Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Site-forest productivity relationships and their management implications in coastal lowland ecosystems… Green, Robert Norman 1989

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

Item Metadata


831-UBC_1989_A6_7 G74.pdf [ 11.88MB ]
JSON: 831-1.0097477.json
JSON-LD: 831-1.0097477-ld.json
RDF/XML (Pretty): 831-1.0097477-rdf.xml
RDF/JSON: 831-1.0097477-rdf.json
Turtle: 831-1.0097477-turtle.txt
N-Triples: 831-1.0097477-rdf-ntriples.txt
Original Record: 831-1.0097477-source.json
Full Text

Full Text

SITE - F O R E S T P R O D U C T I V I T Y RELATIONSHIPS A N D T H E I R M A N A G E M E N T IMPLICATIONS IN C O A S T A L L O W L A N D E C O S Y S T E M S O F E A S T G R A H A M ISLAND, Q U E E N C H A R L O T T E ISLANDS by R O B E R T N O R M A N G R E E N B.S.F., University of B.C., 1976 A THESIS S U B M I T T E D IN P A R T I A L F U L F I L L M E N T O F T H E R E Q U I R E M E N T S F O R T H E D E G R E E O F M A S T E R O F S C I E N C E (6 units) in T H E F A C U L T Y O F G R A D U A T E STUDIES ( D E P A R T M E N T O F F O R E S T R Y ) We accept this thesis as conforming to the required standard T H E U N I V E R S I T Y O F BRITISH C O L U M B I A A P R I L 1989 Robert Norman Green, 1989 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the Head of my Department or by his representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department of Forestry The University of British Columbia 2075 Wesbrook Place Vancouver, Canada V 6 T 1W5 ii A B S T R A C T Relationships between soil, physiographic, floristic, and stand properties were examined in second-growth stands on a range of imperfectly to poorly drained ecosystems on east Graham Island, Queen Charlotte Islands. The major objective was to describe ecological factors associated with variation in tree growth, as expressed by site index of western redcedar. Cedar site index was found to be strongly correlated with soil nutrient content, particularly total N and exchangeable Mg, expressed on a kg/ha basis. Decreasing site index was associated with decreasing rooting depth, due to slowly permeable horizons, and poor soil aeration, reflected by high volumetric moisture content. A simple model using total N content and volumetric moisture content summarized the relationship between cedar site index and soil properties, and accounted for 78% of the site index variation. Three site index classes used as sampling strata (redcedar site index <. 15, 16-20, and >20 m/50 yrs. b.h. age) could be successfully differentiated by soil properties using discriminant analysis. Natural structure in the soils data revealed through principal components analysis and cluster analysis also reflected, with minor overlap, these three site index classes. Understory vegetation could be used to differentiate the poorest site index class, however the remaining two classes could not be floristically differentiated. A simple model relating cedar site index to reciprocal averaging scores derived from vegation data, and the frequency of the nutrient-medium indicator species group only explained 44% and iii 31% respectively of site index variation. Site index of Sitka spruce, western hemlock, and lodgepole pine were all highly correlated with redcedar site index. Hemlock and pine showed similar height growth patterns to cedar, while spruce was markedly different, showing slower height growth than cedar on poorer sites and greater height growth on better quality sites. The sensitivity of spruce to limiting site conditions was also reflected in its increasing stand volume composition with improving site quality. The sampled sites generally supported high stand volumes given the site limitations, with M A I averaging 5 m3ha"1yr"1 o n the poorest sites (site index class 1) to 13 n^ ha'V"1 on the best sites (site index class 3). Management strategies were recommended based on relationships observed in the study. Preferred tree species to manage are: for site index class 1 - cedar, pine, and hemlock; for site index class 2 - cedar, hemlock, pine, (with minor spruce); for site index class 3 - hemlock, spruce, cedar. Site productivity of these ecosystem may be improved by increasing the volume of aerated soil exploitable by tree roots. This could be achieved by site preparation which created relatively low, but wide mounds. iv C O N T E N T S A B S T R A C T ii T A B L E O F C O N T E N T S . iv LIST O F T A B L E S vii LIST O F F I G U R E S ix LIST O F A P P E N D I C E S xi LIST O F S Y M B O L S xii A C K N O W L E D G E M E N T S xiv 1.0 I N T R O D U C T I O N 1 2.0 S I T E / F O R E S T P R O D U C T I V I T Y RELATIONSHIPS - A R E V I E W . . . 4 2.1 Approaches 4 2.2 Limitations 7 2.3 Results 10 3.0 S T U D Y A R E A 13 3.1 Climate 13 3.2 Physiography and Bedrock Geology 16 3.3 Surficial Materials and Soils 16 3.4 Disturbance History and Stands 19 4.0 M E T H O D S 21 4.1 General Approach 21 4.2 Sampling Design 23 4.3 Sampling Methods 24 4.3.1 Vegetation and physiography 25 4.3.2 Soil 25 4.3.3 Stand 28 4.3.4 Soil water quality 29 4.3.5 Soil oxidation zone index 31 4.4 Laboratory Analysis and Data Preparation 32 4.4.1 Soil chemical analysis 32 4.4.2 Determination of soil physical properties 33 4.4.3 Conversion of soil chemical concentration to mass/area basis 36 4.4.4 Stand properties 38 4.6 Data Analysis . . . 41 5.0 R E S U L T S A N D DISCUSSION 44 5.1 Description of the Data 44 5.1.1 Selection of principle variable sets 46 5.1.2 Within-plot variability 51 5.2 Stand Properties 59 5.2.1 Stand age 59 5.2.2 Site index 60 5.2.3 Stand volume 65 5.2.4 Tree species composition 68 5.3 Ecological Properties of Site Index Classes 69 5.3.1 Soils 69 5.3.2 Vegetation 75 5.4 Natural Structure in the Data 80 5.4.1 Soils 80 5.4.2 Vegetation 89 5.5 Functional Relationships Between Site Index and Ecological Properties 94 5.5.1 Site index/soil relationships 95 5.5.2 Site index/vegetation relationships 107 5.6 Soil Water Quality 110 5.6.1 Specific conductivity I l l 5.6.2 Dissolved oxygen 114 5.7 Soil Oxidation Zone Index 116 5.8 A Simplified Model for Field Recognition of Site Index Classes 120 5.9 General Discussion 126 5.9.1 Soil Nutrient Status 126 5.9.2 Soil Aeration 129 5.9.3 Rooting Depth 134 5.10 Management Implications 135 5.10.1 Tree species selection 136 5.10.2 Site amelioration 138 6.0 S U M M A R Y 142 7.0 R E F E R E N C E S 147 vii T A B L E S 1 Climate summary for Hell , Sandspit, and Gospel Point 15 2 Site index classes 24 3 Crown thinning and stem form classes 28 4 Height/diameter equations for species and site index classes 40 5 Summary of available variables 44 6 Pearson correlation matrix for the 14 selected variables 48 7 Coefficients of variation and required sample size to estimate soil property means 52 8 Comparison of means and C V (%) of soil chemical properties of plot 24 based on 3 composite samples and 15 individual samples 54 9 Means (and standard deviations) of site index (m/50yrs) for the three site index classes 62 10 Means and standard deviations of M A I (m3/ha/yr, total live volume) for the three site index classes 67 11 Means and standard deviations of soil variables for site index classes 70 12 Standardized discriminant function coefficients for dependent variables 72 13 Frequency of plots according to actual and predicted site index class membership, based on discriminant analysis 73 14 Species composition of the three site index classes 76 15 Eigenvalue and variance extracted by the first three principal components 82 16 Component correlations with original soil variables 82 17 Means and standard deviations of soil and productivity viii variables for soil-derived clusters 87 18 Variance extracted by R A axes 1 to 3 90 19 Species by sample matrix, ranked according to the first R A axis scores 91 20 Continuous and categorical variables which are readily described in the field 102 21 Contigency tables for selected categorical variables showing frequency by site index class 104 22 Pearson correlation matrix for oxidized zones and CWSI plus selected soil properties 117 23 Means (and standard deviations) of oxidized zones on steel rods for site index classes 119 24 Recommended tree species composition for sites on east Graham Island 138 ix F I G U R E S 1 Study area and sample plot locations 14 2 Frequency histogram of Cw site index in the inventory update data . . . . 23 3 Daily precipitation for April 1 to August 31, 1987 36 4 Correlation structure diagram for soil properties and CWSI 49 5 Boxplots of CVs for forest floor properties, calculated for each plot from composite samples 55 6 Boxplots of CVs for mineral soil properties, calculated for each plot from composite samples 56 7 Boxplots of CVs for soil morphological properties, calculated for each plot from four soil pits 57 8 Boxplots of CVs for soil physical properties, calculated for each plot from four soil pits 58 9 Sample plot distribution by stand age 60 10 Sample plot distribution by Cw site index 61 11 Relationship between Hw and Cw site index 63 12 Relationship between PI and Cw site index 63 13 Relationship between Ss and Cw site index 64 14 Relationship between stand volume and Cw site index 65 15 Relationship between M A I and Cw site index 67 16 Relative tree species composition for the three site index classes 68 17 Plot of discriminant scores for 24 plots in relation to discriminant functions 1 and 2 74 18 Soil moisture spectra for the three site index classes 78 19 Soil nutrient spectra for the three site index classes 79 20 Ordinations of sample plots in relation to soil P C A axes 1 and 2 83 X 21 Cluster analysis on T N , V M C , D2ROOTS, MINPH 85 22 Ordination of samples in relation to R A axes 1 and 2 92 23 Relationships between soil nutrient contents and Cw site index 96 24 Relationships between p H and C / N ratios of the forest floor and mineral soil, and Cw site index 97 25 Relationships between soil physical properties and Cw site index 98 26 Relationship between Cw site index and R A axis 1 108 27 Relationship between Cw site index and the nutrient medium indicator species group .°. 109 28 Summary of field-recognizable ecological properties in relation to site index classes 121 xi APPENDICES A Alphabetical list of plant species 157 B Methods for converting soil chemical concentrations to content in kg/ha 158 C Soil chemical data - kg/ha in the forest floor 159 D Soil chemical data - kg/ha in the mineral soil to the secondary rooting depth 160 E Soil chemical data - kg/ha in the combined forest floor plus mineral soil to the secondary rooting depth 161 F Soil chemical data - p H and C / N ratios for the forest floor and root zone mineral soil 162 G Soil physical properties 163 H Data for categorical variables 164 I Soil oxidation zone index and soil water quality 165 J Site index data 166 K Mensuration data - combined for all species 167 L Mensuration data - by species 168 M Explanation of boxplots 169 N Procedure for checking assumptions of parametric multiple comparison test 170 xii LIST O F S Y M B O L S 1. Soil chemical properties: C A exchangeable calcium: kg/ha except where indicated H U M C N carbon/nitrogen ratio in the forest floor H U M P H p H (CaCl 2) in the forest floor K exchangeable potassium: kg/ha except where indicated M G exchangeable magnesium: kg/ha except where indicated M I N C N carbon/nitrogen ratio in the root zone mineral soil M I N P H p H in the root zone mineral soil MINN mineralizable nitrogen: kg/ha except where indicated P available phosphorus: kg/ha except where indicated T C total carbon: kg/ha except where indicated T N total nitrogen: kg/ha except where indicated 2. Soil physical and morphological properties: AIR air filled porosity in the root zone mineral soil (%/100) C A P thickness of the eolian cap (cm) C F P M coarse fragment content in parent material (% by volume) C L A Y percent clay in the mineral soil D 1 R O O T M primary rooting depth from the mineral surface (cm) D 1 R O O T S primary rooting depth from the ground surface (cm) D 2 R O O T M secondary rooting depth from the mineral surface (cm) D 2 R O O T S secondary rooting depth from the ground surface (cm) D H 2 0 M depth to free water from the mineral surface (cm) D H 2 0 S depth to free water from the ground surface (cm) F F D E P forest floor thickness (cm) M O T L depth to distinct or prominent mottles from mineral surface (cm) P A N depth to a "pan" from mineral surface (cm) P O R total porosity in the root zone mineral soil (%/100) S A N D percent sand in the mineral soil SILT percent silt in the mineral soil S L O P E slope (%) V M C volumetric moisture content in the root zone mineral soil (%/100) C E M presence of cemented horizon (2 classes) CONS consistence of mineral soil (2 classes) C H R O M A chroma of mineral soil (Munsell notation) H 2 0 S T free water status (4 classes) H U E hue of mineral soil (Munsell notation) P M parent material type (5 classes) R O O T rooting density in mineral soil (4 classes) SHAPE " ground surface shape (5 classes) T E X T P M field texture of parent material (CSSC textural classes) 3. Soil water properties: D O dissolved oxygen (ppm) SC specific conductance (dS/m) 4. Soil oxidation zone index: A Z 1 depth of 1st zone, method A (cm) A Z 2 depth of 2nd zone, method A (cm) A Z 3 depth of 3rd zone, method A (cm) BZ1 depth of 1st zone, method B (cm) BZ2 depth of 2nd zone, method B (cm) 5. Stand properties: Cw western redcedar (Thuja plicata) Hw western hemlock (Tsuga heterophylla) PI lodgepole pine (Pinus contorta) Ss Sitka spruce (Picea sitchensis) CWSI site index of Cw (m/50 yrs b.h. age) HWSI site index of Hw (m/50 yrs b.h. age) PLSI site index of PI (m/50 yrs b.h. age) SSSI site index of Ss (m/50 yrs b.h. age) M A I mean annual increment, total volume (m3ha"1yr~1) 6. Data analysis: D A discriminant analysis R A reciprocal averaging xiv A C K N O W L E D G E M E N T S I wish to express my sincere gratitude to the thesis committee members, Dr. T . M . Ballard, Dr. H . Schreier, Dr. P.L Marshall, and Dr. K. Klinka, for their advice and encouragement throughout this project. Special thanks are extended to P. Courtin, G . Shishkov, and G . Tamassay for their cheerful endurance of long field days, and to G . Shishkov particularly, for soil sample preparation and bulk density measurements. I am also grateful for the discussions with A . Banner and R. Pollack in the initial stages of the study. Appreciation is extended to the B.C. Ministry of Forests for financial support, particularly P. Ackhurst, for providing the opportunity to pursue this graduate program. Finally, I wish to sincerely thank my wife, Leigh Hussey, for her constant encouragement, patience, and assistance throughout this project; and my new son, Spencer William Hussey Green, for delaying his arrival until the second draft was completed. 1 1.0 I N T R O D U C T I O N A n extensive area of Crown land along the east side of Graham Island, Queen Charlotte Islands, has recently been the focus of considerable interest for potential forest management. The main area of interest comprises some 35000 hectares representing the southern portion of the Graham Island Crown Land Plan area (Ministry of Lands, Parks, and Housing 1980). A n additional 15000 hectares in the northern portion of the Plan area (north of Port Clements, east of Masset Inlet, west of Naikoon Park) are also recognized as having forestry potential. In the past, the forestry value of this area was considered low. Holland (1964) characterized the area as being comprised largely of muskeg with sparse timber cover, with Alley and Thompson (1978) characterizing it similarly. Together with an old and inaccurate forest inventory, this information has led to uncertainties about the extent and quality of the forest resource. However, an inventory update conducted in 1983-84 revealed considerable area of good quality stands occupying sites apparently capable of sustaining long term forest production (Pollack 1985). This recent data, along with increasing pressure to expand local forest harvesting activities, has prompted the Ministry of Forests to include the area in a proposed provincial forest. However, prior to initiating forestry development, some basic information is required to assist in management planning. Earlier studies (Day 1957) and recent observations made during the inventory update identified a considerable range of productivity in the predominantly second-growth, fire-regenerated stands. This variation has been attributed largely to rather subtle differences 2 in site properties; however, specific relationships have only been speculated on and not studied. The existing, floristic-based ecological classification (Banner et al. 1987) has been of limited use in site identification because of difficulties associated with poor understory development. A majority of the area has been excluded from A A C calculations and timber supply forecasts because of uncertainties regarding the potential productivity of these sites (B.C. Forest Service 1982). In addition to extensive second-growth stands, a relatively large area of poor quality old-growth western redcedar (Thujaplicata - Cw) and western hemlock (Tsuga heterophylla -Hw) stands occur in the study area. Successional relationships between these ecosystems and the more productive second-growth ecosystems are poorly understood. There is uncertainty whether these represent distinctly different sites, or different successional stages on essentially the same site. Understanding site properties associated with tree growth expressed in second-growth stands will be useful in estimating the potential of sites currently supporting decadent old-growth stands. Appropriate management regimes for the study area are also unclear, particularly in regard to suitable tree species and site preparation. Information is required on the role of Sitka spruce (Picea sitchensis - Ss) in stand composition over the range of sites in the area because of its high value (Siurtain 1981). There is also interest in the suitability of lodgepole pine (Pinus contorta - PI) as a crop species on these sites (R. Pollack, pers. comm. 1987). The occurrence of past fires have had a major impact on stand development and possibly productivity through most of the study area (Day 1957). Thus there is considerable interest in the potential role of prescribed burning as a site preparation tool, particularly 3 following harvesting of old-growth stands. Finally, the operable forest land base in the Queen Charlotte Islands has been reduced from the recent creation of the South Moresby National Park. There is interest in silvicultural treatments aimed at improving the natural productivity potential of remaining ecosystems available for forest management. Increased understanding of the relationships between ecological properties and potential forest productivity will contribute substantially to the recognition of sites suitable for forest management, and to the development of appropriate silvicultural treatments for these sites. This information would also be applicable to similar coastal lowland ecosystems occurring within the Hecate Depression (Holland 1964). The basic objectives of the study are: i) to examine relationships between ecological factors and forest productivity, expressed by site index of western redcedar, in even-aged second-growth stands on a range of coastal lowland ecosystems on east Graham Island. ii) to examine stand properties, including relative performance of commercial tree species, associated with the above range of ecosystems. iii) to interpret the above information in relation to suitable silvicultural strategies for these sites. 4 2.0 S I T E / P R O D U C T I V I T Y 1 RELATIONSHIPS - A R E V I E W 2.1 Approaches Relationships between tree growth and site quality have been the subject of intense research worldwide. In the United States alone more than 160 papers were published between 1929 and 1974 (Carmean 1975). While the number of studies has decreased since the peak in the late 1960's, enthusiasm for research on site/productivity relationships remains intense. In general, the two major objectives of these studies have been: i) to predict productivity, typically inferred through site index, and ii) to determine which ecological factors influence tree growth. The first objective primarily seeks an alternative means of estimating site index for stands which are unsuitable for direct use of site index curves. Reliable estimates of site index from age and height measurements can only be obtained from stands that i) contain the species of interest; ii) are even-aged and have a closed canopy; iii) are free of damage or disease; and iv) are between a certain minimum and maximum age (Hagglund 1981). Estimation of site index from age and height relationships is often limited in its applicability because of these restrictions. For example, site index curves can only be used reliably on 30 to 40% of the total Swedish forest land base (Haggland 1977); a similar proportion was estimated for major British Columbia coastal Timber Supply Areas (Green 1987). 1 refers to forest productivity 5 The second objective aims at increasing knowledge of the basic site factors that influence forest productivity. While such information is useful for the prediction of productivity, perhaps its greatest value is in developing management practices with a better understanding of their impacts on tree growth (Campbell 1978, Adams and Jack 1970). Quantification of relationships between productivity and site properties, in combination with knowledge of the effects of various management practices on these properties, will assist foresters in designing treatment regimes which successfully improve site quality (White 1982, Schmidt and Carmean 1988). A number of approaches have been investigated to try to meet these objectives. The use of understory vegetation is an intuitively logical approach to estimating site index since plants act as "phytometers" of site quality by reflecting the integrated effect of complex environmental properties (Carmean 1975). The most common method uses classification of stable (climax or late-seral) plant communities to derive "site types" which have characteristic site qualities. For example, site index has been related to habitat types in the western U.S. (Monserud 1984, Mathiasen et al. 1986), and to site units of the biogeoclimatic ecosystem classification system in B.C. (Green et al. 1989). Rather than focussing on plant communities, another approach uses individual understory species as indicators of site quality. This method recognizes that some species demonstrate a relatively narrow amplitude in relation to certain site factors such as soil moisture regime and soil nutrient regime. The relative frequency or occurrence of important indicator species in an ecosystem can therefore be used to estimate site quality and in turn, site index (Hodgkins 1970, MacLean and Bolsinger 1973, Green et al. 1989). 6 The more common approach to studying site/productivity relationships has focussed on physiographic and soil factors. Typically these studies examine a wide variety of topographic and soil physical, morphological, and chemical variables which are related to site index through correlation, multiple regression, or multivariate data analyses (Carmean 1975). In many cases, the studied factors are indirect indices of the more basic growth-controlling factors of climate, soil moisture, soil nutrients, and soil aeration. Quantification of these basic site factors should provide the best basis for studying site/productivity relationships. However this has proved difficult because of limitations in measurement techniques and in knowledge of ecologically meaningful ways of expressing these factors. Many studies which aim at predicting site index have focussed only on variables which are easily recognized in the field, in order to meet practical needs. Generally these are limited in their precision and application, and may lead to spurious results when any variable is included regardless of its ecological relevance to tree growth (Hagglund 1977, Ford 1983). Other studies have recognized the importance of soil chemical properties for tree growth and have combined these with other soil morphological, physical, and physiographic properties for a more comprehensive analysis (Pritchett and Fisher 1987). While the results are more limited in field application because of measurement difficulties, they provide more insight into site/productivity relationships. In many cases the strength of the relationships is improved by including chemical properties (Johnson et al. 1987). More recently, attempts have been made to relate site index to quantitative measures of soil moisture and soil nutrients to derive a more direct model of tree growth. Klinka and Carter (1989) found promising results in relating Douglas-fir (Pseudotsuga menziesii) site 7 index to actual evapotranspiration, water deficit, and mineralizable nitrogen in the root zone, although the results were inferior to more conventional approaches using categorical variables defined by vegetation-derived site units and qualitative assessments of soil moisture and nutrients. Further improvement is required in the reliable quantification of site factors. In general, the selection of variables to include in site/productivity studies involves a compromise between intuition, the desire to be comprehensive, and the ease with which measurements can be made (Ford 1983). For efficiency in deriving meaningful results, the variables should be based on existing ecological knowledge, and should reflect the basic site factors important to tree growth (Hagglund 1977). 2.2 Limitations In spite of the considerable effort in studying site/productivity relationships, some difficulties have been encountered in obtaining reliable results, leading Ford (1983) to suggest soil/productivity studies are "blunt instruments" for detecting how environmental factors control tree growth. There are several common problems which can limit the success and application of these studies. Relationships between productivity and soil factors have been found to vary with different geographic areas, often reflecting different effects of climate, physiography, or major parent materials and bedrock (Schmidt and Carmean 1988). For example Cook et al. (1977) found that relationships between Scots pine (Pinus sylvestris) 8 growth and site factors differed considerably in two different areas of northeast Scotland. Area-specific relationships were also found between Sitka spruce yield and site properties in northern Scotland (Blyth and McLeod 1978). To accommodate such variation, large heterogeneous study areas should be subdivided initially into more uniform strata within which relationships are examined separately (Carmean 1975). Hagglund (1977) divided Swedish data into more homogeneous "processing groups" based on major tree species, soil, and forest types, for deriving site/productivity relationships for Scots pine and Norway spruce (Picea abies). Relationships between southern U.S. hardwoods and site properties were also improved considerably when the sample population was subdivided into more uniform strata (Broadfoot 1969). Because of the area-specific nature of many site/productivity relationships, studies examining such relationships should have a precise statement defining the population studied, and results should only be extrapolated beyond this population with caution (Carmean 1975). Site/productivity studies may also be limited because variables selected for analysis do not reflect adequately the properties really important for tree growth (Broadfoot 1969). This is particularly true of studies focussing on field-recognizable variables which only indirectly imply basic site factors, often with limited precision (Carmean 1975). In more comprehensive approaches, selected variables may be in an unsuitable form to uncover meaningful relationships, or more variables than those measured may influence growth (Hagglund 1977). Site quality and hence, site productivity is a function of the integrated effect of many complex environmental factors. As a result, many of theses factors are interrelated in their effects on growth; separating their effects in quantitative studies has 9 been difficult (Mader 1977). Data analysis should pay close attention to intercorrelation of variables to try to minimize this effect on resulting models (Carmean 1975). Unsatisfactory results of site/productivity studies may be a result of sampling limitations. Inherent high spatial variability of soil properties (particularly chemical properties), weaknesses in consistently assessing qualitative variables, and possible errors in measurement can all contribute to unexplained residual error in the analysis (White 1982). Sampling design has also limited the success of some studies where the middle portion of the site productivity range is oversampled and the extremes undersampled. It is important to balance the sampling distribution across the full range encountered in the study area (Carmean 1975). Finally, error in the estimation of the common dependent variable, site index, can limit the success of these studies. Haggland (1977) estimated that five to six percent error in site index estimates from available site index curves was added to their site/productivity models. The suitability of site index as a productivity index has been questioned; however, most of the problems involved inaccuracies associated with older anamorphic site index curves. Improved precision is now being realized with newer polymorphic site index curves based on stem analysis and nonlinear regression models (Schmidt and Carmean 1988). Studies relating productivity to site quality inferred from understory vegetation may be limited by floristic variation associated with non-site factors. Successional stage, understory light conditions, disturbance history, and chance can influence species composition and abundance on any given site (Spurr and Barnes 1980, Spies and Barnes 1985). This is 10 particularly true of approaches utilizing indicator species. More successful results are generally obtained when productivity is related to site units derived from stable (late serai, near-climax, or climax) plant communities representing "biologically-equivalent" sites (Klinka and Carter 1989). 2.3 Results The majority of site/productivity studies confirm Coile's (1954) observations that site productivity "is largely determined by soil properties or other site features that influence the quality and quantity of growing space for tree roots". Added to this is climate, which can have an overriding impact on site productivity (Schmidt and Carmean 1988). Most of the successful site/productivity studies reviewed by Carmean (1975) explained 65 to 85% of the variation in site index using physiographic and soil properties. Surface soil depth, subsoil texture, aspect, and slope were the most common properties related to site index, although few of the studies examined soil nutrient properties. Schmidt and Carmean (1988) explained 65 to 83% of the variation in jack pine (Pinus banksiana) site index in Ontario using physical properties such as depth to a root-restricting layer, thickness of the A horizon, coarse fragment content, slope, as well as soil pH. The relationships differed in the four parent material-defined strata. Similar properties explained 66% of the variation in red maple (Acer rubrum) site index in Wisconsin and Michigan (Johnson et al. 1987). The inclusion of soil nutrient properties improved the fit of these models; however, they were not utilized because of the focus on field-application of results. 11 Certain soils of the southeast coastal plain in the U.S. share some similar properties with those of the east Graham Island study area. These soils are characterized by shallow fluctuating water tables because of low relief, high precipitation inputs relative to evapotranspiration, and slowly permeable soils (Buol 1978). Soil chemical properties, and thickness of the aerated root zone, are important to the productivity of the predominantly pine forests. Features such as topographic position, physical properties of the least permeable soil layer, and depth to mottling or a watertable are therefore commonly related to site index (Klawitter 1978). For example, Klawitter (1978) found that loblolly pine (Pinus elliottii) site index increased steadily as the surface soil depth increased from 15 cm to 60 cm. As long as the watertable was greater than 50-60 cm from the surface during the growing season, soil aeration was not seriously reduced and sites remained productive. Sitka spruce tolerates imperfectly to poorly drained sites in Scotland, Ireland, and Britain, however growth varies widely. Blyth and MacLeod (1981b) found that a small number of variables, particularly total P in the humus form and mottling depth, accounted for 60 to 86% of local yield class variation across several different study areas. In a related study they also found total N in the humus form to relate strongly with growth (Blyth and MacLeod 1981a). Growth generally decreased as drainage became poorer from brown earth through to podzol, gleyed podzol, and finally gley soils. Worrel (1987) found that restricted rooting depth due to indurated, excessively clayey, very wet, or very shallow soils had a negative impact on potential yield class of Sitka spruce in northern Britain. Better growth of Sitka spruce was also related to improved soil aeration in peaty gley soils in Britain, with a relatively small increase in depth to an anaerobic layer associated with a large increase 12 in growth (Armstrong et al. 1976). In ecosystems which feature limited rooting volume due to unfavourable subsoils, rooting volume generally shows a curvilinear relationship to growth. Available rooting volume becomes increasingly important as rooting depth becomes more limited (Carmean 1975). This is particularly true in soils featuring high watertables where the quality of an already limited root zone can be negatively affected by reduced aeration (Pritchett and Fisher 1987). According to Day's (1957) observations of soils on east Graham Island, the predominantly fine-textured tills with thick humus forms or shallow peat have marked drainage impedance which in combination with physical soil restrictions, limits the rooting space and decreases soil aeration which negatively affects tree growth. 13 3.0 S T U D Y A R E A The study area occurs on the east side of Graham Island, the northern and largest island of the Queen Charlotte archipelago. It encompasses the area of Crown land north of Miller Creek, east of T F L 39, and south of Naikoon Park, as well as the area between Masset Inlet and Naikoon Park to approximately 8 km north of Port Clements (Figure 1). 3.1 Climate The study area falls entirely within the Submontane Wet Hypermaritime Coastal Western Hemlock (CWHwhl) biogeoclimatic variant which occupies lower elevations on the drier, leeward side of the Queen Charlotte Mountains (Banner et al. 1987). The climate of the C W H w h l can be described as a hypermaritime wet, cool, mesothermal type (Klinka et al. 1989). Summers are cool and moderately wet with occasional dry warm spells. Late fall and early winters are very wet, while winters are cool, mild, and rainy. Strong winds occur with high frequency (Banner et al. 1987). The study area occurs within the drier range of the CWHwhl variant because of the more pronounced rainshadow effect in the easterly localities. A climatic summary from long term climate stations at Hel l and Sandspit is shown in Table 1. These stations may be considered relatively representative of much of the study area, although there is a tendency towards slightly cloudier, moister conditions Figure 1. Study area (within "Graham Forest") and sample plot locations. 15 in the northwestern portion. The low-lying coastal plain has a relatively low precipitation gradient, with annual precipitation means varying from 1200 mm along the coast to 1500 mm near the boundary with the Skidegate Plateau (Karanka 1986). The rainshadow effect occurring across Graham Island can be appreciated by comparing the eastern stations to one at a similar elevation on the windward side of the Queen Charlotte Mountains (Gospel Point, north of Skidegate Channel - Table 1). Less than half the amount of precipitation falls in the eastern localities, while temperatures remain similar. However because of the relatively low potential evapotranspiration experienced in the study area relative to warmer climates in coastal B.C., no water deficit occurs during the growing season. Table 1. Climate summary for Tlell, Sandspit, and Gospel Point (abridged from Banner et al 1987) Variable T l e l l Sandspit Gospel Po No. years record 23-24 32-34 7 Elevation (m) 5 5 34 P r e c i p i t a t i o n : Mean annual ppt. (mm) 1152 1281 2340 Mean mo. May-Sept ppt. 64 57 98 Mean mo. Oct-April ppt. 119 142 264 Driest mo. t o t a l ppt. 51 43 60 Wettest mo. t o t a l ppt. 172 194 355 Temperature: Mean annual T (°C) 7.4 7.9 7.0 Mean mo. May-Sept T 12.0 12.4 11.1 Mean T warmest mo. 14.2 14.7 13.4 Mean T coldest mo. 1.3 2.0 1.2 Extreme max. T 32 . 2 27.8 — Extreme min. T -16.7 -13.9 — Grow. deg. days (>5°C) 1283 1385 — 16 3.2 Physiography and Bedrock Geology The study area is situated within the Queen Charlotte Lowland physiographic region which encompasses the majority of the northeastern half of Graham Island (Holland 1964). It is characterized by flat to very gently undulating terrain with elevations mainly below 160 m. The southern portion of the study area near Lawn Hil l and Miller Creek begins to rise gradually towards the Skidegate Plateau further to the west. The northern extremity of the study area occupies the Argonaut Plain, a very subdued, extensive deposit of post-Pleistocene outwash sands and gravels which dominates the northeastern tip of Graham Island (Holland 1964). Most of the study area is underlain by late Tertiary sedimentary bedrock of the Skonun Formation which includes sandstone, siltstone, and shale as well as some conglomerates. A relatively small area of Masset Formation basalt outcrops in the Lawn Hill area, while a small post-tectonic pluton of granodiorite occurs in the extreme southern portion of the study area near Miller Creek (Sutherland Brown 1968). 3.3 Surficial Materials and Soils The predominant surficial material throughout most of the study area is till of Pleistocene origin (Alley and Thomson 1978). Tills are fine textured (typically sandy clay loams) with less than 25% coarse fragment content, and have high bulk densities. Glaciofluvial sands and gravels associated with the Argonaut Plain characterize the northern extremity of the study area. Marine deposits ranging from fine silts and clays to gravelly 17 sands occur adjacent to the coastline. Because of isostatic rebound following the last glaciation, these may occur up to 15 m above the present sea level (Alley and Thomson 1978, J. Clague 1988, pers. comm.). Sand-textured dunes also occur near the coast, particularly near Tlell. Alluvial silts occur along the main rivers and streams, while localized fluvial deposits may occur within the till plain. Organic deposits overlying predominantly morainal and glaciofluvial materials form extensive and localized peat bogs and fens. Of the sites sampled, the majority (92%) were on morainal material, while the remainder occurred on glaciofluvial and alluvial deposits. A n interesting feature observed while sampling was a uniform surface capping of fine-textured sediments which occurred throughout the study area. This deposit was typically silt loam textured, free of any coarse fragments, of low bulk density, and distinctly different from the underlying parent material. Its thickness averaged 16 cm and varied little, ranging from 11-20 cm in 70% of the 84 soil pits in which it was found. Thicker deposits were generally associated with secondary accumulation from slopewash on gentle concave slopes or depressions. The principal source of this deposit is not certain;, however, based on its characteristics it is most likely eolian (J. Clague, H . Luttmerding 1988, pers. comm.). The probable origin is the extensive glaciofluvial plain to the north which, before vegetation establishment, would have generated considerable windblown sediments during periods of strong dry northwesterly winds (T. Lewis 1988, pers. comm.). The sediments are not as well sorted as typical eolian sediments (J. Clague 1988, pers. comm.); however, this is consistent with the relatively short distance they would have travelled. A particulary interesting feature of these sediments is the occurrence of partly decomposed plant fibers 18 and occasional small logs which were distributed within the capping or buried at its base. This was most prevalent on the wettest sites occurring on subtly depressional terrain. It suggests vegetation was present during initial deposition of the sediments, or secondary slopewash took place after vegetation became established. This eolian capping is very important to forest productivity in the study area as it represents the principal mineral soil rooting substrate. Soils in the sample population are strongly influenced by poor aeration. More than 80% of soil profiles examined had distinct or prominent mottles.or strong gley colours within 50 cm of the mineral surface. This results from slowly permeable soils, high precipitation relative to evapotranspiration, and subdued topography which, in combination, limit free drainage. Soils are predominantly Humic Podzols with true Gleysols occurring less frequently. Dark coloured Bh horizons averaging 10-12 cm thick typically occur between the surface eolian capping and a slowly permeable layer which is either cemented or inherently dense. Pyrophosphate-extractable iron and aluminum averaged 0.1% and 1.4% respectively, with 7.6% total carbon in several horizons tested. This type of soil is typically found on wet sites and gleying is therefore common within the Great Group, but is not explicitly stated in Subgroup names (CSSC 1978). The Bh horizons found in the study area were generally coarser textured than both adjacent horizons and were the location of most soil water concentration. Orstein, or near-orstein (moderately strong cementation) horizons were frequently encountered, with placic horizons occasionally forming in the upper few centimeters. Better aerated Ferro-Humic Podzols were encountered rarely and only in association with better quality sites. Gleyed subhorizons were also present in these soils. 19 Humus forms were predominantly Residuohumimors (Klinka et al. 1981). 3.4 Disturbance History and Stands The most significant factor affecting stand establishment has been fire. The majority of stands in the study area regenerated naturally following an extensive fire approximately 110 to 130 years ago. A more recent fire 70 to 80 years ago swept through the extreme southern portion near Miller Creek, while the northern extremes of the study area (7-8 km north of Port Clements) burned approximately 160-170 and as early as 290 years ago. Evidence of the oldest fire was found in the form of common charcoal fragments mixed through the eolian capping under a decadent stand of cedar and hemlock with a salal understory. Fire is apparently a relatively frequently recurring feature of the ecology of east Graham Island. This is somewhat unusual for the Queen Charlotte Islands and other central and northern coastal ecosystems where wind, and to a lesser extent, mass wasting are the major disturbance factors (Banner et al 1987). Fire-regenerated stands are predominantly even-aged, densely stocked, and comprised of mixtures of western redcedar, western hemlock, lodgepole pine, and Sitka spruce. The relative proportions of these species vary, depending largely on the site characteristics. Charred cedar snags commonly occur throughout the fire-regenerated stands, indicating that cedar formed a major component of the original old-growth forest. These snags often have strong taper and candelabra branching characteristic of the "spike-top" cedar present in existing old-growth forests. 20 A considerable area [approximately 5000 ha (Pollack 1985)] of these old-growth forests occurs within and adjacent to the fire-regenerated stands, apparently having been missed by the fires. These stands are dominated by cedar and hemlock of poor form and very slow growth. Density is low and most of the cedar have dead tops. A dense, tall cover of salal (Gaultheria shallori) typically, but not always, dominates the understory. 21 4.0 M E T H O D S 4.1 General Approach Based on field observations and previous studies, particularly Day's (1957) insightful observations on possible productivity relationships occurring on Graham Island, a working hypothesis regarding ecological properties important for tree growth can be proposed. Forest productivity can be considered a function of climate, soil moisture, soil nutrients, and soil oxygen (Ralston 1964, Bakuzis 1969). In the east Graham Island study area, climate is relatively uniform and may be excluded as a factor contributing to productivity variation. Shortage of soil water is not limiting, with soil moisture ranging from fresh to wet regimes (Klinka et al. 1984). Soil nutrients are most likely important, given the variation in potential nutrient reserve defined by available soil rooting zone and soil chemical properties, as well as possible soil water nutrient inputs. Soil oxygen ("soil aeration") is very important to tree growth in the study area, interacting with the above factors in defining the quantity of soil exploitable by tree roots, and the quality of this soil in relation to root development and function. Al l of these factors are influenced largely by the subdued physiography and soil physical properties such as fine-textures and compacted or cemented layers which limit soil drainage and rooting depth. The approach of the study is to examine forest productivity in relation to a number of ecological variables which express as closely as possible these key site factors. In many respects this approach is similar to those discussed in Section 2.1. Forest productivity is 22 inferred through site index. Although site index has some limitations (Monserud 1984), it is still the best available index of growth which is easy to estimate and relatively free of stand density effects (Hagglund 1981). The most recent site index equations derived from stem analysis data (Mitchell and Polsson 1988) were used in the study, so site index estimates should be as precise as currently possible, short of actually using stem analysis. The sample population is restricted to fully stocked, second-growth stands which are essentially even-aged and within the acceptable range for reliable site index estimation (20 to 150 yrs, Mitchell and Polsson 1988). Although mean annual increment (MAI) varied somewhat with density and age within these stands, it should be a reasonable reflection of natural potential productivity and is used as an additional descriptive variable. The sample population is also restricted to mineral soils to limit variation in major parent material. The focus of the study is to improve understanding of site/productivity relationships in the area, and emphasis is placed on a comprehensive array of continuous environmental and floristic variables to achieve this. Both continuous and categorical variables which can be readily identified in the field were also be examined for practical application of site/productivity relationships. The full range of productivity found in the study area will be sampled with a balanced distribution of plots to maximize the strength of any relationships which may occur. Data analysis is of an exploratory nature, employing a number of techniques to assess structure in the data, and to examine relationships between the various properties studied. 23 4.2 Sampling Design The sample population is represented by fully stocked, even-aged fire-regenerated stands established on mineral soil. Sampling strata were defined to obtain a balanced sample across the productivity range of the population. Site index of western redcedar was used to infer forest productivity because it forms a major component in most stands. Based on the frequency distribution of cedar site index in the 1983/84 inventory data (Figure 2), three site index classes were recognized (Table 2). 25 20 15 10 Frequency (%) n = 72 m r n m r n 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 Cw Site Index (m/50yrs) Figure 2. Frequency histogram of Cw site index in the inventory update data. Site index is m/50 yrs. b.h. age. 24 Table 2. Site index classes. Strata Cw Si t e Index 1 1 < 15 2 16 - 20 3 > 20 1 m/50 yrs. b.h. age The classes essentially divide the observed distribution into thirds representing the lower, middle, and upper ranges of site index. As Cw site index was approximately normally distributed in the inventory sample, stratum 2 was the most strongly represented, with stratum 1 and particularly stratum 3 occurring less frequently. More refined stratification was not considered, given the relatively limited range of site index (14 m), the difficulty in establishing a greater number of meaningful breaks in the sample distribution, and the greater difficulty in distributing plots evenly in a larger number of classes. Eight stands from each strata were selected for sampling, following a reconnaissance of the study area. A single plot was established in each stand, yielding a total of 24 sample plots. 4.3 Sampling Methods Square, 400 m 2 (20 m x 20 m) sample plots were established within portions of each stand which were as homogeneous as possible in stand structure, understory vegetation, and site properties. Plots were located within these portions using a random bearing and distance. 25 4.3.1 Vegetation and physiography Vegetation was described following the methods of Walmsley et al (1980). A l l vascular plants and bryophytes growing on the principal substrate (forest floor and decaying wood) were listed and their percentage cover estimated. Nomenclature of vascular species followed that of Taylor and MacBryde (1977), mosses and hepatics follow Crum et al. (1973), and lichens follow Hale and Culberson (1970). One deviation from this, Kindbergia oregana, follows Ochyra (1982). Physiographic description included elevation, slope, aspect, and slope shape. Slope shape refers to the general physiography influencing the plot and includes depressional concave, receiving concave, elevated (convex), level, and sloping. A l l these slope shapes are expressed rather subtly in the study area because of the subdued topography. 4.3.2 Soil Within each sample plot, 16 subsample points were located systematically using a 4 row X 4 row grid centered over the plot (rows were located at 2.5, 5, 5, 5, and 2.5 m intervals along two adjacent plot sides). If a subsample fell on an "unsuitable" substrate such as rocks, stumps, and decaying logs, it was relocated to the nearest suitable location or to a new random point. One subplot was randomly rejected to leave 15 subplots for soil chemical sampling. In addition, one subplot was randomly selected from each of the four 26 north-south rows to provide 4 locations for detailed soil description. At each of the detailed subplots, a soil pit was excavated to a depth of one meter or less depending on the presence of very dense layers. Principal layers were identified which reflected obvious pedological discontinuities such as parent material, density, particle size, organic matter content, etc. Since the focus was not on soil classification, these layers did not necessarily coincide with soil horizons. On average, three to four layers were recognized in mineral soil profiles. The following morphological properties were described for each soil profile: 1. for each layer; the layer thickness, coarse fragment content, soil texture, structure, consistence, colour, mottles, roots, cementation, and presence of charcoal fragments were described or measured. The forest floor thickness, selected morphological properties, and humus forms were also described. 2. for the profile, the depth from the mineral soil and ground surface to the following properties was measured: - free water present in the soil - pan (cemented or high density). - primary root zone (to bottom of zone of majority of rooting) - secondary root zone (to bottom of all rooting). Often the primary and secondary root zones coincided because of abrupt root restrictions. - mottles (distinct and prominent). 27 In addition, the "free water status" for the profile was described using the following classes: 1 - absent 2 - slow: very slow seepage into pit; no appreciable standing water. 3 - moderate: seepage water moving slowly but steadily into the pit; standing water present but does not accumulate to substantial, depth. 4 - rapid: seepage water moves quickly into the pit, filling it to substantial depth. At each detailed subplot, samples of the forest floor and root zone mineral soil were collected for bulk density measurement. Rectangular blocks of forest floor were extracted using a pruning saw and subsequently trimmed with shears. The dimensions were measured on all sides and averaged to give values for the three axes from which the sample volume was calculated. A 500-cm3 bulk density core was used to collect mineral soil samples of known volume since the root zone was generally restricted to the coarse fragment-free eolian capping. If large roots interfered with the core a new sample was taken. In a few cases where rooting extended into underlying materials containing appreciable coarse fragments, the bulk density sample was collected using the glass bead displacement technique (Klinka et al 1981). At each of the 15 subplots, forest floor thickness was measured, and a sample of forest floor and root zone mineral soil was collected. Mineral soil samples were randomly bulked in the field in groups of five to produce three composite samples per plot. Forest floor samples were bulked in a similar manner following drying and preliminary grinding in the lab. In two plots, all subsamples of forest floor and mineral soil were individually retained for, analysis to provide information on within-plot variability of chemical properties. 28 4.3.3 Stand In each plot, all trees greater than 7.5 cm dbh were tallied according to species, dbh, and pathological symptoms (B.C. Forest Service 1983). In addition, "stem form" was assessed for lodgepole pine to provide information on the relative frequency of stem irregularities, while "crown thinning" (Table 3) was assessed for all species to detect any apparent decline in crown vigour. Table 3. Crown thinning and stem form classes C l a s s D e s c r i p t i o n a. Crown t h i n n i n g : F u l l f u l l crown; no i n d i c a t i o n o f t h i n n i n g f o l i a g e Moderate crown b e g i n n i n g t o t h i n out; l o s s o f f o l i a g e t h i n n i n g apparent Severe crown ve r y t h i n ; poor f o l i a g e r e t e n t i o n , dead t h i n n i n g branches, e t c . b. Stem form: Smooth Moderately i r r e g u l a r Very i r r e g u l a r no obvious s u r f a c e i r r e g u l a r i t i e s n o t i c e a b l e bumps, f l u t e s , e t c . but not pronounced v e r y pronounced bumps, f l u t e s , e t c . ; l i k e l y t o a f f e c t wood q u a l i t y s i g n i f i c a n t l y . 29 Height and breast-height age measurements for site index estimation were made on six dominant trees of redcedar and four dominant trees of the remaining species on the plot. These "site trees" had to be free of damage and disease. Trees which exhibited suppressed early growth were rejected. Occasionally insufficient site trees were found within the plot, in which case adjacent trees outside the plot boundaries, but on the same site, were selected. Site trees were also measured for dbh and were supplemented with height and dbh measurements of several trees from a range of diameters. This information was used for the derivation of species-specific height/diameter equations. 4.3.4 Soil Water Quality Because of the important role of soil water in the study area, two aspects of water quality were assessed to attempt to understand its relationship to tree growth and other site properties. Specific conductivity and dissolved oxygen were measured in situ in saturation zone water through a 2 inch diameter aluminum standpipe with a solid bottom and a number of 5 mm holes drilled along the sides. The standpipe was inserted in a bore hole drilled with a soil auger to a depth of approximately 80 cm (or less, depending on the density of the soil). This was located in a representative portion of the plot, avoiding decaying logs or other atypical microsites. Once inserted, the stand pipe was capped with a rubber stopper and allowed to settle for a period of two hours. The water thus collected was felt to represent the water within the general ecosystem boundaries (Binkley 1980). In six plots, insufficient water was present in the soil to collect a sample and measurements were not taken. 30 Specific conductivity Specific conductivity was measured in situ using a YSI Model 33 conductivity meter and cell. The cell was inserted into the stand pipe and submerged fully in the soil water present. Measurements were taken according to the manufacturer's instructions, with the resulting conductivity values expressed in dS/m 2 at 25°C. The specific conductivity meter and cell were checked four times during the sampling period using a standard solution of saturated C a S 0 4 [2.63 g/1 C a S 0 4 yields a specific conductance of 2.205.dS/m @ 25°C (Richards 1954)]. The average meter reading of 1.365 dS/m was considerably lower than the actual value and is most likely due to contamination of the cell electrode. When the platinum-coated electrode becomes contaminated with use, less "active" electrode surface area is exposed to the sample resulting in a lower cell constant. Meter readings can be corrected by applying a factor based on the ratio of the actual and measured values (B. von Spindler, 1988, pers. comm.). In this case, all readings were corrected with a factor of 1.62. Dissolved Oxygen Dissolved oxygen in the soil water was measured in situ using a YSI Model 5 IB dissolved oxygen meter. This employs a polarographic method with a membrane-covered electrode. Dissolved oxygen diffuses across the membrane from the sample and is reduced 2 1 dS/m equals 10"3 umhos/cm 31 and subsequently read at the polarized electrode. The method is well-suited to field testing because of its simplicity, the fact samples can be measured in situ rather than requiring collection, and the lack of interference by suspended sediments and other impurities (Franson 1976). To take a measurement, the probe was attached to a 1 m long rod and lowered into the stand pipe so the electrode was fully submerged in the water. Readings were taken according to the manufacturer's instructions. To avoid localized depletion of dissolved oxygen around the electrode, it was gently moved up and down within the water sample while the probe stabilized and a reading was taken. The resulting readings were expressed in parts per million. The probe's membrane and electrolyte were replaced regularly at the end of each day. 4.3.5 Soil Oxidation Zone Index One of the difficulties in studying site/productivity relationships in ecosystems featuring waterlogged soils is the quantification of soil aeration. In this study, air-filled porosity, and volumetric moisture content of the soil, as well as dissolved oxygen in the soil water were measured for this purpose. Some measure of oxygen content in the soil would be desirable; however, currently available platinum electrode techniques generally yield results of questionable reliability (Mclntyre 1970). As an alternative, McKee (1978) tested a simple technique for estimating the depth of oxygenated soil in sites influenced by high water tables. The length of rusting on iron rods driven into waterlogged soils proved to be a 32 reliable indicator of the depth to the average water table level over a 16-week period in coastal South Carolina. The basis for the method is that an iron rod will oxidize in the aerated zone of soil but will not oxidize in the non-aerated zone. This was tested in this study by driving six mild steel rods (6 mm x 1 m) into the soil to a depth of 80 cm (less if a barrier was encountered). These were distributed in each plot adjacent to the four detailed soil pits plus two additional randomly selected subplots. The rods were removed and measured on August 25-26, 1987, leaving a 4- to 6-week period for the rods to oxidize. By spanning the period from July through August, the depth of oxidation should provide an index of the average maximum zone of aerated soil when water tables are at their lower levels. At the time of removal, the length of rod featuring signs of oxidation was measured from the ground surface. 4.4 Laboratory Analysis and Data Preparation 4.4.1 Soil chemical analysis Following air-drying, mineral soil samples were crushed with a rolling pin and passed through a 2 mm sieve to remove coarse fragments, then stored in air-tight plastic containers. Forest floor samples were air-dried, coarsely ground with a Waring blender for compositing, then fine-ground with a Wiley mill and stored in air-tight containers prior to analysis. Mineral soil and forest floor p H was determined potentiometrically using a Radiometer p H meter on a 1:2 soil to C a C l 2 slurry. Total carbon was determined using a L E C O induction furnace and C analyzer. Total nitrogen was measured colorimetrically with a Technicon y 33 Autoanalyzer on a semi-micro Kjeldahl digest. Mineralizeable N was measured colorimetrically from a 1 N KC1 extractant following a Waring-Bremner 2-week anaerobic incubation (30° C). Initial ammonium N was not deducted from the total. Available phosphorus was measured colorimetrically using the ascorbic acid method, on a 1:10 soil:solution Mehlich extract (Mehlich 1978). Exchangeable calcium, magnesium, and potassium were determined with the 1 M ammonium acetate (pH 7) method, using a Perkin-Elmer atomic absorption spectrophotometer. Subsamples of the three composite samples of mineral soil were further bulked to form one sample per plot for which particle size analysis was undertaken using the pipette method. The sand content was determined by wet sieving. The above methods are described by Lavkulich (1978). 4.4.2 Determination of soil physical properties In the laboratory, bulk density samples were placed in paper bags, weighed, then dried in an oven at 105°C for at least 24 hours. Depending on moisture content, forest floor samples occasionally required over 36 hours drying. Dry weights were recorded and samples were then passed through a 2 mm sieve. Mineral soil aggregates and coarse, but weathered organic debris were gently crushed with a rolling pin. The remaining coarse fragments >2 mm in size were separated into mineral and organic, and weighed. The volume of mineral coarse fragments was determined through water displacement. The volume of organic coarse fragments was determined by approximating individual pieces to three-dimensional 34 figures (cylinders, rectangular parallelepipeds, prisms, or frustums), measuring them, and summing their volumes. Although there were some measurement errors, it was felt to be more precise than using water displacement as organic fragments would absorb water, and were difficult to displace accurately because of their tendency to float. Coarse-fragment-free bulk density was calculated using the following equation (Klinka et al. 1981): BDcff = (Md - Mcf)/ (Vh - Vcf) where BDcff is the coarse-fragment-free bulk density, Md is the dry weight of the sample, Mcf is the coarse fragment weight, Vh is the volume of the excavated sample, and Vcf is the coarse fragment volume. The volumetric composition of the soil in relation to porosity and water content was determined for the root zone mineral soil, corrected for coarse fragment (>2 mm diameter) content. Coarse fragment volume was excluded from the total soil volume since coarse fragments may occupy appreciable volume in a soil sample without making a commensurate contribution to the porosity or water capacity of the soil (Gardner 1965). Porosity represents the proportion of a volume of soil occupied by water and air. Porosity was calculated using the following formula (Brady 1974): POR = 1 - (BDcff/particle density) 35 where P O R is the total pore space (proportion) of the <2 mm soil fraction, and B D is the coarse-fragment-free bulk density (Mg/m 3). Because of the relatively high proportion of organic matter in some soils, a weighted particle density was calculated as follows: W P D = (1.30M) + (2.65(l-OM)) where W P D is the weighted particle density (Mg/m 3), O M is the organic matter content calculated from [1.7carbon(%)]/100 (Brady 1974), and 1.3 and 2.65 are the approximate particle densities of organic matter and mineral soil respectively (Brady 1974). Volumetric water content represents the proportion of the soil volume occupied by water. It was calculated as follows (Black 1988): V M C = (Ww/Dw)(Ds/Ws) where V M C is the volumetric water content (proportion), Ww is the mass of water (wet sample mass - oven-dried mass), Dw is the density of water (1.0), Ds is the coarse-fragment-free soil bulk density (Mg/m 3), and Ws is the mass of soil (<2 mm fraction). Air-filled porosity is the proportion of the soil volume occupied by air-filled pores, and was calculated as follows (Vomocil 1965): AIR = P O R - V M C 36 Moisture content and air-filled porosity obviously vary with the actual soil moisture conditions at the time of sampling. In this study they reflect mid-growing season conditions during a climatically stable period in which limited precipitation fell (Figure 3). In spite of the variation in absolute values of porosity and water content, the relative differences between sites at the time of sampling should provide useful insight into productivity relationships. Ppt (mm) 4/1 4/15 5/1 5/15 6/1 6/15 7/1 7/15 8/1 8/15 8/31 Date (mo./day) Figure 3. Daily precipitation for April 1 to August 31, 1987. 4.4.3 Conversion of soil chemical concentrations to mass/area basis Soil chemical properties (other than pH) were initially expressed as concentrations by mass in the <2 mm fraction. A more ecologically meaningful expression of soil nutrients is the total quantity (in kg/ha) present within the available root zone. This integrates 37 concentrations of soil nutrients with soil physical properties such as bulk density and thickness, resulting in a better representation of the total nutrient reserve potentially available to trees (Lewis 1976). Several site/productivity studies have examined soil nutrients expressed on a mass/area basis, often finding better relationships with site index compared to expressions of nutrient concentrations (Adams et al. 1970, Blyth and McLeod 1981a, White 1982, Johnson et al. 1987). The conversion of concentration data to quantities expressed as kg/ha was done using the procedure described by Kabzems [1985 - based on Lewis (1976) and Roy (1984)]. The initial nutrient concentration values used in the conversion are means from the three composite samples of forest floor and mineral soil for each plot. The coarse-fragment-free bulk density values used for the forest floor and mineral soil are means of the four samples taken within each plot. The forest floor thickness is a mean of the 15 subplot measurements. The rooting depths are means of the four measurements made in each plot. The quantity of nutrients was calculated separately for the forest floor, mineral soil in the primary root zone, and mineral soil in the secondary root zone. These were then combined to give two expressions of soil nutrients - total kg/ha in the forest floor plus mineral soil in the primary root zone, and in the secondary root zone. Formulas used for converting concentration data to kg/ha are given in Appendix B. 38 4.4.4 Stand Properties Site index Site index was estimated for all site trees using recent polymorphic site index equations derived from stem analysis data. Various definitions of site height were used in these models, while dominant height was used in the selection of site trees in this study. These discrepancies in tree selection procedures should, however, have little effect on site index estimates (Mitchell and Polsson 1988). Tree ages are at breast height in all cases. Western redcedar site index was estimated using the equation of Kurucz [1985 - in Mitchell and Polsson (1988)]. Site index was calculated for each site tree using the following formula derived from the original model (R. Drummond 1986, pers. comm.): SI = 2500 (H-1.37)(0.05027 + 0.01411A + 0.000097667A2^ A 2 - (H-1.37)(-3.11785 - 0.02465A + 0.00174A2). + 1.37 if A>50 then H = H - A D J , where A D J = .02379545H - 0.000475909A(H), A = breast height age, and H = total height. Western hemlock site index was estimated using Wiley's [1978 - in Mitchell and Polsson (1988)] equation, with the Kurucz (1981) correction for B.C. The B.C. correction causes the height growth curves to fall off more quickly in older ages compared to Washington and Oregon where the original curves were developed. Thus, for use in B.C., this correction improves the curves for older ages , while biasing them for younger (<50 year) ages. As the 39 stands in the study area ranged from 75-120 years of age, and occurred at considerably higher latitudes than Washington and Oregon, it was felt the B.C. correction would provide a more accurate estimate (K. Mitchell, pers. comm. 1988). Site index was estimated for each tree, using computer-generated tables of site index for a range of ages and heights, determined from the original equations. Sitka spruce site index was estimated using the equations of Barker and Goudie [1987 -in Mitchell and Polsson (1988)]. This model was developed from data collected in the Queen Charlotte Islands and is directly applicable to the study area. Site index was estimated for each site tree in a manner similar to that used for hemlock. Site index for lodgepole pine was estimated using the equations developed by Goudie (1984). Two caveats must be recognized in using this source for PI growing in the study area. First, Goudie's models are based on trees growing in continental climates of B.C. and Alberta. No site index data for coastal PI is available. Goudie's model does however, include a "wet site coefficient" which was estimated from site trees growing on wetter than average sites. This coefficient results in height growth being sustained longer than on drier sites. Applying this coefficient should provide a reasonably accurate estimate of site index in the study area where soil moisture deficits do not occur. The second caveat is that some site trees in the study area were outside the upper range of the data used to derive the model. However, the model has been found to perform well at the edges of the data (J. Goudie 1988, pers. comm.). Site index was estimated for each site tree using a similar method to that used for Ss and Hw. 40 Mensuration Prior to calculating stand volumes, height/diameter equations were derived for each tree species. A quadratic model was selected to best represent the height/diameter relationship expressed in scatter plots. A comparative evaluation indicated better fits of the quadratic models relative to linear models. A procedure for testing for different slopes and intercepts, described by Cunia (1971), was used to determine whether the data for each species should be pooled, or fitted to separate equations for each site index class. For Cw and Hw, it was found that separate site index class equations would best represent the data. For PI, separate equations were fitted to class 1, and pooled classes 2 and 3. A l l data were pooled for Ss to produce a single equation. Table 4 summarizes the final equations selected for subsequent volume calculations. Table 4. Equations for calculating tree height from dbh, according to tree species. Fit of the equations indicated by coefficient of multiple determination (R 2) and standard error of the estimate (Se). Species S i t e Index Equation 1 Class Cw 1 HT=5.112 + 0.854DBH - 0.011DBH2 (R2=0.606 Se=1.81 n=68) 2 HT=11.223 + 0.547DBH - 0.004DBH2 (R2=0.693 Se=1.98 n=66) 3 HT=12.113 + 0.459DBH - 0.002DBH2 (R2=0.715 Se=2.42 n=46) Hw 1 HT=2.349 + 1.139DBH - 0.015DBH2 (R2=0.712 Se=2.09 n=48) 2 HT=6.458 + 0.896DBH - 0.007DBH (R2=0.797 Se=2.27 n=51) .2 41 Table 4. Continued Species S i t e Index Class Equation Hw 3 HT=9.312 + 0.842DBH - 0.006DBH2 (R2=0.700 Se=3.01 n=36) PI 1 HT=6.431 + 0.870DBH - 0.012DBH2 (R2=0.530 Se=1.27 n=41) 2&3 HT=10.812 + 0.661DBH - 0.005DBH2 (R2=0.554 Se=2.28 n=63) Ss 1-3 HT=10.923 + 0.624DBH - 0.003DBH2 (R2=0.820 Se=3.48 n=77) 1 HT=total tree height, D B H = diameter at breast height Tree volumes were calculated using whole stem cubic meter volume equations from B.C. Forest Service (1976). Only one utilization standard, total stem volume to a 7.5 cm minimum dbh, was used. No deductions were made for decay, waste, and breakage. Cruise data for each plot were processed through a BASIC computer program to calculate volume/hectare, basal area/hectare, stems/hectare, and quadratic mean diameter by species and for live and dead trees. 4.5 Data Analysis The general approach to data analysis was to i) examine ecological and stand properties in relation to the three site index classes defined for sampling; ii) assess the inherent structure defined by the ecological properties themselves; and iii) examine functional relationships between site index and ecological properties. The selection of variables used in the analyses was made from an initial screening of the complete variable set which 42 identified the most promising ones. These variables were assessed for within-plot variability using coefficients of variation. This provided some relative measure of the variability contained in the plot means which were used in all subsequent analyses. Sample size calculations (Freese 1962) were made to express within-plot variability in the more practical terms of sampling requirements. Differences between site index classes expressed in categorical variables was assessed using contingency tables. The Tukey H S D multiple comparison test was used to assess differences expressed in continuous variables. A multivariate examination of the relationship of site index classes to soil properties, in effect a "test" of these classes, was done using discriminant analysis. Inherent structure present in the soil data, irrespective of site index classes, was examined using a combination of principal components analysis to produce a geometrical ordination of samples, and cluster analysis to impose an "objective" classification on the data. The vegetation data were examined using reciprocal averaging which produces a simultaneous ordination of samples and species on coordinated scales, and is useful for observing structural relationships between species and samples (Hill 1973). Functional relationships between site index and selected ecological properties were examined using multiple linear regression. A number of models were evaluated which related site index to original soil variables, field-recognizable variables, and understory vegetation expressed as reciprocal averaging scores and indicator species groups. Model 43 evaluation and variable selection procedures described by Chatterjee and Price (1977) and Zar (1984) were used to determine the best models for the data. More detailed descriptions of these and other statistical analyses outlined previously will be presented in Section 5.0, preceding discussion of the relevant results. 44 5.0 R E S U L T S A N D DISCUSSION 5.1 Description of the Data A large number of continuous and categorical variables were initially generated from the raw data (Table 5). Table 5. Summary of available variables Code Variable Type Expr. Source1 1. Soil chemical properties: T C total carbon C P M CS T N total nitrogen C P M CS MINN mineralizeable nitrogen c P M CS P available phosphorus c P M CS C A exchangeable calcium c P M CS M G exchangeable magnesium c P M CS K exchangeable potassium c P M CS H U M C N carbon/nitrogen ratio in forest floor c P M CS H U M P H p H in forest floor c P M CS M I N C N carbon/nitrogen in mineral soil c P M CS M I N P H p H in the mineral soil c P M CS 2. Soil physical properties: D 1 R O O T M 1° rooting depth from mineral surface c P M ISP D 1 R O O T S 1° rooting depth from ground surface c P M ISP D 2 R O O T M 2° rooting depth from mineral surface c P M ISP D 2 R O O T S 2° rooting depth from ground surface c P M ISP D H 2 0 M depth to free water from mineral surf. c P M ISP D H 2 0 S depth to free water from ground surf. c P M ISP P O R total porosity in mineral soil2 c P M ISP AIR air filled porosity in mineral soil c P M ISP V M C volume, moisture cont. in mineral soil c P M ISP F F D E P forest floor thickness c P M SP P A N depth to a "pan" from mineral surface c P M ISP C A P thickness of eolian cap c P M ISP M O T L depth to mottles from mineral surface c P M ISP C L A Y percent clay in mineral soil c P M CS 45 Table 5. Continued Code Variable Type Expr. Source1 SILT percent silt in mineral soil C P M CS S A N D percent sand in mineral soil C P M CS C F P M coarse frag, content of par. material3 C P M ISP S L O P E slope C P CP H 2 0 S T free water status K P M ISP R O O T rooting density in mineral soil K P M ISP H U E hue of mineral soil K P M ISP C H R O M A chroma of mineral soil K P M ISP T E X T P M field texture of parent material K P M ISP CONS consistence of mineral soil K P M ISP C E M presence of cemented horizon K P M ISP S H A P E ground surface shape K P CP P M parent material type K P CP 3. Soil water properties: SC specific conductance C P CP D O dissolved oxygen C P CP 4. Soil aeration index: A Z 1 depth of 1st zone, method A C P M SP A Z 2 depth of 2nd zone, method A C P M SP A Z 3 depth of 3rd zone, method A C P M SP BZ1 depth of 1st zone, method B C P M SP BZ2 depth of 2nd zone, method B C P M SP 5. Stand properties: CWSI site index of Cw (m/50 yrs) C P M ST HWSI site index of Hw (m/50 yrs) C P M ST PLSI site index of PI (m/50 yrs) C P M ST SSSI site index of Ss (m/50 yrs) C P M ST M A I mean annual inc., standing gross vol. C P CP 1 Type: C = continuous, K=categorical Expression: PM=plot mean, P=plot (no subsamples) Source: CS = composite samples, ISP = intensive soil pits, SP = soil subplots, CP = complete plot, ST=site tree 2 Mineral soil in the root zone 3 Predominant subsoil below the root zone 46 The soil nutrient properties (TC,TN,MINN,P,CA,MG,K) were expressed in the following ways: 1. concentrations in the forest floor and mineral soil 2. quantity (kg/ha) in the forest floor 3. quantity in the primary root zone mineral soil 4. quantity in the secondary root zone mineral soil 5. quantity in the combined forest floor plus primary root zone mineral soil 6. quantity in the combined forest floor plus secondary root zone mineral soil Plot "means" of categorical variables represent the predominant class described for the four intensive soil pits in each plot. Analysis of categorical soils data was undertaken separately (see section 5.5.1) following the principal analyses which focussed only on continuous data. Soil water properties and soil aeration index were examined independently from the remaining site variables. 5.1.1 Selection of principle variable sets. The first step in the analysis of site/productivity relationships is generally to reduce the number of variables to those which show some trend with the dependent tree growth variable (Broadfoot 1969, Campbell 1973, Carmean 1975, McQuilkin 1976). The complete set of continuous soil variables (except as indicated above) was therefore screened using scatter plots and Pearson correlation analysis to examine relationships with site index of 47 redcedar (CWSI). Variables which showed no or poor relation to site index were deleted from the list. A process similar to that of White (1982) was used to determine the best method for expressing soil nutrient properties. The six expressions described in the previous section (p. 46) were each examined in relation to CWSI using correlation analysis, with the one showing the strongest overall relationships being selected for all subsequent analyses. A total of 14 variables were selected, with soil nutrients best expressed as the content (kg/ha) in the combined forest floor plus secondary root zone. Water table depth and secondary rooting depth measured from the ground surface were selected to be consistent with the expression of soil nutrient content. Table 6 shows the correlation coefficients amongst the selected variables while Figure 4 graphically summarizes their correlation structure. Table 6. Pearson correlation matrix for the 14 selected variables. CWSI VMC AIR DH20S D2ROOTS TC TN MINN CA MG IC HUMCN HUMPH MINCN MINPH CUS I 1.000 VMC -0.632 1.000 AIR 0.627 -0.966 1.000 DH20S 0.556 -0.721 0.690 1.000 D2R00TS 0.712 -0.547 0.575 0.623 1.000 TC 0.642 -0.519 0.551 0.453 0.705 1.000 TN 0.800 -0.452 0.476 0.453 0.578 0.781 1, .000 MINN 0.665 -0.526 0.503 0.539 0.314 0.478 0, .795 1.000 CA 0.689 -0.660 0.659 0.542 0.551 0.605 0. .639 0.619 1. 000 MG 0.796 -0.666 0.689 0.513 0.730 0.749 0. .755 0.619 0. 822 1.000 K 0.596 -0.846 0.824 0.693 0.640 0.692 0. .649 0.665 0. .683 0.771 1.000 HUMCN -0.457 -0.099 0.059 0.020 -0.102 -0.099 -0. .554 -0.577 -0. .316 -0.269 -0.040 1.000 HUMPH 0.480 -0.312 0.319 0.356 0.231 0.165 0. .528 0.546 0. .699 0.532 0.350 -0.509 1.000 MINCN -0.543 0.159 -0.157 -0.137 -0.217 -0.145 -0. .573 -0.546 -0, .253 -0.351 -0.211 0.532 -0.271 1.000 MINPH 0.569 -0.330 0.381 0.236 0.287 0.172 0. 560 0.627 0, .507 0.614 0.366 -0.630 0.745 -0.609 1.000 4^  OO 49 VMC DH20S D2ROOTS HUMPH MINPH HUMCN MINCN SOIL PHYSICAL PROPERTIES SOIL CHEMICAL PROPERTIES Figure 4. Correlation structure diagram for selected soil variables and CWSI. The thin line connects variables with r > 0.62 (significant at alpha = 0.999); the thick line connects variables with r > 0.75; the variables indicated with shaded boxes are the key variables. Al l soil properties were positively correlated with site index with the exception of the volumetric moisture content and C / N ratios. Nitrogen and magnesium showed the strongest relationship to site index, while p H and C / N ratios of the forest floor and mineral soil, the depth to free water, and air-filled porosity were the weakest. A large number of intercorrelations occurred amongst the variables, with the stronger relationships shown in the correlation structure diagram. Volumetric moisture content and air-filled porosity have 50 a relatively strong negative correlation which is expected since the former is used in the calculation of the latter. Depth to free water is correlated to both V M C and AIR, with shallower depths associated with increasing moisture and decreasing air-filled porosity. Most of the selected soil nutrients were correlated with each other, with T N and M G showing strong correlation with the largest number of nutrients. The p H and C / N ratios tend to stand apart from the remaining nutrients, except for H U M P H which shows relatively strong correlation with C A , and MINPH which is related positively with MINN and M G . MINPH is negatively correlated with both M I N C N and H U M C N , and positively correlated with its counterpart, H U M P H . The general trend expressed amongst the soil variables is that higher p H values are associated with lower C / N ratios, which in turn, are associated with greater nutrient contents. Nutrient contents tend to vary up or down together, and decrease consistently with increasing volumetric moisture content or decreasing air-filled porosity. Most nutrient contents tend to increase with increasing rooting depth since the latter is used in the calculation of nutrient quantity. However the strength of this relationship varies amongst nutrients. The intercorrelation amongst many of the selected variables poses a potential problem in the multiple regression and multivariate analyses which are subsequently used in the data analysis. For example, strong linear relationships amongst independent variables in multiple regression result in ambiguous, unstable models in which the unique effects of individual variables cannot be estimated (Chatterjee and Price 1977). In discriminant analysis, strong correlation among discriminating variables renders canonical coefficients unstable and hard to interpret (Morrison 1969). To reduce these potential multicollinearity problems, the 14 51 selected variables were further screened to yield 4 key variables which were correlated with CWSI, and were not strongly intercorrelated themselves. These were T N , D2ROOTS, V M C , and M I N P H (see Figure 4). MINPH was included as it represented well the p H and C / N properties, although it was not as strongly correlated with CWSI as the other three. T N was selected to represent the remaining soil chemical variables with which it is correlated, as it is strongly correlated with CWSI, and is not strongly correlated with the remaining key variables. 5.1.2 Within-plot variability Because of inherent lateral variability, characterizing soil properties with a high degree of precision is commonly a problem in forest ecology studies (Mader 1963, Courtin et al. 1983). While composite sampling is an efficient way of characterizing soil chemical properties with reasonable precision at an acceptable cost, it is useful to have some knowledge of the variability these samples represent. In this study, all 15 subsamples of forest floor and mineral soil from one plot (#24) were individually analyzed for chemical properties. Within-plot variability was assessed through comparison of coefficients of variation (CV) for each property, and by calculation of the number of samples necessary to estimate the mean value of the property with a specified allowable error and degree of confidence. A computer program called S A S C A L (Marshall 1987) was used to compute sample sizes for two allowable errors (10 and 20%) and two confidence levels (80 and 90%). A summary of the CVs and required sample sizes are shown in Table 7. 52 Table 7. Coefficients of variation and required sample size to estimate soil property means at .80 and .90 confidence levels, with 10, and 20% allowable error. Calculations based on concentration data. Allowable Error and PH1 TC TN MINN CA MG K P Confidence Limits Sample siz e Mineral i s o i l CV (%) 4 42 45 39 46 28 28 24 20% @ .80 1 9 10 8 10 5 5 4 20% @ .90 1 14 16 12 16 7 7 6 10% @ .80 1 30 35 26 36 14 14 11 10% @ .90 1 49 58 43 59 23 23 17 Forest f l o o r CV (%) 5 2 14 24 27 19 22 24 20% @ .80 1 1 3 4 5 4 4 4 20% @ .90 1 1 4 6 7 5 5 6 10% § .80 1 1 5 11 13 7 10 11 10% @ .90 1 1 7 18 21 12 15 17 1 TC-total C; TN-total N; MINN-mineralizeable N; C A , M G , K- exchangeable Ca, Mg, K respectively; P-available P The variability trends in mineral soil properties show P H being the least variable, T N , T C , and C A the most variable, and M G , K, P, and MINN having intermediate variability. While the low and high variability of P H and C A respectively is consistent with most variability studies, the trend of higher T C and T N variability relative to other bases is not (Lewis 1976, Slavinski 1977, Kabzems 1985). The C V values were compared to those of two fresh to moist sites from Courtin et al. (1983) and two fresh sites from Kabzems (1985). While the CVs covered similar ranges there were some notable differences. M G , K, and P were all considerably less variable in this study than at either the Courtin et al. (1983) or 53 Kabzems (1985) sites. The other properties tended to be less variable than Courtin et al. (1983) but more variable than Kabzems (1985). The exception was T N which was more variable in both these studies. Sample size calculations show that few properties can have means estimated with a high degree of precision using 15 subsamples. This intensity of sampling can be expected to provide estimates with 20% error and 90% confidence level for all properties. Forest floor properties were substantially less variable than mineral soil properties, with the most variable property, C A , having only 27% C V . T C was very stable across all subsamples, showing only 2% C V which was even lower than P H . T N varied 14% while the remaining properties varied from 19 to 27%. These values compared very favourably with CVs of forest floor properties from two fresh sites in Kabzems' (1985) study, and with comparable properties of forest floors from Carter (1982). Sample size calculations indicate that means of all properties can be estimated with 20% error and 90% confidence, or 10% error and 80% confidence using 15 subsamples or fewer. In fact, most properties can be estimated with 10% error and 90% confidence using this sampling intensity. The effect of composite sampling on means and variability was assessed by comparing means and C V s of chemical properties based on the 15 individual subsamples from plot 24, and from the 3 composite samples derived from these subsamples (Table 8). Generally the means from the two sets of samples are very similar. Carter (1983) also found strong correlation between means of forest floor properties calculated from 15 subsamples and their composite sample. Variability of means from composite samples is generally 54 substantially lower than means from the individual subsamples. This is a common feature of composite samples since each sample actually represents a mean of its component subsamples (Slavinski 1977). Composite sampling offers a method of lowering sampling costs, while obtaining means close to the individual subsample means with lower variability. However, degrees of freedom are lost which widens confidence intervals of means and weakens statistical tests on the data. Table 8. Comparison of means and C V (%) of soil chemical properties of plot 24 based on 3 composite samples and 15 individual samples. Mineral s o i l Forest f l o o r Property Mean Mean CV CV Mean Mean CV CV 15 3 15 3 15 3 15 3 PH 3.5 3.5 4 2 3.5 3.5 5 2 TC(%) 15.2 13.2 42 18 45.8 46.0 2 1 TN(%) 0.37 0. 37 45 31 0.93 0.90 14 8 MINN(ppm) 41 46 39 13 220 194 24 5 P(ppm) 2.6 2.3 24 17 35.9 31.0 24 20 CA(meq) 0.60 0.59 46 8 10.92 10.53 27 20 MG(meq) 1.68 1.77 28 17 11.84 11.80 19 10 K(meq) 0.17 0.17 28 12 1.49 1. 37 22 8 Within-plot variation of soil properties was examined across all 24 sample plots. Boxplots of CVs for forest floor properties (expressed as concentrations) calculated from composite samples in each plot are shown in Figure 5. The trend in variability amongst the different properties is similar to that expressed in the 15 subsamples from plot 24. Variability of P H and T C is very low and consistent across all plots. T N and M G are also very consistent, with the majority of plots showing less than 14% C V for either property. The greatest variability is expressed in MINN, P, and C A , with a greater range in C V 55 occurring across all plots. However, the variation is still relatively modest with most plots falling below 20% C V for all three properties. > O 1.0 0 . 8 0 . 6 0 . 4 0 . 2 0 . 0 JL PH O TC TN C N MINN C A MG K Figure 5. Boxplots of CVs for forest floor properties, calculated for each plot from composite samples. Explanation of boxplots in Appendix M . TC=total C; T N = total N C N = C / N ratio; MINN = mineralizeable N; P = available P; C A , M G , K = exchangeable Ca, Mg, K respectively Figure 6 shows boxplots of CVs for mineral soil chemical properties calculated from composite samples for all plots. A similar trend to the plot 24 subsamples occurs amongst properties, except P and M G variability in the subsamples appear to be in the lower range of the overall data set. P and C A show the broadest range in C V across all plots, with two outlier plots exceeding 60% C V for C A . P H is extremely consistent, while K also shows a relatively narrow range in C V across the 24 plots. The remaining properties exhibit similar 56 ranges in CV. The majority of plots show less than 30% CV in all properties except P and CA. > O 1.0 0.8 0.6 0.4 0.2 -0.0 PH TC TN C N MINN C A MG K Figure 6. Boxplots of CVs for mineral soil properties, calculated for each plot from composite samples. Explanation of boxplots in Appendix M. TC = total C; TN = total N CN = C/N ratio; MINN=mineralizeable N; P = available P; CA, MG, K = exchangeable Ca, Mg, K respectively The within-plot variability was also examined for a number of soil physical and morphological properties which were measured in four soil pits in each plot. Figure 7 shows boxplots of CVs from all 24 plots calculated for soil morphological properties. > o 1 . 0 0 . 8 0 . 6 0 4 0 2 0 0 T T T Figure 7. Boxplots of CVs for soil morphological properties, calculated for each plot from four soil pits. Explanation of boxplots in Appendix M. FFDEP=forest floor depth; D2ROOTM and D2ROOTS=rooting depth (secondary) from mineral surface and ground surface, respectively; DH20M and DH20S = depth to soil free water from mineral surface and ground surface, respectively; CAP = thickness of eolian surface capping. FFDEP showed relatively high variability (20 - 40%) however the variability was quite consistent across all plots except one which exceeded 70% CV. The two properties which were measured from both the ground surface and mineral soil surface (D2ROOT, DH20) displayed lower and more consistent variability when measured from the ground surface. This may result from a small increase in standard deviation from the addition of forest floor thickness, relative to the increase in total thickness. The majority of plots had less than 30% CV for D2ROOTS which is an important variable in later analyses. The eolian cap shows a similar pattern of variability to D2ROOTS and DH20S except for 2 highly variable plots which featured some disturbance of surface horizons. 58 Boxplots of within-plot C V for soil physical properties calculated for all 24 plots are shown in Figure 8. I 1 r o I V M C AIR BDMIN BDFF CWSI Figure 8. Boxplots of CVs for soil physical properties, calculated for each plot from four soil pits. Explanation of boxplots in Appendix M . VMC=volumetric moisture content, AIR = air-filled porosity, BDMINN-mineral soil bulk density, BDFF=forest floor bulk density, and CWSI = site index of redcedar Within-plot variability of V M C is relatively low; less than 20% for most plots, and is quite consistent across the 24 plots. B D F F variability also covers a relatively narrow range over the data set, and averages about 20% C V . One highly variable plot featured a relatively thin Moder humus form that was difficult to sample and which may have contributed to its variability. The relatively modest and narrowly ranging variability of this > O 0.8 0.6 0 4 0 2 59 forest floor physical property is consistent with its chemical properties. Greater variability was exhibited in B D M I N which may reflect sampling techniques. Roots occasionally interfered with collecting a clean sample with the fixed volume bulk density core. AIR shows the highest variability across all plots, which may reflect compounding errors in the calculation. AIR is calculated by subtracting V M C , determined from moisture content and soil volume, from total porosity, determined from bulk density and weighted particle density. For comparative purposes, the C V of CWSI calculated from the 6 site trees in each plot is also shown in Figure 8. It varies little within a plot (5 - 8% CV) , and is remarkably stable in its variability across all 24 plots. 5.2 Stand Properties 5.2.1 Stand age The distribution of sample plots by stand age is shown in Figure 9. The majority of plots were in stands 95 to 115 years of age. These are representative of the most widespread stand type which dominated the area between approximately Lawn Hill and Port Clements. The relative uniformity in stand age and structure suggests that one or several closely spaced fires swept through this area. The younger stands around 75 years old were restricted to the southern portion of the study area in the vicinity of Miller Creek, the location of the most recent fires. One plot was established in a 160-year-old stand representative of the older wildfire history which characterizes the northern portion of the 60 study area. 25 20 15 10 Frequency (%) 70 75 80 85 90 95 100105110115 120 125130 135140 145150155 160 Stand Age (years) Figure 9. Sample plot distribution by stand age (age in 5 year classes - eg. 75=75-79 yrs.) 5.2.2 Site index Cedar site index averaged 18.1 m over the whole sample, and ranged from 10.9 m to 25.1 m. The distribution of samples in relation to CWSI is shown in Figure 10. There was an apparent clustering of plots within the three site index classes which defined the sampling strata. Most plots in class 1 were in the 13-to 14-m CWSI range. One plot had a CWSI of 10 m, which was atypically low for the sample population. The majority of plots 61 in class 2 ranged from 16 to 18 m, while 21 m was the most common CWSI for class 3 plots. This pattern of site index shown in the sample was in fact quite representative of the natural distribution of site index observed throughout the study area during the Frequency (%) 25 20 15 10 5 -Class 1 Class 2 _1 I rar-m t Class 3 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 Cw Site Index (m/50 yrs) Figure 10. Sample plot distribution by Cw site index. reconnaissance phase. In terms of geographical distribution, the most widespread stands growing on mineral soil throughout the study area were from site index class 2, with class 1 stands also occurring relatively frequently but in a more localized pattern. Stands from class 3 were uncommon, and were encountered only in the southern portion of the study area near Lawn Hill and Miller Creek. This is likely a reflection of the slightly greater relief in this area where the lowlands begin to rise up to the Skidegate Plateau. The mean site indexes of cedar, hemlock, pine, and Sitka spruce for the three site index classes are shown in Table 9. 62 Table 9. Means (and standard deviation) of site index (m/50yrs b.h. age) for the three site index classes. Site Index Class Variable 1 2 3 n 8 8 8 CWSI 14.0 17.8 22.3 (1.4) (1.1) (1.3) HWSI 12.0a1 16.8b 21. 6c (1.5) (1.8) (3.5) PLSI2 13.4a 18.6b 22 . 2c (1.5) (2.0) (1.4) SSSI3 9. la 16. lb 25. 6c (2.5) (1.7) (4.2) 1 classes not followed by a common letter are significantly different using the Tukey H S D test (P<0.05). See Appendix N for procedure used to check test assumptions. 2 n for class 3 is 7 3 n is 5, 6, 8 for classes 1, 2, 3 respectively Al l species follow a similar trend across the three classes, and are all significantly different between all pairs of classes (P<0.05). This is not unexpected since all species are responding together to differences in site quality across the classes. The generally low within-class variability contributes to the significant differences observed between all classes. However, class 3 appears to contain the most variability, at least in relation to hemlock and Sitka spruce. The relatively high variability expressed by Sitka spruce in class 1 is likely a reflection of the smaller sample size. A more informative picture of relative species performance is revealed in the relationships between site index of cedar and the remaining three species. Figures 11 through 13 show site index of Hw, PI, and Ss respectively, in relation to the site index of Cw. Hw Site Index (m/50 yrs) 25 15 10 5 r=.89 ^^^^^^^ / 1 / 1 1 1 A T * i i i i i i i i 7 9 11 13 15 17 19 21 23 25 27 29 Cw Site Index (m/50 yrs) Figure 11. Relationship between Hw and Cw site index: Dashed line is a 1:1 line; solid line is a least-squares fitted line. PI Site Index (m/50 yrs) 5 10 15 20 25 30 Cw Site Index (m/50 yrs) Figure 12. Relationship between PI and Cw site index. Solid line is a least-squares fitted line. 64 Ss Site Index (m/50 yrs) 35 30 25 20 15 10 5 5 10 15 20 25 30 35 Cw Site Index (m/50 yrs) Figure 13. Relationship between Ss and Cw site index. Dashed line is a 1:1 line; solid line is a least-squares fitted line Site indexes of Hw and PI are both highly correlated with Cw site index and demonstrate a very similar height growth pattern. Hw appears to have slightly slower height growth than Cw on poorer sites, reflecting cedar's better adaptation to the waterlogged soils typical of these sites. Pine appears to follow an identical height growth pattern to Cw. Sitka spruce displays markedly different height growth than Cw, although site index is highly correlated between the two species. On poorer sites (less than 20 m CWSI) Ss shows substantially slower height growth than Cw, with the difference getting greater as the site gets poorer. The opposite occurs as the site quality improves (above 20 m CWSI). This reflects the sensitivity of Ss to limiting site conditions and its ability to grow rapidly when site conditions become adequate for its requirements. 65 Barker (1985) examined site index relationships between Ss, Hw, and Cw in a wide range of second-growth ecosystems on Moresby and adjacent islands to the south of Graham Island. He also found significant correlations between Cw and Ss, and Cw and Hw, although they were weaker than in this study. This likely reflects the small number of individual stem analyzed trees sampled over a wider range of sites, compared to plot means of site index sampled over a narrower range of sites in this study. 5.2.3 Stand volume The range of total volume per hectare (live stems) sampled and its relationship to CWSI is shown in Figure 14. 2000 Volume (m3/ha) 10 14 16 18 20 22 Cw Site Index (m/50 yr) Figure 14. Relationship between total live stand volume (m3/ha) and Cw site index. 66 Plots in the lowest site index class supported from about 400 to 700 m 3/ha, averaging approximately 530 m 3/ha. The intermediate sites (class 2) supported from 670 to 1230 m 3 /ha with an average of about 920 m 3/ha, while the best sites (class 3) supported from about 900 to 1800 m 3/ha, averaging approximately 1200 m 3/ha. Variation in age must be considered when viewing these volumes, particularly in the class 3 plots. The four plots with the lowest volumes in this class (Figure 14) are 75 to 78 years old. If their growth were projected to an age similar to the rest of the data they should fall into the linear trend exhibited by the rest of the data. The total standing volumes described are slightly higher than merchantable volumes because of the allowance for top and stump. Barker's (1985) data showed that close utilization merchantable volume (10 cm top, 30 cm stump) ranged from 11% to 3% less than total volumes from 300 to 1700 m 3 /ha respectively. Mean annual increment is a more meaningful expression for comparing volumes amongst the sites as it reduces the differences due to age in stands approaching or somewhat past culmination age. It was found to be highly correlated with CWSI in the study plots (Figure 15). Table 10 shows mean M A I for the three site index classes - all classes were significantly different (P<0.05). These values are generally in line with those presented by Barker (1985), although he apparently did not sample sites as poor as those from site index class 1. The comparative precision of the M A I values shown for each site index class should be viewed with a certain amount of caution due to the younger aged stands on some of the class 3 sites. Depending on culmination age, M A I for these stands may be slightly higher or lower if projected to a similar age as the rest of the data. — - - — MAI (m3/ha/yr) •x r=.87 * * * * * * * i i i i i 10 12 14 16 18 20 22 24 26 Cw Site Index (m/50 yrs) Figure 15. Relationships between MAI and CWSI Table 10. Means and standard deviations of MAI (m3 ha"1 yr"1 total live volume) for the three site index classes. Site Index Class Variable 1 2 3 n 8 8 8 MAI 5. Oa1 8.8b 12. 6c Std. dev. (1. 0) (1.9) (2 • 0) 1 classes not followed by a common letter are significantly different using the Tukey HSD test (P<0.05). See Appendix N for procedure used to check test assumptions. The volume and MAI data indicate that these sites are capable of supporting relatively high volumes and densities given their inherent limitations. Barker (1985) found substantially higher basal area levels occurring in older second-growth stands in the Queen 68 Charlotte Islands relative to published normal yield tables. While this was partly attributed to sampling deficiencies in the older normal yield table data, it is believed that mortality differences do occur such that northern latitude coastal stands are characterized by greater stem densities (Meyer 1937). It is possible that these sites are capable of supporting higher densities because transpirational demands are lower relative to warmer southern latitudes, water deficits do not occur in most sites, and tree species are moderately to strongly shade tolerant, with the exception of lodgepole pine. Such ecological conditions would reduce inter-tree competition and allow greater stand densities. 5.2.4 Tree species composition Relative tree species composition for the three site index classes is shown in Figure 16. Figure 16. Relative tree species composition (proportion of total live volume). 69 The poorest sites are dominated by Cw, followed by Hw and PI in approximately equal proportions. Ss is all but absent because of its inability to tolerate the limiting site conditions. The intermediate site index class supports approximately equal proportions of Cw, Hw, and PI. Ss is present in most of these stands, but only averages about 7% of the total volume. It increases substantially in the composition of site index class 3 stands, to about equal proportions with Cw and Hw. This reflects spruce's ability to perform very well on sites which are able to adequately supply its growth requirements. PI drops substantially as a component of these better quality sites, likely a reflection of its shade intolerance and inability to compete with the increased composition of Ss. Examination of stem and crown conditions revealed no significant trends in relation to site index observed. Only PI showed a marked increase in the frequency of thinning crowns in site index class 3 which again is a reflection of its decreasing competitive ability on the better quality sites. 5.3 Ecological Properties of Site Index Classes 5.3.1 Soils Means and standard deviations of soil properties for the three site index classes are shown in Table 11. Class 1 stands apart from the other classes in its higher soil moisture content. This is a reflection of the depth to free water which is also significantly shallower in class 1 plots relative to classes 2 and 3. While air-filled porosity follows 70 Table 11. Means and standard deviations of soil variables for site index classes. S i t e Index Class Variable 1 2 3 n 8 8 8 Mean SD Mean SD Mean SD VMC (%/100) .73al . 09 • 56b .07 .53b .08 AIR (%/100)2 • 09a • 14a • 15a DH20S (cm) 30a 6 61b 27 70b 33 D2ROOTS (cm) 26a 5 35b 5 41b 6 TC (kg/ha) 127102a35331 152260ab 24597 189833b 42031 TN (kg/ha) 2517a 624 2917a 571 4832b 1310 MINN(kg/ha) 44a 12 59ab 16 98b 53 P (kg/ha) 13a 6 16a 4 14a 6 CA (kg/ha) 310a 93 394a 75 530b 111 MG (kg/ha) 209a 70 330b 71 461c 96 K (kg/ha) 130a 45 178ab 35 203b 36 HUMPH 3.2a .1 3.3a . 1 3.4a .2 HUMCN 54. 6ab 4.9 59. 0a 5.9 49 . 3b 7 . 4 MINPH 3.3a .2 3.4ab .1 3.6b .2 MINCN 46.2a 6.4 47.7a 10.2 34.8b 5.1 SAND (%) 20. 4a 7.2 33.9a 11.7 31. 7a 14.0 SILT (%) 49. 9a 9.9 47. 8a 8.1 45.2a 9.8 CLAY (%) 26. 3a 8.3 18. 3a 4.6 23 . l a 5.6 HUMDEP (cm) 13a 3 16a 2 15a 5 1 classes not followed by a common letter are significantly different using the Tukey H S D test (P<0.05). See Appendix N for procedure used to check test assumptions. 2 test done on a logarithmic transformation; means reported as antilog of transformed means a related trend of lower values in class 1, the difference from other classes is not significant. Rooting depth differs significantly between classes 1 and 2, and 1 and 3, and shows an increasing trend from class 1 through 3. Soil nutrient properties show a shift in class similarities, with class 3 standing apart from classes 1 and 2 in its significantly higher 71 content of T N and C A , and its lower MINCN. Although classes 1 and 2 do not differ significantly in these properties, there is a trend of increasing nutrient content, decreasing C / N ratio, and increasing p H from class 1 through 3. T C , MINN, K, and M I N P H all show an increasing trend from class 1 through 3, although classes 1 and 2, and 2 and 3 do not differ significantly. Magnesium is the only chemical property which differs significantly between all three classes. Mineral soil fine fractions did not differ significantly amongst the site index classes, although S A N D was considerably lower in class 1 sites. This may be due to a difference in the deposition of the surface capping in the class 1 sites which may have involved some settling in periodic standing water. The previous section discusses the differences in individual soil properties among the three site index classes. It would be useful to examine the combined ability of these variables to discriminate between these classes - in effect, testing the validity of the site index classes in relation to soil properties. This can be accomplished with discriminant analysis (Omi et al. 1979). Discriminant analysis is a multivariate technique which statistically distinguishes between several predetermined groups on the basis of a set of discriminating variables. This is accomplished by weighting and linearly combining the discriminating variables in some fashion so the groups are forced to be as statistically different as possible (Klecka 1975). The results can be used to determine which variables are good at discriminating the groups, how well these variables discriminate the groups, and how the groups are spatially related as defined by these variables (Morrison 1969). 72 Discriminant analysis was performed for site index classes 1 through 3 using all 24 plots and the following key soil variables; T N , D2ROOTS, V M C , and MINPH. The canonical correlations of the two derived discriminant functions were 0.90 and 0.55, indicating that the first function is strongly correlated with the site index classes while the second function is only moderately correlated. Standardized discriminant function coefficients are presented in Table 12. The first discriminant function is largely influenced by D2ROOTS and to a lesser extent, V M C which is negatively correlated. T N and M I N P H appear to contribute little. The second discriminant function is primarily a reflection of T N and V M C , while D 2 R O O T S and MINPH contribute essentially nothing. Table 12. Standardized discriminant function coefficients for dependent variables. Variable Function 1 Function 2 VMC -0.579 0.636 TN 0.376 0.725 D2ROOTS 0.662 -0.132 MINPH 0.331 0.165 The adequacy of the discriminant functions was tested by classifying the original set of plots using the probability of a plot's class membership calculated from discriminant scores (Klecka 1975). Table 13 summarizes the classification success by comparing predicted and actual class memberships. The classification routine was able to identify correctly 92% of the plots as members of the class they actually belonged to. This large proportion of successful classifications suggests that the soil properties are good discriminators of the three site index 73 Table 13. Frequency of plots according to actual and predicted site index class membership, based on discriminant analysis Actual SI Class n 1 Predicted SI 2 Class 3 C l a s s i f i c a t i o n Success (%) 1 8 8 0 0 100 2 8 0 8 0 100 3 8 0 2 6 75 Total 92 classes. Two class 3 plots (#01 and #12) were misclassified into class 2 by a narrow margin of probability. The combined effect of slightly shallower D2ROOTS, higher V M C , and lower T N values relative to class 3 plots contributed to the misclassification of plot 1, while slightly lower T N , MINPH, and shallower D 2 R O O T S influenced plot 12. Spatial relationships among the groups were examined by plotting discriminant scores for the 24 plots in the space defined by the two discriminant functions (Figure 17). Site index classes 1 and 2 are clearly separated with very minor overlap, with class 1 plots forming a relatively tight group. Classes 2 and 3 are also clearly distinguished; however, some overlap occurs with two class 3 plots showing similarity to class 2. Class 3 appears to have more variability than the other classes, as indicated by its greater dispersion among plots. Based on the discriminant function coefficients, V M C decreases while D2ROOTS increases from class 1 through to class 3. The increasing T N along factor 2 is primarily reflecting plots 23 and 24 which are located at the top right portion of the ordination and which had the highest T N contents for the class. 2.5 1.5 -1 -0.5 -- 0 . 5 v - 1 - 1 .5 -• 1 1. • 22 \ O 7 ° N 1 I 1 6 / / + -UL X In V19 / • 9 • / 13 / + 151 24 / O \ ",1 \ 1 ' ^ + \ \ 6 X \ \ \ \ \ \ \ 2 3 » 20. - 2 • SI C lass 1 Funct ion 1 SI C lass 2 1 r" 4 SI C lass 3 74 Figure 17. Plot of discriminant scores for 24 plots in relation to discriminant functions 1 and 2. Plots displayed according to site index class. The classification success described above may be biased upwards since the classification was undertaken on the same sample plots that were used to derive the discriminant functions. One method of avoiding this is to fit a discriminant function to part of the data and test this function by classifying the remaining data (Morrison 1969). The data set was therefore randomly divided in half, with each site index class split into two sets of four plots. Discriminant analysis was performed on the "training" set of plots using the same four key soil variables. The discriminant functions derived from these data were then used to classify the remaining 12 plots. This "chance" model correctly classified 83% of the plots, 75 with one plot being misclassified in each of site index classes 2 and 3, and all class 1 plots correctly classified. This is good classification success, particularly considering the small sample size of the training and test data sets. This verifies the interpretation of the original model which suggests the soil properties are relatively good discriminators of the three site index classes. Ordinations of the test plots and training plots in relation to the discriminant functions showed similar spatial relationships among site index classes to the original analysis of all plots. 5.3.2 Vegetation The frequency and mean cover of understory species found in the three site index classes are shown in Table 14. Several species showed a definite affinity to class 1, particularly Sphagnum girgensohnii, Blechnum spicant, and Bazzania denudata. These species are all indicators of moist to wet soils which were characteristic of these sites (Klinka et al. 1989). Gaultheria shallon was also associated with class 1 sites, although it also occurred in other site index classes under openings in the canopy. Hylocomium splendens and Rhytidiadelphus loreus showed a greater affinity to class 1, as indicated by their higher cover. This contrasts with Kindbergia oregana, which showed a lower affinity with class 1 . 76 Table 14. Species composition of the three site index classes j 1 | 2 j 3 l 8 i 8 I 8 Species Presence class and mean percent cover Bazzania denudata jIII .6' Blechnum spicant V 2 .5 II .5 Blepharostoma trichophyllum II .3 II .3 II .3 Calypogeja muelleriana II .8 ! IV .9 II .6 Calypogeja trichomanis I .6 II .8 II .6 Carex anthoxanthea I .3 Cephalozia bicuspidata I .1 Cephalozia l u n u l i f o l i a I * 1 II .4 II .4 Coptis a s p l e n i f o l i a I ' 3 Dicranum fuscescens II .4 I .1 II .3 Dicranum scoparium I .1 II .5 II .8 Dryopteris expansa I . 1 Gaultheria shallon III 1 .9 Gyrothyra underwoodiana I .3 I .3 Hookeria lucens I .4 I .3 III 2.5 Hylocomium splendens V 36 .3 V 15 .6 V 9.1 Hypnum c i r c i n a l e III 1 .0 II 1 .3 II .4 Kindbergia oregana IV 12 .2 V 44 . 1 V 35.9 Lepidozia reptans IV 1 * 3 IV 2 . 5 IV 1.4 L i s t e r a caurina II • 3 II 1 .1 L i s t e r a cordata IV 1 .4 III 1 .3 II .4 Lobaria oregana II .3 II .4 Lophozia i n c i s a I .1 I . 1 Luzula p a r v i f l o r a I . 1 Lysichitum americanum II .9 Maianthemum dilatatum I .3 I . 1 II .3 Malus fusca II .9 Menziesia ferruginea IV 1 .4 II .4 Moneses u n i f l o r a I .1 I . 1 P e l l i a neesiana I * 1 P e l t i g e r a aphthosa I . 1 P l a g i o c h i l a porelloides III 1 .8 II .6 III 1.4 Plagiothecium undulatum III 4 .'3 V 9 . 1 V 10.9 Pteridium aquilinum I . 1 Rhizomnium glabrescens V 6 .3 V 9 .5 V 13 .8 Rhytidiadelphus loreus V 28 .1 V 7 .7 V 12.0 Riccardia m u l t i f i d a I .3 II .4 I . 1 Scapania bolanderi V 9 .8 V 14 .4 V 5.6 Sphagnum girgensohnii IV 5 .0 Sphagnum squarrosum I .4 Streptopus amplexifolius II .3 Vaccinium alaskaense I .3 Vaccinium parvifolium v 2 •8! IV 2 .5 IV 1.1 Veratrum v i r i d e ! i .0 1 1 I i 1 Presence classes express percentage frequency: I = 1-20%, II = 21-40%, III=41-60%, IV = 61-80%, V = 81-100%. S i t e Index Class n 77 relative to the remaining two classes, although it was generally present on most sites. This is likely a reflection of the more humid microclimate of forest floors in the wetter class 1 sites. Hylocomium splendens and Rhytidiadelphus loreus are known to show greater preference than Kindbergia oregana for forest floors with greater water holding capacity and therefore, a more humid microclimate (Klinka and Krajina 1986). Hookeria lucens occurred more commonly in class 3, which likely reflects the higher pH, lower C / N ratio, and greater nutrient content of these sites (Klinka et al. 1989). The vegetation data were transformed and expressed in the form of spectra of five soil moisture and three soil nutrient indicator species groups (ISG) according to Klinka et al. (1989). Spectral values for each sample plot were calculated from the sum of percent cover of all species included in an ISG (eg. nutrient-poor ISG), divided by the sum of percent cover of all ISGs for that attribute (eg. all soil nutrient ISGs), multiplied by 100. These values give the relative frequencies of ISGs present in a sample plot. Mean spectral frequencies were calculated for each site index class from the component plot's ISG frequency data. Figure 18 shows the soil moisture spectra for the three site index classes. The very dry to moderately dry ISG is poorly represented in all classes, which reflects the lack of water deficits in these sites. Site index class 1 has a lower frequency of moderately dry to fresh indicator species, but had more fresh to very moist and slightly more very moist to wet species relative to other classes. Again this reflects its wetter soils. Generally classes 2 and 3 were quite similar, although class 3 did show a very slight tendency towards moister soils. 78 Frequency (%) 80 60 •10 20 0 VD-MD M D - F F - V M VM-W W-VW S o i l M o i s t u r e ISG I I SI Class 1 I - 1 SI Class 2 SI Class 3 Figure 18. Soil moisture spectra for the three site index classes ( V D - M D : very dry to moderately dry; M D - F : moderately dry to fresh; F - V M : fresh to very moist; V M - W : very moist to wet; W-VW: wet to very wet). Figure 19 shows the soil nutrient spectra for the three site index classes. There is a clear trend of decreasing frequency of the nutrient poor ISG and increasing frequency of the nutrient medium ISG from site index class 1 through 3. This is consistent with the trend observed in soil properties. Similar relationships between soil ISGs and soil nutrient status have been observed by Kabzems (1985) and Green et al. (1989). Al l sites are, however, dominated by the nutrient-poor ISG which reflects the thick, acid Mor humus forms commonly occurring in most plots. The occurrence of a few nutrient-rich species in site index class 1 is associated with one plot (#11) which was atypical in the occurrence of several wet, nutrient-rich indicator species such as Veratrum viride, Lysichitum americanum, and Maianthemum dilatatum. Frequency (%) 100 i POOR MEDIUM RICH Soil Nutrient ISG ] SI Class 1 I I SI Class 2 SI Class 3 Figure 19. Soil nutrient spectra for the three site index classes. 80 5.4 Natural Structure in the Data In the previous section, predetermined site index classes were characterized and analyzed in relation to soil and vegetation properties. This section assesses the natural structure which may exist in the soil and vegetation data, and the underlying factors which contribute to these observed relationships. The consistency of any natural groupings with the original site index classes is also examined as a further test of how well these classes reflect meaningful ecological patterns in the study area. 5.4.1 Soil properties Soil properties were examined using principal components analysis and cluster analysis. These are both exploratory multivariate techniques which can be used to uncover natural groupings or structure in the data. While principal components analysis produces a geometrical ordination of sample plots which can be visually assessed for natural patterns, cluster analysis imposes a more objective classification on the data. As the relative strengths and weaknesses of these two methods are complementary, it is profitable to combine them in analyses to examine whether partitioning the data using clustering indicates similar structure as the geometrical representation (Gordon 1981). Principal components analysis was undertaken on the data set defined by the 14 selected soil variables described in section 5.1.1. The analysis was done on a correlation matrix of all 24 plots using the MIDAS statistical package (Fox and Guire 1977). Only the first three 81 axes were extracted since axes beyond the first three or four are generally not readily interpretable (Nichols 1977). A single outlier (plot 23) severely distorted the ordination so it was removed and the analysis repeated. To assist in the interpretation of components, component scores were correlated with the original variables. Cluster analysis was undertaken on the data set defined by the same 14 soil variables. Euclidean distance was used as a measure of proximity between cases. A l l data were standardized to Z-scores3, as Euclidean distance is greatly affected by different variable scales (Everitt 1980). Clusters were identified using Ward's method (also known as sum-of-squares method). This is a hierarchical clustering algorithm which aims to partition the set of cases into a number of groups so as to minimize the total within-group sum-of-squares about group centroids. Ward's method is widely used, and is an intuitively reasonable approach to deriving compact clusters which avoids the "chaining" problem common to some of the other methods (Gordon 1981). The output includes a dendrogram of cases displaying hierarchically arranged groups, with branches lined up so that similar cases are closest to each other in their order. The relative branch lengths reflect the joining distance between pairs of clusters (Wilkinson 1987). The three extracted principal components accounted for 75% of the variation in the data (Table 15). The majority of this was in the first component. Component correlations with the original variables are shown on Table 16. 3 Z scores are calculated by subtracting the sample mean from a case, then dividing by the standard deviation. The mean of a set of Z scores is 0, and the variance is 1 (Zar 1984). 82 Table 15. Eigenvalue and variance extracted by the first three principal components. Component Eigenvalue % Variance 1 Cumulative % variance 1 6.50 46.4 46.4 2 2.34 16.7 63 .2 3 1.62 11.6 74.8 1 percent variance explained by a component is the proportion of its eigenvalue in relation to the sum of all component eigenvalues (14.02 in this case) Table 16. Component correlations with original soil variables. Values are r. Variable 1 Component 2 3 VMC -.77 .43 -.22 AIR .62 -.25 .22 DH20S . 66 -.40 . 19 D2ROOTS .88 .03 -.00 TC .80 -.03 -.46 TN .79 .31 -.47 HUMPH .39 .42 .64 HUMCN -.05 -.85 .10 MINPH .39 .63 .52 MINCN -.24 -.59 .21 MINN .75 .09 -.40 CA .81 .11 .19 MG .88 .11 . 01 K .82 -.41 -.08 All variables except p H and C / N of both the forest floor and mineral soil show high loadings on the first component. A l l variables are positively correlated with the exception of V M C . The quantity of nutrients in the root zone, the rooting depth, and the root zone aeration as inferred by moisture content, clearly characterize this component. The second component is interesting in that the only variables loading highly are p H and C / N ratio, the same variables that load weakly on the first component. This component could be 83 interpreted as an index of nutrient mineralization rates, with higher p H values and lower C / N ratios associated with higher positions along the component. Figure 20 shows an ordination of samples in relation to the first two components. CO < u CL 2 -•1 --2 -- 3 / 11 \ I • V 1 9 21 / / 2 \ / \ 22 / + 1 6 6 ^ J 3 /O , 5 2 4 y \ \ o \ \ + 1 7 14 O / 1 0 / \ \ \ - 3 SI C lass 1 - 1 PCA Axis 1 SI C lass 2 1 SI C lass 3 Figure 20. Ordination of sample plots in relation to soil P C A axes 1 and 2 (plot 23 deleted). Dashed lines show the three main clusters identified in cluster analysis. Plots are arranged along the first component according to a trend of increasing site index from the lower to upper range of the axis. Site index class 1 plots are generally restricted to the lower range while site index class 3 plots occur in the upper range. Some overlap occurs in the central area of the ordination, with three plots from site index class 1 occurring together with plots from site index class 2. The outlier deleted from the 84 analysis (plot 23) would be associated with the extreme upper right range of the ordination. Plots 3, 5, 24, and 11 stand apart in the upper range of the second axis because of their higher p H and lower C / N values. Presumably these soils could be interpreted to have slightly more active mineralization rates. Plot 11 appears to represent a slightly better quality soil than the other low productivity sites, based on its higher p H and lower C / N ratio. This is reflected to some extent in the vegetation which contains more herbaceous species, including several nutrient-medium to rich indicator species such as Lysichitum americanum, Maianthemum dilatatum, and Veratrum viride (Klinka et al. 1989). Plots 13, 18, and 20 tend to stand apart from the main body of data along the lower range of the second axis, a result of their higher C / N and lower p H values. Cluster analysis identified two groups at the highest level, one representing plots from site index classes 1 and 2, the other containing predominantly site index class 3 plots (Figure 21). A n important secondary division occurred in the first group, where five plots from SI class 1 were split into a separate cluster from remaining plots. A similar secondary division of the second main group identified plot 23 as an outlier. The cluster analysis has identified three principal clusters which very closely match the structure revealed in the principal components analysis ordination (Figure 20), including the single outlier. This "agreement" would suggest the structure uncovered in both analyses could be considered a reasonable reflection of actual structure present in the soils data. SI Class Plot Distance (scaled) 1 1 9 — 1 21—J 1 22 1 7 1 11 2 6 2 18 2 8 2 15 2 4 1 1 9 1 13 1 16 1 2 2 3 10 3 24 3 3 3 5 3 12 3 l — i 2 17— 2 20 3 14 3 23 Figure 21. Cluster analysis on the 14 soil variables 86 These three "natural" groups identified on the basis of soil properties reflect quite well the three site index classes. Only 17% of cases could be considered "misclassified" in relation to the site index classes. The exceptions are site index class 1 plots 9, 13, and 16 which are edaphically more closely related to site index class 2 because of their relatively lower V M C values, and plot 16's higher T N value. Site index class 2 plot 17 appears more closely related to class 3 because of its relatively deeper rooting depth, and higher T N content. Plot 23 stands out from site index class 3 to which it is most closely related because of its high MINN content and high mineral soil p H . One means of checking how useful natural clusters are is to assess whether they convey information about other related variables which were not used in the analysis. The Tukey H S D multiple comparison test was used to compare all possible pairs of the three soil-derived clusters in relation to the complete set of 14 soil variables as well as CWSI, HWSI, PLSI, SSSI, MAI , SAND, SILT, C L A Y , H U M D E P , and S L O P E (Table 17). Al l three clusters differ significantly in relation to MAI, while clusters 1 and 3, and 2 and 3 differ in CWSI, HWSI, PLSI, and SSSI. As expected, clusters show an increasing trend in productivity from 1 through to 3. V M C was important in differentiating cluster 1 from 2 and 3, with cluster 1 having more limited aeration. Clusters 2 and 3 were similar in this property. AIR only differed significantly between clusters 1 and 3. Rooting depth (D2ROOTS) differed between clusters 1 and 3, and 2 and 3, but not 1 and 2. DH20S was important only between the extreme clusters (1 and 3), although the trend was one of greater depths with increasing 87 productivity. Cluster 1 had significantly less sand in the fine fraction of upper mineral soil than the other clusters, while cluster 3 had significantly greater slope than cluster 1. Table 17. Means and standard deviations of soil and productivity variables for soil-derived clusters. S o i l - D e r i v e d C l u s t e r V a r i a b l e 1 2 3 n 5 10 9 Mean SD Mean SD Mean S MAI (nfaa^yr1) 4.9al 0.9 7.5b 2.2 12. 5C 2.0 CWSI (m/50yrs) 14. 4a 0.8 16. 4a 2.6 21. 8b 1.9 HWSI (m/50yrs) 12. 2a 1.0 15. 2a 3.1 21. lb 3.5 PLSI (m/50yrs) 13. 6a 1.5 17. 0a 3.3 21. 8b 1.9 SSSI (m/50yrs) 10. 3a 1.8 13. 7a 4.8 24. 5b 5.1 VMC (%/l00) .79a .05 • 58b . 08 . 54b .08 AIR (%/100) • 09a .02 • 14ab .04 .16b .07 DH20S (%/100) 31a 8 50ab 27 71b 31 D2ROOTS (%/100) 25a 6 32a 5 41b 5 TC (kg/ha) 108253a29522 151495b21510 188593c39492 TN (kg/ha) 2234a 572 2849a 449 4718b 1273 MINN (kg/ha) 40a 10 55a 15 95b 50 P (kg/ha) 9a 3 16b 5 14ab 6 CA (kg/ha) 272a 71 377a 72 527b 104 MG (kg/ha) 174a 38 307b 75 451c 94 K (kg/ha) 99a 13 176b 30 204b 34 HUMPH 3.3a .1 3 .2a . 1 3.4a .2 HUMCN 51. 5a 1.7 59. 6b 5.3 49.9a 7.1 MINPH 3.4a .2 3 . 3a . 1 3 . 6a .2 MINCN 45. 5a 6.2 47. 9a 9.6 35.8b 5.6 SAND 16. 9a 6.6 3 2 . 8° 10.5 30.6ab13.5 SILT 48. la 12.6 48. 7a 7.4 46.3a 9.7 CLAY 29.7a 9.0 18. 6b 4.0 23.1ab 5.2 HUMDEP 12a 2 16a 2 15a 5 SLOPE2 0a 0 2a 2 5ab 4 1 classes not followed by a common letter are significantly different using the Tukey H S D test (Pj<0.05). See Appendix N for procedure used to check test assumptions 2 class differences using to a non-parametric analogue of the Tukey H S D test (Zar 1984) 88 In relation to soil chemical properties, all three clusters differed in TC and MG, with a strong increasing trend from cluster 1 through to 3. A similar trend occurred for other nutrients, with TN, MINN, and CA showing a strong difference between clusters 2 and 3, and 1 and 3. P was significantly lower in cluster 1 than the others. HUMPH and MINPH did not differ significantly amongst clusters. MINCN is significantly lower in cluster 3, than 1 and 2, while HUMCN is highest in cluster 2. In general, the three clusters identified in the soil data reflect distinctly different forest productivities. The first and least productive cluster appears particularly limited in soil aeration, as reflected in moisture content and air-filled porosity. Base content is important in differentiating all three clusters, and nitrogen differences are particularly strong between the last two clusters. The overall trend is one of greater nutrient reserves in more productive clusters. Depth of rooting is an important differentiating property, but particularly so between clusters 2 and 3 where soil aeration is not as strongly limiting as in the first cluster. Properties of the soil-derived clusters did not differ substantially from the site index classes because of the similarity in their component plots (see section 5.3.1). In general, productivity indices were more variable within soil-derived clusters than site index classes, as expected. The exception was the first soil-derived cluster which was smaller and more uniform than its counterpart site index class. Variability of soil physical and chemical properties was slightly lower overall within soil-derived clusters since these properties were the basis for stratification. There were some minor changes in properties which significantly 89 differed between soil-derived clusters compared to site index classes. In the latter, all productivity indices but only one soil chemical property (MG) differed between all classes because of the relative shifts in within-group variability comparing the two classifications. D2ROOTS differed significantly between site index classes 1 and 2, but not 2 and 3, whereas clusters 2 and 3 differed, but not 1 and 2. This reflects the inclusion of site index class 1 plots into cluster 2. Aside from some minor differences in pH and C / N ratios, the two classifications were otherwise quite similar. 5.4.2 Vegetation An objective examination of natural structure in the vegetation data was undertaken using reciprocal averaging (RA). R A ("correspondence analysis") may be considered a doubly standardized, non-centered eigenvector analysis (Noy-Meir and Whittaker 1978). The algorithm iteratively calculates sample scores from the weighted average of species scores and vice versa, beginning with species which are weighted along a rough initial gradient, according to percent cover. This continues until a stable solution is reached (Gauch et al. 1977, Pielou 1984). Although similar to principal components analysis, R A differs in its double standardization of the initial data matrix, and its use of chi-square rather than covariance or correlation matrices (Noy-Meir and Whittaker 1978). R A has been shown by many authors to be a superior technique to principal components analysis for indirect ordination of vegetation data (Hill 1973, Gauch et al 1977, and Whittaker 1978). It is particularly good at extracting the first major axis of vegetation and environment variation from a data set (Whittaker 1978). 90 The vegetation data matrix was trimmed of rare species (species present in fewer than 5% of sample plots) and submitted to reciprocal averaging analysis in the O R D I F L E X statistical package (Gauch 1977). Sample and species scores for the first two R A axes were used to plot samples in species space. The initial analysis using all plots was strongly affected by a single outlier (# 11). This plot had atypical cover of nutrient-rich indicator species (see Section 5.3.1). Since outliers cause severe distortion in the ordination, it was removed and the R A analysis repeated (Whittaker and Gauch 1978). The first three R A axes resulting from this analysis accounted for 58% of the variance in the data (Table 18). The proportion of variance extracted drops off substantially in subsequent axes. Most of the variance is accounted for by the first axis. Table 18. Variance extracted by R A axes 1-3 . Axis Eigenvalue % Variance 1 Cumulative % variance 1 0.29 32.2 32 . 2 2 0.13 13.9 46.1 3 0.11 11.9 58.0 1 percent variance explained by an axis is the proportion of its eigenvalue in relation to the sum of eigenvalues for all axes (0.90 in this case). A species by sample matrix with both species and samples ranked by the first R A axis scores reveals much information about the floristic structure in the data (Table 19). The most obvious feature of this matrix is that all but one plot from site index class 1 (plot # 13) are ranked at the low end of the axis, where they are characterized by higher coverage of Hylocomium splendens, Rhytidiadelphus loreus, the nearly unique occurrence oi Sphagnum girgensohnii, Gaultheria shallon, Blechnum spicant, and Menziesii ferruginia, and lower coverage of Kindbergia oregana and to a lesser extent, Plagiothecium undulatum. Table 19. Species by sample matrix, ranked according to the first R A axis scores. Matrix values are deciles, scaled by the matrix maximum (code 9 = 45% cover) S i t e index: 1 1 1 1 1 1 3 2 2 2 3 2 3 1 3 2 2 3 3 2 3 2 3 clas s Plot number: 1 1 2 0 0 2 2 1 0 0 0 1 0 1 0 0 0 1 2 1 1 2 1 LYSIAME 1 9 6 2 9 7 1 3 5 4 6 5 8 1 3 3 2 8 2 4 7 4 0 0 + SPHAGIR 1 + + - + GAULSHA - + + + STREAMP + + BAZZDEN - + + - + BLECSPI + + + + + + + - - - - - - - + - - - - - - - -MENZFER + + + + -HYLOSPL 5 7 5 7 5 4 2 3 1 2 2 1 + 1 + 1 1 1 1 + + 1 -RHYTLOR 5 5 3 4 3 2 3 1 2 1 1 1 3 1 2 1 + 1 + + + - -MALUFUS + + MONEUNI + PLAGPOR + - + - + + - - + + + - + - + + - - + - - - -VACCPAR + + + + + - + + - - - + + + + + + + - + + + + HYPNCIR + + + - + - - + - - - + - - - - - - - + + - + DICRFUS - - - + - - - - - - - - - - + - - - - + + - -LISTCOR + - - + + + - - - + + + - + + - + + - - - + -HOOKLUC - - - - - + 1 - + - - - + - + - - + - - - - -SCAPBOL + 1 1 1 + 1 + 1 1 2 1 1 + 1 1 1 1 + + 1 + 1 + LEPIREP + + + - + + + + + + + + - - - - + - + + + - + BLEPTRI - - + - - - - - + + + + DICRSCO - - - - - - + - + + - - - - - + - + - - - - -RHIZGLA + - + 1 + + 1 + 3 1 1 + 3 1 2 2 + 1 1 + 1 1 + CALYMUE - - - - + + - - + + + + + - - + - - - + - - + LOPHINC - - - - - - - - - - - + - - - - + - - - - - -KINDORE 1 1 1 - - 3 3 6 3 4 5 5 5 5 5 5 6 5 9 9 + 5 + GYROUND - - - - - - - - - - - - + - - + - - - - - - -LISTCAU - + - + - - - - - - - + - - - - - - - + •- + -PLAGUND - + - - 1 1 1 1 1 1 1 1 + 1 + + + 1 1 1 3 3 2 LOBAORE - - - - - - - - - + - - - - + - - + - + - - -MAIADIL + + - - + - - - -CALYTRI - - - - - - - - + - + - - + - - - - + - + + -CEPHLUN - - - - - - - - - - - - + — - + - - - - - + + 1 see Appendix A for species names. 92 There does not appear to be a distinct floristic trend among site index class 2 and 3 plots. A n ordination of samples in relation to the first two R A axes (Figure 22) shows the general structure in the data. 100 90 80 A 1 4 70 -60 -50 40 30 20 10 • 7 • O h 9 • 21 15 o 5 • 13 A g 1 O 2 0 o 17 A 2 4 23 20 I 40 60 8 0 100 SI C lass 1 Axis 1 SI C lass 2 SI C lass 3 Figure 22. Ordination of samples in relation to R A axes 1 and 2 (plot 11 deleted). The differentiation of the majority of site index class 1 plots from the rest of the data is evident. The outlier plot which was excluded from the analysis (# 11) is closely associated 93 with site index class 1 based on other characteristics. Plot 21 has floristic affinities to the growth class 2 and 3 plots due to its lower and higher cover of Hylocomium splendens and Kindbergia oregana respectively (Table 19). Plots 10, 14, 20, 17, and 24 stand out somewhat from the bulk of the data because of the high cover of Plagiothecium undulatum in the first three, and very high cover of Kindbergia oregana in the latter two. They are separated from each other due to this difference. Two anomalies stand out. Plot 13 is floristically similar to the bulk of site index class 2 and 3 plots; yet it has distinctly poorer productivity and soil properties. O n the other hand, plot 23 appears floristically related to the poorer sites with its relatively high cover of Hylocomium splendens and Rhytidiadelphus loreus, yet it has the highest productivity and the most favourable soil properties of all plots. In these cases, floristic features do not appear to reflect site properties. In general, the patterns revealed in the R A analysis were similar to those discussed in Section 5.3.1. Site index class 1 plots appear to stand out as having the most distinct floristic characteristics which reflect the wetter soils of these sites. Site index class 2 and 3 plots feature greater cover oi Kindbergia oregana but cannot be easily differentiated floristically. 94 5.5 Functional Relationships Between Site Index and Ecological Properties In the previous sections, examination of relationships between redcedar site index and ecological properties has focussed on data structure and classes of sample plots defined by site index. In order to better understand the underlying ecological process influencing tree growth on these sites, the functional relationships between site index and selected ecological variables were investigated using multiple linear regression. The principle of model selection was to choose the smallest number of meaningful independent variables that explains the most substantial part of the variation in redcedar site index (Chatterjee and Price 1977). The procedure for model selection followed the methods for evaluating regression models outlined by Chatterjee and Price (1977) and Zar (1984). Initially, a model using all selected variables in the particular example is fit to the data. The adequacy of fit is assessed using the coefficient of multiple determination (R 2) arid the residual mean square (RMS). The model assumptions are then evaluated for serious violations using graphical examination of residuals. A plot of residuals which shows a random scatter is indicative of a well specified model (Chatterjee and Price 1977). If serious violations are evident, variable transformations may be undertaken to improve the model. The partial regression coefficients are then evaluated to identify non-essential variables. Non-essential variables are those which are not significantly different from 0 using a student's t test (P<0.05). Including these variables in a model leads to a loss of precision in estimation and prediction (Chatterjee and Price 1977). Non-essential variables are dropped and the model refit to the 95 data, with the new model being assessed as described above. Since all models evaluated had relatively few predictor variables, all combinations of meaningful variables were evaluated for a given example. 5.5.1 Site index/soil relationships Soil chemical and physical properties Figures 23 to 25 show scatterplots for the initial set of 14 soil variables. A l l variables show positive linear relationships with CWSI except V M C , H U M C N , and M I N C N which are negatively related. T N displays a potentially curvilinear relationship which is addressed in the following discussion on model selection. The influence of each case on the correlation between a variable and CWSI was assessed using influence plots (Wilkinson 1988). Al l variables except MINPH, MINCN, H U M C N , and H U M P H showed no single case(s) having undue influence on the relationship. One plot (#23) stood out from the rest in terms of MINN content yet it did not influence the correlation coefficient. This plot did however have undue influence on the p H and C / N ratio of the forest floor and mineral soil, being atypically high and low in p H and C / N ratio, respectively. These atypical properties of plot 23 are consistent with its humus form which was the only Moder encountered in the study area, and which reflected greater biological activity and nutrient mineralization rates. A n additional plot (#11) had an atypically high M I N P H for its CWSI value, which influenced the correlation. This was also reflected in its atypical vegetation TN (kg/ha) (Thousandt) r-0.78 * * # * * *** * * * * IS 20 CWSI (m/50yrs) 250 200 h MINN (kg/ha) 15 20 CWSI (m/50yrs) 300 250 150 100 TC (kg/ha) (Thousands) 15 20 CWSI (m/50yrs) 25 MS (kg/ha) r-0.61 * * * * - * * * * * * * * * * # * * * * * * 15 20 CWSI (m/50yrs) CA (kg/ha) 600 400 * r-0.70 * * * * * * * * * * * * * ** * * * * * < 15 20 CWSI (m/50yrs) 25 K (kg/ha) 250 h 200 I I ! i 15 20 CWSI (m/50yrs) Figure 23. Relationships between soil nutrient contents and Cw site index. 97 3.8 3.6 HUMPH !| HUMCN 3.2 70 r=.48 ** * i: 80 * * * # * r=-.46 * * * * * * J 50 * * * * * - * * * * ** * * * 40 * * l 1 i i i 30 1 * 1+ 16 18 20 22 CWSI (m/50yrs) 24 26 10 12 14 16 18 20 22 24 26 CWSI (m/50yrs) r=.57 * * * * * * *** * * * * ** * * * * * * * * • 1 14 16 18 20 22 CWSI (m/50yrs) 30 r=-.54 * * * * ** * * * * * * * * *** * * * * * 24 26 ..._jL 14 16 18 20 CWSI (m/50yrs) 22 24 26 Figure 24. Relationships between p H and C / N ratios of the forest floor and mineral soil, and Cw site index. 98 VMC (%) BO -* * * * * r=-.63 * * * * -* * ¥ ** * * * * * r 1 i 30 25 20 16 10 5 AIR(%) r-0.51 * * * * * * * - * * * * * * ** * * * * * 14 16 18 20 22 CWSI (m/50yrs) 24 26 10 16 20 CWSI (m/50yrs) 25 D2R00TS (cm) DH20S (cm) 10 12 14 16 18 20 22 CWSI (m/50yrs) * r=.56 * -* * * * * * -* * * * * * * * * * i i i 10 12 14 16 18 20 CWSI (m/50yrs) 22 24 26 Figure 25. Relationships between soil physical properties and Cw site index. 99 which contained several nutrient-rich indicator species not present in other plots with similar site index values (see section 5.3.1). Because of the influence of these cases, a certain amount of caution must be used in interpreting the correlations, although the degree of influence was low (10 - 19%) in all cases. Only the four key variables (TN, D2ROOTS, MINPH, V M C ) were selected to represent soil properties in the regression analysis because of the problems associated with multicollinearity. A number of combinations of these independent variables were fit to the data. Three models were identified in which all coefficients were significant - CWSI as a function of 1) V M C and T N ; 2) V M C and D2ROOTS; and 3) D 2 R O O T S and T N . The first model was the most promising based on the above assessments. A single outlier (#09) was identified based on its high studentized residual (Wilkinson 1988). This plot was deleted and the model refit to the remaining 23 plots. The new model showed an improved fit with very minor change in the coefficients, indicating the model is stable and not sensitive to the one outlier. Plot 09 represents a 150 year old stand on outwash sand in the northern study area. It may be considered atypical from the bulk of the population because it has the lowest CWSI encountered, it is in the northern extremity of the study area, and it is the only plot on this parent material. On this basis, its deletion from the data for purposes of this model may be justified. A n evaluation of residuals from this model revealed no distinct pattern, indicating serious departures from the assumptions did not occur. 100 Two logarithmic transformations of T N were tested to determine if a real curvilinear relationship existed between T N and CWSI. No obvious changes occurred in the fit or form of residuals using the transformed variables, therefore, T N was left in its original form in the final model. The form of the final model is: [1] CWSI = 19.47 - 11.02VMC + 0.002TN R2=0.780 Se4= 1.69 m where V M C is volumetric moisture content expressed as a proportion, and T N is total N content in the forest floor plus root zone mineral soil expressed as kg/ha. Model [1] effectively summarizes the important soil properties associated with redcedar height growth, accounting for 78% of the variation in CWSI. T N is representative of the total nutrient reserve potentially available to trees. It is strongly correlated with most other nutrients, particulary MINN and M G , and incorporates soil physical properties (bulk density) and rooting depth which influence soil nutrient content. While other nutrients show a similar trend in relation to CWSI, T N is the best overall expression, with every 1000 kg/ha increase in T N associated with a 2 m increase in site index (with V M C remaining constant). V M C on the other hand, reflects the soil aeration status in the mineral soil rooting zone. It affects productivity in a negative manner, with each 11% increase in moisture content associated with a lm drop in site index (with T N remaining constant). In essence, this model indicates that the quantity of root zone nutrient reserve and root zone quality in terms of aeration, are most important to forest productivity in the study area. 4 standard error of the estimate 101 The two other models fitted to the the same data set are: [2] CWSI = 18.247 - 11.284VMC + 0.202D2ROOTS R 2 = 0.600 Se=2.26 m where D 2 R O O T S is the secondary rooting depth measured from the ground surface, expressed in cm. [3] CWSI = 8.097 + 0.002TN + 0.147D2ROOTS R 2 = 0.740 Se = 1.84 m A n examination of residuals indicated that no serious departures from the assumptions occurred in either model. These models essentially reflect the importance of rooting depth to variation in Cw site index. Model [2] emphasizes rooting depth and aeration status in the root zone, as reflected by moisture content. Model [3], which explains nearly as much variation in CWSI as model [1], emphasizes the combined importance of rooting depth and soil nutrient content. While T N content partly reflects the available rooting depth, these two variables were not strongly correlated and therefore, do not create multicollinearity problems in the model. 102 Field-recognizable properties The above relationship provides insight into the major processes influencing site index; however, it has limited utility for site index prediction because of difficulties in measuring the independent variables. Therefore, regression analysis was applied to a set of variables which can be readily identified in the field. Because field description of forest ecosystems frequently requires qualitative assessment, a mix of both continuous and categorical variables was initially considered for analysis (Table 20). Table 20. Continuous and categorical variables which are readily described in the field. Symbol Property D 2 R O O T S secondary rooting depth from ground surface (cm) D H 2 0 S depth to free water from ground surface (cm) S L O P E slope (%) H 2 0 S T free water status (absent, slow, moderate, rapid) R O O T rooting density in the soil root zone (very few, few, plentiful, abundant) T E X P M field texture of parent material (C, SC, C L , SCL, SIL, SIL+, SL, LS 1 ) CONS consistence of root zone soil (soft, firm-friable) C E M presence of cemented horizon (present, absent) H U E hue of root zone soil (Munsell colour notation) C H R O M A chroma of mineral soil (Munsell colour notation) 103 Table 20. Continued. Symbol Property S H A P E surface shape (depression, level, sloping, convex, receiving) P M parent material (alluvial, fluvial veneer over morainal, lacustrine veneer over morainal, morainal, glaciofluvial) 1 texture codes from CSSC (1987). SIL+ denotes a high clay content within the range for silt loam. The categorical variables were initially screened to identify those which were most promising in relation to CWSI. This was done by examining plot frequencies according to site index class and categorical variable class using contingency tables. Only H20ST, CONS, R O O T 1 , and S H A P E , showed any apparent association with site index class (Table 21). Most site index class 1 plots had free water moving rapidly or moderately into the soil pits, with no plots lacking free water. On the other hand, none of the class 2 or 3 plots featured rapid input of free water. Several plots from these classes showing moderate flow of free water were located in the Lawn Hill area where seepage water seemed to be more prevalent, particularly in class 3 sites. The majority of class 1 plots featured a soft, buttery-like consistence in the upper mineral soil because of prolonged saturation. A l l class 2 and 3 plots had upper mineral soil with a firm or friable consistence and subangular blocky structure. Most class 2 and 3 plots had abundant or plentiful root density within the root zone, while only one class 1 plot had abundant roots, reflecting the better soil aeration conditions of the more productive sites. Site index class 1 plots were predominantly located in depressional terrain which contributes to their stagnant soil water regimes. Due to the subdued topography characteristic of the study area, these surface shapes are rather subtle. 104 However, since slowly permeable soil layers tend to be close to the surface, a relatively minor change in surface topography can impact the available aerated root zone Table 21. Contingency tables for selected categorical variables showing frequency by site index class Variable Category S i t e 1 n 8 Index 2 8 Class 3 8 H20ST absent 0 3 3 slow 1 3 2 moderate 4 2 3 rapid 3 0 0 CONS sof t 6 0 0 f i r m - f r i a b l e 2 8 8 ROOT very few 1 0 0 few 3 0 0 p l e n t i f u l 3 3 3 abundant 1 5 5 SHAPE depression 7 0 1 l e v e l 1 2 1 sloping 0 3 2 convex 0 3 0 receiving 0 0 4 sufficiently to affect tree growth. Many site index class 1 sites occurred in subdued depressions only a few decimeters lower than adjacent sites with 3 - 4 m greater site index (Cw). Boggie (1972) also found that minor variation in microtopography strongly influenced lodgepole pine growth on waterlogged soils. While parent material texture wasn't included, there was one interesting trend. Site index class 3 plots tended to have coarser textured soils (5 plots SL or coarser), while 88% of class 1 and 2 plots had S C L or heavier textures. This may reflect improved porosity, aeration, and permeability of the coarser underlying soils. 105 This set of continuous and categorical variables were fitted to a series of multiple regression models to determine the best expression of CWSI. Categorical variables were expressed as dummy variables where x-1 variables of 0 or 1 value were used to distinguish x categories of each variable. In this manner, dummy variables only influence the intercept in the regression (Chatterjee and Price 1977). Fifteen models incorporating various combinations of the above categorical variables plus D2ROOTS, DH20S, and S L O P E were evaluated using previously described procedures. The S H A P E variable consistently showed only the depression category having any significant influence therefore this variable was changed to D E P with two categories; depressional, and not depressional. The only variables which showed any significant influence in the various models were D2ROOTS, S L O P E , CONS,.and D E P . The most promising models based on these variables were: [4] CWSI = 11.589 + 0.214D2ROOTS - 3.212CONS R 2 = 0.61 Se=2.40 m where CONS takes a value of 1 for soft and 0 for friable-firm. [5] CWSI = 10.174 + 0.257D2ROOTS - 2.528DEP R 2 = 0.60 Se=2.41 m where D E P takes a value of 1 for depressional and 0 otherwise. [6] CWSI = 8.565 + 0.232D2ROOTS + 0.565SLOPE R 2 = 0.73 Se = 1.98 m . Models 4 and 5 are very similar in fit, and in what they represent, since plots which have soft soil consistence often are in depressional terrain. These models reduce CWSI by 106 3.2 m and 2.5 m, respectively, for sites with soft consistency and depressional terrain, reflecting the negative effect of high soil water content and reduced aeration on tree growth. The model which explains most of the variation in CWSI (73%) is that based on rooting depth and slope. Each of the categorical variables (CONS and DEP) were separately added to this model; however, neither contributed significantly when combined with D 2 R O O T S and S L O P E . One plot (#23) was identified as having high leverage in model 4 because of its relatively greater slope (12%). Based on a refitting of the model with plot 23 deleted, this case appears to have minimal effect on the model except for pulling up the S L O P E coefficient slightly and improving the fit slightly. It was decided not to delete this case as the relationship between tree growth and the steeper slopes it expresses is important in the study area. It appears that rather minor changes in slope can influence tree growth, likely a reflection of improved soil water movement and subsequent enhanced soil aeration. In effect, this model indicates that CWSI increases 2.3 m for every 10 cm increase in D2ROOTS, if S L O P E is held constant, while a 2% increase in S L O P E is associated with a i m increase in CWSI while holding D 2 R O O T S constant. This relationship has to be interpreted with some caution, however. Slopes were small and did not vary a great deal over the sample. There is some measurement error at such small slopes, which may influence the model and its application. Intuitively, steeper slopes should be associated with better tree growth because of improved soil drainage and aeration. There is some question as to whether the relationship described in the model is realistic for the whole study area. 107 5.5.2 Site index/vegetation relationships To examine functional relationships between CWSI and understory vegetation, the original data set containing 46 species had to be reduced to a manageable size for regression analysis, which was done in two ways. The first method used the three reciprocal averaging axes as new variables to summarize the information contained in the original data set. Each plot is characterized for each plot by 3 scores for the 3 axes, rather than by cover of individual species. This, in effect, is a linear transformation of the original species into 3 new variables. The second method used the five soil moisture and three soil nutrient indicator species groups to summarize the vegetation. Each plot is characterized by a frequency value for each of the 8 ISGs. The plot scores of the three R A axes were regressed against CWSI. Only the first R A axis (RAI) had significant coefficients in any of the models, therefore a simple linear model of CWSI as a function of R A I was used to represent the relationship. The form of the model was: [7] CWSI = 13.355 + 0.083RA1 R 2 = 0.44 Se=2.81 m where R A I is the plot's score along the first R A axis. A plot of this relationship is shown in Figure 26. Based on the results from Section 5.4.2, model 7 actually represents a relationship between two groups of plots - those with low site index dominated by Hylocomium splendens, Rhytidiadelphus loreus, and the remaining, floristically indistinguishable plots dominated by Kindbergia oregana. Although the regression represents 108 a significant relationship (P<0.05) it explains only 44% of the variation in CWSI, and does not really reflect a continuous linear function. CWSI (m/50yrs) ' 0 10 20 30 40 50 60 70 RA Axis 1 Score 80 90 100 Figure 26. Relationship between Cw site index and R A axis 1. Various combinations of the 8 ISG variables were regressed against CWSI and it was quickly apparent that only the nutrient-poor (NP), nutrient-medium (NM), and moisture-fresh to very moist (FVM) ISG's showed a significant (P<0.05) effect on CWSI. This effect was only expressed singularly, with no combinations of two or more variables having more than one significant coefficient. The best model expressed CWSI as a function of the N M group. The form of the model is: [8] CWSI = 14.806 + 19.363NM R2=0.31 Se=3.03 m J 109 where NM is the frequency of the nutrient-medium ISG present in a plot. A plot of this relationship is shown in Figure 27. Figure 27. Relationship between CWSI and the nutrient-medium ISG. As expected, CWSI increases with increasing frequency of nutrient-medium indicator species, which is consistent with the trend of improving soil nutrient status. Although the relationship is more continuous than that with RAI scores, it is weaker, explaining only 31% of the variation in CWSI. In neither case does vegetation show a strong functional relationship with CWSI. The composition and cover of understory vegetation present on a site depend to some extent on non-site features such as disturbance, overstory density and 110 composition, and chance. High stand density characteristic of these stands severely limits light reaching the understory, resulting in low floristic diversity, often dominated by shade-tolerant mosses. Overbrowsing by Sitka black-tailed deer (Odocoileus hemionus sitkensis) has also apparently altered the floristic structure of many ecosystems in the study area (Banner et al. 1987). Such "non-site" variation in understory vegetation development masks its ability to clearly reflect changes in site quality. 5.6 Soil Water Quality Soil water represents an important property of ecosystems, potentially influencing their function and productivity. In most sites in the study area, the presence of soil water near the surface at various times during the year has a potentially negative impact on site productivity. While factors such as the depth, timing, and duration of this excess soil water are important in determining its impact, so too is the quality of the water. Dissolved oxygen in soil water can prevent hypoxia if present in sufficient concentration. Moving soil water tends to be well mixed and contains more dissolved oxygen than stagnant water. As a result, "aeration-stress" in trees is offset and roots can still function within the saturated zone (Zaerr 1983, Kozlowski 1984, Pritchett and Fisher 1987). This is believed to be a major factor in determining productivity of Sitka spruce in soils affected by excess soil water (Blyth and McLeod 1981a, 1981b). I l l The dissolved chemical loading in soil water has an important influence on ecosystem productivity as it contributes to the available nutrient pool. Supplemental inputs of dissolved nutrients in seepage water have been associated with improved tree growth (Bourgeois 1969). 5.6.1 Specific conductivity Specific conductivity (electrical conductivity) is a numerical expression of the ability of a water sample to carry an electric current, which is dependent on the concentration of dissolved ions in the water and its temperature (Franson 1976). It is an index of the total loading of soluble salts and has been found to correlate well with the sum of cations ( C a + + , M g + + , N a + , K + ) and anions (HC0 3", S04", Cl") determined chemically (Richards 1954, Bourgeois 1969, Feller and Kimmins 1979). Specific conductivity values for the 18 measured plots are shown in Appendix I. Values averaged 0.151 dS/m and ranged from 0.081 to 0.312. The coefficient of variation across all plots was 39%. Stream water sampled from Lawn Creek which drains the southern portion of the study area had a conductivity of 0.160 dS/m which is similal to the study plots in the vicinity. From the limited data available on conductivity in south coastal ecosystems, the soil water in the study area appears to have above normal concentrations of dissolved ions. Data from several small streams in the Malcolm Knapp Research Forest in Haney showed average conductivities of approximately 0.020 dS/m (Feller and Kimmins 1979). Soil leachate from C horizons of tills in the southern part of the Research Forest had 112 similar conductivity values (Bourgeois 1969), as did soil seepage water sampled from mature stands throughout the research forest (Klinka et al. 1986). The supply of elements to soil water depends on a number of factors, including mineralogy of the parent material, weathering rate, organic matter decomposition, precipitation, and movement of water through the soil. However, the main source of several ions to the soil solution is the weathering of soil minerals (Richards 1954, Slavinski 1977). Thus the dissolved chemical load in soil water is a direct function of the soil material with which the water has been in contact (Walmsley 1974). The rate of water moving through the soil can also affect ion concentration with fast moving water creating a shorter equilibrium time of the soil-water system, thus lowering ion concentration (Dreibelbis and McGuinness 1957). Lower ion concentration in soil water can therefore be expected in Haney ecosystems because of the base-poor granitic bedrock from which parent materials are derived, and the relatively fast moving subsurface seepage associated with the characteristic steep slopes. In the east Graham Island lowlands, soils are derived primarily from volcanic and sedimentary lithology which yield a relatively greater quantity of bases upon weathering compared to grantic material. Also, because of the subdued topography, water accumulates and moves slowly if at all through the soil, maximizing the equilibrium of the soil-water system. The relationship between specific conductivity and tree growth (CWSI) was poor (r = 0.237). Apparently the concentration of dissolved ions in the soil water is not as important as the total soil nutrient reserve potentially available to plants. However, 113 variation in non-nutrient, or luxury-consumption nutrient ions may contribute to this poor correlation with tree growth. There was some evidence of a distributional pattern in specific conductance within the study area. Most (4 out of 6) plots from the Lawn Hill area had relatively high conductivities, averaging 0.200 dS/m. A n isolated area of Masset Formation bedrock outcrops in the Lawn Hill area and is characterized by base-rich basalt (Sutherland-Brown 1968). This may influence the mineralogical composition of the local surficial materials and in turn, soil water quality. Other plots between Lawn Hil l and Mayer Lake also tended to have high conductivities (above 0.140 dS/m), while all plots located north of Port Clements had "low" conductivities (below 0.110 dS/m). These differences may also reflect variation within the mineralogical composition of the parent materials which would have mixed lithologies depending on the source bedrock. Plots in the Miller Ck. area tended to have lower conductivities which may reflect the relatively base-poor granodiorite pluton that outcrops in this vicinity (Sutherland-Brown 1968). Specific conductivity values were correlated against all soil chemical variables (both content and concentrations), and plotted against all categorical soil and physiographic variables. Only Mg concentration in the forest floor showed a significant (P<0.05) although weak relationship to conductivity (r = 0.463). The likely reason for this correlation is uncertain. 114 5.6.2 Dissolved oxygen Dissolved oxygen (DO) values averaged 4.7 ppm and ranged from 3.2 to 8.4 ppm (Appendix I). The coefficient of variation across all measured plots was 30%. To put these in perspective, fast and slow moving stream water from Lawn Creek had values of 9.2 and 8.3 ppm respectively, reflecting the degree of aeration due to mixing and stirring of flowing water. Other researchers have recorded D O values of 10.2-10.6 ppm (Walmsley 1974), 10-11 ppm (Baker 1974), and 9-13 ppm (Feller and Kimmins 1979) in flowing streams. No comparative data are available on D O measurements of saturation zone soil water. While D O in the soil water was clearly lower than that of well-mixed flowing water, no relationship between tree growth and D O was evident. Also, no significant relationships were evident with any of the soil chemical properties (contents and concentrations), or the categorical soil and physiographic variables. This reflects the inherent variability of dissolved oxygen in soil water, or problems in the measurement technique. In the latter case, a number of factors affect precision and repeatability of measurements. One of the more significant potential problems is the disruption of the natural equilibrium by inserting the stand pipe into the saturation zone. By disturbing the water in the presence of air, diffusion of atmospheric oxygen into the sample may change the actual gas content (Walmsley 1974, Franson 1976). This is logistically difficult to overcome when sampling ground water, and generally requires complex techniques employing sealed sample extraction equipment (W.H. Black 1987, pers. comm.). Another factor potentially contributing to variability is that the volume of free water and the rate at which the stand pipe filled varied between sites. This 115 may affect the quality of the D O measurement, particularly in cases where smaller volumes of water were present in the stand pipe. Finally, point samples of dissolved water may not be meaningful in relation to forest productivity. The duration and timing of oxygen supply is important to tree growth. For example, the duration and magnitude of a growing season depression in D O may be a better indication of growing conditions (H. Schreier 1989, pers. comm.). Monitoring dissolved oxygen in soil water over a full year, particularly in relation to water table level, may reveal more physiologically meaningful trends in oxygen availability. One sample plot was of particular interest, however. Plot 03 showed the highest dissolved oxygen value recorded (8.4 ppm). The measurement was repeated for verification with the same result. The soil profile contained a 17 cm thick layer of porous gravelly loamy sand (30% coarse fragments), over a slowly permeable layer of clayey till. The sand layer was saturated with a perched water table above the dense till. As the site was on a uniformly gentle slope, the seepage water likely moves easily within the coarse layer because of its high hydraulic conductivity. This would result in higher dissolved oxygen content because of the mixing effect. This was reflected in the live roots present within this saturated layer. This site supported one of the more productive stands in the study area. 116 5.7 Soil Oxidation Zone Index Initial attempts to measure the oxidized zone of steel rods revealed that iron oxidation was manifested in different morphological features related to the degree of oxidation. A single zone of "rusted" steel was lacking on all but the wettest sites. Two methods of distinguishing zones of oxidation were therefore tested. The first method focussed on the presence of rust accumulations on the rod surface. Three zones of oxidation were recognized: zone 1 - many patches of bright orange to shiny red accumulations of oxidized material creating a rough surface; surface of metal shiny; soil material often adheres to accumulations. Assumed to reflect good aeration. zone 2 - patches of oxidized accumulations fewer, with dull dark reddish colours rather than bright orange or shiny red as in zone 1; surface of metal generally shiny but may be dull; soil material adheres to accumulations. Assumed to reflect moderate aeration. zone 3 - no reddish oxidized accumulations; patches of dark (black) discoloured material present but not accumulated to any appreciable thickness; surface of metal dull. Assumed to reflect poor aeration. Below zone 3 the metal surface is dull and free of any discolouration, due to the more-or-less complete lack of oxygen. The second method focussed on the presence of the shiny surface on the steel rod. Two zones were recognized: zone 1 - surface of metal bright and shiny; reddish oxidized accumulations often present. Assumed to reflect moderate to well aerated conditions. zone 2 - surface of metal dull; no reddish oxidized accumulations; patches of dark (black) discoloured material present but not accumulated to any appreciable thickness. Assumed to reflect poor aeration. (Like zone 3 above). 117 Measurements were made according to the two methods, with the depths recorded from the ground surface to the bottom of each zone. Using the first method, zone 1 was generally easy to distinguish, having a sharp lower boundary. The lower boundary of zone 3 was also easily recognized. The boundary between zone 2 and 3 was somewhat more difficult as it often represented a transition associated with decreasing oxygen supply. In the second method of measurement, the lower boundaries of both zones were relatively easy to distinguish. In 26 of the 144 rods measured, a lower boundary of the last zone using both methods was not reached within the depth of rod inserted in the ground. In these cases, the maximum depth reached by the rod was recorded. Data representing this lower zone must be interpreted with caution. Table 22 shows the Pearson correlation matrix of the depth of oxidized zones with CWSI and selected soil properties. Table 22. Pearson correlation matrix for oxidized zones and CWSI plus selected soil properties. Zone CWSI VMC AIR DH20S D2ROOTS DUFF Method 1: AZ11 0.355 -0.421 0.413 0.439 0.652 0.871 AZ2 0.581 -0.710 0.575 0.561 0.791 0.510 AZ3 0.563 -0.794 0.682 0.589 0.721 0.383 Method 2: BZ1 0.627 -0.833 0.808 0.829 0.708 0.350 BZ2 0.563 -0.793 0.683 0.589 0.721 0.384 1 A Z 1 , AZ2 , A Z 3 : method 1, zones 1 to 3 respectively BZ1, BZ2: method 2, zones 1 and 2. According to method 1 measurements, the depth to the bottom of all three zones was relatively weakly correlated with CWSI, with A Z 2 and A Z 3 showing the stronger relationships. Increasing site index is associated with increasing thickness of "oxidized" soil. 118 The first zone was highly correlated with the forest floor thickness measured where the rods were inserted. This was obvious in the field, and reflects the much greater porosity of the forest floor relative to mineral soil, and its closer proximity to the atmosphere. The greatest density of rooting also occurs within this well-aerated zone on all sites, reflecting the importance of good aeration to root development (Brady 1974). V M C was well correlated with both A Z 2 and AZ3 , while AIR was correlated with A Z 3 only. Greater moisture content and lower air-filled porosity associated with shallower depths of "oxidized" soil. The depth to the bottom of A Z 3 showed the strongest correlation because it usually had a very abrupt, shallow lower boundary on sites featuring saturated soils. A Z 2 and A Z 3 were also well correlated with D2ROOTS, showing that shallower rooting is associated with thinner zones of "oxidized" soil. In the second measurement method, the lower zone is the same as A Z 3 therefore relationships are identical (except for rounding). The upper zone (BZ1) generally shows stronger correlations with most properties than either A Z 1 or A Z 2 from the first method. It is also highly correlated with DH20S, indicating shallow water tables are associated with shallow zones of "oxidized" soil. Nearly 40% of the variation in CWSI is associated with BZ1. It thus appears that measuring the depth of a "shiny" metal surface on steel rods provides the most meaningful index of the thickness of "oxidized" soil. It shows the best correlation to tree growth, and to soil properties which are affected by, or which directly influence the depth of aerated soil. Finally, it is relatively easy to identify the lower boundary of this zone. The mean depths to the bottom of the oxidized zones for the three site index classes are shown in Table 23. 119 Table 23. Means (and standard deviations) of oxidized zones on steel rods for site index classes. S i t e Index Class Zone 1 2 3 n 8 8 8 AZ1 10al 14a 14a (2) (2) (5) AZ2 22a 33b 38b (6) (6) (6) AZ3 31a 45b 50b (9) (9) (6) BZ1 20a 33b 38b (6) (10) (10) BZ2 31a 45b 50b (9) (9) (6) 1 classes not followed by a common letter are significantly different using the Tukey H S D test (P<0.05) Site index classes 1 and 2 differ significantly for all measured oxidized zones except A Z 1 , while classes 2 and 3 do not differ. The class 1 and 2 differences reflect the abrupt change to non-oxidized steel at shallow depths in the class 1 sites featuring saturated surface soils. Soil properties which influence this, V M C and DH20S, also differ significantly between class 1 and 2 but not between classes 2 and 3 (Table 11). While class 2 and 3 do not differ significantly, there is a trend of increasing thickness of oxidized soil from class 2 to 3. The mean depths of BZ1, considered the most useful of the measurements, are similar but slightly less than the mean rooting depths (D2ROOTS) for the site index classes. This would suggest that this simple measurement is a reasonable index of the zone of soil receiving sufficient aeration for root development. 120 5.6 A Simplified Model for Field Recognition of Site Index Classes While the model previously described in Sec. 5.5.1 and based on field-recognizable properties showed a reasonably good relationship to CWSI, there were some potential limitations in its application associated with the S L O P E variable. In light of this, an attempt was made to synthesize all relevant, field-recognizable information revealed in the analysis into a simple system for characterizing and differentiating the three site index classes. This would provide field staff with a method for estimating the expected range in productivity indices for a given site based on important and easily measured or described ecological features. This system is summarized in Figure 28. SITE INDEX CLASS Cw SITE INDEX (m/50yr8) MAI (m^ha/yr gross standing volume) 13-15 1 4.2-5.9 1 17-19 21-23 7.3-10.3 10.9-14.3 SURFACE SHAPE PARENT MATERIAL SOIL WATER STATUS SOIL CONSISTENCE SLOPE ROOTING DEPTH (ground surface) SOIL WATER DEPTH (ground surface) VEGETATION ROOTING DEPTH 2 (mineral surface) SOIL WATER DEPTH 2 (mineral surface) PARENT MATERIAL COARSE FRAGEMENTS PARENT MATERIAL TEXTURE depression (level) organic rapid soft <1% < 30cm < 35cm Hylocomium splendens Rhytidiadelphus loreus Blechnum Bpicant < 16cm < 23cm convex, sloping (level) receiving, sloping (level) clay alluvial morainal absent or slow moderate firm/friable 1-5% 3-9+% 30-38cm 36-48+cm > 35cm Kindbergia oregana 14-24cm 21-30+cm > 23cm 0-35% > 35% clay loam to loamy sand 195% confidence interval potentially applicable in old-growth stands Figure 28. Summary of field-recognizable ecological properties in relation to site index classes 122 A number of site properties can interact in a variety of ways to affect the principal factors influencing variation in productivity - the depth and quality of the aerated root zone. Thus the best way to apply this information in identifying a site index class is to describe the appropriate site properties, assess where these fit on the summary figure, and determine which class is most strongly reflected. Class 1 is easy to identify because of the persistent water table, saturated upper soil, and very shallow rooting. Sites from class 1 occur commonly throughout the lowlands, often grading into adjacent poorer bog forests. Classes 2 and 3 are more difficult to differentiate, and emphasis should be placed on rooting depth, slope, surface shape, and to some extent coarse fragment content in the parent material. However, it should be noted that sites from class 3 were relatively uncommon in the lowlands, occurring mainly in the Lawn Creek and Miller Creek areas. Even in Lawn Creek, they were fairly localized. Because of the fluctuations in water tables in relation to high rainfall periods, properties relating to soil free water characteristics are most relevant during the mid-late summer when the water tables have dropped and more-or-less stabilized. The data were collected from June 24 to July 17, 1987, a period when precipitation was infrequent. Rooting depth and depth to soil free water should be measured in several soil pits in an area and averaged. There is considerable variation in these properties on many sites. To apply this information in identifying sites supporting old-growth stands, some adjustments must be made. Understory species composition is substantially different from second-growth stands and the forest floor is usually thicker, thus affecting total depths 123 measured from the ground surface. Rooting depth and free water depth measured from the mineral soil surface should be used. These properties showed slightly weaker relationship to Cw site index and therefore show more overlap between productivity classes in the summary. The vegetation characteristics shown in the summary for second growth stands do not apply in old-growth. The relationships shown in the summary have not been tested in old-growth stands aside from cursory observations, and should be the subject of future research. Terms used in Figure 28 are explained in the following: S U R F A C E S H A P E : the general topographic expression of the site in question. Because of the subdued topography characteristic of the lowlands, surface shapes tend to be rather subtle, yet they can have a significant impact on productivity. Since water tables tend to be close to the surface, minor changes in surface topography can result in enough change in available aerated root zone to affect tree growth. Depression: concave, depressional topography relative to surrounding area. While generally localized, depressional terrain can also be on a somewhat broader scale. It is associated with poor drainage and persistently high water tables. Level: flat topography. Sloping: sloping topography. In the lowlands, slopes are typically less than 10% except for localized stronger slopes (eg. gulleys, creeks, etc.), and the more sloping terrain rising to the Skidegate Plateau in the south (eg. Miller Ck., Lawn Hill). Convex: somewhat "elevated" convex topography relative to the surrounding area. Again this tends to be a subtle expression, often representing less that 1 m elevation difference from an adjacent depressional area. Receiving: generally concave shaped, lower positions of gentle slopes which receive additional inputs of seepage water. Often the moving water is better oxygenated than stagnant soil water resulting in better productivity. Such sites were found at Lawn Hill and Miller Cr. 124 P A R E N T M A T E R I A L : the unweathered material from which soils are derived (below the upper "developed" horizons). Organic: parent material comprised of organic sediments. These are poorly to well decomposed "peat" or "muck" type materials that are saturated with water. Typically these are more than 40cm thick below the forest floor but thinner layers may be considered under this category. Organic soils were not sampled in the study but were observed under low productivity stands during reconnaissance. Alluvial: older, usually inactive post-glacial alluvial sediments associated with creeks and rivers. Morainal: glacial till, generally fine-textured and dense in lowland ecosystems. P A R E N T M A T E R I A L C O A R S E F R A G M E N T S : estimated volumetric content of gravel, cobbles and stones in the parent material. SOIL W A T E R STATUS: characteristics of "free water" present in the soil profile. Absent: none visible in the soil pit. Slow: very slow seepage into pit; no appreciable standing water. Moderate: seepage water moving slowly but steadily into pit; standing water present but does not accumulate quickly to appreciable depth. Rapid: seepage water moving rapidly into pit, accumulating quickly to appreciable depth (bailing bucket required). SOIL C O N S I S T E N C E : the apparent consistency or structure of soil in the root zone. Soft: a soft, "squishy" consistency similar to soft butter. Firm-friable: soil is firm and/or friable; appears to have some structure. R O O T I N G D E P T H (ground surface): depth from the ground surface to the bottom of the zone of visible living roots. SOIL W A T E R D E P T H (ground surface): depth from the ground surface to the presence of free water in the soil profile. 125 V E G E T A T I O N : predominant understory species, the principal differentiating ones being Hylocomium splendens, Kindbergia oregana, Rhytidiadelphus loreus, and Blechnum spicant. R O O T I N G D E P T H A N D SOIL W A T E R D E P T H (mineral surface): as above but measured from the surface of the mineral soil (below forest floor). 126 5.9 General Discussion Several approaches have been used to examine the data, from the perspective of site index-defined classes, classes recognized in the data structure itself, to functional relationships among various properties. In all the analyses, essentially the same key features were revealed regarding the important ecological variables associated with variation in tree growth. Site index of redcedar is associated with the quantity and to some extent, availability of root zone nutrient reserves, the aeration status of the available rooting zone, and the effective rooting depth. This conforms to the majority of previous site/productivity studies where forest productivity is largely correlated with soil properties that reflect the status of soil moisture, soil nutrients, and soil aeration (given a uniform climate) (Ralston 1964, Broadfoot 1969, Carmean 1975, Pritchett and Fisher 1987). However, the precise nature of these relationships varies with the tree species, soil, physiography, climate, and region under investigation. 5.9.1 Soil nutrient status In this study, soil nutrient content, expressed on a kg/ha basis, particularly total N and exchangeable Mg, was found to be strongly associated with tree growth. Similar results have been observed in other studies relating soil properties to tree growth. For example, total N in the forest floor was found to have the highest and most consistent correlation of all site variables with growth of Sitka spruce on a range of soils in northeast Scotland (Blyth and MacLeod 1981a, 1981b). Because rooting was confined mainly to the forest floor in 127 these sites, N content of the mineral soil did not contribute much to the relationship. In contrast to the Graham Island sites, extractable (by acetic acid) Mg showed a significant although weak negative correlation with Ss growth in these Scottish soils. It was believed this reflected the influence of drainage on growth as Glentworth and Muir (1963, as cited by Blyth and MacLeod 1981a) found that exchangeable Mg increased in the subsoil as drainage deteriorates. Total N and exchangeable Mg content in the mineral soil (Kabzems 1985) and mineral soil plus forest floor (Roy 1984) were found to correlate strongly with Douglas-fir site index in ecosystems of south coastal B.C. In both these studies, however, mineralizeable N accounted for the greatest variation in Fd site index. This contrasts with the Graham Island study area where mineralizable N showed a positive but less pronounced trend with Cw site index. Most of these sites featured well-developed, acid Mor humus forms which are known to be associated with slow decomposition and mineralization rates (Wilde 1958, Klinka et al. 1981). Given these conditions, a more subdued trend with increasing site quality can be expected in comparison to the Douglas-fir ecosystems described by Roy (1984) and Kabzems (1985) where humus forms varied through Mors, Moders, to Mulls from low to high productivity, respectively. In fact, the one plot featuring a Moder humus form in the Graham Island sample also had the highest level of mineralizable N. Mineral soil and forest floor p H and C / N ratios showed an increasing and decreasing trend respectively in relation to tree growth. These trends, particularly the latter, are generally associated with improving mineralization rates (Powers 1980, Pritchett and Fisher 128 1987) and do in fact coincide with increasing mineralizable N content in the study plots. However, C / N ratios are high and p H levels very acid in most plots, indicating a relatively low availability of N in these soils (Heilman 1979). This again explains the more subdued relationship between mineralizable N and tree growth relative to Douglas-fir ecosystems. A n interesting feature of the Graham Island sites was the lack of any relationship between phosphorus and tree growth. In Britain, total P is one of the most frequent nutrients found important for tree growth (White 1982). In northeast Scotland, total P content in the forest floor was consistently correlated with Sitka spruce growth (Blyth and MacLeod 1981a). Phosphorus deficiencies, manifested as retarded growth, are widespread in conifer plantations on poorly drained soils. Reactions between P and soluble Fe, A l , and Mn in strongly acid and poorly aerated soils form hydroxyphosphates which render P insoluble and unavailable to plants (Pritchett and Fisher 1987).This is particularly true of soils featuring large amounts of extractable aluminum (James et al. 1978). In the Graham Island sites, soils are very acid with high (>1%) extractable A l concentrations, and most are poorly aerated for varying durations, therefore available P levels are likely to be low across most of the study area. In addition, the lack of any relationships which have been shown in other studies may reflect the form in which P was expressed. Blyth and MacLeod (1981a) found available P to be poorly correlated with Ss growth while total P was consistently correlated over a wide range of soils in Scotland. Similar findings have been reported by Adams et al. (1970) and James et al. (1978). While total P content may be high, the majority of it is not available to plants, being bound in organic form in the forest floor 129 and high organic matter content characteristic of these mineral soils (James et al. 1978). The lack of correlation between available P and Cw site index in the Graham Island sites is consistent with these findings. To place the Graham Island soils into perspective with other coastal B.C. soils, they were assessed in relation to the quantitative classification of soil nutrient r e g i m e s proposed by Courtin et al. (1985). This classification is based on a population of mesothermal soils from coastal southwestern B.C., and uses C / N ratio and p H of the forest floor, and total N and the sum of exchangeable bases in the mineral soil root zone (in kg/ha) to characterize classes. Soils from site index classes 1 and 2 were most similar to the "poor" nutrient regime, while site index class 3 soils were within the range of the "medium, N low" nutrient regime. The C / N ratios in the Graham Island soils tended to be higher than those characterizing the nutrient regime classes, likely reflecting slower forest floor decomposition in the cooler soils in the study area. The sum of exchangeable bases tended to be slightly lower in the Graham Island sites which may reflect greater leaching losses from these soils. In soils featuring poor aeration, Fe and M n in reduced form displaces cations from exchange sites rendering them more susceptible to leaching, particularly in high precipitation climates (Ponnamperuma 1984). 5.9.2 Soil aeration Poor soil aeration, reflected in moisture content, air-filled porosity, and oxidation of iron rods was consistently associated with poorer tree growth in the Graham Island sites. 130 Similar relationships have been found in other areas where soils feature periodic or seasonal saturation. Armstrong et al. (1976) found improved growth of Sitka spruce associated with better soil aeration indicated by soil oxygen flux measured in the profile of peaty gley soils in Scotland. Moisture content of the surface 15 cm of soil, a reflection of aeration, was significant in regression models using soil physical and morphological properties to predict Sitka spruce growth in Britain (Page 1970). Blyth and MacLeod (1981a, 1981b) found depth to mottling to be highly correlated with Sitka spruce yield in forests of Scotland where poor drainage and soil aeration are important. Growth of loblolly pine and other tree species has been found to be associated with soil properties related to aeration status such as depth to a water table and mottling in coastal lowland ecosystems of the southeast U.S. (Klawitter 1978). In general, soil aeration "stress" can have a profound effect on forest productivity as a good supply of oxygen is necessary for root growth and function (Pritchett and Fisher 1987). Poor aeration affects soil chemical properties by setting in motion a series of "redox" reactions according to the degree of anaerobism. Denitrification of nitrate takes place, which may lead to net loss of N. As previously mentioned, Fe and M n become soluble in their reduced forms, displacing cations from exchange sites where they may be subject to increased leaching losses (Ponnamperuma 1984). Ethylene, an important plant growth regulator, is produced in oxygen-deficient soils, and may reach concentrations which are injurious to plant growth (Smith and Renstall 1971). Poor soil aeration also directly affects the plant. Reduced root respiration in an oxygen-deficient environment is believed to stimulate abcissic acid (ABA) production in leaves 131 which causes stomata to close and photosynthesis to drop (Ray 1972). This has been shown to occur in Sitka spruce (Coutts 1980). Other plant growth regulators are affected by signals from aeration- stressed roots, with the general trend being lower production of gibberellic acids and cytokinins, and increased contents of auxins and ethylene in above-ground plant parts (Reid and Bradford 1984). The exact nature of the physiological effects of these changes is poorly understood, although ethylene accumulation has been shown to accelerate leaf abscission and senescence (Sexton 1985) and is associated with several morphological adaptations for tolerating poorly aerated soils such as formation of adventitious roots (Hook 1984) and enhanced internal oxygen transport (Jackson 1985). Nutrient uptake is reduced in plants influenced by poor soil aeration, largely due to reduced root respiration. This inhibits active transport of nutrients, particularly N, P, and K. Magnesium and Ca are less affected as they rely more on passive uptake (Kozlowski and Pallardy 1984). Lower transpirational demand associated with reduced plant vigour also contributes to lower rates of nutrient uptake. Lower nutrient uptake, particularly P, is exacerbated by reduced mycorrhizal development on roots in poorly aerated soil (Jackson and Drew 1985). Since most mycorrhizae are strongly aerobic, they are inhibited in poorly aerated soil. This has been demonstrated in many tree species, including Douglas-fir (Gadgil 1971), lodgepole pine (Boggie 1972, Coutts and Philipson 1978b) and Sitka spruce (Coutts and Philipson 1978b). The above physiological effects of poor soil aeration generally interfere with providing adequate supplies of water, nutrients, and balanced plant growth regulators to the crowns 132 of plants. The obvious result is reduced growth of the entire plant. Root growth not only slows or stops entirely, but dieback of major portions of the root network may occur. This results in development of a shallow root system, limited to the zone of adequate aeration. The extent to which these effects take place in a given ecosystem depends on the nature of soil saturation, including its depth, its duration, its timing during the year, and the quality of the water. Water quality, particularly dissolved oxygen content is believed to play an important role in the impact of soil saturation. Moving soil water contains higher concentrations of oxygen because of mixing and can offset aeration stress in trees (Pritchett and Fisher 1987). The dissolved oxygen data from the Graham Island sites did not show a definite relationship with tree growth, largely a result of measurement limitations. However, one plot did demonstrate enhanced dissolved oxygen content in what appeared to be a moving perched water table. It is likely that this factor plays a role in other site index class 3 sites featuring periodic soil saturation which are located on gentle slopes or adjacent to streams. Perhaps the most important of soil saturation conditions is the timing of saturation events. Trees are most sensitive to soil saturation during the growing season when respiratory demand for oxygen is high and oxygen solubility is low due to higher temperatures (Gill 1970). For example, Coutts and Philipson (1978a) found substantially higher survival of lodgepole pine and Sitka spruce roots that were dormant when waterlogged, compared to those that were active. It was also observed that initially dormant roots resumed growth shortly after aeration was reestablished, while initially active roots took considerably longer to regenerate new growth. The point samples of indices of soil 133 aeration used in this study are somewhat limited in that they do not provide any information on temporal variation in soil saturation activity over the range of sites, although they do address relative differences between sites during the mid-growing season. During my three years experience in the area, there has been considerable variation in the state of water tables on different sites, depending on the season, the precipitation events in the days preceding inspection, and in the growing season precipitation patterns. Obvious evidence of dieback of well-established roots in several soil pits indicates that the nature of soil saturation varies over the years, and that annual growth likely varies according to these conditions. In addition to soil saturation conditions, the effects of poor soil aeration on tree growth also vary with the species. Different species have different capabilities for tolerating poor aeration. In general, angiosperms are more tolerant than gymnosperms. Western redcedar, lodgepole pine, and black spruce (Picea mariand) are examples of B.C. gymnosperms with relatively high tolerance. Since the principal limitation of saturated soil is oxygen deficiency, trees can follow two adaptation strategies. They can improve the internal supply of oxygen to the roots by way of morphological adaptations, or they can improve the ability of root cells to function in periods of low oxygen through metabolic adaptation. Usually a combination of these is adopted by most tolerant species (Tripepi and Mitchell 1984). Examples of morphological adaptations are the formation of hypertrophied lenticels to enhance gas exchange into the phloem (Hook and Scholtens 1978), formation of adventitious roots above the poorly aerated zone, and formation of special "water adapted" roots which feature enhanced internal transport of oxygen to the active root tips. Such roots 134 develop in lodgepole pine in response to soil saturation and feature large stelar cavities which allow oxygen to move from aerated regions to root tips within poorly aerated environments (Coutts and Philipson 1978a, 1978b). 5.9.3 Rooting depth The effective rooting depth was associated with tree growth in the Graham Island study (accounting for 51% of CWSI variation), with perched water tables and dense, heavy-textured, and/or strongly cemented soil layers limiting total rooting depth from the ground surface to under 50 cm for most sites. Root growing space, measured as the depth of rooted soil, has been used as an indicator of site productivity more often than most soil physical properties in other site/productivity studies. It can be measured with some accuracy, particularly if a distinct restricting layer occurs, and it integrates many other factors important to tree growth such as water storage capacity and nutrient reserve (Pritchett and Fisher 1987). Rooting depth has been correlated with tree growth in a wide range of studies in North America (Carmean 1975) and specifically for Sitka spruce in Britain (Worrel 1987). Effective rooting depth in the study area, while restrictive in itself because of limitations in the volume of soil exploitable by trees, becomes more critical when periodic soil saturation reduces the quality of the already limited root zone. These root zone limitations will likely affect the long term productivity of stands on these sites. As stands develop, root systems extend and more fully exploit the fixed volume of soil. Thus the effective "quantity" 135 of soil decreases in relation to the supply to trees (Armson 1977). Day (1963) suggested these factors would lead to eventual decline in stand volume and quality as stands age on lowland ecosystems of east Graham Island. He speculated that as stands age, they may reach a point where the demand for growing space and nutrients exceeds the supply capacity of these soils featuring shallow root zones. This would lead to decline in height growth, die-back of root systems, opening of the canopy due to death of the least tolerant trees (Sitka spruce), and invasion of understory shrub species such as salal, Vaccinium spp., and Menziesia ferruginea. A steady, slow decline in standing volume would follow, resulting in decadent old-growth stands such as those common in the study area. While his observations were based on limited data, Day's (1957, 1963) theories were intuitively logical and hold up in view of closer examination of these ecosystems. 5.10 Management Implications Understanding some of the important ecological factors associated with tree growth provides a basis for developing management strategies which take advantage of these factors to improve forest productivity or prevent its degradation (Stone 1977). In the Graham Island study area, analysis of ecological relationships has provided some insight into what may be considered appropriate silvicultural practices, particularly regarding tree species selection and site amelioration. 136 5.10.1 Tree species selection One of the most important, and least expensive silvicultural decisions to make is selecting suitable tree species to manage on a site. In the study area, the choice is limited to only four commercial species, however considerable variation was observed in relative species performance between different sites. The species of most concern is Sitka spruce. Not only is it the most sensitive species to degrading site conditions, but it is also the most preferred from a management perspective. Because of generally higher volumes and value of stands containing greater spruce components, managers favour spruce wherever possible (Farr and Harris 1983). In the study area, spruce showed the greatest sensitivity of all species to poor soil aeration, and decreasing soil nutrient content. In light of this it is unsuitable for site index class 1 ecosystems. In site index class 2 sites, soil aeration improves sufficiently so that spruce can form a component of mixed species stands. Based on existing stand conditions, it should be limited to less than 10 to 15% of the stand composition. One of the factors which enables spruce to perform reasonably well on these sites, in spite of their limitations in aerated rooting zone depths, may be the presence of lodgepole pine in the stand composition. It has been commonly observed in Britain that the performance of Sitka spruce on poorly aerated gley soils is enhanced significantly when grown together with lodgepole pine rather than in pure stands. While the mechanism for this synergistic effect is not clear, it is believed to be related to N supply (Carlyle and Malcolm 1985). The ability of lodgepole pine to aerate its rhizosphere in poorly aerated soil is one possible factor contributing to this effect. As previously discussed, pine can adapt to limited soil aeration by forming special "water adapted" roots which contain large cavities in the stele to allow 137 enhanced internal transport of oxygen to the growing root tips (Coutts and Philipson 1978a, 1978b). This enables pine roots to function and, in fact, grow down to 20 cm and more into an anaerobic soil. This enhanced oxygen transport system was found to be absent in Sitka spruce roots, thus explaining its lower tolerance to aeration stress. Sitka spruce root systems would conceivably benefit from improved rhizosphere aeration in intimate association with lodgepole pine root systems in mixed stands. Soil conditions improve sufficiently in site index class 3 ecosystems that spruce can be managed as a major component of mixed species stands. In such stands it generally achieves dominance in the canopy, with individual trees having large diameters relative to other species. However, it should not be managed as a monoculture on these sites as the soils may still be sufficiently limited, and unable to meet the demands of such stands. Lodgepole pine plays an interesting role in these ecosystems as it is unusual to find it forming a major component of second-growth stands within wet maritime climates of B.C. Its establishment, no doubt, is a reflection of the fire history. Because of pine's high tolerance of poorly aerated soil, it has an important role in the management of these ecosystems. Specifically, it can be managed as a component in mixed species stands on site index class 1 and 2 sites. It is not well suited to long term production on site index class 3 sites as it is unable to successfully compete with Sitka spruce. Careful selection of planting stock is important for successful management of pine on these sites. The stock must be selected from local provenances adapted to the ecological conditions of Graham Island and it must also be capable of developing well-formed roots which can readily exploit the 138 shallow rooting zones. To ensure good lateral root development from the root collar down, copper-coated containers should be used. Redcedar and western hemlock appear suited to the full range of site conditions encountered and form components of all stands, although cedar is more tolerant of very poorly aerated soils (Hook and Scholtens 1978). A summary of the recommended species composition for management of these sites is shown in Table 24. Table 24. Recommended tree species composition for sites on east Graham Island. Site index class Species composition 1 Cw (leading), PI, Hw 2 Cw, Hw, PI, (Ss) 3 Hw, Ss, Cw 5.10.2 Site amelioration Since the major factor limiting productivity in these sites is the shallow depth of well aerated root zones, one method of improving productivity potential is to increase the available rooting space. Two approaches to this problem are to reduce the effects of perched water tables on the root zone through forest drainage, or to raise the level of aerated soil above the restricting layer through mounding. Drainage of poorly drained soils for forestry has been widely used in the U.S. Southeast and Gulf coastal plains, as well as 139 Britain, and European and Scandinavian countries. Substantial increases in stand yield have been realized on many sites through this practice, with the major benefits being derived from increased rooting depth, improved soil aeration, and reduced windthrow (Pritchett and Fisher 1987). However the effectiveness of surface water draining varies considerably depending on the conditions of the site, the design and construction of the drainage system, and the follow-up maintenance. Soil hydraulic conductivity is of prime importance as it affects the draw-down effect of ditches. In heavy textured soils with slow conductivity, the draw-down of the water table may only extend a few meters from the ditch. Substantially greater density of drainage ditches is required in such cases to effectively remove water from the site (Boelter 1978, Pritchett and Fisher 1987). Drainage systems must be engineered with a continuous ditch grade to prevent ponding. Therefore, uniform, gently sloping topography is most suitable. To be effective, the drainage system must have an outlet which will allow water to move off the site in a short period of time (Terry and Hughes 1978). Drainage systems must also be maintained for long-term effectiveness, to prevent ditches from plugging and allowing the site to revert to its original condition (Thomasson 1975). Another form of drainage involves ripping strongly cemented soil horizons which impede water movement. Such techniques are only applicable where subsoils are deep and freely drained. Rapid reconsolidation of ripped trenches may limit long-term effectiveness. In the Graham Island study area, drainage has relatively limited application. The majority of soils are heavy textured tills with low hydraulic conductivity. Limited draw-down 140 of surface water would require high drainage density. Because of the dense subsurface layers, root development would still remain limited to a shallow zone. However, successful drainage would improve the aeration conditions within this limited root zone. Drainage design would pose problems over much of the area due to a lack of continuous slope. Also, natural drainage systems are infrequent, making it difficult to construct effective drainage -outlets. There is some potential for drainage in the northern portion of the area on glaciofluvial sand deposits. Portions of this area feature gently sloping terrain suitable for drainage systems. The occurrence of deep, coarse textured, freely-drained sand below the impermeable layer in some areas provides an outlet for drainage, however maintenance of such an outlet is essential to prevent build up of finer-textured sediments with lower hydraulic conductivity. Because of the strongly cemented horizons common in these soils, the increased volume of aerated soil provided by drainage will remain limited. A n alternative to surface water drainage is mounding. Row mounding (bedding) has been used in the U.S. Southeast coastal plain to improve growth of slash pine (Pinus elliottii) on water logged soils (Pritchett and Fisher 1987). Haines and Haines (1978) observed improved soil conditions from bedding, resulting in substantially larger and more extensive slash pine root systems compared to untreated soils. The density of branching and frequency of mycorrhizal associations was particularly improved through bedding. One of the limitations of this form of bedding is that the early advantages from the treatment may not persist through the rotation. The mounds are created by bedding harrows or ploughs, and represent a relatively small volume of "raised" soil. While such volumes enhance early stand performance, trees would fully exploit the improved rooting medium well before they 141 reached maturity. Mounding offers the most promise as a site modification treatment in the Graham Island area, however a specific form of mound is required to achieve the greatest long term benefits. Based on the results of this study, a relatively modest increase in the depth of aerated soil is associated with substantial increases in tree growth. Therefore mounds can be relatively low (eg. 50 cm). Mounds should, however, be wide (eg. 2-3 m) to provide a large volume of aerated soil for root systems to exploit as the trees age. Such mounds would closely mimic the type of blowdown disturbance which is associated with productive tree growth within the study area. To accelerate the weathering and amelioration of recently overturned subsoils, a cover of grasses and legumes can be established on the mounds following planting. This type of mounding treatment can be undertaken with medium-sized excavators, and has been found to be operationally feasible in a trial area north of Port Clements. While costs are relatively high, expected long term gains should be substantial, both in terms of stand yield and value, and in expanding the forest land base by incorporating land currently excluded from the yield analysis. 142 6.0 S U M M A R Y Twenty-four sample plots were established in fire regenerated second-growth stands on the east side of Graham Island, Queen Charlotte Islands. These were distributed over a range of site productivities as indicated by site index of western redcedar, and represented a number of imperfectly to poorly drained mineral soils characteristic of the coastal lowland ecosystems. Soil chemical and physical properties, soil water quality, floristic composition, mensuration characteristics, and site index of major tree species were determined for each plot. The available data were screened to select the most promising variables for examining site/productivity relationships. Three classes defined by site index of cedar were characterized according to stand properties and ecological properties. Discriminant analysis was used to test objectively the validity of these classes in relation to soil properties. Inherent structure in the soils data was assessed using a combination of principal components analysis and cluster analysis. Reciprocal averaging was used to examine structure present in the vegetation data. Inherent data structure was compared with the three site index classes to assess how well these classes reflect ecological patterns in the study area. Functional relationships between cedar site index and ecological properties were investigated using multiple linear regression. Relevant information revealed in all the analyses was incorporated into a simplified model for field recognition of site index classes to assist practitioners. Results of the study were interpreted in relation to suggested silvicultural practices for these ecosystems. The main results of the study are summarized as follows: 143 1. Tree growth, as expressed by cedar site index, was correlated with nutrient content in the available root zone, the depth of this root zone, and the mineral soil root zone aeration, as reflected by moisture content, air-filled porosity, and oxidation of iron rods. Of the nutrients examined, total nitrogen and exchangeable magnesium showed the strongest relationships. A simple model using total N content and volumetric moisture content in the mineral soil root zone summarized the relationship between site index and soil properties, accounting for 78% of the site index variation. This model indicates that the quantity of root zone nutrient reserve available to trees, and the aeration status of the root zone, as reflected by moisture content, are most important to tree growth. Using field-recognizable variables alone, 73% of the variation in site index could be accounted for in a simple model using total rooting depth from the ground surface and slope. There is, however, some question as to whether this relationship is realistic for the whole study area because of some limitations in the slope variable. Rooting depth alone accounted for 51% of site index variability. 2. The effectiveness of understory vegetation showed some limitation in accurately reflecting site conditions due to non-site related variation. While the wettest, lowest productivity sites clearly differed from remaining sites largely on the basis of moss cover, the site index class 2 and 3 sites could not be floristically differentiated. In spite of their similar understory development, these sites ranged considerably in productivity (7 to 14 m 3 ha" 1 yr"1 MAI), and soil properties. In such dense stands with poor floristic diversity, emphasis on soil and physiographic features is required to effectively recognize important differences in site quality. 144 3. Three site index classes were originally designed as sampling strata to obtain a balanced distribution of plots. However, these classes appeared to reflect natural groupings occurring in the study area. Cedar site index observed in second-growth stands tended to cluster within the ranges of each class, with 13-14 m, 17-18 m, and 21 m/50yrs b.h. age being the most frequent site indices encountered. These site index classes are reasonably well discriminated by soil properties using discriminant analysis, and are also reflected, with minor overlap, in the natural structure present in the soil data. Site index classes and soil-defined clusters do not differ substantially in terms of stand and soil variables. 4. Subtle differences in topography can have a big impact on tree growth. Many of the study sites could be considered sensitive in relation to changing environmental conditions. Soils have limiting conditions in the form of slowly permeable, root restricting layers close to the surface. Minor physiographic variation can increase or decrease the volume of well-aerated rooting zone sufficiently to affect productivity. Tree growth is sensitive to relatively minor changes in the major limiting factor, as on most ecologically "limiting" sites. 5. Sitka spruce is the most sensitive of the tree species present in these stands to deteriorating site quality. Hemlock, lodgepole pine, and cedar followed similar height growth patterns based on site index, although Hw tends to have slightly slower growth than cedar on the poorest sites. Spruce exhibits substantially poorer growth than cedar 145 on the poorer sites, but as site conditions improve, its performance exceeds that of cedar. This reflects spruce's ability to grow very rapidly when site conditions become adequate for its requirements. These relative differences in tree species tolerance to site conditions are reflected to some extent in stand composition. Poorest sites support Cw, with slightly less Hw, and PI proportions; intermediate sites support approximately equal proportions of Cw, Hw, and PI, with a minor component of Ss; the better sites support equal proportions of Cw, Hw, and Ss, with PI dropping out because of its inability to compete with the dominant Ss. 6. Some management strategies were recommended on the basis of the ecological relationships observed in the study,. Tree species selection generally follows the patterns of species composition observed above. The recommended mixtures are: for site index class 1 - Cw, PI, and Hw; for site index class 2 - Cw, Hw, PI, (Ss); for site index class 3 - Hw, Ss, Cw. Site modification treatments aimed at increasing the depth of aerated root zone have good potential for improving productivity of these ecosystems. Creating relatively low but wide mounds would provide a reasonable volume of aerated soil which can be exploited as trees age. Such mounds would closely mimic the type of blowdown disturbance associated with productive tree growth within the study area. To accelerate the weathering and amelioration of recently overturned subsoils, a cover of grasses and legumes may be established on mounds following planting. 147 7.0 R E F E R E N C E S Adams, S.N., W . H . Jack, and D A . Dickson. 1970. The growth of Sitka spruce on poorly drained soils in Northern Ireland. Forestry 43: 125-133. Alley, N.F., and B. Thompson. 1978. Aspects of environmental geology, parts of Graham Island, Queen Charlotte Islands. Bulletin No. 2. B.C. Min. Env., Resource Analysis Branch, Victoria, B.C. Armson, K . A . 1977. Forest soils. Properties and processes. Univ. of Toronto Press, Toronto, Ont. Armstrong, W., T .C . Booth, P. Priestley, and D J . Read. 1976. The relationship between soil aeration, stability, and growth of Sitka, spruce (Picea sitchensis (Bong.) Carr.) on upland peaty gleys. J. Appl. Ecol. 13: 585-591. Baker, T . E . 1974. The major soils of the Tofino area of Vancouver Island and implications for land use planning and management. Ph.D. thesis, Dept. of Soil Sci., Univ. of B.C., Vancouver, B.C. Bakuzis, E . V . 1969. Forestry viewed in an ecosystem perspective. In The ecosystem concept in natural resource management. G . M . van Dyne (ed.). Academic Press, New York, pp. 189-258. Banner, A. , J. Pojar, R. Trowbridge, and W.J. Beese. 1987. Ecosystem classification of the Coastal Western Hemlock Zone, Wet Hypermaritime Subzone (CWHwh), Queen Charlotte Islands, British Columbia. B.C. Min. For., Smithers, draft rep. Barker, J. 1985. Productivity of major ecological associations of Tree Farm License 24, Queen Charlotte Islands, B.C. Final report, B.C. Min. For. Sec. 88 research programme. Binkley, D . E . 1980. Water chemistry profile comparisons of early and mid-successional forests in coastal British Columbia. M.Sc. thesis, University of B.C., Faculty of Forestry Black, T . A 1988. Soil science 414, Biometeorology, lecture handouts. Univ. of B.C. Soil Science Dept., Vancouver, B.C. Blyth, J.F., and D A . McLeod. 1978. The significance of soil variability for forest soil studies in northeast Scotland. J. Soil Sci. 29: 410-430. Blyth, J.F., and D A . McLeod. 1981a. Sitka spruce (Picea sitchensis) in northeast Scotland.!: Relationships between site factors and growth. Forestry 54:41-62. Blyth, J.F. and D.A. McLeod. 1981b. Sitka spruce in Northeast Scotland II. Yield prediction by regression analysis. Forestry 54: 63-73 148 Boelter, D . H . 1978. Drainage experience in Lake States and northern European peatlands. TnProceedings, Soil moisture-site productivity symposium. W . E . Balmer (ed.), Nov. 1-3, 1977, Myrtle Beach, South Carolina. U S D A For. Serv., pp. 167-175 Boggie, R. 1972. Effect of watertable height on root development of Pinus contorta on deep peat in Scotland. Oikos 23: 304. Bourgeois, W.W. 1969. A study of soils and leachates from two forest sites using tension lysimeters. M.Sc. thesis, Univ. of B.C. Dept. of Soil Science Brady, N.C. 1974. The nature and properties of soil. 8th edition. MacMillan Publishing Co., Inc., New York. Broadfoot, W . M . 1969. Problems in relating soil to site index for southern hardwoods. For. Sci. 15: 354-364. British Columbia Forest Service. 1976. Whole stem cubic metre volume equations and tables. B.C. Min. For., Victoria, B.C. British Columbia Forest Service. 1982. Queen Charlotte Timber Supply Area report. B.C. Min. For., Planning Branch, Victoria, B.C. British Columbia Forest Service. 1983. Forest and range inventory manual. B.C. Min. For., Inventory Branch, Victoria, B.C. Buol, S.W. 1977. Soil moisture characteristics in the southern states. In Proceedings, Soil moisture - site productivity symposium. W.E . Balmer (ed.), Myrtle Beach, South Carolina, Nov. 1-3, 1977. U S D A For. Serv., pp. 30-39. Campbell, R . G . 1978. The Wayerhauser land classification system. In Proceedings, Soil moisture - site productivity symposium. W.E . Balmer (ed.), Myrtle Beach, South Carolina, Nov. 1-3, 1977. U S D A For. Serv., pp. 17-27 Carmean, W . H . 1975. Forest site quality evaluation in the United States. Adv. in Agron. 27: 209-269. Carter, R . E . 1983. Forest floors under second-growth Douglas-fir stands: their chemical variability and some relationships tp productivity. M.Sc. thesis, Dept. of Soil Sci., Univ. of B.C., Vancouver, B.C. Chatterjee, S. and B. Price. 1977. Regression analysis by example. John Wiley and Sons, New York. Cook, A. , M . N . Court, and D.A. McLeod. 1977. Theprediction of Scots pine growth in northeast Scotland using readily accessible site characteristics. Scott. For. 31: 251-264. 149 Courtin, P.J., M . C . Feller, and K. Klinka. 1983. Lateral variability in some properties of disturbed forest soils in southwestern British Columbia. Can. J. Soil Sci. 63: 529-539. Courtin, P.J., K. Klinka, M . C . Feller, and J.P. Demaerschalk. 1985. A n approach to quantitative classification of nutrient regimes of forest soils. Paper presented at the symposium: Community organization and ecosystem conservation - a contemporary synthesis, June 23-29, 1985, Canadian Congress of Biology, London, Ont. Coutts, M.P. 1980. Effects of waterlogging on water relations of actively growing and dormant Sitka spruce seedlings. Ann. Bota. 47: 746-753. Coutts, M.P. and J.J. Philipson. 1978a. The tolerance of tree roots to waterlogging. I. Survival of Sitka spruce and lodgepole pine. New Phytol. 80: 63-69. Coutts, M.P., and J.J. Philipson. 1978b. The tolerance of tree roots to waterlogging. II. Adaptation of Sitka spruce and lodgepole pine to waterlogged soil. New Phytol. 80: 71-77. Crum, H.A. , W.C. Steere, and L . E . Anderson. 1973. A new list of mosses of North America north of Mexico. The Bryologist 76:85-130. Cunia, T. 1973. Dummy variables and some of their uses in regression analysis. Proc. I U F R O Subject Group S4.02. Vol. 1:1-146. Day, W.R. 1957. Sitka spruce in British Columbia - a study in forest relationships. Forestry Commission Bulletin No. 28, Imperial Forestry Institute, Oxford. Day, W.R. 1963. The development of Sitka spruce on shallow peat. Scottish Forestry 17:219-236. Driebelbis, F.R. and J.L. MeGuiness. 1957. Plant nutrient losses in lysimeter percolates. Agron. J. 49:523-527. Everitt, B. 1980. Cluster analysis. 2nd edition. Halsted Press, New York. Farr, W.A. and A.S. Harris. 1983. Site index research in the western hemlock - Sitka spruce forest type of Coastal Alaska. In I U F R O symposium on on forest site and continuous productivity. R. Ballard, S.P. Gessel (eds.). August, 1982, Seattle, Wash. U S D A Forest Service, Gen. Tech. Rep. PNW-163, pp. 87-93. Feller, M . C . and H . Kimmins. 1979. Chemical characteristics of small streams near Haney in southwestern British Columbia Water. Res. Res. 15:247-258. Ford, E .D. , and A J . Fraser 1968. The concept of actual and potential production as an aid to forest management. Forestry 41: 175-181. 150 Fox, D.J. and K . E . Guire. 1976. Documentation for MIDAS. Stat. Res. Lab., Univ. of Michigan. Gadgil, P.D. 1971. Effect of waterlogging on mycorrhiza of radiata pine and Douglas-fir. N.Z. Jour. For. Sci. 2: 222-226. Gardner, W . H . 1965. Water content. In Methods of soil analysis. Part 1. C A . Black, D.D. Evans, L . E . Ensminger, J.L. White, and F . E . Clark, (ed.). Amer. Soc. of Agronomy. Madison, Wise, pp. 82-153. Gauch, H . G . Jr., R . H . Whittaker, and T.R. Wentworth. 1977. A comparative study of reciprocal averaging and other ordination techniques. J. Ecol. 65: 157-174. Gill, C.J. 1970. The flooding tolerance of woody species - a review. For. Abstracts. 31: 671-688. Goodman, L . A . 1960. On the exact variance of products. Jour. Amer. Stat. Assoc. 55: 708-713. Gordon, A . D . 1981. Classification. Methods for the exploratory analysis of multivariate data. Chapman and Hall, New York. Goudie, J.W. 1984. Height growth and site index curves for lodgepole pine and white spruce andinterim managed stand yield tables for lodgepole pine in British Columbia. B.C. Min. For. Res. Branch, Int. Rep. FY-1983-84. Green, R.N. 1987. Ecological basis for growth and yield estimation. Paper presented at the Growth and Yield Workshop, Nov. 26, 1987, Nanaimo, B.C. Green, R.N., P.L. Marshall, and K. Klinka. 1989. Estimating site index of Douglas-fir (Pseudotsuga menziesii [Mirb.] Franco) from ecological variables in southwestern British Columbia. Forest Science 35:50-63. Hagglund, B. and J-E . Lundmark. 1977. Site index estimation by means of site properties. Scots pine and Norway spruce in Sweden. Studia Forestalia Suecicia, Nr. 138, Swedish College of Forestry, Stockholm. Hagglund, B. 1981. Evaluation of forest site productivity. Comm. For. Bur. For. Abstr., Rev. Art. 42(ll):515-527. Haines, L .W. and S.G. Haines. 1978. Site preparation for regeneration. In Proceedings, Soil moisture-site productivity symposium. W.E . Balmer (ed.), Nov. 1-3, 1977, Myrtle Beach, South Carolina. U S D A For. Serv., pp. 176-195. Hale, M . E . and W.L. Culberson. 1970. A fourth checklist of the lichens of the continental United States and Canada. The Bryologist 73:499-543. 151 Heilmari, P.E. 1979. Minerals, chemical properties, and fertility of forest soils. In Forest soils of the Douglas-fir region. P.E. Heilman, H.W. Anderson, and D . E . Baumgartner (ed.) . Wash. State Univ., Coop. Ext. Serv., Pullman, Wash., pp. 121-136. Hill, M . O . 1973. Reciprocal averaging. J. Ecol. 61:237-249. Hodgkins, E J . 1970. Productivity estimation by means of plant indicators in the longleaf pine forests of Alabama. In Tree growth and forest soils. C T . Youngberg and C.B. Daveys (eds.). Oregon State Univ. Press. Corvallis, Oregon. Holland, S.S. 1964. Landforms of British Columbia - A physiographic outline. B.C. Dept. Mines and Petroleum Resources, Bull. No. 48. Hook, D .D . 1984. Adaptation to flooding with fresh water. In Flooding and plant growth. T .T . Kozlowski (ed.),Academic Press Inc., Orlando, Florida. Hook, D.D. and J.R. Scholtens. 1978. Adaptation and flood tolerance of tree species. In Plant life in anaerobic environments. D.D. Hook and R . M . M . Crawford (ed.), Ann Arbor Science Publishers Inc., Ann Arbor, Mich. Jackson, M.B. 1985. Ethylene and the response of plants to excess water in their environment - a review. In Ethylene and plant development. J.A. Roberts and G.A. Tucker (ed.). Butterworths, London. Jackson, M.B. and Drew, M . C . 1984. Effects of flooding on growth and metabolism of herbaceous plants. In Flooding and plant growth. T .T. Kozlowski (ed.). Academic Press Inc., Orlando, Florida. James, H . , M.N. Court, D .A. McLeod, and J.W. Parsons. 1978. Relationships between growth of Sitka spruce (Picea sitchensis), soil factors, and mycorrhizal activity on basaltic soils in western Scotland. Forestry 51: 105-119. Johnson, J.E. , C L . Haag, J .G. Bockheim, and G . G . Erdmann. 1987. Soil - site relationships and soil characteristics associated with even-aged red maple (Acer rubrum) stands in Wisconsin and Michigan. For. Ecol. and Mgmt 21: 75-89. Kabzems, R.D. 1985. Quantitative classification of soil nutrient regimes of some mesothermal Douglas-fir ecosystems. M.Sc. thesis, Fac. of Forestry, Univ. of B.C., Vancouver, B.C. Karanka, E J . 1986. Trends and fluctuations in precipitation and stream runoff in the Queen Charlotte Islands. B.C. Min. For., Land Mgmt. Rep. No. 40, Victoria, B.C. Klawitter, R .A. 1978. Growing pine on wet sites in the southeast coastal plain. In Proceedings, Soil moisture - site productivity symposium. W . E . Balmer (ed.), Myrtle Beach, South Carolina, Nov. 1-3, 1977. U S D A For. Serv., pp. 49-62. 152 Klecka, W.R. 1975. Discriminant analysis. In SPSS. Statistical Package for the Social Sciences. 2nd ed. Nie, N.H. , C H . Hull, J .G. Jenkins, K. Steinbrenner, and D . H . Bent (ed.). McGraw-Hill, New York. Klinka, K., R .N. Green, P.J. Courtin, and F .C . Nuszdorfer. 1984. Site diagnosis, tree species selection, and slashburning guidelines for the Vancouver Forest region. B.C. Min. For. Land Mgmt. Rep. 25, Victoria, B.C. Klinka, K. and R . E . Carter. 1989. Ecological classification and relationships between site quality and forest productivity in coastal Douglas-fir ecosystems, (submitted for publ. in Can. J. For. Res.) Klinka, K. and V.J . Krajina. 1986. Ecosystems of the University of British Columbia Research Forest, Haney, B.C. The University of British Columbia, Faculty of Forestry, Vancouver, B.C. Klinka, K., V.J . Krajina, A . Ceska, and A . M . Scagel. 1989. Indicator plants of coastal British Columbia. Accepted for publication by the Univ. of B.C. Press. Kozlowski, T .T. 1984. Responses of woody plants to flooding. In Flooding and plant growth. T.T. Kozlowski (ed.). Academic Press Inc., Orlando, Florida. Lavkulich, L . M . 1978. Methods manual (3rd edition). Pedology laboratory. Dept. of Soil Sci., Univ. of B.C., Vancouver, B.C. Lewis, T. 1976. The till-derived podzols of Vancouver Island. Ph.D. thesis. Dept. of Soil Sci., Univ. of B.C., Vancouver, B.C. Mclntyre, D.S. 1970. The platinum microelectrode method of soil aeration measurement. Adv. Agron. 22:235-283. McKee, W . H . Jr. 1978. Rust on iron rods indicate depth of soil water tables. P. 265-294 In Soil moisture - site productivity symp. proc, W.E . Balmer (ed.). U S D A For. Serv. Southeast. Area, State and Priv. For., Myrtle Beach S.C. MacLean, C D . and C L . Bolsinger. 1973. Estimating Dunning's site index from plant indicators. U S D A Forest Service, Research Note PNW-197, Portland, Oregon. McQuilkin, R A . 1976. The necessity of independent testing of soil-site equations. Soil Sci. Soc. Amer. J. 40: 783-785. Mader, D . L . 1963. Soil variability - a serious problem in soil-site studies in the Northeast. Soil Sci. Soc. Am. Proc. 27: 707-709. Marshall, P.L. 1987. A microcomputer program for assisting in the design of simple random samples. The For. Chron. 63:422-425. 153 Mathiasen, R.L. , E . A . Blake, and C.B. Edminster. 1986. Estimates of site potential for Douglas-fir based on site index for several southwestern habitat types. Great Basin Naturalist 46: 277-280. Mehlich, A . 1978. New extractant for soil test evaluation of phosphorous, potassium, magnesium, calcium, sodium, mangenese, and zinc. Comm. Soil Sci. and Plant Anal. 9: 477-492. Meyer, W . H . 1937. Yield of uneven-aged stands of Sitka spruce and Western hemlock. U S D A Tech. Bull. 544. Miller, R . G . 1986. Beyond A N O V A , basics of applied statistics. John Wiley and Sons, New York. Ministry of Lands, Parks, and Housing. 1980. Graham Island Crown Land Plan. Prov. of B.C., Int. Rep., Victoria, B.C. Mitchell, K.J. and K.R. Polsson. 1988. Site index curves and tables for British Columbia: Coastal species. Canada-B.C. Min. For. F R D A Report 037. Monserud, R.A. 1984. Problems with site indices: an opinionated review. In Forest land proc, J. Bockheim (ed.). Dep. Soil Sci., Univ. Wisconsin, Madison, pp. 167-180. Morrison, D . G . 1969. On the interpretation of discriminant analysis. Jour, of Marketing Res. 6: 156-163. Morrison, D.F. 1976. Multivariate statistical methods. 2nd edition. McGraw-Hill, New York. Nichols, S. 1977. On the interpretation of principal components analysis in ecological contexts. Vegetatio 34: 191-197. Noy-Meir, I. and R . H . Whittaker. 1978. Recent developments in continuous multivariate techniques. In Ordination of plant communities. R . H . Whittaker (ed.). Dr. W. Junk bv Publishers, The Hague, The Netherlands. Ochyra, R. 1982. Kindbergia (Brachytheciaceae, Muslci). A new name for Stokesiella. Lindbergia 8:53-54. Omi, P.N., L . C . Wensel, and J.L. Murphy. 1979. A n application of multivariate statistics to land-use planning: classifying land units into homogenous zones. Forest Sci. 25:399-414. Pielou, E . C . 1984. The interpretation of ecological data. John Wiley and Sons, New York. Pollack, R. 1985. Queen Charlotte Islands Forest District forest inventory report no. 1, east coast Graham Island. B.C. Min. For., Int. Rep. Ponnamperuma, F.N. 1984. Effects of flooding on soils. In T .T. Kozlowski (ed.), Flooding 154 and plant growth. Academic Press Inc., Orlando, Florida. Powers, R.F. 1984. Sources of variation in mineralizable soil nitrogen. In Proceedings of the nitrogen asessment workshop. May 19-20, 1982. Pack Forest, Wash. University of Wash. pp. 25-31. Pritchett, W.L. and R.F. Fisher. 1987. Properties and management of forest soils. Second edition. John Wiley and Sons, New York. Reid, D . M . and K J . Bradford. 1984. Effects of flooding on hormone relations. In Flooding and plant growth. T.T. Kozlowski (ed.). Academic Press Inc., Orlando, Florida. Ralston, C W . 1964. Evaluation of forest site productivity. In Int. Rev. For. Res. Vol. 1. J.A. Romberger and P. Mikola (ed.). Academic Press, New York, pp. 171-201. Ray, P .M. 1972. The living plant. 2nd edition. Holt, Rinehart, and Winston, Inc., New York. Richards, L A . (ed.) 1954. Diagnosis and improvement of saline and alkaline soils. U.S. Salinity Lab. Staff. U S D A Agric. Handbook No. 60. Robinson, B., M . C . Feller, and K. Klinka. 1982. Four common mosses as indicators of forest floor acidity in the Coastal Western Hemlock Zone of southwestern British Columbia. Syesis 15:17-23. Roy, R . J J . 1984. Ordination and classification of immature forest ecosystems in the Cowichan Lake area, Vancouver Island. M . Sc. thesis, Fac. of Forestry, Univ. of B.C., Vancouver, B.C. Schmidt, M . G . and W . H . Carmean. 1988. Jack pine site quality in relation to soil and topography in north central Ontario. Can. J. For. Res. 18:297-305. Sexton, R. 1984. Ethylene and abscission. In Ethylene and plant development. J.A. Roberts and G.A. Tucker (ed.). Butterworths, London. Siurtain, M . 1981. Distribution, management, variability, and economics of Sitka spruce in coastal British Columbia. MSc thesis, Fac. of Forestry, Univ. of British Columbia. Slavinski, H . C . 1977. Soil variability along a topographic sequence in the U B C Endowment Lands. MSc thesis, Soil Science Dept., Univ. of British Columbia. Smith, K A . and S.W.F. Restall. 1971. The occurrance of ethylene in anaerobic soil. Jour. Soil. Sci. 22:430-443. Spath, H . 1980. Cluster analysis and algorithms for data reduction and classification and objects. Ellis Horwood Limited, Chichester. 155 Spies, T .A. and B.V. Barnes. 1985. A multifactor ecological classification of the northern hardwood and conifer ecosystems at Sylvania Recreation Area, Upper Peninsula, Michigan. Can. J. For. Res. 15:949-960. Spurr, S.H. and B.V. Barnes. 1980. Forest ecology. 3rd edition. Wiley, New York. Stone, E . L . 1978. A critique of soil moisture - site productivity relationships. In Proceedings, Soil moisture - site productivity symposium. W.E . Balmer (ed.), Myrtle Beach, South Carolina. U S D A For. Serv., pp. 377-387. Sutherland-Brown, A . 1968. Geology of the Queen Charlotte Islands, British Columbia. B.C. Dept. of Mines and Petroleum Res. Bulletin No. 54. Taylor, R . L . and B. MacBryde. 1977. Vascular plants of British Columbia. Tech. Bull. No. 4 Univ. B.C. Press, Vancouver, B.C. Terry, T .A . and J .H. Hughes. 1978. In Proceedings, Soil moisture-site productivity symposium. W . E . Balmer (ed.), Nov. 1-3, 1977, Myrtle Beach, South Carolina. U S D A For. Serv., pp. 148-166. Thomasson, A J . (ed.) 1975. Soils and field drainage. Soil Survey Technical Monograph No. 7. Soil Survey of England and Wales, Harpenden. Titus, S J . 1987. A graphic approach to summarizing forestry data. The For. Chron. 63: 347-350. Tripepi, R.R. and C A . Mitchell. 1984. Stem hypoxia and root respiration of flooded maple and birch seedlings. Physiol. Plant. 60: 567-571. Vomocil, J.A. 1965. Porosity. In Methods of soil analysis. Part 1. C A . Black, D.D. Evans, L . E . Ensminger, J.L. White, and F . E . Clark, (ed.). Amer. Soc. of Agronomy. Madison, Wise, pp. 299-314. Walmsley, M . 1973. Soil-water chemistry relationships and characteristics on a sequence of soil environments. M.Sc. thesis, Soil Science Dept., Univ. of British Columbia. Walmsley, M . , G . Utzig, T. Void, D. Moon, and J. van Barneveld. 1980. Describing ecosystems in the field. R A B Tech. Paper No.2, Min. of Environment and Min. of Forests, Province of B.C. White, E J . 1982. Relationship between height growth of stand and open grown single trees of Scot's pine (Pinus sylvestris) and site factors in Great Britain. Forest Ecol. and Manage. 4:247-259. Whittaker, R . H . 1978. Evaluation of ordination techniques. In Ordination of plant communities. R . H . Whittaker (ed.). Dr. W. Junk bv Publishers, The Hague, The Netherlands. 156 Wilde, S.A. 1958. Forest soils. Ronald Press, New York. Wilkinson, L . 1988. S Y S T A T . The system for statistics. S Y S T A T Inc., Evanston. 111. Worrel, R. 1987. Predicting the productivity of Sitka spruce on upland sites in northern Britain. For. Comm. Bull. No. 72. Zaerr, J.B. 1983. Short-term flooding and net photosynthesis in seedlings of three conifers. For. Sci. 29: 71-78. Zar, J .H. 1984. Biostatistical analysis. 2nd edition. Prentice-Hall, Inc., Englewood Cliffs, N.J. A P P E N D I X A . Alphabetical list of plant species. B A Z Z D E N Bazzania denudata (Torrey ex Gott. et al.) BLECSPI Blechnum spicant (L.) Roth B L E P T R I Blepharostoma trichophyllum (L.) Dum. C A L Y M U Calypogeja muelleriana (Schiffn.) K. Mull. C A L Y T R I Calypogeja trichomanis (L.) Corda C A R E A N T Carex anthoxanthea (Presl) C E P H B I C Cephalozia bicuspidata (L.) Dum. C E P H L U N Cephalozia lunulifolia (Dum.) Dum. C O P T A S P Coptis aspleniifolia Salisb. C Y S T F R A Cystopteris fragilis (L.) Bernh. in Scrad. D I C R F U S Dicranum fuscescens Turn. D I C R S C O Dicranum scoparium Hedw. D R Y O E X P Dryopteris expansa (Presl) Fraser-Jenkins E Q U A R V Equisetum arvense L . G A U L S H A Gaultheria shallon Pursh G Y R O U N D Gyrothyra underwoodiana M . A . Howe H O O K L U C Hookeria lucens (Hedw.) Sm. H Y L O S P L Hylocomium splendens (Hedw.) B.S.G. H Y P N C I R Hypnum circinale Hook. K I N D B E R Kindbergia oregana (Sull.) Ochyra L E P I R E P Lepidozia reptans (L.) Dum. L I S T C A U Listera caurina Piper LISTCOR Listera cordata (L.) R. Br. in Ait. L O B A O R E Lobaria oregana (Tuck.) Mull. Arg. L O P H I N C Lophozia incisa (Schrad.) Dum. L U Z U P A R Luzula parviflora (Ehrh.) Desv. L Y S I A M E Lysichitum americanum Hult. & St. John M A I A D I L Maianthemum dilitatum (How.) Nels. & Macbr. M A L U F U S Malus fusca (Raf.) Schneid. M E N Z F E R Menziesia ferruginea Sm. M O N E U N I Moneses uniflora (L.) Gray P E L L N E E Pellia neesiana (Gott.) Limpr. P E L T A P H Peltigera aphthosa (L.) Willd. P L A G P O R Plagiochila porelloides (Torrey ex Ness) Lindenb P L A G U N D Plagiothecium undulatum (Hedw.) B.S.G. P T E R A Q U Pteridium aquilinum (L.) Kuhn in Decken R H I Z G L A Rhizomnium glabrescens (Kindb.) Kop. R H Y T L O R Rhytidiadelphus loreus (Hedw.) Warnst. S C A P B O L Scapania bolanderi Aust. SPHAGIR Sphagnum girgensohnii Russ. S P H A S Q U Sphagnum squarrosum Crome S T R E A M P Streptopus amplexifolius (L.) D C . V A C C A L A Vaccinium alaskaense How. V A C C P A R Vaccinium parvifolium Sm. in Rees V E R A V I R Veratrum viride Ait. 158 A P P E N D I X B. Conversions from nutrient concentration to kg/ha (modified from Kabzems (1985). To convert soil nutrients expressed as concentrations (percent, ppm, meq/lOOg) to weight (kg) per hectare for a given soil layer. To calculate total nutrient content for a site, combine the kg/ha calculated for the forest floor, with that calculated for the mineral soil to the appropriate rooting depth. 1. Determine the proportion (P) of soil (<2mm fraction) consisting of nutrient "N": i) for ppm: P = ppm N x 106 ii) for %: P = % N x IO"2 iii) for meq/lOOg: P = (meq N/100) x 103 x equivalent wt. x 106 2. Determine the weight of soil (<2 mm fraction) in 1 hectare for the thickness of the soil layer. kg soil/ha = BDcff x IO'3 x 108 x T H x (1-CF) where BDcff is coarse-fragment-free bulk density (Mg/m 3), T H is layer thickness (cm) and C F is the proportion of coarse fragments, by volume. 3. Calculate the weight of nutrient "N" in the soil layer. kg/ha N = P x kg soil/ha These calculation steps can be reduced as follows: For nutrient N: i) expressed in %: kg/ha N = 103 x %N x BDcff x T H x (1-CF) ii) expressed in ppm: kg/ha N = 10"1 x ppmN x BDcff x T H x (1-CF) For exchangeable bases (expressed in meq/lOOg): kg/ha C A : kg/ha C A = 20.04 x meqCA x BDcff x T H x (1-CF) kg/ha M G : kg/ha M G = 12.16 x meqMG x BDcff x T H x (1-CF) kg/ha K: kg/ha K = 39.096 x meqK x BDcff x T H x (1-CF) 159 A P P E N D I X C. Soil chemical data - kg/ha nutrients in the forest floor. Note: See page xii for explanation of codes. Class refers to site index class. PLOT CWSI C l a s s TC TN MINN P CA MG K 1 21 3 119685 2023 48 18 407 245 113 2 20 2 87814 1819 31 12 391 231 88 3 22 3 51566 1102 28 13 333 136 70 4 18 2 82737 1403 53 17 304 166 111 5 21 3 79679 1814 36 10 290 198 68 6 16 2 76057 1242 23 13 229 206 119 7 14 1 74368 1496 26 12 203 188 96 8 18 2 70881 1116 28 7 273 172 96 9 11 1 86391 1411 27 11 347 181 . 124 10 23 3 123831 2366 37 10 449 285 137 11 15 1 46701 920 14 5 205 117 52 12 22 3 75248 1438 34 11 276 159 82 13 15 1 108284 1739 32 21 435 238 ' 139 14 22 3 66242 1265 45 7 413 120 117 15 18 2 75153 1364 44 8 296 142 107 16 15 1 74951 1364 36 11 216 146 116 17 18 2 65824 1213 41 10 283 113 98 18 18 2 93380 1401 37 8 361 226 110 19 15 1 46751 916 27 6 242 85 68 20 17 2 72912 1178 36 12 343 173 129 21 14 1 65797 1305 31 11 355 156 78 22 15 1 48503 892 17 6 190 88 66 23 25 3 78222 2240 64 5 605 192 70 24 23 3 93491 1831 39 6 429 292 109 160 A P P E N D I X D . Soil chemical data - kg/ha nutrients in the mineral soil to the secondary rooting depth Note: See page xii for explanation of codes. Class refers to site index class. Plot CWSI Class TC TN MINN P CA MG K 1 21.2 3 77207 2167 30 8 122 217 70 2 19.9 2 67607 1333 21 5 74 87 39 3 21.9 3 94956 2457 25 5 269 388 119 4 17.6 2 38945 856 17 5 29 57 41 5 21.1 3 115713 3135 51 8 144 229 77 6 16.4 2 59097 1477 12 5 54 200 72 7 13.7 1 4493 109 2 0 3 4 2 8 18. 3 2 53761 1213 14 3 81 108 57 9 10.9 1 54650 1228 14 4 36 42 39 10 23.3 3 114915 3075 30 2 28 184 97 11 14.6 1 57596 1558 23 1 77 87 33 12 21.8 3 88978 2650 50 2 54 156 121 13 14.5 1 64318 1130 12 3 39 118 55 14 21.5 3 71273 1944 45 4 214 227 92 15 18. 0 2 88901 2263 37 5 69 115 76 16 14.7 1 86950 2090 32 5 50 81 71 17 17.8 2 112852 2590 32 8 224 260 109 18 18.0 2 55571 1179 19 3 35 143 60 19 14.9 1 36052 751 9 1 30 35 18 20 16. 6 2 116591 1683 29 5 108 239 109 21 13.5 1 61133 1210 15 2 28 48 33 22 15.3 1 99877 2014 34 2 25 60 48 23 25.1 3 103762 4573 161 2 60 392 199 24 22.5 3 163894 4583 57 3 147 266 82 161 A P P E N D I X E . Soil chemical data - kg/ha nutrients in the combined forest floor plus mineral soil to the secondary rooting depth. Note: See page xii for explanation of codes. Class refers to site index class. PLOT CWSI Class TC TN MINN P CA MG K 1 21 3 196892 4190 78 25 529 462 184 2 20 2 155421 3152 52 17 465 319 127 3 22 3 146523 3559 54 18 602 524 189 4 18 2 121682 2259 70 22 333 223 152 5 21 3 195392 4949 86 18 433 427 146 6 16 2 135154 2719 35 18 283 405 191 7 14 1 78861 1604 28 12 207 193 98 8 18 2 124642 2329 41 10 354 280 152 9 11 1 141041 2639 40 15 384 223 162 10 23 3 238746 5441 68 12 477 468 234 11 15 1 104297 2478 37 6 282 203 85 12 22 3 164227 4088 84 13 331 315 203 13 15 1 172601 2869 44 25 473 357 194 14 22 3 137515 3209 90 11 627 347 209 15 18 2 164054 3627 80 13 365 256 183 16 15 1 161901 3455 69 16 266 227 187 17 18 2 178677 3804 73 18 507 374 207 18 18 2 148951 2580 57 11 396 368 170 19 15 1 82803 1667 36 7 272 120 86 20 17 2 189503 2862 65 16 452 412 238 21 14 1 126929 2515 47 13 383 204 111 22 15 1 148379 2906 52 8 215 148 113 23 25 3 181984 6813 225 6 . 665 584 269 24 23 3 257385 6413 96 9 576 558 191 162 A P P E N D I X F. Soil chemical data - p H and C / N ratios for the forest floor and root zone mineral soil. Note: See page xii for explanation of codes. Class refers to site index class. PLOT CWSI CLASS HUMPH HUMCN MINPH MINCN 1 21.2 3 3.2 59.5 3.4 35.2 2 19.9 2 3.3 48. 6 3.3 51.8 3 21 .9 3 3.6 46.9 3.8 3.8.6 4 17.6 2 3.2 59.3 3.3 44.4 5 2 1 . 1 3 3.2 44 .1 3.6 37.2 6 16.4 2 3.2 61.5 3.5 40 .0 7 13.7 1 3 .1 50.7 3.4 41.3 8 18.3 2 3 . 3 63 . 6 3.5 43.9 9 10.9 1 3.2 61.7 3.3 44.4 10 23.3 3 3.2 52.4 3.5 37.3 11 14.6 1 3.5 50.9 3.7 36.8 12 21.8 3 3.2 52.9 3 . 3 33.9 13 14 . 5 1 3.2 62 . 4 3 .1 56.5 14 21 .5 3 3.6 52.7 3.5 37 .1 15 18. 0 2 3.2 55.3 3 . 3 39.2 16 14.7 1 3.1 55.0 3.3 41 .6 17 17.8 2 3.3 54.3 3.4 43.9 18 18 .1 2 3.2 66.8 3.3 47.3 19 14.9 1 3.3 51 .1 3.3 48 .0 20 16.6 2 3.4 62.2 3.4 70.9 21 13.5 1 3 . 3 50. 5 3 . 3 51.4 22 15.3 1 3.2 54.4 3.2 49 . 7 23 25 .1 3 3.7 34.9 4 . 1 22.7 24 22 .5 3 3.5 51. 3 3 . 5 36.6 163 A P P E N D I X G . Soil physical properties. Note: See page xii for explanation of codes. Class refers to site index class; B D M I N and B D F F refer to bulk density of root zone mineral soil and forest floor respectively, expressed in M g / m 3 . V M C and AIR are expressed as proportions; DH20S, D2ROOTS, F F D E P , and C A P in cm; and SAND, SILT, C L A Y in percent. PLOT GC CUSI VMC AIR BDMIN BDFF DH20S D2R00TS FFDEP CAP SAND SILT CLAY 1 3 2 1 . 2 0 . 5 8 0 . 0 7 0 .90 0 .12 47 37 22 4 0 . 8 3 4 . 0 2 5 . 2 2 2 19 .9 0 . 6 9 0 .10 0.51 0.11 34 34 18 18 25 .2 5 3 . 7 21.1 3 3 2 1 . 9 0 . 5 6 0 .14 0 . 7 7 0.12 26 39 10 12 2 6 . 6 53 .2 20 .2 4 2 17 .6 0 .58 0 .12 0 . 7 8 0.12 47 29 16 21 52 .4 3 2 . 6 15.0 5 3 21.1 0 .59 0 .18 0 .59 0.14 32 37 13 17 18.5 55 .2 26 .3 6 2 16.4 0 .60 0 .15 0.65 0 .10 36 36 16 19 2 5 . 8 5 1 . 4 2 2 . 8 7 1 1 3 . 7 0 .84 0 . 0 9 0.14 0.11 34 35 15 42 9 . 9 5 1 . 8 3 8 . 3 8 2 18 .3 0 .50 0 .23 0 .70 0.12 57 31 13 13 4 8 . 3 3 9 . 2 12.5 9 1 10 .9 0 .65 0 .12 0 .55 0.11 29 26 18 12 3 0 . 0 5 0 . 2 19 .8 10 3 23 .3 0 . 4 8 0 . 2 7 0 .63 0 .13 88 50 21 15 2 4 . 3 4 8 . 9 2 6 . 8 11 1 14.6 0.81 0 . 0 6 0.31 0.11 21 22 10 15 15.2 53.1 3 1 . 7 12 3 2 1 . 8 0.54 0 . 1 8 0 . 9 9 0 .13 84 37 14 13 2 6 . 9 50 .2 2 2 . 9 13 1 14.5 0 .60 0.11 0.72 0.14 26 25 18 12 2 4 . 6 55 .2 2 0 . 2 14 3 21 .5 0 .42 0 .22 0 .92 0 .13 121 45 12 6 2 . 7 2 6 . 8 10.5 15 2 18.0 0 .54 0 .08 0 . 9 7 0.12 42 30 14 11 2 3 . 9 5 3 . 4 2 2 . 7 16 1 1 4 . 7 0 . 6 8 0 .12 0 .50 0 .13 30 30 13 16 24 .5 5 3 . 2 2 2 . 3 17 2 17 .8 0 . 5 7 0 .14 0.74 0 .10 79 41 14 13 2 1 . 9 5 4 . 7 23 .4 18 2 18.1 0 .52 0 . 1 6 0 .83 0.11 106 34 19 9 3 9 . 3 4 5 . 6 15.1 19 1 14 .9 0 .74 0.11 0 . 3 8 0 . 0 9 29 21 11 18 26 .5 26 .5 2 0 . 0 20 2 16.6 0 . 4 7 0 . 1 8 0 . 8 8 0.11 90 42 15 13 3 4 . 0 52 .0 14.0 21 1 13.5 0 .73 0 .09 0 .42 0.12 42 20 12 17 20 .3 59 .2 20 .5 22 1 15 .3 0.81 0 .09 0.22 0.11 30 29 10 19 12.4 4 9 . 7 3 7 . 9 23 3 25.1 0 .45 0 .24 0 .84 0.21 96 34 9 15 28 .2 4 5 . 6 2 6 . 2 24 3 22 .5 0 . 6 6 0 .14 0 . 4 6 0.11 69 48 19 20 2 5 . 2 4 7 . 8 2 7 . 0 -164 A P P E N D I X H . Categorical variables. Note: See page xii for explanation of codes. CLASS refers to site index class. Categories are as follows: H 2 0 S T - 1= absent, 2=slow, 3=moderate, 4=rapid; R O O T - l=very few, 2=few, 3= plentiful, 4 = abundant; CONS - 0=firm or friable, l=soft; C E M -0=strongly cemented horizon absent, 1 = cemented horizon present; T E X P M - 1 = C, 2 = SC, 3 = C L , 4 = SCL, 6 = SIL, 7 = SIL+ (high clay), 8 = SL, 9 = LS; S H A P E -1 = depression, 2=level, 3 = sloping, 4 = convex, 5=receiving; P M - 1 = alluvial, 2=fluvial veneer over morainal, 3=lacustrine veneer over morainal, 4=glaciofluvial, 5=morainal; H U E and C H R O M A according to Munsell notation. PLOT CWSI CLASS H20ST ROOT CONS CEM TEXPM SHAPE PM HUE CHROMA 1 21.2 3 2 3 0 0 6 1 1 2.5 2 2 19.9 2 2 3 0 0 4 4 5 3.5 2 3 21.9 3 2 3 0 0 4 5 5 3.5 2 4 17.2 2 1 3 0 0 3 1 2 4.5 2 5 21.1 3 2 4 0 0 8 5 5 3.5 2 6 16.4 2 2 4 0 0 4 1 3 3.0 2 7 13.7 1 3 1 1 0 4 2 2 3.0 2 8 18.3 2 1 4 0 0 8 3 5 4.5 1 9 10.9 1 2 3 1 1 7 2 4 4.0 1 10 23.3 3 1 4 0 0 7 2 5 3.0 2 11 14.6 1 3 2 1 0 1 2 5 4.0 2 12 21.8 3 0 3 0 1 9 4 5 3.5 2 13 14.5 1 2 4 0 0 2 1 5 4.0 1 14 21.5 3 0 4 0 1 8 4 5 5.0 2 15 18.0 2 1 4 0 1 3 4 5 4.0 2 16 14.7 1 2 3 0 1 4 2 5 4.0 2 17 17.8 2 0 3 0 0 1 4 5 4.0 2 18 18.1 2 0 4 0 1 4 3 5 5.0 1 19 14.9 1 2 2 1 1 4 2 5 4.0 1 20 16.6 2 0 4 0 1 4 3 5 5.0 2 21 13.5 1 1 3 1 1 4 2 5 4.0 1 22 15.3 1 3 2 1 1 4 2 5 3.0 2 23 24.8 3 0 4 0 1 9 5 2 3.0 2 24 22.5 3 1 4 0 0 4 5 5 3.5 2 165 APPENDIX I. Soil oxidation zone index and soil water quality. Note: See page xii for explanation of codes. CLASS refers to site index class. DO in ppm, and SC in mS/m >LOT CWSI CLASS AZ1 AZ2 BZ1 BZ2 BZ3 DO SC 1 21.2 3 25 45 16 36 45 3.2 0.194 2 19.9 2 30 40 17 35 39 4.3 0.191 3 21.9 3 29 53 8 39 53 8.4 0.144 4 17.6 2 24 34 12 25 34 4 .5 0.095 5 21 .1 3 29 47 12 35 47 3.4 0.191 6 16.4 2 31 48 14 38 48 4 .7 0.194 7 13.7 1 13 18 9 15 18 3 .5 0.113 8 18. 3 2 34 1 4 4 11 30 44 6.7 0.312 9 10.9 1 27 \ 50 14 35 50 3 .6 0.081 10 23.3 3 50 63 24 45 63 11 14. 6 1 17 28 8 21 28 3.6 0.113 12 21.8 3 38 45 17 32 45 13 14.5 1 29 39 10 24 39 4.6 0.207-14 21.5 3 49 51 14 44 51 15 18.0 2 24 37 10 28 37 4 . 3 0. 081 16 14.7 1 17 30 10 23 30 4 .9 0. 084 17 17.8 2 29 49 14 33 49 5.9 0.102 18 18 .1 2 40 49 13 34 49 19 14.9 1 16 28 9 18 28 4 .9 0.162 20 16.6 2 54 62 17 45 62 21 13.5 1 21 29 8 23 29 7.2 0.178 22 15.3 1 17 28 10 19 28 3.4 0.145 23 25 .1 3 45 47 7 29 47 4.3 0.132 24 22.5 3 39 50 14 42 50 A P P E N D I X J. Site index data. Note: See page xii for explanation of codes. CLASS refers to site index class. PLOT CLASS CWSI HWSI SSSI PLSI 1 3 21.2 22.5 27.0 2 2 19.9 20.2 19.1 22.7 3 3 21.9 13.7 20.8 23.2 4 2 17.2 17.1 16.5 19.7 5 3 21.1 21.2 21.9 23.2 6 2 16.4 15.7 18.1 7 1 13 .5 13 .4 13.6 8 2 18.3 16.7 17.7 9 1 10.9 8 . 8 5.2 11.0 10 3 23.3 25.5 30.7 23.7 11 1 14.6 13.0 11.4 16.2 12 3 21.8 22.7 20.5 22.2 13 1 14.5 12 . 9 14.3 14 3 21.5 23.5 24.5 20.5 15 2 18.0 17.5 15.7 18.8 16 1 14.7 13 . 3 9.4 14.1 17 2 17.8 17 . 6 15.7 18.4 18 2 18.1 15.5 14.0 16.3 19 1 14.9 11.5 8.3 12.5 20 2 16. 6 14.0 15.8 17.1 21 1 13.5 11.5 12.8 22 1 15.3 11.5 11.3 12.8 23 3 24.8 22.1 28.0 22.9 24 3 22.5 22 . 0 31.0 . 20.0 167 A P P E N D I X K. Mensuration data - combined for all species. Note: See page xii for explanation of codes. CLASS refers to site index class. M A I in m3ha" V ; V O L (volume) in m 3/ha; B A (basal area) in m 2/ha; Q M D (quadratic mean diameter) in cm; S T H A (stems/ha). Al l volumes are total standing volume > 7.5cm dbh of live trees. PLOT CLASS CWSI AGE MAI VOL BA QMD STHA 1 3 21.2 123 9.9 1216 92 38.9 775 2 2 19.9 107 10.6 1136 99 37.4 900 3 3 21.9. 129 14.3 1849 149 44.1 975 4 2 17.6 98 8.2 808 75 27.6 1250 5 3 21.1 103 11.7 1203 99 40.4 825 6 2 16.4 102 10.9 1112 102 25.5 2000 7 1 13.7 111 4.6 507 60 20.3 1900 8 2 18.3 99 7.3 727 69 24.5 1475 9 1 10.9 165 3.9 643 88 22.2 2275 10 3 23.3 75 14.3 1073 85 29.7 1225 11 1 14.6 73 6.5 472 58 19.9 1875 12 3 21.8 78 11.5 899 80 26.5 1500 13 1 14.5 103 5.4 561 67 20.0 2125 14 3 21.5 78 13 . 0 1013 81 35.3 825 15 2 18.0 113 7.0 796 71 27.2 1225 16 1 14.7 111 6.5 723 86 20.9 2500 17 2 17.8 110 11.2 1234 111 28.5 1750 18 2 18.0 103 8.8 907 85 24.9 1750 19 1 14.9 98 4.0 396 53 17.3 2225 20 2 16.6 106 6.3 670 65 22.5 1625 21 .1 13.5 107 4.8 514 62 20.3 1925 22 1 15.3 106 4.4 461 57 18 . 4 2150 23 3 25.1 75 11.8 886 77 35.7 775 24 3 22.5 104 12.9 1340 103 41.1 775 168 APPENDIX L. Mensuration data - by species. Note: See page xii for explanation of codes. CLASS refers to site index class. VOL, BA, QMD, STHA as above for each species indicated by preceding two letter code. .OT CLASS CWVOL CUBA CWQMD CUSTHA HUVOL HWBA HUQMD HWSTHA PLVOL PLBA PLQMD PLSTHA SSVOL SSBA SSQMD SSSTHA 1 3 141 13 40 .9 100 609 44 34.5 475 0 0 0. .0 0 466 35 46.9 200 2 2 457 45 40 .6 350 500 40 32.1 500 0 0 0. .0 0 179 13 57.5 50 3 3 708 64 45 .3 400 151 11 30.8 150 297 24 39, .0 200 694 50 53.0 225 4 2 318 32 26 .8 575 261 24 25.2 475 220 18 36. .2 175 9 1 21.0 25 5 3 463 44 36 .2 425 92 7 33.1 75 320 25 46. .1 150 327 24 49.3 125 6 2 376 38 25 .0 775 282 28 19.4 950 455 36 40. .9 275 0 0 0.0 0 7 1 191 23 20 .6 700 181 22 18.1 850 135 15 25. .0 300 0 0 0.0 0 8 2 324 33 24 .5 700 263 24 24.0 525 139 13 25. .2 250 0 0 0.0 0 9 1 259 43 26 .5 775 227 28 16.9 1250 157 17 29. .5 250 0 0 0.0 0 10 3 88 9 21 .6 250 463 37 25.8 700 15 1 26. ,2 25 506 38 43.8 250 11 1 266 33 20 .7 975 39 6 11.5 575 154 18 31. .8 225 12 1 13.3 100 12 3 409 41 27 .7 675 421 33 26.0 625 30 3 21. .6 75 39 4 24.8 75 13 1 237 29 20 .6 875 105 15 14.5 875 216 23 29. .0 350 2 0 12.0 25 14 3 152 16 25 .6 300 347 25 35.8 250 49 4 32, .5 50 466 36 45.1 225 15 2 145 15 27.4 250 443 39 25.7 750 121 10 32. .2 125 86 7 30.2 100 16 1 304 38 20 .4 1150 188 23 18.0 900 232 25 26. .6 450 0 0 0.0 0 17 2 352 36 28 .1 575 363 32 25.5 625 439 37 31. .6 475 78 7 33.3 75 18 2 286 29 25 .5 575 152 16 18.6 575 413 36 28. .2 575 56 4 46.8 25 19 1 243 33 19 .9 1050 94 13 13.4 925 59 7 18.6 250 0 0 0.0 0 20 2 194 20 23 .1 475 294 29 20.4 875 158 14 27.9 225 24 2 23.5 50 21 1 266 32 22 .7 800 154 20 16.0 975 95 10 29.3 150 0 0 0.0 0 22 1 285 35 19 .7 1150 69 10 12.7 800 108 12 27. .1 200 0 0 0.0 0 23 3 430 41 36 .2 400 147 12 29.0 175 77 6 39. .6 50 232 19 39.6 150 24 3 305 29 35 .1 300 400 29 36.4 275 58 5 35. .1 50 577 41 58.7 150 169 A P P E N D I X M . Explanation of boxplots. A boxplot is a simple graphical summary of a batch of data. Based on the median and other percentiles of data distribution, boxplots provide more information about the nature of one or several data sets than traditional summary statistics such as the mean and standard deviation. (Titus 1987). The following figure compares a histogram and boxplot for within-plot CVs for exchangeable Ca in the mineral soil, across the 24 plots. In the boxplot, the vertical line within the box represents the median. The left and right sides of the box represent the 25 and 75 quartiles, respectively. The lower (left) and upper (right) whiskers extend to the lower and upper 1/5 percentiles respectively, or to extreme values. Values outside the inner fence (the 25 quartile minus 1.5 times the interquartile range, and the 75 quartile plus 1.5 times the interquartile range) are plotted with an asterisk. Empty circles represent values beyond three times the interquartile range on either side of the box (Wilkinson 1987). z: CO o 0.0 0.2 0.4 0.6 0.8 1.0 170 A P P E N D I X M . Procedure for checking assumptions of parametric multiple comparison test. The Tukey H S D test is relatively robust to violations of the basic assumptions for parametric analysis, however substantial departures from the assumptions can affect its performance (Zar 1984). Examination of residuals from a one-way A N O V A was the basic method used to detect potentially serious problems of heteroscedasticity (Wilkinson 1987). The following procedure was used to check the assumptions. 1. If residuals do not display a problematic pattern, use the parametric test. 2. If residuals show a problematic pattern, transform the data. If transformation improves the residuals, and yields the same outcome as the untransformed data, assume the test is robust, and use the untransformed data with the parametric test. 3. If transformation improves the residuals but yields a different result from the untransformed data, assume the test is not robust, and use the transformed data with the parametric test. 4. If transformation does not improve the residuals, use a nonparametric test on the original data. If this yields the same result as the parametric test, assume the original test is robust, and use the parametric test on the untransformed data. 5. If transformation does not improve the residuals, and the nonparametric test results on the original data differ from the parametric test, assume the parametric test is not robust, and use the nonparametric test. 


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items