegetative biomass partitioning of seed plantleaves, stems, and roots. The American Naturalist 159(5):482–497.Nishio, M. and S. Okano. 1991. Stimulation of the growth of alfalfa and infection of rootswith indigenous vesicular–arbuscular mycorrhizal fungi by the application of charcoal.Bull. Natl. Grassi. Res. Inst 45:61–71.Norland, M. R. and D. L. Veith. 1995. Revegetation of coarse taconite iron ore tailing usingmunicipal solid waste compost. Journal of Hazardous Materials 41(2):123–134.Novak, J. M., I. Lima, B. Xing, J. W. Gaskin, C. Steiner, K. C. Das, M. Ahmedna,D. Rehrah, D. W. Watts, and W. J. Busscher. 2009. Characterization of designer biocharproduced at different temperatures and their effects on a loamy sand. Annals of Environ-mental Science 3:195–206.Noyd, R. K., F. L. Pfleger, and M. P. Russelle. 1995. Interactions between native prairiegrasses and indigenous arbuscular mycorrhizal fungi: Implications for reclamation oftaconite iron ore tailing. New Phytologist 129(4):651–660.Noyd, R. K., F. L. Pfleger, and M. R. Norland. 1996. Field responses to added organicmatter, arbuscular mycorrhizal fungi, and fertilizer in reclamation of taconite iron oretailing. Plant and Soil 179(1):89–97.Odum, E. P. 1969. The strategy of ecosystem development. Science 164(877):262–270.Oelmann, Y., W. Wilcke, V. M. Temperton, N. Buchmann, C. Roscher, J. Schumacher,E. D. Schulze, and W. W. Weisser. 2007. Soil and plant nitrogen pools as related toplant diversity in an experimental grassland. Soil Science Society of America Journal71(3):720–729.Ojala, R. and V. Huhta. 2001. Dispersal of microarthropods in forest soil. Pedobiologia45(5):443–450.137BibliographyOjima, D. S., B. O. M. Dirks, E. P. Glenn, C. E. Owensby, and J. O. Scurlock. 1993.Assessment of C budget for grasslands and drylands of the world. Water, Air, and SoilPollution 70(1):95–109.Okada, H., H. Harada, and I. Kadota. 2005. Fungal–feeding habits of six nematode isolatesin the genus Filenchus. Soil Biology and Biochemistry 37(6):1113–1120.Ontario-Biodiversity-Council. 2010. State of Ontario’s Biodiversity 2010. A report of theOntario Biodiversity Council, Peterborough, ON.Oue´draogo, E., A. Mando, and N. P. Zombre´. 2001. Use of compost to improve soil propertiesand crop productivity under low input agricultural system in West Africa. Agriculture,Ecosystems & Environment 84(3):259–266.Parisi, V., C. Menta, C. Gardi, C. Jacomini, and E. Mozzanica. 2005. Microarthropodcommunities as a tool to assess soil quality and biodiversity: a new approach in Italy.Agriculture, Ecosystems & Environment 105(1):323–333.Parton, W. J., J. M. O. Scurlock, D. S. Ojima, T. G. Gilmanov, R. J. Scholes, D. S. Schimel,T. Kirchner, J. C. Menaut, T. Seastedt, and E. G. Moya. 1993. Observations and modelingof biomass and soil organic matter dynamics for the grassland biome worldwide. GlobalBiogeochemical Cycles 7(4):785–809.Pascual, J. A., C. Garc´ıa, and T. Hernandez. 1999. Lasting microbiological and biochemicaleffects of the addition of municipal solid waste to an arid soil. Biology and Fertility ofSoils 30(1):1–6.Picard, N., L. Saint-Andre´, and M. Henry. 2012. Manual for building tree volume andbiomass allometric equations: From field measurement to prediction. Food and Agricul-tural Organization of the United Nations, Rome and and Centre de Coopration Interna-tionale en Recherche Agronomique pour le Dveloppement , Montpellier.Pimm, S. L. 1991. The balance of nature?: Ecological issues in the conservation of speciesand communities. University of Chicago Press.Piper, J. K. and S. L. Pimm. 2002. The creation of diverse prairie–like communities. Com-munity Ecology 3(2):205–216.Pokorny, M. L., R. L. Sheley, T. J. Svejcar, and R. E. Engel. 2004. Plant species diversity ina grassland plant community: Evidence for forbs as a critical management consideration.Western North American Naturalist 64(2):219–230.138BibliographyPoorter, H. and O. Nagel. 2000. The role of biomass allocation in the growth responseof plants to different levels of light, CO2, nutrients and water: A quantitative review.Functional Plant Biology 27(12):1191–1191.Prach, K. and R. J. Hobbs. 2008. Spontaneous succession versus technical reclamation inthe restoration of disturbed sites. Restoration Ecology 16(3):363–366.Pu¨schel, D., J. Rydlova´, R. Sudova´, and M. Gryndler. 2008a. Cultivation of flax in spoil bankclay: Mycorrhizal inoculation vs. high organic amendments. Journal of Plant Nutritionand Soil Science 171(6):872–877.Pu¨schel, D., J. Rydlova´, and M. Vosa´tka. 2008b. Does the sequence of plant dominantsaffect mycorrhiza development in simulated succession on spoil banks? Plant and Soil302(1):273–282.Pu¨schel, D., J. Rydlova´, R. Sudova´, M. Gryndler, and M. Vosa´tka. 2011. The potential ofmycorrhizal inoculation and organic amendment to increase yields of Galega orientalisand Helianthus tuberosus in a spoil bank substrate. Journal of Plant Nutrition and SoilScience 174(4):664–672.Pywell, R. F., J. M. Bullock, A. Hopkins, K. J. Walker, T. H. Sparks, M. J. W. Burke, andS. Peel. 2002. Restoration of species rich grassland on arable land: Assessing the limitingprocesses using a multi site experiment. Journal of Applied Ecology 39(2):294–309.R-Core-Team. 2013. R: A language and environment for statistical computing. R Foundationfor Statistical Computing. Available at http://www.R-project.org/.Rao, A. V. and R. Tak. 2002. Growth of different tree species and their nutrient uptake inlimestone mine spoil as influenced by arbuscular mycorrhizal (AM)–fungi in Indian aridzone. Journal of Arid Environments 51(1):113–119.Rayburn, E. and J. Lozier. 2003. A falling plate meter for estimating pasture forage mass.West Virginia University, Morgantown, West Virginia Extension. Fact Sheet. Nov. 2003 .Rayburn, E. B. and S. B. Rayburn. 1998. A standardized plate meter for estimating pasturemass in on–farm research trials. Agronomy Journal 90(2):238–241.Reynolds, H. L., A. Packer, J. D. Bever, and K. Clay. 2003. Grassroots ecology: Plant–microbe–soil interactions as drivers of plant community structure and dynamics. Ecology84(9):2281–2291.139BibliographyRichter, B. S. and J. C. Stutz. 2002. Mycorrhizal inoculation of big sacaton: Implicationsfor grassland restoration of abandoned agricultural fields. Restoration Ecology 10(4):607–616.Rillig, M. C. 2004. Arbuscular mycorrhizae and terrestrial ecosystem processes. EcologyLetters 7(8):740–754.Risser, P. G. and W. J. Parton. 1982. Ecosystem analysis of the tallgrass prairie: Nitrogencycle. Ecology 63(5):1342–1351.Rodger, L. 1998. Tallgrass communities of southern Ontario: A recovery plan. WorldWildlife Fund and Ontario Ministry of Natural Resources.Rondon, M., J. Lehmann, J. Ramirez, and M. Hurtado. 2007. Biological nitrogen fixationby common beans (Phaseolus vulgaris L.) increases with bio–char additions. Biology andFertility of Soils 43(6):699–708.Ros, M., M. T. Hernandez, and C. Garcia. 2003. Soil microbial activity after restoration ofa semiarid soil by organic amendments. Soil Biology and Biochemistry 35(3):463–469.Ros, M., S. Klammer, B. Knapp, K. Aichberger, and H. Insam. 2006. Long–term effects ofcompost amendment of soil on functional and structural diversity and microbial activity.Soil Use and Management 22(2):209–218.Roub´ıc˘kova´, A., O. Mudra´k, and J. Frouz. 2013. Soil fauna plant interactions during suc-cession at post–mining sites. Pages 250–264 in J. Frouz, editor. Soil biota and ecosystemdevelopment in post mining sites. CRC Press, Boca Raton, FL.Rowe, H. I., C. S. Brown, and V. P. Claassen. 2007. Comparisons of mycorrhizal responsive-ness with field soil and commercial inoculum for six native montane species and Bromustectorum. Restoration Ecology 15(1):44–52.Ruiz-Jaen, M. C. and M. Aide. 2005. Restoration success: How is it being measured?Restoration Ecology 13(3):569–577.Rydlova´, J., D. Pu¨schel, M. Vosa´tka, and K. Charva´tova´. 2008. Different effect of mycor-rhizal inoculation in direct and indirect reclamation of spoil banks. Journal of AppliedBotany and Food Quality 82(1):15–20.Sala, O. E., W. J. Parton, L. A. Joyce, and W. K. Lauenroth. 1988. Primary production ofthe central grassland region of the United States. Ecology 69(1):40–45.140BibliographyScheffer, M., S. Carpenter, J. A. Foley, C. Folke, and B. Walker. 2001. Catastrophic shiftsin ecosystems. Nature 413(6856):591–596.Scherber, C., N. Eisenhauer, W. W. Weisser, B. Schmid, W. Voigt, M. Fischer, E.-D.Schulze, C. Roscher, A. Weigelt, and E. Allan. 2010. Bottom–up effects of plant diversityon multitrophic interactions in a biodiversity experiment. Nature 468(7323):553–556.Scheu, S. and E. Schulz. 1996. Secondary succession, soil formation and development of adiverse community of oribatids and saprophagous soil macro–invertebrates. Biodiversityand Conservation 5(2):235–250.Schielzeth, H. 2010. Simple means to improve the interpretability of regression coefficients.Methods in Ecology and Evolution 1(2):103–113.Schimmelpfennig, S., C. Mu¨ller, L. Gru¨nhage, C. Koch, and C. Kammann. 2014. Biochar,hydrochar and uncarbonized feedstock application to permanent grassland – effects ongreenhouse gas emissions and plant growth. Agriculture, Ecosystems & Environment191:39–52.Schmidt, H.-P., C. Kammann, C. Niggli, M. W. H. Evangelou, K. A. Mackie, and S. Abiven.2014. Biochar and biochar-compost as soil amendments to a vineyard soil: Influences onplant growth, nutrient uptake, plant health and grape quality. Agriculture, Ecosystems& Environment 191:117–123.Schmidt, M. W. I. and A. G. Noack. 2000. Black carbon in soils and sediments: Anal-ysis, distribution, implications, and current challenges. Global Biogeochemical Cycles14(3):777–793.Schulz, H. and B. Glaser. 2012. Effects of biochar compared to organic and inorganic fer-tilizers on soil quality and plant growth in a greenhouse experiment. Journal of PlantNutrition and Soil Science 175(3):410–422.Schwartz, M. W., J. D. Hoeksema, C. A. Gehring, N. C. Johnson, J. N. Klironomos, L. K.Abbott, and A. Pringle. 2006. The promise and the potential consequences of the globaltransport of mycorrhizal fungal inoculum. Ecology Letters 9(5):501–515.Seastedt, T. R. 1984. The role of microarthropods in decomposition and mineralizationprocesses. Annual Review of Entomology 29(1):25–46.Se´re´, G., C. Schwartz, S. Ouvrard, C. Sauvage, J. Renat, and J. Morel. 2008. Soil construc-tion: A step for ecological reclamation of derelict lands. Journal of Soils and Sediments8(2):130–136.141BibliographyShiralipour, A., D. B. McConnell, and W. H. Smith. 1992. Physical and chemical propertiesof soils as affected by municipal solid waste compost application. Biomass and Bioenergy3(3):261–266.Shrestha, G., S. J. Traina, and C. W. Swanston. 2010. Black carbon’s properties and rolein the environment: A comprehensive review. Sustainability 2:294–320.Singh, B., B. P. Singh, and A. L. Cowie. 2010. Characterisation and evaluation of biocharsfor their application as a soil amendment. Soil Research 48(7):516–525.Singh, J. S., W. K. Lauenroth, and R. K. Steinhorst. 1975. Review and assessment of varioustechniques for estimating net aerial primary production in grasslands from harvest data.Botanical Review 41(2):181–232.Six, J., H. Bossuyt, S. Degryze, and K. Denef. 2004. A history of research on the linkbetween (micro)aggregates, soil biota, and soil organic matter dynamics. Soil and TillageResearch 79(1):7–31.Skjemstad, J. O., D. C. Reicosky, A. R. Wilts, and J. A. McGowan. 2002. Charcoal carbonin US agricultural soils. Soil Science Society of America Journal 66(4):1249–1255.Smith, R. S., R. S. Shiel, R. D. Bardgett, D. Millward, P. Corkhill, G. Rolph, P. J. Hobbs,and S. Peacock. 2003. Soil microbial community, fertility, vegetation and diversity astargets in the restoration management of a meadow grassland. Journal of Applied Ecology40(1):51–64.Smith, S. E. and D. J. Read. 2008. Mycorrhizal Symbiosis, 3rd edition. Academic Press,Elsevier, Amsterdam.Smith, V. C. and M. A. Bradford. 2003. Litter quality impacts on grassland litter de-composition are differently dependent on soil fauna across time. Applied Soil Ecology24(2):197–203.Sohi, S., E. Lopez-Capel, E. Krull, and R. Bol. 2009. Biochar, climate change and soil: Areview to guide future research. CSIRO Land and Water Science Report 05/09. 64pp.Stahl, P. D., S. E. Williams, and M. Christensen. 1988. Efficacy of native vesicular–arbuscular mycorrhizal fungi after severe soil disturbance. New Phytologist 110(3):347–354.Steel, H., B. Vandecasteele, K. Willekens, K. Sabbe, T. Moens, and W. Bert. 2012. Nema-tode communities and macronutrients in composts and compost–amended soils as affectedby feedstock composition. Applied Soil Ecology 61:100–112.142BibliographySteel, H., D. Buchan, S. De Neve, M. Couvreur, T. Moens, and W. Bert. 2013a. Nematodeand microbial communities in a rapidly changing compost environment: How nematodeassemblages reflect composting phases. European Journal of Soil Biology 56:1–10.Steel, H., F. Verdoodt, A. Cˇerevkova´, M. Couvreur, P. Fonderie, T. Moens, and W. Bert.2013b. Survival and colonization of nematodes in a composting process. InvertebrateBiology 132(2):108–119.Steinbeiss, S., G. Gleixner, and M. Antonietti. 2009. Effect of biochar amendment on soilcarbon balance and soil microbial activity. Soil Biology and Biochemistry 41(6):1301–1310.Steiner, C., W. Teixeira, J. Lehmann, T. Nehls, J. Luis, J. de Maceˆdo, W. E. H. Blum, andW. Zech. 2007. Long term effects of manure, charcoal and mineral fertilization on cropproduction and fertility on a highly weathered Central Amazonian upland soil. Plant andSoil 291(1):275–290.Streit, B., A. Buehlmann, and P. Reutimann. 1985. Mite succession in compost communities:Studies with Oribatei, Gamasina, and Uropodina. Pedobiologia 28(1):1–12.Suding, K. N., K. L. Gross, and G. R. Houseman. 2004. Alternative states and positivefeedbacks in restoration ecology. Trends in Ecology and Evolution 19(1):46–53.Swift, M. J., O. W. Heal, and J. M. Anderson. 1979. Decomposition in terrestrial ecosystems,vol. 5. Blackwell Scientific Publication, Oxford, UK.Symonds, J. E. and J. Pither. 2012. Multi–scale analyses of exotic and native plantspecies diversity within Canada’s endangered antelope–brush ecosystem. Plant Ecology213(8):1263–1275.Taheri, W. I. and J. D. Bever. 2010. Adaptation of plants and arbuscular mycorrhizal fungito coal tailings in Indiana. Applied Soil Ecology 45(3):138–143.Termorshuizen, A. J., S. W. Moolenaar, A. H. M. Veeken, and W. J. Blok. 2004. The valueof compost. Reviews in Environmental Science and Biotechnology 3(4):343–347.Thies, J. E. and M. C. Rillig. 2009. Characteristics of biochar: Biological properties. Pages85–105 in J. Lehmann, and S. Joseph, editors. Biochar for environmental management:Science and technology. Earthscan, London.Thorne, M., L. Rhodes, and J. Cardina. 2013. Effectivity of arbuscular mycorrhizal fungicollected from reclaimed mine soil and tallgrass prairie. Open Journal of Ecology 3(3):224–233.143BibliographyTiessen, H., J. O. Moir, and M. R. Carter. 1993. Characterization of available P by sequentialextraction. Pages 75–86 in M. R. Carter, editor. Soil sampling and methods of analysis.Lewis Publishers, Boca Raton, FL.Tilman, D. and J. A. Downing. 1994. Biodiversity and stability in grasslands. Nature367(6461):363–365.Tilman, D., J. Knops, D. Wedin, P. Reich, M. Ritchie, and E. Siemann. 1997. The influenceof functional diversity and composition on ecosystem processes. Science 277:1300–1302.Tremblay, A. and J. Ransijn. 2013. LMERConvenienceFunctions: A suite of func-tions to back–fit fixed effects and forward–fit random effects, as well as othermiscellaneous functions. R package version 2.5. Available at http://CRAN.R-project.org/package=LMERConvenienceFunctions .Tscherko, D., J. Rustemeier, A. Richter, W. Wanek, and E. Kandeler. 2003. Functionaldiversity of the soil microflora in primary succession across two glacier forelands in theCentral Alps. European Journal of Soil Science 54(4):685–696.van der Heijden, M. G., R. D. Bardgett, and N. M. van Straalen. 2008. The unseen majority:Soil microbes as drivers of plant diversity and productivity in terrestrial ecosystems.Ecology Letters 11(3):296–310.van der Heijden, M. G. A., J. N. Klironomos, M. Ursic, P. Moutoglis, R. Streitwolf-Engel,T. Boller, A. Wiemken, and I. R. Sanders. 1998. Mycorrhizal fungal diversity determinesplant biodiversity, ecosystem variability and productivity. Nature 396:69–72.van Veen, J. A. and E. A. Paul. 1979. Conversion of biovolume measurements of soil or-ganisms, grown under various moisture tensions, to biomass and their nutrient content.Applied and Environmental Microbiology 37(4):686–692.Verheijen, F. G. A., E. R. Graber, N. Ameloot, A. C. Bastos, S. Sohi, and H. Knicker. 2014.Biochars in soils: New insights and emerging research needs. European Journal of SoilScience 65(1):22–27.Visser, S. and D. Parkinson. 1992. Soil biological criteria as indicators of soil quality: Soilmicroorganisms. American Journal of Alternative Agriculture 7(1):33–37.Vittoz, P. and A. Guisan. 2007. How reliable is the monitoring of permanent vegetationplots? A test with multiple observers. Journal of Vegetation Science 18(3):413–422.144BibliographyWahba, M. M. and K. M. Darwish. 2008. Micro–morphological changes of sandy soilsthrough the application of compost manure. Journal of Applied Biological Sciences2(3):95–98.Wali, M. K. 1999. Ecological succession and the rehabilitation of disturbed terrestrial ecosys-tems. Plant and Soil 213(1):195–220.Walker, L. R., J. Walker, and R. del Moral. 2007. Forging a new alliance between successionand restoration. Pages 1–18 in L. R. Walker, J. Walker, and R. J. Hobbs, editors. Linkingrestoration and ecological succession. Springer New York, New York.Wallwork, J. A. 1976. The distribution and diversity of soil fauna. Academic Press NewYork.Walter, I., F. Martinez, and G. Cuevas. 2006. Plant and soil responses to the application ofcomposted MSW in a degraded, semiarid shrubland in central Spain. Compost Scienceand Utilization 14(2):147–154.Wardle, D. A. 1999. How soil food webs make plants grow. Trends in Ecology and Evolution14(11):418–420.Wardle, D. A. 2006. The influence of biotic interactions on soil biodiversity. Ecology Letters9(7):870–886.Wardle, D. A., G. W. Yeates, R. N. Watson, and K. S. Nicholson. 1995. The detritusfood–web and the diversity of soil fauna as indicators of disturbance regimes in agro–ecosystems. Pages 35–43 in H. P. Collins, G. P. Robertson, and M. J. Klug, editors. Thesignificance and regulation of soil biodiversity. Springer Netherlands.Wardle, D. A., O. Zackrisson, and M. C. Nilsson. 1998. The charcoal effect in boreal forests:Mechanisms and ecological consequences. Oecologia 115(3):419–426.Wardle, D. A., K. I. Bonner, and G. M. Barker. 2002. Linkages between plant litter de-composition, litter quality, and vegetation responses to herbivores. Functional Ecology16(5):585–595.Wardle, D. A., R. D. Bardgett, J. N. Klironomos, H. Seta¨la¨, W. H. van der Putten, and D. H.Wall. 2004. Ecological linkages between aboveground and belowground biota. Science304(5677):1629–1633.Warnock, D. D., J. Lehmann, T. W. Kuyper, and M. C. Rillig. 2007. Mycorrhizal responsesto biochar in soil – concepts and mechanisms. Plant and Soil 300(1):9–20.145BibliographyWarnock, D. D., D. L. Mummey, B. McBride, J. Major, J. Lehmann, and M. C. Rillig.2010. Influences of non–herbaceous biochar on arbuscular mycorrhizal fungal abundancesin roots and soils: Results from growth–chamber and field experiments. Applied SoilEcology 46(3):450–456.Weber, J., A. Karczewska, J. Drozd, M. Licznar, S. Licznar, E. Jamroz, and A. Kocowicz.2007. Agricultural and ecological aspects of a sandy soil as affected by the application ofmunicipal solid waste composts. Soil Biology and Biochemistry 39(6):1294–1302.Wold, S., M. Sjo´stro´m, and L. Eriksson. 2001. PLS–regression: A basic tool of chemometrics.Chemometrics and Intelligent Laboratory Systems 58(2):109–130.Wolkowski, R. P. 2003. Nitrogen management considerations for landspreading municipalsolid waste compost. Journal of Environmental Quality 32(5):1844–1850.Wortmann, C. S. and D. T. Walters. 2007. Residual effects of compost and plowing onphosphorus and sediment in runoff. Journal of Environmental Quality 36(5):1521–1527.Wu, F. Y., Y. L. Bi, and M. H. Wong. 2009. Dual inoculation with an arbuscular mycorrhizalfungus and Rhizobium to facilitate the growth of Alfalfa on coal mine substrates. Journalof Plant Nutrition 32(5):755–771.Xu, G., L. L. Wei, J. N. Sun, H. B. Shao, and S. X. Chang. 2013. What is more importantfor enhancing nutrient bioavailability with biochar application into a sandy soil: Director indirect mechanism? Ecological Engineering 52:119–124.Yeates, G. W., T. Bongers, R. G. M. De Goede, D. W. Freckman, and S. S. Georgieva.1993. Feeding habits in soil nematode families and generaan outline for soil ecologists.Journal of Nematology 25(3):315.Yeniay, O. and A. Goktas. 2002. A comparison of partial least squares regression with otherprediction methods. Hacettepe Journal of Mathematics and Statistics 31(99):99–111.Zhao, J. and D. A. Neher. 2013. Soil nematode genera that predict specific types of distur-bance. Applied Soil Ecology 64:135–141.Zuur, A. F., E. N. Ieno, and G. M. Smith. 2007. Analysing ecological data. Springer NewYork, New York.146AppendicesAppendix AR Code for AM Fungal Plant PlugRoot Colonization in the PlantPlug Triall i b r a r y ( ggp lot2 )c o l . d f <− read . csv ( ”C:\\ Users \\Ohsowski\\Documents\\PhD\\D i s s e r t a t i o n \\Data\\AMF Colon i za t i on \\Plant Plug AMF Colon i za t i on . csv ” )names ( c o l . d f ) <− c ( ”abb” , ” fun ” , ” rep ” , ”myco” , ”ac” , ”vc” )c o l . d f $ to t <− c o l . d f $ac + c o l . d f $vccolMn . df <− aggregate ( c o l . d f $ tot , by=l i s t ( c o l . d f $abb , c o l . d f $myco , c o l . d f $ fun ), FUN = mean , na . rm = TRUE)names ( colMn . df ) <− c ( ”abb” , ”myco” , ” fun ” , ”totMn” )colTotSD . df <− aggregate ( c o l . d f $ tot , by=l i s t ( c o l . d f $abb , c o l . d f $myco , c o l . d f $fun ) , FUN = sd , na . rm = TRUE)names ( colTotSD . df ) <− c ( ”abb” , ”myco” , ” fun ” , ”totSD” )colTotSD . df $ totSE <− colTotSD . df $totSD / s q r t (10)f i n a l . d f <− merge ( colMn . df , colTotSD . df , by = c ( ”abb” , ”myco” , ” fun ” ) )147Appendix A. R Code for AM Fungal Plant Plug Root Colonization in the Plant Plug Trial# Graphsgraph <− ggp lot ( f i n a l . df , aes ( x = f a c t o r ( abb ) , y = totMn , f i l l = myco , ) )l i m i t s <− aes (ymax = totMn + totSE , ymin = totMn − totSE )dodge <− p o s i t i o n dodge ( width = 0 . 9 )l egendLabe l s <− c ( ”No Inoc ” , ” Inoc ” )legendBreaks <− c ( ”N” , ”Y” )graph output <− graph +theme bw( ) +geom bar ( p o s i t i o n = dodge , s t a t = ” i d e n t i t y ” ) +geom er ro rba r ( l i m i t s , width = 0 . 5 , c o l o r = ” black ” , p o s i t i o n = dodge ) +labs ( x = ” Plant Spec i e s ” ,y = ” Total % Co lon i za t i on ” ) +s c a l e f i l l grey ( breaks = legendBreaks , l a b e l s = legendLabels , s t a r t = 0 . 3 ,end = 0 . 7 5 ) +f a c e t g r id (˜ fun ) +theme ( panel . g r i d . major = element l i n e ( co l ou r = ’ grey85 ’ ) ) +theme ( panel . g r i d . minor = element l i n e ( co l ou r = ’ grey85 ’ ) ) +theme ( legend . p o s i t i o n = c ( 0 . 8 8 , 0 . 88 ) ) +theme ( a x i s . t ex t . x = element text ( s i z e = 13) ) +theme ( a x i s . t ex t . y = element text ( s i z e = 13) ) +theme ( a x i s . t i t l e . x = element text ( s i z e = 12 , v ju s t = 0 . 1 ) ) +theme ( a x i s . t i t l e . y = element text ( s i z e = 12 , ang le = 90) ) +theme ( legend . t i t l e = element blank ( ) ) +theme ( s t r i p . background = element r e c t ( f i l l = ’ grey85 ’ ) ) +theme ( s t r i p . t ex t . x = element text ( f a c e = ’ bold ’ , s i z e = 14) )graph output148Appendix BR Code for AM Fungal RootColonization in the Plant PlugTrial Field Plotsl i b r a r y (RODBC)l i b r a r y ( ggp lot2 )l i b r a r y (glmmADMB)# Data Transformation Sec t ionCube . Tns <− f unc t i on ( x ) { x ˆ 3 }Square . Tns <− f unc t i on ( x ) { x ˆ 2 }Raw. Tns <− f unc t i on ( x ) { x }Sqrt . Tns <− f unc t i on ( x ) { s q r t ( x ) }Log . Tns <− f unc t i on ( x ) { l og10 ( x + 1) }RecipRoot . Tns <− f unc t i on ( x ) { −1 / s q r t ( x ) }Recip . Tns <− f unc t i on ( x ) { −1 / ( x ) }InvSquare . Tns <− f unc t i on ( x ) { −1 / ( x ˆ 2) }# Back Transformationsunsca l e . fn <− f unc t i on ( x ) { as . data . frame ( unsca l e (x , unsca l e . ob j e c t ) ) }backLog . Tns <− f unc t i on ( x ) { 10 ˆ ( x ) − 1 }backSqrt . Tns <− f unc t i on ( x ) { ( x ) ˆ 2 }backRaw . Tns <− f unc t i on ( x ) { ( x ) }149Appendix B. R Code for AM Fungal Root Colonization in the Plant Plug Trial Field Plots# Pred i c t i on Data Centered and Sca led to Unity## This un i t y f unc t i on compares the c o l l e c t e d t e s t p l an t s to the p l s r dataana lyzed## x = t e s t v a r i a b l e data ; y = main data ; r e qu i r ed to have same mean /var iancePredU <− f unc t i on (x , y ) {cente r <− x − mean( y )sdCenter <− sd ( y − mean( y ) )c en te r / sdCenter}# Pearson ’ s Method ( Parametric Test )PearsonsMethod <− f unc t i on ( x ) {cor (x , use = ” complete . obs” , method = ” pearson ” )}# Spearman ’ s Method (Non−Parametric Test )SpearmansMethod <− f unc t i on (x , y ) {cor (x , use = ” complete . obs” , method = ”spearman” )}# Plot Informat ion Data Framechannel <− odbcConnectAccess ( ”C: / Users /Ohsowski/Documents/PhD/ D i s s e r t a t i o n /Data/ d i s s e r t a t i o n data exp 1 13 aug2” )p l o t In f oCo l . df <− sqlQuery ( channel , ”SELECT plot , b iochar rate , compost rate ,amf , hgt , treatment FROM DATA Q WHERE harvestOne = FALSE” )c l o s e ( channel )# Training Data , Sum and Mean Stem Length Organizat ionchannel <− odbcConnectAccess ( ”C: / Users /Ohsowski/Documents/PhD/ D i s s e r t a t i o n /Data/ d i s s e r t a t i o n data exp 1 13 aug2” )ac . df <− sqlQuery ( channel , ”SELECT plot , data , dataType , season FROM AMF WHEREdataType = ’ acCol ’ AND harvestOne = FALSE” )vc . df <− sqlQuery ( channel , ”SELECT plot , data , dataType , season FROM AMF WHEREdataType = ’ vcCol ’ AND harvestOne = FALSE” )c l o s e ( channel )totCol . df <− merge ( ac . df , vc . df , by = c ( ” p l o t ” , ” season ” ) )totCol . df $ to t <− totCol . df $ data . x + totCol . df $ data . y150Appendix B. R Code for AM Fungal Root Colonization in the Plant Plug Trial Field Plotsf i n a l C o l . df <− unique ( merge ( p l o t In f oCo l . df , totCol . df , by = c ( ” p l o t ” ) ) )f i n a l C o l . df <− data . frame ( f i n a l C o l . df $ plot , f i n a l C o l . df $ treatment , f i n a l C o l . df$hgt , f i n a l C o l . df $compost rate , f i n a l C o l . df $ b iochar rate , f i n a l C o l . df $amf ,f i n a l C o l . df $ season )f i n a l C o l . df $ data . x , f i n a l C o l . df $ data . y , f i n a l C o l . df $ to t )names ( f i n a l C o l . df ) <− c ( ” p l o t ” , ” treatment ” , ”hgt ” , ”compost ra t e ” , ” b iocharra t e ” , ”amf” , ” season ” , ”arb” , ” ves ” , ” to t ” )f i n a l C o l . df $ season <− i f e l s e ( f i n a l C o l . df $ season == ”2011” , ”A” , ”B” )f i n a l C o l . df $ p l o t <− as . f a c t o r ( f i n a l C o l . df $ p l o t )f i n a l C o l . df $ treatment <− as . f a c t o r ( f i n a l C o l . df $ treatment )f i n a l C o l . df $amf <− as . f a c t o r ( f i n a l C o l . df $amf )f i n a l C o l . df $ season <− as . f a c t o r ( f i n a l C o l . df $ season )f i n a l C o l . df $ ra t e <− f i n a l C o l . df $compost ra t e + f i n a l C o l . df $ b iochar ra t emean( f i n a l C o l . d f $ to t )var ( f i n a l C o l . df $ to t )f i n a l C o l . df $ totT <− Raw. Tns ( f i n a l C o l . d f $ to t )qqnorm ( f i n a l C o l . df $ totT )q q l i n e ( f i n a l C o l . df $ totT )TC. n u l l <− lmer ( totT ˜ 1 + ( 1 | p lo t ) , data = f i n a l C o l . df )TC1 <− lmer ( totT ˜ treatment ∗ season ∗ hgt ∗ amf + ( 1 | p lo t ) , data = f i n a l C o l. df )summary(TC1)anova (TC. nu l l , TC1)r e l L i k (TC. nu l l , TC1)mcp . fnc (TC1)pamer . fnc (TC1)TC2 <− update (TC1 , . ˜ . −treatment : season : amf : hgt )summary(TC2)anova (TC2)anova (TC1, TC2)r e l L i k (TC1, TC2)TC3 <− update (TC2 , . ˜ . −season : hgt : amf )summary(TC3)anova (TC3)151Appendix B. R Code for AM Fungal Root Colonization in the Plant Plug Trial Field Plotsanova (TC2, TC3)r e l L i k (TC2, TC3)TC4 <− update (TC3 , . ˜ . −treatment : season : hgt )summary(TC4)anova (TC4)anova (TC3, TC4)r e l L i k (TC3, TC4)# FINAL MODELTC5 <− update (TC4 , . ˜ . −treatment : hgt : amf )summary(TC5)anova (TC5)anova (TC4, TC5)r e l L i k (TC4, TC5)anova (TC5)summary(TC5)pamer . fnc (TC5)TC5. ph <− mcposthoc . fnc ( model = TC5, var = l i s t ( ph1 = ” treatment ” ) )summary(TC5. ph)#Wireframe Graphs c a t t e r 3 d ( f i n a l C o l . df $ totT ˜ f i n a l C o l . d f $ hgt + f i n a l C o l . d f $ rate ,bg=” white ” , a x i s . s c a l e s=TRUE, g r id=TRUE, id . method=” i d e n t i f y ” ,e l l i p s o i d=FALSE, xlab=” biochar ra t e ” , ylab=” cover ” , z lab=”compostra t e ” ,groups = as . f a c t o r ( f i n a l C o l . df $amf ) )wireframe ( totT ˜ ra t e + hgt , data=f i n a l C o l . df , x lab = ”Compost Rate” , ylab = ”Biochar Rate ) ” ,drape = TRUE,co lo rkey = TRUE)p <− wireframe ( v a r i a b l e ˜ ra t e ∗ hgt , data=f i n a l . d f )npanel <− c (4 , 2)rotx <− c (−50 , −80)ro t z <− seq (30 , 300 , l ength = npanel [ 1 ]+1)update (p [ rep (1 , prod ( npanel ) ) ] , l ayout = npanel ,panel = func t i on ( . . . , s c r e en ) {panel . wire frame ( . . . , s c r e en = l i s t ( z = ro t z [ cur r ent . column ( ) ] ,x = rotx [ cur rent . row ( ) ] ) )# Graph Set−up152Appendix B. R Code for AM Fungal Root Colonization in the Plant Plug Trial Field PlotstotCol <− aggregate ( f i n a l C o l . df $ tot , by = l i s t ( f i n a l C o l . df $ season , f i n a l C o l . df$ treatment , f i n a l C o l . df $amf ) , FUN = mean , na . rm = TRUE)names ( totCol ) <− c ( ” season ” , ” treatment ” , ”amf” , ” data ” )totColSD <− aggregate ( f i n a l C o l . df $ tot , by = l i s t ( f i n a l C o l . df $ season , f i n a l C o l .df $ treatment , f i n a l C o l . df $amf ) , FUN = sd , na . rm = TRUE)names ( totColSD ) <− c ( ” season ” , ” treatment ” , ”amf” , ” sd” )totCol . gr <− merge ( totCol , totColSD , by = c ( ” season ” , ” treatment ” , ”amf” ) )# GGPLOT Graphinggraph <− ggp lot ( data = totCol . gr , aes ( x = treatment , y = data , f i l l = amf ) )l i m i t s <− aes (ymax = totCol . gr $ data + totCol . gr $sd , ymin = totCol . gr $ data −totCol . gr $ sd )dodge <− p o s i t i o n dodge ( width = 0 . 9 )graph output <− graph +theme bw( ) +geom bar ( p o s i t i o n = ”dodge” , s t a t = ” i d e n t i t y ” ) +geom er ro rba r ( l i m i t s , width = 0 . 5 , c o l o r = ” black ” , p o s i t i o n = dodge , s t a t =” i d e n t i t y ” ) +f a c e t g r id ( . ˜ season ) +labs ( x = ”Carbon Amendment” ,y = ” Estimated Biomass\n dry mass ( g ) ” ) +theme ( panel . g r i d . major = element l i n e ( co l ou r = ’ grey85 ’ ) ) +theme ( panel . g r i d . minor = element l i n e ( co l ou r = ’ grey85 ’ ) ) +theme ( legend . p o s i t i o n = c (0 , −0.40) ) +theme ( a x i s . t ex t . x = element text ( s i z e = 11 , ang le = 60) ) +theme ( a x i s . t ex t . y = element text ( s i z e = 13) ) +theme ( a x i s . t i t l e . x = element text ( s i z e = 12 , v ju s t = 0 . 1 ) ) +theme ( a x i s . t i t l e . y = element text ( s i z e = 12 , ang le = 90) ) +theme ( legend . t i t l e = element blank ( ) ) +theme ( s t r i p . background = element r e c t ( f i l l = ’ grey85 ’ ) ) +theme ( s t r i p . t ex t . x = element text ( f a c e = ’ bold ’ , s i z e = 14) )graph output153Appendix CR Code for Plant Plug TrialBiomass Predictions and Plot MassCalculationsl i b r a r y (RODBC)l i b r a r y ( reshape )l i b r a r y ( t c l t k )l i b r a r y ( t c l t k 2 )l i b r a r y ( Hmisc )l i b r a r y ( p l s )l i b r a r y (DMwR)l i b r a r y ( ggp lot2 )l i b r a r y ( p ly r )l i b r a r y ( s c a t t e r p l o t 3 d )l i b r a r y ( rg l , pos=4)l i b r a r y (mgcv , pos=4)l i b r a r y (MuMIn)# Data Transformation Sec t ionCube . Tns <− f unc t i on ( x ) { x ˆ 3 }Square . Tns <− f unc t i on ( x ) { x ˆ 2 }Raw. Tns <− f unc t i on ( x ) { x }Sqrt . Tns <− f unc t i on ( x ) { s q r t ( x ) }Log . Tns <− f unc t i on ( x ) { l og10 ( x + 1) }RecipRoot . Tns <− f unc t i on ( x ) { −1 / s q r t ( x ) }Recip . Tns <− f unc t i on ( x ) { −1 / ( x ) }InvSquare . Tns <− f unc t i on ( x ) { −1 / ( x ˆ 2) }154Appendix C. R Code for Plant Plug Trial Biomass Predictions and Plot Mass Calculations# Back Transformationsunsca l e . fn <− f unc t i on ( x ) { as . data . frame ( unsca l e (x , unsca l e . ob j e c t ) ) }backLog . Tns <− f unc t i on ( x ) { 10 ˆ ( x ) − 1 }backSqrt . Tns <− f unc t i on ( x ) { ( x ) ˆ 2 }backRaw . Tns <− f unc t i on ( x ) { ( x ) }# Pred i c t i on Data Centered and Sca led to Unity## This un i t y f unc t i on compares the c o l l e c t e d t e s t p l an t s to the p l s r dataana lyzed## x = t e s t v a r i a b l e data ; y = main data ; r e qu i r ed to have same mean /var iancePredU <− f unc t i on (x , y ) {cente r <− x − mean( y )sdCenter <− sd ( y − mean( y ) )c en te r / sdCenter}# Pearson ’ s Method ( Parametric Test )PearsonsMethod <− f unc t i on ( x ) {cor (x , use = ” complete . obs” , method = ” pearson ” )}# Spearman ’ s Method (Non−Parametric Test )SpearmansMethod <− f unc t i on (x , y ) {cor (x , use = ” complete . obs” , method = ”spearman” )}################################## 2011 Plant Pred i c t i on Data ################################### Plot Informat ion Data Framechannel <− odbcConnectAccess ( ”C: / Users /Ohsowski/Documents/PhD/ D i s s e r t a t i o n /Data/ d i s s e r t a t i o n data exp 1 13 aug2” )155Appendix C. R Code for Plant Plug Trial Biomass Predictions and Plot Mass Calculationsp l o t 1 1 I n f o . df <− sqlQuery ( channel , ”SELECT plot , po s i t i on , funGroup , spAbbr ,amf , treatment , b iochar rate , compost rate , season , hgt FROM DATA Q WHEREseason = ’2011 ’ ” )c l o s e ( channel )# Training Data ( Var iab l e Creat ion )channel <−odbcConnectAccess ( ”C: / Users /Ohsowski/Documents/PhD/ D i s s e r t a t i o n /Data/d i s s e r t a t i o n data exp 1 13 aug2” )mainTrain11 . df <− sqlQuery ( channel , ”SELECT plot , po s i t i on , data , dataType ,spAbbr , p l s r , t e s t FROM DATA Q WHERE season = ’2011 ’ AND t e s t = 0 AND p l s r= 1” )stemLenTrain11 . df <− sqlQuery ( channel , ”SELECT plot , po s i t i on , data , spAbbr ,p l s r FROM DATA Q WHERE season = ’2011 ’ AND dataType = ’ stemLen ’ AND p l s r =1 AND t e s t = 0” )## Ca l cu l a t e s the Mean o f stem l en g t h f o r the t r a i n i n g p l an tstemLenMeanTrain11 . df <− aggregate ( stemLenTrain11 . df $data , by=l i s t (stemLenTrain11 . df $ plot , stemLenTrain11 . df $ po s i t i on , stemLenTrain11 . df $spAbbr , stemLenTrain11 . df $ p l s r ) , FUN = mean , na . rm = TRUE)names ( stemLenMeanTrain11 . df ) <− c ( ” p l o t ” , ” p o s i t i o n ” , ”spAbbr” , ” p l s r ” , ”stemLenMean” )## Sum of stem l en g t h s f o r p r e d i c t i on p l an t sstemLenSumTrain11 . df <− aggregate ( stemLenTrain11 . df $data , by=l i s t (stemLenTrain11 . df $ plot , stemLenTrain11 . df $ po s i t i on , stemLenTrain11 . df $spAbbr , stemLenTrain11 . df $ p l s r ) , FUN = sum , na . rm = TRUE)names ( stemLenSumTrain11 . df ) <− c ( ” p l o t ” , ” p o s i t i o n ” , ”spAbbr” , ” p l s r ” , ”stemLenSum” )## Sum of stem l en g t h s f o r p r e d i c t i on p l an t sstemLenCountTrain11 . df <− aggregate ( stemLenTrain11 . df $data , by=l i s t (stemLenTrain11 . df $ plot , stemLenTrain11 . df $ po s i t i on , stemLenTrain11 . df $spAbbr , stemLenTrain11 . df $ p l s r ) , FUN = func t i on ( x ) c ( count = length ( x ) ) )156Appendix C. R Code for Plant Plug Trial Biomass Predictions and Plot Mass Calculationsnames ( stemLenCountTrain11 . df ) <− c ( ” p l o t ” , ” p o s i t i o n ” , ”spAbbr” , ” p l s r ” , ”stemLenCount” )## Mean o f p l an t biomass f o r p r e d i c t i on p l an t smassTrain11 . df <− sqlQuery ( channel , ”SELECT plot , po s i t i on , data , spAbbr , p l s rFROM DATA Q WHERE season = ’2011 ’ AND dataType = ’ mass ’ AND p l s r = 1 ANDt e s t = 0” )massMeanTrain11 . df <−aggregate ( massTrain11 . df $data , by=l i s t ( massTrain11 . df $plot , massTrain11 . df $ po s i t i on , massTrain11 . df $spAbbr , massTrain11 . df $ p l s r ), FUN = mean , na . rm = TRUE)names ( massMeanTrain11 . df ) <− c ( ” p l o t ” , ” p o s i t i o n ” , ”spAbbr” , ” p l s r ” , ”mass” )c l o s e ( channel )# Data ( Var iab l e Creat ion ) −channel <−odbcConnectAccess ( ”C: / Users /Ohsowski/Documents/PhD/ D i s s e r t a t i o n /Data/d i s s e r t a t i o n data exp 1 13 aug2” )data11 . df <− sqlQuery ( channel , ”SELECT plot , po s i t i on , data , dataType , spAbbr ,core FROM DATA Q WHERE season = ’2011 ’ AND core = 1 AND harvestOne = 0” )stemLenData11 . df <− sqlQuery ( channel , ”SELECT plot , po s i t i on , data , spAbbr ,core FROM DATA Q WHERE season = ’2011 ’ AND dataType = ’ stemLen ’ AND core =1 AND harvestOne = 0” )c l o s e ( channel )stemLenMeanData11 . df <−aggregate ( stemLenData11 . df $data , by=l i s t ( stemLenData11 .df $ plot , stemLenData11 . df $ po s i t i on , stemLenData11 . df $spAbbr , stemLenData11. df $ core ) , FUN = mean , na . rm = TRUE)names ( stemLenMeanData11 . df ) <− c ( ” p l o t ” , ” p o s i t i o n ” , ”spAbbr” , ” core ” , ”stemLenMean” )stemLenSumData11 . df <−aggregate ( stemLenData11 . df $data , by=l i s t ( stemLenData11 .df $ plot , stemLenData11 . df $ po s i t i on , stemLenData11 . df $spAbbr , stemLenData11. df $ core ) , FUN = sum)157Appendix C. R Code for Plant Plug Trial Biomass Predictions and Plot Mass Calculationsnames ( stemLenSumData11 . df ) <− c ( ” p l o t ” , ” p o s i t i o n ” , ”spAbbr” , ” core ” , ”stemLenSum” )## Sum of stem l en g t h s f o r p r e d i c t i on p l an t sstemLenCountData11 . df <− aggregate ( stemLenData11 . df $data , by=l i s t (stemLenData11 . df $ plot , stemLenData11 . df $ po s i t i on , stemLenData11 . df $spAbbr ,stemLenData11 . df $ core ) , FUN = func t i on ( x ) c ( count = length ( x ) ) )names ( stemLenCountData11 . df ) <− c ( ” p l o t ” , ” p o s i t i o n ” , ”spAbbr” , ” core ” , ”stemLenCount” )# AG Training Data ImportvisAG <− subset ( mainTrain11 . df , spAbbr == ’AG’ )agMassMean <− subset ( massMeanTrain11 . df , spAbbr == ’AG’ )agWPHgt <− subset ( mainTrain11 . df , spAbbr == ’AG’& dataType == ’sWPHgt ’ , s e l e c t= c ( plot , po s i t i on , spAbbr , p l s r , data ) )names (agWPHgt) <− c ( ” p l o t ” , ” p o s i t i o n ” , ”spAbbr” , ” p l s r ” , ”WPHgt” )agLv4Hgt <− subset ( mainTrain11 . df , spAbbr == ’AG’& dataType == ’ s4LvHgt ’ ,s e l e c t = c ( plot , po s i t i on , spAbbr , p l s r , data ) )names ( agLv4Hgt ) <− c ( ” p l o t ” , ” p o s i t i o n ” , ”spAbbr” , ” p l s r ” , ” lv4Hgt ” )agTrain11 . tmp <− merge ( merge (agWPHgt , agLv4Hgt , by = c ( ” p l o t ” , ” p o s i t i o n ” , ”spAbbr” , ” p l s r ” ) ) , agMassMean , by = c ( ” p l o t ” , ” p o s i t i o n ” , ”spAbbr” , ” p l s r ” ))# AG Training Transformation Sec t ionagTrain11 . tmp$massT <− Log . Tns ( agTrain11 . tmp$mass )agTrain11 . tmp$lv4HgtT <− Log . Tns ( agTrain11 . tmp$ lv4Hgt )agTrain11 . tmp$WPHgtT <− Sqrt . Tns ( agTrain11 . tmp$WPHgt)158Appendix C. R Code for Plant Plug Trial Biomass Predictions and Plot Mass CalculationsagTrain11 . tmp$massUT <− s c a l e ( agTrain11 . tmp$massT)agTrain11 . tmp$lv4HgtUT <− s c a l e ( agTrain11 . tmp$lv4HgtT )agTrain11 . tmp$WPHgtUT <− s c a