CHEMICAL TERRAIN VARIABILITY: A GEOMORPHOLOGICAL APPROACH USING NUMERICAL AND REMOTE SENSING TECHNIQUES b V HANSPETER SCHREIER B.A., Un iver s i t y Df Colorado, 1970 D ip. Photo I n te rp . , I .T.C. D e l f t , 1972 M.Sc., Un iver s i t y of S h e f f i e l d , 1973 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE-DF DOCTOR OF PHILOSOPHY i n THE FACULTY OF GRADUATE STUDIES DEPARTMENT OF GEOGRAPHY We accept t h i s thes i s as conforming tD the requ i red standard THE UNIVERSITY OF BRITISH COLUMBIA December, 1976 (o) Hanspeter Schreier, 1976 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of Brit ish 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 The University of Brit ish Columbia 2075 Wesbrook P l a c e Vancouver, Canada V6T 1W5 Date ABSTRACT The v a r i a b i l i t y of chemical parameters over the landscape was examined i n t h i s research. A t e r r a i n hierarchy based on genetic geo-morphological unit concepts was developed i n two Quaternary landscapes i n the Fraser Valley and i n the Peace River area i n B r i t i s h Columbia. The r e l a t i v e v a r i a b i l i t y within and betueen d i f f e r e n t h i e r a r c h i c a l units ranging from " s i t e s " to "landform units" to "landform unit types" was compared. The v a r i a b i l i t y was large i n a l l units but was smaller at the s i t e scale than at the landform unit scale, within single landform units chemical parameters were shown to be c l o s e l y r e l a t e d to type of drainage. A v a i l a b l e Ca, Mg, Na, H, and S i were found to be the most important d i f f e r e n t i a t i n g parameters for a l l u n i t s . S i t e categories which r e f l e c t e d units of s i m i l a r parent material, form, and i n f e r r e d genesis were determined by a p p l i c a t i o n of a c l u s t e r a n a l y s i s procedure. S i t e s grouped by t h i s method were not coextensive with i n d i v i d u a l land-form units, thus suggesting that the environmental imprint on the.chemical conditions was not as strong as that of the genesis. The best grouping was obtained with the Peace River data where more natural conditions p r e v a i l e d . A data screening through f a c t o r analysis p r i o r to the grouping improved the landform unit type c l a s s i f i c a t i o n i n the Fraser V a l l e y where the chemical conditions were complicated by a more complex and intensive land use pattern. M u l t i s p e c t r a l remote sensing techniques were used to assess the p o t e n t i a l of p r e d i c t i n g chemical ground conditions from s p e c t r a l measure-ments. M u l t i s p e c t r a l photography combined with density s l i c i n g and addi t i v e color viewing techniques were used to quantify chemical ground conditions on the photographic image. Areas'of d i f f e r e n t s o i l moisture and percent Carbon content could r e a d i l y be i d e n t i f i e d and quantified by t h i s means. Exchangeable Ca, Mg, and Na could p a r t i a l l y be d i f f e r e n -t i a t e d probably as a r e s u l t of d i r e c t c o r r e l a t i o n with Carbon. The wider band photography (color k00-700 nm and color IR 500-900 nm wavelength range) produced better r e s u l t s than the narrow band black and white images i i (50D-6QD nm and 6DD-7DD nm wavelength range). The trends i n detecting chemicals were consistent f o r both vegetated and non-vegetated surfaces; the s l i c e d c o l o r f i l m image was s l i g h t l y more useful f o r analyzing ex-posed s o i l surfaces, while the s l i c e d color IR image proved to be more useful f o r the i n t e r p r e t a t i o n of vegetated surfaces. Direct d i g i t a l r e f l e c t i o n measurements were made with a m u l t i -channel spectrometer from the a i r , and D n s o i l samples on the ground and i n the laboratory. In the f i e l d the A-DO-1000 nm s p e c t r a l wavelength range was used and the laboratory analysis was extended to the 35D-25D0 nm wavelength range. Only bare s o i l surfaces were investigated. Corre-l a t i o n and regression analysis revealed that % Carbon, % Fe, exchangeable Mg, and exchangeable K could be predicted from s p e c t r a l r e f l e c t i o n values. There i s evidence thattthe spectral-chemical r e l a t i o n s h i p follows a c u r v i l i n e a r function, but adequate predictions were obtained with l i n e a r r e l a t i o n s h i p s at low chemical concentration l e v e l s . Despite differences i n measuring techniques s i m i l a r regression trends were obtained f or a l l three methods and the 50D-11D0 nm wavelength range was found to be most useful i n t h i s a n a l y s i s . The t o t a l s p e c t r a l reflectance curve was found to be of importance since s o i l s from s i m i l a r parent materials produced c h a r a c t e r i s t i c curves which could r e a d i l y be d i f f e r e n t i a t e d by a l l three types of measurements. i i i TABLE OF CONTENTS Page Abstract i L i s t of Figures v i i L i s t of Tables i x L i s t of Plates x L i s t of Symbols and Abbreviations x i Acknowledgements x i i INTRODUCTION 1 A. Aims of Study B. Organisation 3 C. Basic P r i n c i p l e s ^ CHAPTER I METHOD AND TECHNIQUES 9 A. H i e r a r c h i c a l Framework 9 1. H i e r a r c h i c a l Structure 2. I d e n t i f i c a t i o n of Units 3. S e l e c t i o n of Parameters B. Choice and Description of F i e l d Area 13 1. Salmon Basin Study Area 2. Peace River Study Area 3. Comparison between Genesis of Sample Units k. F i e l d Targets f o r Airborne S p e c t r a l Mission C. E s s e n t i a l Sampling and Laboratory Techniques . . . . 2h 1. Sampling Scheme 2. Laboratory Techniques D. Numerical Methods 28 1. Euclidean Grouping Method 2. Factor Analysis E. Remote Sensing Method 31 1. M u l t i s p e c t r a l Photography 2. Spectral R e f l e c t i o n Measurements i n the Laboratory 3. F i e l d Spectral Analysis i v Page CHAPTER II HIERARCHICAL TERRAIN UNITS . kZ A. The Internal V a r i a b i l i t y Df Units at D i f f e r e n t H i e r a r c h i c a l Levels h3 1. Internal S i t e V a r i a b i l i t y 2. Comparison of V a r i a b i l i t y between H i e r a r c h i c a l Units B. The S i t n i f i c a n c e of Selected Parameters i n Characterizing Units 52 1. Parameter Si g n i f i c a n c e at S i t e Level 2. Parameter Si g n i f i c a n c e at d i f f e r e n t H i e r a r c h i c a l Levels C. The Internal Homogeneity of the Basic Units and t h e i r S u i t a b i l i t y f o r Numerical Treatment 56 1. Numerical Grouping using Single Horizon Data 2. Numerical Grouping using Data from A l l Horizons CHAPTER III NUMERICAL APPLICATIONS 61 A. Numerical Treatment of Fraser Valley Data 61 1. Direct Numerical Grouping using Parameters Selected through S i g n i f i c a n c e Test 2. Numerical Grouping a f t e r Parameter Screening through Factor Analysis B. Numerical Treatment of Peace River Area Data . . . . 67 1. Direct Numerical Grouping 2. Direct Numerical Grouping using C Horizon Data Only 3. Numerical Grouping a f t e r Parameter Screening through Factor Analysis C. Comparison of the Two Areas and Assessment of Results . 75 1. Comparison of the Two Areas 2. Comparison of Results 3. Implications and Conclusions CHAPTER IV FACTORS AFFECTING PARAMETER VARIABILITY . . . . 82 A. Macro- to Meso-Scale Analysis 82 1. Parent Material 2. Time Factor 3. Process Oriented Factors B. Meso-Scale Investigations 89 1. Correlations amongst Parameters 2. Mean Values V Page C. Meso- to Micro-Scale Investigation 93 1. The Use of Multiparameter Data i n Assessing Single Landform Units 2. The Addition of New Parameters to the Numerical Grouping CHAPTER V REMOTE SENSING: MULTISPECTRAL PHOTOGRAPHY . . . 101 1. Remote Sensing A p p l i c a t i o n 2. M u l t i s p e c t r a l Photography B. Image Qu a n t i f i c a t i o n through Density S l i c i n g . . . . 105 1. Vegetated Surface 2. Bare S a i l Surface 3. Discussion and Results of Procedure C. Image Enhancement through Additive Color Viewing . . . 114 1. Introduction 2. Vegetated Surface 3. Bare S o i l Surface k. Discussion of Results and Procedure D. Conclusion 120 CHAPTER VI REMOTE SENSING: DIGITAL MULTISPECTRAL SENSING . 122 A. Introduction 122 B. Airborne Spectral Analysis 123 1. Assessment of Ground Control and Spectral Values 2. Contrasting Average per F i e l d Values 3. Analysis of S i t e S p e c i f i c Data k. Summary of Airborne Mission C. Ground Spectral Measurements 13B 1. Instrument Assessment 2. The E f f e c t of Parameter A l t e r a t i o n on Spectral Reflectance 3. Comparison between Spect r a l and Chemical Data k. Summary of Ground Measurements D. Laboratory Measurements 148 1. Extractions of Exchangeable Cations 2. Parent Material E f f e c t 3. Comparison between Spectral and Chemical Data k. Summary of Laboratory Analysis D. Conclusion 100 A. Introduction 101 v i Page E. Comparison between Airborne, Ground, and Laboratory Spectral Measurements 157 1. Spectral Curves 2. Correlations 3. Regressions F. Summary 162 1. Predictions 2. Relative Importance of Individual Parameters 3. Spectral Curves k. Types of Measurements CONCLUSIONS 166 APPENDIX I LITERATURE REVIEW - SOIL CHEMICAL VARIABILITY AS RELATED TO GEOMORPHOLOGY. . 173 A. Introduction B. S o i l Geomorphological Relationship C. S o i l Landscape V a r i a b i l i t y D. Conclusions APPENDIX II A REVIEW OF MULTISPECTRAL REMOTE SENSING AS APPLIED TO CHEMICAL TERRAIN ANALYSIS 186 A. Introduction B. Multispectral Photography C. Multispectral Sensing D. Conclusions APPENDIX;III SUPPLEMENTARY INFORMATION ON NUMERICAL GROUPING . 200 APPENDIX IV SUPPLEMENTARY DATA ON DENSITY SLICING AND ADDITIVE COLOR VIEWING 207 APPENDIX V SUPPLEMENTARY DATA ON CHEMICAL PREDICTIONS FROM SPECTRAL REFLECTION MEASUREMENTS . . . . 214 BIBLIOGRAPHY 218 v i i LIST DF FIGURES Figure Page 1 Organisational flow diagram 5 2 Overview of d i f f e r e n t operational t e r r a i n analysis schemes 7 3 H i e r a r c h i c a l framework 10 4 Comparison between d i f f e r e n t h i e r a r c h i c a l schemes 10 5 Geomorphological units i n the Salmon Basin 15 6 Geological cross section of Peace River study area 19 7 Geomorphological units i n the Peace River area •19 8 Landform unit comparison i n the two study"areas 22 9 Sampling design 25 ID Euclidean grouping procedure 28 11 Projections i n f a c t o r a n a l y s i s 30 12 Remote sensing procedure 32 13 Spectral s e n s i t i v i t y range for d i f f e r e n t methods used 32 14 Spectral transmission properties of f i l t e r s 33 15 Sample d i s t r i b u t i o n i n d e t a i l e d s i t e analysis 43 16 Comparison of s i t e v a r i a b i l i t y using s i g n i f i c a n c e t e s t 45 17 . Range of s i t e v a r i a b i l i t y 46 IB H i e r a r c h i c a l d i s t r i b u t i o n 47 19 Elements showing a s i g n i f i c a n t l y d i f f e r e n t d i s t r i b u t i o n 48 when compared at landform unit l e v e l 2D S i g n i f i c a n t difference between type of landform units 48 21 C o e f f i c i e n t of v a r i a t i o n f o r d i f f e r e n t h i e r a r c h i c a l units 50 22 C o e f f i c i e n t of v a r i a t i o n f o r d i f f e r e n t h i e r a r c h i c a l units 51 23 Dendrogram r e s u l t i n g from numerical grouping of 57 A horizon data 24 Dendrogram f o r samples from B horizons 58 25 Dendrogram f o r samples from C horizons 59 26 Dendrogram f o r the combined A , B, and C horizon 60 samples 27 S p a t i a l grouping of s i t e s a f t e r f a c t o r analysis 68 28 Units and sample d i s t r i b u t i o n for the Peace River study area 69 29 Grouping of s i t e s using C horizon data 73 30 Grouping of s i t e s a f t e r f a c t o r analysis 74 31 Results of s i g n i f i c a n c e t e s t 84 32 Results of s i g n i f i c a n c e t e s t between chemical parameters of d i f f e r e n t landform units 85 33 Correlations with slope values 9D 34 Correlations r e l a t e d to elevation 91 35 T i l l unitT-Z, chemical s i t e types r e s u l t i n g from 94 numerical grouping 36 T i l l unit T2, chemical d i s t r i b u t i o n 95 37 Grouping of s i t e s i n landform unit LT3 98 v i i i 38 S i t e l o c a t i o n and grouping 99 39 Range and v a r i a b i l i t y of d i f f e r e n t groups 99 40 Influence of vegetation of r e f l e c t i o n 124 41 Spectral r e f l e c t i o n range of f i e l d Gl to G4 125 kZ Mean target r e f l e c t i o n curve 127 43 Spectral r e f l e c t i o n curves f or s o i l s from d i f f e r e n t parent materials 128 kk Correlations of average target values 131 45 Correlations amongst parameters using s i t e s p e c i f i c data 132 46 S i g n i f i c a n t c o r r e l a t i o n s amongst s i t e s p e c i f i c parameters 133 kl S i g n i f i c a n t c o r r e l a t i o n s amongst parameters using average target data 133 48 Spectral and chemical c o r r e l a t i o n s of target G and H 13k 49 Regression of exchangeable Mg with r e f l e c t i o n values at 650. nm wavelength 135 50 a & b Reproduc i b i l i t y of measurements 139 51 Relationship between percent Carbon and spe c t r a l reflectance 141 52 a & b E f f e c t of p a r t i c l e s i z e on sp e c t r a l r e f l e c t i o n 142 53 R e f l e c t i o n before and a f t e r extraction 143 54 a & b Parent material e f f e c t 145 55 S i g n i f i c a n t c o r r e l a t i o n s between chemical and s p e c t r a l reflectance 146 56 S i g n i f i c a n t c o r r e l a t i o n amongst chemical parameters 146 57 Comparison between USGS and USDA measurements 149 58 a & b Spectr a l response before and a f t e r extraction 150-51 59 a & b Parent material e f f e c t (USGS and USDA measure-ments) 153-54 60 Co r r e l a t i o n between chemical and laboratory s p e c t r a l measurements 155 61 a %• d Comparison between d i f f e r e n t types of measure-ments 159-60 62 Co r r e l a t i o n trends 161 63 Standard error 164 64 Location of s i t e s i n Fraser Valley study area 200 65 Dendrogram of d i r e c t grouping of s i t e s 201 66 S p a t i a l grouping of s i t e s i n Fraser V a l l e y area 202 67 Dendrogram of s i t e grouping a f t e r f a c t o r analysis 203 68 Dendrogram of d i r e c t grouping of s i t e s 204 69 Dendrogram of d i r e c t grouping of s i t e s using C horizon data 205 70 Dendrogram of numerical grouping a f t e r factor a n a l y s i s 206 71 a - d Regression and p r e d i c t i o n from airborne data 214-15 72 A - d Regression and pr e d i c t i o n from ground data 216-17 i x LIST DF TABLES Table Page 1 Target conditions f o r f i e l d r e f l e c t i o n measurements 23 2 Summary of sampling techniques 26 3 Parameters most s i g n i f i c a n t i n characterizing s i t e units 52 4 S i g n i f i c a n t parameters useful f o r characterizing landform units 54 5 C h a r a c t e r i s t i c parameters f o r d i f f e r e n t i a t i n g out-wash from marine units at d i f f e r e n t l e v e l s of the t e r r a i n hierarchy 55 6 D r i g i n of group membership from d i r e c t numerical grouping 62 7 Factor loadings 66 8 Dri g i n of group membership a f t e r f a c t o r analysis 66 9 Group members a f t e r numerical c l a s s i f i c a t i o n 70 ID Grouping of s i t e s with C horizon data only 71 11 Factor loadings for the Peace River area data 72 12 Grouping of s i t e s a f t e r f a c t o r analysis 75 13 Differences between study areas 77 14 Mean values f o r l a c u s t r i n e units 92 15 Mean values f o r l a c u s t r o - t i l l units 92 16 Ph y s i c a l properties associated with s i t e s 94 17 Color categories r e l a t e d to ground conditions 107 IB Density classes and associated ground conditions 11D 19 Density classes and associated ground conditions 111 20 Color categories and associated ground conditions 116 21 S p e c i f i c a t i o n s used to create a d d i t i v e color image 117 22 Color categories and associated ground conditions 118 23 Comparison of density s l i c i n g and c o l o r additive techniques 120 24 Average chemical conditions of f i e l d s 129 25 Results of predictions of exchangeable Mg from s p e c t r a l r e f l e c t i o n values at 650 nm wavelength 136 26 Chemical conditions of samples used i n Figure 53 143 27 Regression equations f o r selected chemical parameters 147 28 Comparison of regression equations from d i f f e r e n t laboratory measurements 156 29 Comparison of regression l i n e s between d i f f e r e n t measurements 163 3D Density classes and associated ground conditions 207 31 Density classes and associated ground conditions 208 32 Density c l a s s e s and associated ground conditions 209 33 Density classes and associated ground conditions 210 34 Density classes and associated ground conditions 211 35 Chemical ranks f o r density l e v e l s from c d o r image 212 36 Chemical ranks from add i t i v e color image 212 37 Chemical ranks f o r density l e v e l s from color image 213 38 Chemical ranks f o r density l e v e l s from color-IR image 213 39 Chemical ranks from add i t i v e color image 213 LIST OF PLATES Plate, Page I Landsat image of Fraser Valley study area 15 II Landsat mosaic of the Peace River area 17 III R o l l i n g t i l l upland i n western part of study area 21 IV Beatton River Canyon i n eastern part of study area 21 V M u l t i s p e c t r a l photographic set-up 3k VI Density s l i c i n g set-up 35 VII Spectrometer set-up kO VIII Part of T2 landform unit 96 IX Four band m u l t i s p e c t r a l photography, grass covered surface 103 X Four band m u l t i s p e c t r a l photography, bare s o i l surface 104 XI Color enhanced density l e v e l s 106 XII Ground sample l o c a t i o n , vegetated surface 106 XIII Color enhanced equal density area f o r color-IR f i l m 107 XIV Exposed s o i l surface, i n d i c a t i n g reference s i t e l o c a t i o n 109 XV Density enhancement from color photography 110 XVI Density enhancement from i n f r a r e d color photography 111 XVII Color a d d i t i v e image ( c o l o r and IR band) 116 XVIII Color a d d i t i v e image ( a l l bands) 118 XIX Image comparison 119 XX Enhanced density l e v e l s (500-600 nm band) 207 XXI Enhanced density l e v e l s (600-700 nm band) 208 XXII Enhanced density l e v e l s (600-700 nm band) 209 XXIII Enhanced density l e v e l s (500-600 nm band) 210 XXIV Enhanced density l e v e l s ( c olor image) 211 x i LIST OF SYMBOLS AND ABBREVIATIONS ex. = exchangeable elements av. = ava i l a b l e elements F e n = Fe - Oxalate extraction F e n = Fe - D i t h i o n i t e extraction Fep = Fe - Pyrophosphate extraction nm = nanometers ppm = parts per m i l l i o n = s i g n i f i c a n c e l e v e l M = Marine 0 = Outuash GM = Glacio-Marine T = T i l l LT = L a c u s t r o - T i l l L " = Lacustrine S = Sandstone ifiS = Manufacturer's name f a r additive color viewing IR = Infrared x i i ACKNOWLEDGEMENTS The suggestions and help from a number of,people are g r a t e f u l l y acknowledged: Prof. •. Slaymaker, my supervisor, f o r advising and r e -viewing t h i s work; Prof. L. Lavkulich of the Dept. of S o i l Science at UBC f o r h i s i n s p i r a t i o n and generous support both f i n a n c i a l l y and through making a v a i l a b l e the laboratory f a c i l i t i e s f o r the chemical analysis; the s t a f f of the S o i l Science Laboratory at UBC f o r t o l e r a t i n g my i n t r u s i o n s ; Prof. M. Church f o r h i s c r i t i c a l review; and the members of my committee for reviewing the f i r s t d r a f t . The m u l t i s p e c t r a l photography was obtained with the help and f a c i l -i t i e s of Parker Williams of Integrated Resources Photography Ltd., Vancouver, whose technical advice and ingenuity have been of great help and i n s p i r a t i o n . The additive color images were kindly produced by Gary Washburn of the University of C a l i f o r n i a , R iverside. The use of the density s l i c i n g f a c i l i t i e s at the Alberta Centre for Remote Sensing i n Edmonton (courtesy of Mr. C. Brinker) i s also g r a t e f u l l y acknowledged. Dr. Jim Gower of Ocean and Aquatic A f f a i r s i n P a t r i c i a Bay kindly arranged the use of the airborne spectrometer, and Dr. B. N e v i l l e of the same I n s t i t u t i o n supervised the airborne data gathering and data e x t r a c t i o n . Their t e c h n i c a l support was greatly appreciated. S p e c i a l thanks go to Dr. H. Hunt of the- U.S. Geological Survey i n Denver, Colorado, and to Dr. C. Wiegand and Associates at the U.S. Department of Agriculture Laboratory i n Weslaco, Texas, f o r t h e i r generous help i n producing the laboratory s p e c t r a l r e f l e c t i o n curves of the s o i l samples. A hearty thanks i s extended to Arthur and John f o r t h e i r h e l p f u l discussions and f o r reviewing the t h e s i s . F i n a l l y , I would l i k e to express my deepest gratitude to P a t r i c i a f o r her i n s p i r a t i o n , tolerance, and support i n a l l phases of t h i s work. The Fraser Valley research was p a r t i a l l y supported by grants from N.R.C., and Environment Canada v i a Westwater Research Centre kindly made ava i l a b l e by Prof. Slaymaker. The Peace River Project was supported by a grant from the Defense Research Establishment i n Ottawa, courtesy of Prof. L. Lavkulich. INTRODUCTION A. AIMS OF STUDY In studies of terrain analysis chemical variability has not been considered in any rigorous manner. The lack of emphasis on this subject may be attributed to the following: (1) a large number of factors control the chemical distribution; (2) the nature of the system in which these factors interact i s complex and dynamic viewed at any scale; and (3) the analysis of inherently large sampling networks i s labor intensive, costly, and time consuming. The distribution of soluble chemicals in nature i s important at different levels in the natural system. For example, in agriculture certain elements are c r i t i c a l to f e r t i l i t y and plant growth. Similarly, a basic knowledge of the chemical variability over spatial units i s important for the proper management of different types of land use. At the basin scale studies of water quality provide information on chemical distribution which i s essential especially with respect to dissolved sediment load and source area identification. At the site specific scale in order to study processesnrelevant to the evolution of the natural systems, knowledge of the distribution of soluble chemicals in the rego-l i t h zone of the land surface i s a necessity. To tackle the subject of chemical variability a mass balance systems approach seems most effective. Such a study would require the identifica-tion and monitoring of chemical cycling and input-output measurements at various levels in the system. It is f e l t that such an approach requires long term, large scale investigations on a continuous basis which are best accomplished by a team effort. As an alternative, in the present study a more static geomorphological framework was used to analyze chemical var-i a b i l i t y . 2 Genetic geomarphological units, an entity that combines homogeneous surface material, surface form, and interpretations as to origin, were used as a basis for this study. The usefulness of physiographic units in predicting chemical conditions has been investigated by Webster and Beckett (1964 and 1970). They concluded that in intensively cultivated areas and at the physiographic unit scale the chemical variability was far too great to be useful for prediction. The complexity of quantifying chemical conditions in intensively used areas was investigated in a pilot project in the Lower Fraser V/alley. The same approach was pursued in the main study in the Peace River area in Northeastern British Columbia where the geomarphological unit framework was used following the suggestion by Webster and Beckett (1970) that such a procedure would probably yield more profitable results in an undeveloped area. In addition i t was considered worthwhile to investigate the chemical-physiographic association in more detail at a large scale. In examining three-dimensional landform units i t i s essential to know f i r s t the aim of the investigation. This w i l l determine the scale and detail at which the sampling and chemical analysis are to be performed. Because of time constraints, cost factors, and the need for quantitative data, parametric investigations are usually reserved for large scale, detailed studies while the genetic approach is mostly used for medium to small scale analysis. In resources evaluation in the past a descriptive landscape approach (Mabbutt 1968) has been dominant. In the present study the terrain i s classified into a hierarchy of basic landform units using a genetic/parametric approach and the chemical variability i s analyzed within these units. The lower end of the hierarchy is emphasized in this study where the scale of investigation varies from meters to kilometers. Micro-morphological questions generally investigated by the s o i l chemist and macro-chemical analysis pursued by the geachemist are not considered in this analysis. 3 The general aims of this research are: (1) to evaluate the character D f variability o f chemical a t t r i -butes in the solum over the landscape, and (2) to examine different uays in which information on chemical variability can be obtained and analyzed. The specific aims and tasks include: (1) an examination of the factors responsible for the variability, (2) an evaluation of the usefulness of a geomorphological unit framework in classifying and predicting conditions D f soluble chemicals at different levels of a hierarchical landscape, (3) the development of a numerical technique by which chemical conditions can be quantified, and (4) the evaluation of the alternative strategies of multispectral photography and direct spectral measurements and thus the examination of whether chemical variability can be predicted by remote sensing techniques. The data basis was established by using traditional sampling and laboratory techniques. A quantification of chemical conditions was accomplished with numerical classification procedures while the prediction of chemicals was pursued by multispectral sensing techniques. B. ORGANISATION The present research consists essentially of five parts: f i r s t i s a discussion on the methods of analysis; this i s followed by two sections concerned with the basic concepts, and two in which the techniques are evaluated. In Chapter Two a hierarchical terrain unit concept i s developed. The identification of fundamental landform units on a parametric-genetic basis, the assessment of their internal homogeneity, and the role of such units in the development of a hierarchy are the main subjects. In Chapter Three numerical grouping programs such as factor and cluster analysis are applied to determine the usefulness and relative advantages of such techniques in processing a large amount of parametric data. In Chapter Four emphasis i s placed on identifying major factors which control the chemical distribution in the s o i l . S o i l forming factors such as parent material, topography, climate, time, and biota are analyzed and their relative importance in controlling the chemical distribution and chemical processes are assessed. Finally, an attempt is made to evaluate the potential of remote sensing techniques to speedily generate chemical ground information. Multispectral photography and direct spectral observation both i n the laboratory and from the air were used for this endeavor. Density slic i n g and additive color viewing were used as secondary techniques to quantify the analysis of the photographs while numerical techniques were used for the spectral observations. In Figure 1 on page 5 the overall study framework i s portrayed in the form of a flow diagram. BASIC PRINCIPLES In order to analyze terrain a classification scheme must f i r s t be developed to create some understandable and meaningful order in the complexity of terrain. In this process a number of Basic principles have to be considered. 5 TECHNIQUES TERRAIN UNIT HIERARCHY TECHNIQUES Conclusions on terrain variability and terrain hierarchy Comparison of classification and terrain units ^1 PHYSIOGRAPHIC REGION LANDFORM UNIT TYPE T I I LANDFORM UNIT T SITE TYPE SITE L Numerical grouping techniques Aerial Photo Interpretation _ l Remote Sensing Field Sampling of soils for chemical analysis Multispectral Photography Spectral Observation Parametric data from f i e l d and laboratory analysis Correlation regression analysis Conclusions on factors affecting parameter relationship Conclusions on remote sensing methods Figure 1. Organisational flow diagram. 6 1. Parametric vs. Genetic Approach Approaches emphasizing readily accessible and easily measurable parameters are usually referred to as parametric methods while approaches using unit identification are known as genetic methods (Mabbutt 1968). In the parametric approach parameter variability i s of primary concern and units are only identified in a second step by either extrapolation or superimposition of point specific information. As a result the f i n a l units are often somewhat a r t i f i c i a l despite the quantitative data from which they were derived (Mitchell 1974, Riguier 1974). Genetic classifications deal essentially with the identification of units, and essential parameters are measured to verify these units. The identification of genesis i s often only possible after a factual", basis has been established. Spatial units derived genetically are more natural but often less quantitative. A review of the most important operational terrain analysis methods for resources evaluation i s given in Figure 2 and i s based on the differences between the parametric and genetic approach. They are primarily based on geomorphology and vary in emphasis from a purely parametric to a genetic approach. The former are more quantitative and are mainly concerned with surface form measurements. On the other extreme the genetic methods emphasize the identification and description of units. It i s evident from Figure 2 that there i s no method which combines the genetic and parametric approaches. Although numerous researchers (see for example Cline 1949) have suggested the use of process parameters most indicative of genesis, emphasis on morphology and morphometry i s s t i l l prominent. However, morphology should not be used i n i s o l a t i o n s i n c e i t i s only expressive of past processes which more D f t e n than not are polygenetic. In addition different genesis can produce similar landforms. As an alter-native, the use of physical and chemical parameters within genetic units seems to provide a more promising basis for terrain analysis (Croft 1974). APPROACH PRINCIPAL EMPHASIS MAJOR OPERATIONAL SCHEMES PRINCIPAL AUTHORS PARAMETRIC APPROACH GENETIC APPROACH MORPHOMETRY MORPHOLOGY MORPHOGENESIS MORPHO-CHRONO-GENESIS Computer Stimulation Slope Profi le Analysis 7" Landscape Mapping, Morphological Mapping Geomorphological Mapping Terrain Modelling Watershed Analysis Terrain Analogues Von Lopik & Kolb 1959 Wood a Snell 1960 US Army Corps Eng. MEXE Terrain Analysis CSIRO Land Research USSR Landscape Science ITC System French Scheme Parametric Land Analysis . Speight 1968, 1U King 1970 Scott ii Austin 1971 CLI Land Inventory H i l l s a Portelar.CE 1960, Hi l ls 1966 Lacate 1971 Vinogradov 1962 Solntsev 1962 Prokaiev 1962 Hobson 1972 Junkins 1971 Tobler et al 1968 Turner et al 1967 Horton 1945 Strahler 1957 Melton 1957 Beckett 8. Webster 1965,69 " et a l 1972 Webster & Beckett 1969 Mitchell & Perrin 1967 Christian & Stewart 1968 Hebbutt & Steiuart 1965 Msbbut et a l 1963 Trlcart 1971 Veratappen, and Van Zuidam 1963 Figure 2. Overview of different operational terrain analysis schemes. a In the present study a combined genetic/parametric method i s used which i s based on a genetic geomorphological classification of terrain units and a parametric evaluation of chemical va r i a b i l i t y . 2. Landform-Soil Relationship Genetic parameters and s o i l properties are often not readily observable from aerial photographs. As an alternative, emphasis has been placed on landform interpretations which to a certain degree reflect genesis and s o i l properties. The ut i l i z a t i o n of a geomorphological frame-work has been undertaken by several authors ( c f . Troeh 1964, Acton and Stonehouse 1965, Ruhe and Walker 196B, walker and Ruhe 1968, Kleiss 1970, and Daniels et a l 1970, 197']!). A l l found significant relationships between geomorphological parameters and s o i l properties. 3. Hierarchical Framework Terrain i s best investigated in a hierarchical framework. Such a scheme can fallow twa contrasting directions: an analytic or divisive schema, versus a synthetic or ascending approach. The former has tradi-tionally dominated the f i e l d of terrain analysis where for example the physiographic approach i s s t i l l dominant. It has been argued by Uright (1972) that an ascending classification i s i n i t i a l l y more desirable since i t i s based on observed properties of individuals and not, as in the analytical approach, on a basic understanding of genetic variability within the population. Such agglomerative c l a s s i f i -cations are also more compatible to computer analysis than divisive ones. As a science develops, a better understanding of genesis i s accumulated and at this stage a divisive classification becomes more useful. Because we have only a limited understanding of terrain conditions and s o i l genesis the ascending approach is investigated in this study u t i l i z i n g the site as the basic taxonomic unit in the terrain hierarchy. 9 CHAPTER I METHOD AND TECHNIQUES In developing the method used to analyze chemical terrain v a r i -a b i l i t y the following topics were considered: A. Hierarchical framework of terrain units B. Choice and description of f i e l d area C. Essential sampling and laboratory techniques D. Numerical methods E. Remote sensing techniques A. HIERARCHICAL FRAMEWORK Considering the great complexity and variability of terrain con-ditions which exist over the Earth's surface i t is evident that a class-i f i c a t i o n process has to be used that w i l l simplify the comprehension of terrain. Such a procedure should be related as closely as possible to nature while at the same time allowing an understanding of the major features and processes which control terrain conditions. In this study a three-stage process was used: (1) A hierarchical structure was set up, (2) units were identified, and (3) parameters useful in unit dis-crimination were examined. 1. Hierarchical Structure To analyze terrain conditions comprehensively at various scales a hierarchical scheme of geomorphological units was developed. Genetic/ parametric principles were used to identify units of increasing size and these were arranged in an ascending hierarchical structure. The overall scheme used in this study i s outlined in Figure 3 on the following page. The origin of this scheme can be traced to the CSIRO (Christian and Stewart 1968) and MEXE (Mitchell 1971, Beckett and Webster 1972) terrain ID HIERARCHICAL LEVELS DEFINITION SCALE OF ANALYSIS Physiographic Region Comparable topography and lithology having undergone similar geomorphic history Small scale 1:1,000,000 Landform type units An assemblage of landform units which are similar in genesis, form and s u r f i c i a l material properties Landform units Specific landform with respect to genesis, s u r f i c i a l material, form (consists of an assemblage of site types) 1:50,000 to 1:30,000 Site types An assemblage of sites which are identified on the basis of form and processes Site Elementary landform unit which for practical purposes i s homogeneous with respect to internal properties identified on the basis D f form Large scale 1:10,000 Figure 3. Hierarchical framework. DIFFERENCES CSIRO/MEXE APPROACH PRESENT STUDY Scale and hierarchi-cal structure Reconnaissance survey emphasizing higher hierarchical units Detailed survey emphasizing lower hierarchical units Aims Natural land resources inventory Assessment of hier-archy and variability of different units, suitability of scheme for detailed analysis and numerical treat-ment Nature of investiga-tion Descriptive to semi-quanti-tative, emphasizing physical parameters Quantitative empha-sizing physical and chemical parameters Basic approach for unit identification Stresses morphology Stresses morphology and genesis Figure k. Comparison between different hierarchical schemes. 11 analysis approaches. The basic differences between these two and the present study are summarized in Figure k. 2. Identification of Units The hierarchical geomorphological framework as described above was investigated at two levels: (a) the site, and (b) the landform unit. The site being the lowest taxonomic unit was conceived as having the greatest internal homogeneity and as such could be used as the basic landscape unit within this hierarchical system. Sites were identified in the f i e l d as single morphological units which for practical purposes could not be subdivided on the basis of form. Such sites were defined by Bourne (1931) as small areas providing throughout their extent "similar local conditions as to climate, physio-graphy, geology, soi l s and edaphic factors." As such these sites have ecological relevance and changes in the native vegetation are often indicative of site boundaries. Sites are also indicative of discrete micro-climatic conditions which often find expression in the vegetation cover. To characterize the site conditions a number of parameters were analyzed and site data were generated both in the f i e l d and from sample analysis in the laboratory. Landform units are of a higher taxonomic order. They are made up of an assemblage of sites and are therefore larger and less homogeneous. Landform units were identified on aerial photography using form, inferred parent material and genesis as basic c r i t e r i a . Verification of this interpretation was made in the f i e l d and from quantitative data obtained from sample analysis. The units were quantified on the basis of the parametric information of the sites which corresponded to the individual units. This was done numerically by assembling similar sites into groups of site types. The variation within the different hierarchical units was assessed and the usefulness of such a geomorphological scheme was examined. 12 Geomarphological rather than geological, hydrological or edaphic units were chosen for this study. These units contained the essential geological information but were more readily observable on aerial photo-graphs. They were superior to hydrological and edaphic units because they mere at a scale most useful for practical application and provided direct rather than indirect spatial information. 3. Selection of Parameters To compare and classify the identified units i t i s essential to select parameters which are characteristic of a l l units. In this selection the concept of differentiating characteristics was followed. This was f i r s t postulated by Cline (1949) and later restated by Uright (1973) as follows: In general, the principal requirements are that differentiating characteristics should be: 1. Intrinsic properties of the things to be grouped. 2. Observable, measurable, readily accessible, and ideally of a comparatively permanent nature. 3. Indicator properties which contain the maximum number of accessary characteristics. 4. Important for, and capable of application to, the objectives of the classification. 5. Suitable for the construction of a hierarchy. Not a l l of these c r i t e r i a could be followed in selecting chemical parameters and in a number of cases parameters with assumed genetic relevance were used. Physical and chemical properties were used as differentiating characteristics of a l l sites. Soluble chemicals essential to plant growth were considered as primary parameters in this analysis. In addition emphasis was placed on those chemical parameters which were common, which occurred in readily detectable concentrations, and which on the basis of the pilot project proved to haveeuseful differentiating characteristics. Properties referring to the basic s o i l farming factors -13 climate, topography, parent material, biota, and time - were considered. Morphometric and chemical parameters mere measured to assess the possible relationships between these factors. Chemical parameters were measured for the surface A horizon and parent material C horizon. The following properties were measured quantitatively in the laboratory and in the f i e l d : Physical Parameters for both studies Slope angle Slope position Relief Aspect Compaction Soil color Particle size Soil moisture Chemical Parameters for main study Extractable Ca, Mg, Na, K, P, C, and pH for p i l o t study Available Ca, Mg, Na, K, Cu, Cd, Pd, S i , Zn, Ni, PD. , SD^, CI, Mn Time effects on individual elements were assessed over a three-week period while climate was considered in a comparative sense using f i e l d areas in different climatic regions. Finally land use, vegetation type and percent vegetation cover were the biotic parameters measured. B. CHOICE AND DESCRIPTION DF FIELD AREA The research methodology was evaluated in two f i e l d areas: (1) The Lower Fraser Valley in Southwestern British Columbia, and (2) the Fort St. John area in Northeastern British Columbia. The Salman Basin in the Fraser Valley was used essentially as a pilot study area for i n i t i a l testing of most of the techniques. Because of the complex geomarphological history and the substantial influence of man in this area a second test area was selected in the Peace River d i s t r i c t where 14 a more intensive survey was undertaken. Moreover, by studying these two different areas i t was also anticipated that the effects of regional climate could partially be assessed. The two f i e l d areas are described separately below and a comparison between them i s given in a third section. 1. Salmon Basin (a) Location and climate The Salmon Basin covers an area of about 15 km and i s located in the Langley Municipality of the Lower Fraser Valley. The overall setting i s illustrated in Plate I. The coastal mountains strongly influence the maritime climate which prevails in the area. The precipitation pattern i s one of extensive winter rains mainly attributed to frontal activity, while a deficit moisture regime exists during the summer months. Average annual precipi-tation varies from 130-150 cm, the average snowfall i s around 80 cm and the mean annual temperature i s 10°C. (b) Geology and geomorpholoqy A series of unconsolidated sediments of Pleistocene and Recent age dominate the study area. l\!o evidence of structural control or bedrock exposure was observed and according to Armstrong (1957) the sediments reach a thickness of up to 300 meters. The study area can be divided into an upland and a lowland, the latter being part of the Fraser channel and floodplain system near Fort Langley. Al l u v i a l deposits ranging from s i l t y clay to sand are dominant and are interspersed with bogs up to 10 meters in thickness. The deposits which form the upland are of g l a c i a l and marine origin and can be c l a s s i -fied into: outwash, marine, glaciD-marine and beach deposits on the basis of material type and particle size. Plate I. Landsat image af Fraser Valley study area. 16 According to Armstrong (1957) and Ryder (1972) three major glacial and one interglacial periods have been the most important events of the Pleistocene history. The geomorphology as derived from the s u r f i c i a l geology report (Armstrong 1957) and the Soils Survey Report (Luttmerding et a l 1966) i s illustrated in Figure 5. The Abbotsford Outwash, the Cloverdale Marine and the Uhatcom Glacio-Marine units were used for the pilot study. (c) Soils and land use The soils analyzed in the pilot project belong to the Orthic Concretionary Brown (Abbotsford Outwash), Humic Eluviated Gleysols (Cloverdale Marine) and Orthic Acid Brown Wooded (Whatcom Glacio-Marine) subgroups. The units of marine and glacio-marine origin are mostly used for pasture land, with patches of alder forest along the riverbanks. The outwash units being better drained are arable (strawberries) but sections of mixed coniferous and alder forest are also present. Low density residential subdivisions are predominantly found in the outwash area but are present in a l l parts of the basin. 2. Peace River Study Area (a) Location and climate A f i e l d section, 40 x 3 km, 15 km north of Fort St. John was chosen for a more intensive study. The choice of this location was influenced by the fact that a great variety of local landform variation exists in this section of terrain, the human influence i s of relatively recent history and the f i e l d area i s readily accessible. The exact location of the study area i s illustrated on the Landsat mosaic (plate'II). The area borders the Rocky Mountain foothills and i s under the influence df a moderately dry continental climate. Annual precipitation varies from 40-50 cm, about 50% of which occurs during 18 summer convectianal showers. Winters are very cold, the average annual snowfall reaching 170 cm and the mean annual temperature being 1°C. (b) Geology and Geomorpholoqy A series of sandstone and shale layers of Cretaceous age underlie the study area. The strata dip slightly to the east as revealed by exposures an the sides af the deeply dissected canyons and western slopes of the uplands. The geomarphology can be conceived of as a cuesta structure masked by a thick s u r f i c i a l caver of Pleistocene origin. The overall geological setting i s illustrated in Figure 6 in the farm of a longitudinal cross section. The following three landform divisions characterize the f i e l d area: (1) a series of ra i l i n g uplands (to the west), (2) trenches and canyons (central section) (3) gently sloping platform with thick s u r f i c i a l cover (to the east). The geomarphological history has been described by Mathews (1963) and the. effects af glaciatian are most prominantly seen in the h i l l y upland, where the bedrock i s covered by glacial t i l l of up to 40 m in thickness. With the exception of the Tea Creek area at the western ex-tremity, mast t i l l i s attributed to the latest glacial advance of the Laurentian ice sheet. The deeply dissected interglacial valleys have widespread lacustrine deposits which are the result of a series af ice dammed lakes which were formed during glacial retreat. However, these deposits are seldom homogeneous and pebbles, shore line material, gravel and t i l l are found in many locations within the lake deposits. The mare heterogeneous of these deposits were appropriately named l a c u s t r o - t i l l by s o i l surveyors (see Farstad et a l 1965) and are attributed to rafting by floating ice. The s u r f i c i a l cover which was encountered in the study Cretaceous Sandstone I::::::I Cretaceous Shale 0 2 4 6 km 1 i » i • i • • i Scale Figure 6 . Geological cross section of Peace River study area. HH Sandstone H::::) Lacustrine 1_—-I Lacustro-Till Til l V V River Canyon 0 6 km Scale Figure 7 . Geomorphological units i n the Peace River area. 20 area Is represented in Figure 7 and follows the longitudinal geological cross section outlined in Figure 6. The geomorphological units were identified from aerial photographs (photo scale 1:33,000) and were verified during the f i e l d analysis period. The gently sloping platform interrupted by a deeply dissected interglacial valley i s illustrated in Plate III while the t i l l covered rolling upland i s shown in Plate IV. Landform units used in this study f a l l into four categories: t i l l , l a c u s t r o - t i l l , lacustrine, and sandstone. These were determined on the basis of form, relative position, parent material type and particle size distribution. For each category several examples were analyzed so that a later comparison was possible. (c) Soils and land use The soils were identified by Farstad et a l (1965) as belonging to five great groups: Black, Solodized Solonetz, Solod Gray Uooded, Orthic Grey, and Eluviated Gleysol. The s u r f i c i a l deposits were used as primary indicators in the classification of s o i l series and mapping of s o i l asso-ciations. There are essentially three types of land uses: arable agriculture, forest and bogs. The former i s dominant in the f l a t lacustrine units, while virgin aspen, poplar, lodgepole pine forests are most extensive on the t i l l units and canyon side slopes. Pasture land i s very limited and agriculture, although extensive, i s limited by the short growing season. Rape, rye, barley, wheat and seed grasses are planted almost exclusively. The bogs are found in the depressional areas and consist of thick peat layers with open black spruce cover.: 3. Comparison between the Genesis of Sample Units Both f i e l d areas are "Quaternary landscapes" with a similar history of glacial and interglacial periods. Although the deposits have undergone 21 Plate IV. Rolling t i l l upland in western part of the Peace River study area. 22 somewhat different genesis, both areas are composed of thick unconsolidated sediments. A comparison along these lines i s given in Figure 8. Genesis Peace Area Fraser Area Terrigenous G l a c i a l - t i l l Acqueous Lacustrine Fluvial outwash Marine Mixed or complex intermediary between basic units Lacustro-till Glacio-marine Figure 8. Landform unit comparison in the two study areas. The mixed or complex units were of additional interest because they provided some basis for an analysis of the effect of complex genesis on the variability, especially with reference to the basic units studied within each f i e l d area. k. Field Targets for Airborne Spectral Mission \ (a) Fraser Valley spect^al^ sensing The Salmon Basin was also used for the airborne spectral mission. Unfortunately only three fields with \reshly t i l l e d s o i l s could be identified within the basin confines anti^for this reason two adjacent areas in the Fraser Valley were also included in this study. This allowed a comparison of airborne spectral information from more contrasting \ s o i l surfaces. A series of 11 agricultural fields belonging to four types of geomorphological units were selected for this project. In Table I a summary of the different target conditions i s provided: 23 Table 1 Target conditions for f i e l d reflection measurements Field # Geomorphological Units S o i l Series Location Cover A outwash Columbia Salman Basin bare B outwash Columbia Salmon Basin strawberries C marine Berry/Cloverdale Salmon Basin pasture D marine Cloverdale/Milner Langley bare E organic/deltaic Lulu & Lumbum kl-Langley bare F organic/deltaic Uinod kl-Langley bare G l deltaic westham with Delta/Ladner bare G2 minor Blundell C— G3 P 11 and Crescent 11 11 % H 11 11 n 11 In order to assess spectral signatures in a more systematic way freshly exposed soils were given preference. However, to establish the effect of vegetation cover one f i e l d with strawberries (having some 30 cm of s o i l exposed between rows) and one with total grass cover were also analyzed. Sample location and identification were handled as fallows: Simul-taneously with the October 2nd f l i g h t mission 40 s o i l samples were col-lected along the f l i g h t l i n e . A video camera and two motorized 35 mm IMikon cameras were installed in the aircraft to provide ground information and the f i e l d sample location was marked onto conventional aerial photo-graphs and coordinate distances in the f i e l d were paced off. (b) Peace River multispectral photographic mission The study used for the chemical terrain analysis was also used for the multispectral photographic mission (see Section A). To establish a proper data base 100 s a i l samples were collected simultaneously with the 24 aerial mission to allow the quantification of several dynamic parameters. A detailed description of the airborne system used i s given in Section E later in this chapter. C. ESSENTIAL SAMPLING A IMP LABORATORY TECHNIQUES 1. Sampling Schemes The pilot project was undertaken to explore the usefulness of the hierarchical framework and numerical technique and to assess the relative importance of chemical parameters which are potentially useful for terrain quantification. The pilot project results formed the basis for the method and parameters used in the Peace River area. In this study the techniques and their applicability were assessed more rigorously. (a) Pilot study In the pilot study concepts and methods were explored which resulted in the analysis of units at two levels: the site and the landform unit level. In the f i r s t case three sites were selected for a study of the "within site variation". Two sites within a marine unit and one on an outwash landform unit were selected. Within each site 5-8 s o i l pits were dug and samples from A, B and C horizons were collected for laboratory analysis. The pits were chosen randomly within each site and practical considerations limited the number. Landform units were delineated on aerial photographs and by placing a grid over the selected coverage 45 random sites were selected. Some deviation from this strat i f i e d pattern became necessary during fieldwork because several sample points f e l l into privately owned residential land holdings. The sample site location i s given in Appendix I I I - l . 25 (b) Main project Tuo series of samples were collected independently over the same study area. This was done deliberately for three reasons: (a) to confirm the sampling accuracy, (b) to assess the variability of some parameters, and (c) to establish ground control for the remote sensing mission. The f i r s t series of samples was collected over a three-week period. Seventy-two sites were chosen on a random basis within the study area. Sites were analyzed as to form and physical properties, and samples from the A and C horizons were collected for laboratory analysis. A second series of samples (100) was collected within 32 hours of the remote sensing mission. This was done to establish ground control at the time of aerial cover, in order to check the more dynamic variables such as s o i l moisture, etc. In this second scheme samples from only A horizons were collected. An effort was made to duplicate at least 50% of the samples with respect to the sites collected i n i t i a l l y . A section of the sampling design i s illustrated in Figure 9 below: Figure 9. Sampling design. X = First sampling series, 1 - 80. 0 = Second sample series, 200 - 299. (c) Sampling design for airborne spectral remote sensing mission Eight non-vegetated fields were selected in the lower Fraser Valley in and around the Salmon Basin, and 40 s o i l samples (A horizon only) were collected simultaneously during the airborne mission. The samples were 26 collected along a predetermined fli g h t line at varying intervals. The sample sites were recorded on the aerial photography and the within f i e l d position was measured on the ground. In the f i e l d , close attention was paid to s o i l color variations which formed the basis of this subjective sampling. (d) Summary of sampling techniques The different sampling methods can best be summarized in Table 2 below l i s t i n g area, units and sample numbers. Table 2 Summary of Sampling Technique Field Area Type of Analysis # of Landform Units # of Samples Horizons Sampling Design Fraser Ualley Analysis of site variation 3 2D ABC (each) Stratified random Pilot study 5 45 ABC (each) Stratified random Multispectral sensing mission 5 4D A Along line of f l i g h t Peace River Basic ground sampling 13 72 AC (each) Random Remote sensing mission ID 1DD A Random 2. Laboratory Techniques The samples were divided into three parts: the f i r s t was used for the chemical analysis, the second for particle size determination and the third for spectral reflectance measurements. 27 (a) Chemical analysis of s o i l samples The chemical analysis was performed on the < 2 mm fraction of the samples. For the Fraser Ualley study the following chemical analyses were performed: available Ca, Mg, K and Mn were measured by sodium acetate extraction (Greweling and Peech 1965) and atomic absorption spectrometric measurements. Available Na, Cu, Si, Zn were measured by D.l n HC1 extraction and atomic absorption (Black et a l 1965). Free iron oxide and iron-organic complexes were determined by the acid ammonium oxalate extraction (McKeague and Day 1966), sodium pyrophosphate extraction and citrate bicarbonate dithionite extraction (Mehra and Jackson 1960, weaver et a l 196S). Sulphate was determined by turbidity and chlorine was analyzed by the potentiometric ti t r a t i o n method. For the Peace River study exchangeable cations Ca, Mg, Na and H were determined by neutral ammonium acetate extraction and atomic absorption spectrophotometric procedure (Lavkulich 1974). The Leco induction furnace carbon analyzer (Allison et a l 1965) was used to determine carbon content for samples from both study areas. Similarly the ammonium fluoride-HCl extraction method (Jackson 1956) was used to determine phosphorus in both study areas. Finally, pH measurements were carried out in a 1:1 mineral/ water slurry. A description of the above i s given by Lavkulich (1974). Most are considered standard procedures in the Soil Science Laboratory at the University of British Columbia. The accuracies D f the analysis methods diffe r for each set of measurements. Generally the very low and very high chemical concentrations contain larger errors. The former i s c r i t i c a l when reaching the detection limit of the analysis equipment. For high concentrations the extract was diluted t D f i t into the most accurate detection range of the instrument. In the latter case errors due to dilutions are important. Care was taken to use the optimum conditions and highest accuracy and in a l l cases 25% duplicate samples were analyzed to assure reproducibility of results. 28 (b) Physical parameters S a i l moisture was determined in the laboratory by drying the samples at 105°C immediately upon return from the f i e l d . Wet and dry s o i l color were determined using the Munsell s o i l color charts and particle size analysis was accomplished by H2O2 treatment, wet sieving and hydrometer analysis (Day 1950). D. NUMERICAL METHODS The physical and chemical parameters were coded and stored in a f i l e for access by various s t a t i s t i c a l programs. A l l numerical analyses were carried out on an IBM 370 di g i t a l computer. Besides conventional stat i s -t i c s such as correlation and regression, and significance tests, two numerical classification methods were used which merit some discussion: (1) Euclidean grouping method, and (2) factor analysis. 1. Euclidean Grouping Method Sites were classified numerically using multiparameter chemical data. This was achieved by using the hierarchical clustering procedure described by Ward (1963). This technique i s explained by the example illustrated in Figure 10. Mg 200., Ca Mg Na Site A 100 80 10 Site B 150 150 5 Site C 300 20 30 L , , r— 100 200 300 Ca ppm Figure 10. Euclidean grouping procedure. 29 Values for Ca and Mg were platted in twa-dimensianal space and the distance between the positions of these sites was measured using the Pythagorean theorem. It i s obvious that sites A and B are closer to each other than to site C. They form an i n i t i a l cluster and the distance between them can be considered as a degree of similarity. The closer the sites are to one another the more similar their properties. Once the i n i t i a l cluster i s formed the mean value of the two sites becomes the central point and the distance to the next nearest site i s computed at a lower order of similarity so as to minimize the within cluster variance of the new group. In this way a hierarchy can be built up which in this example would look as follows: , 1 , A B C In the above example only two dimensions, Ca and Mg, were considered. However, this scheme can easily be extended to n-dimensions using matrix algebra. The UBC-C-GROUP program (Patterson and Uhitaker 1973) which i s based on Ward's (1963) average grouping technique proved useful for this project. Other Euclidean grouping methods such as nearest linkage, com-plete linkage, and nodal linkage (see Sokal and Sneath 1963) were tried but provided results which were less interpretable. A l l sites were grouped together according to their similarity with respect tD a l l chemical parameters. At least on theoretical grounds such an approach should provide a comprehensive classification, since a l l par-ameters were considered simultaneously and of equal weight. 2. Factor Analysis The grouping procedure previously outlined considers a l l variables equally but i t does not provide information as to the importance of each variable. Individual variables can be assessed with regard to their 3D contribution to the total variance by using a factor analysis. Factor analysis i s a method for describing a complex interrelationship of multiple variables in terms of the smallest number of factors. The scheme is based on the concept that correlated variables are not com-pletely independent and contain redundant information. If variables are sufficiently intercorrelated they can be represented as a cluster of vectors, the cosine of the angle between each vector pair being equal to the correlation coefficient of the two vectors in question. Then a principal axis of best f i t i s passed through the common vertex so that the sum of squares of the projection length i s maximized (the length i s measured along the new axis). Since the vectors are of unit length their projection length on the axis i s equal to the cosine of the new angle formed between vectors and the main axis. After measuring the projections on the f i r s t axis, a second axis at right angles to the f i r s t i s drawn and the new projection is once more measured. This can be illustrated in Figure 11. V = Vector J 1 = Projection length Figure 11. Projections in factor analysis. These projection lengths are called loadings and can be used as weights to combine the original variables into fewer factors. For a detailed discussion see Cattel (1965) and Abler et a l (1971). The' factor loadings, when multiplied with the original data deter-mine the weighted score of the factors. A distance grouping method as VI VI 31 described above can then be used to measure the similarity between the sites. In this case, however, the weighted data information was used. The factor analysis technique was used to analyze the variables and to determine their importance in the subsequent classification. E. REMOTE SENSING METHODS Every abject absorbs, emits, scatters, transmits or reflects elec-tromagnetic energy. With different sensors i t is passible to measure the amount of reflected or emitted energy at different wavelengths and thus the spectral characteristics may be plotted against wavelengths. Knowing the spectral characteristics of different objects i t is passible to assess the properties of the object which are responsible for producing the spectral characteristics of the object as a whole. In the present study a number of remote sensing techniques were explored to examine whether chemical terrain conditions can be predicted from remote spectral measurements. The basic aim of the analysis was to measure variations in surface reflection of terrain units and s o i l samples in various parts of the electromagnetic spectrum from the air , on the ground, and in the laboratory and then to compare these results with the measured chemical parameters. The analysis was limited to multispectral techniques in the visible and near infrared wavelength band and a l i t e r a -ture review on such techniques as they have been applied to detecting chemical conditions i s given in Appendix I I . Two separate projects, one in the Peace River area and one in the Fraser Ualley were undertaken and three basic techniques were used: (1) multispectral photography, (2) laboratory spectral reflectance anal-ysis, (3) f i e l d spectral reflectance measurements. An overall view of the procedure is provided in Figure 12. These methods were used in the 350-2500 nm spectral wavelength regions and the specific sensitivity range for each method is given in Figure 13. In the case of multispectral 32 Peace River Project Fraser Project Multispectral photography Density slici n g Comparison of different bands & den-s i t i e s with chemical conditions Additive color viewing Lab spectral reflectance analysis; Reflection measurements of samples in laboratory Field spectral reflectance analysis Airborne spectral analysis Comparison of different colors with chemical conditions Ground spectral analysis Comparison of spectral reflectance with chemical conditions Figure 12. Remote sensing procedure. UV Visible Near Infrared Spectral Wavelength Range 1 1 400 700 i i 1000 1500 2000 2500 nm 1 — Multispectral photography Field spectral] measurements Laboratory •j "j spectral measurements Figure 13. Spectral sensitivity range for different methods used. 33 photography relatively large spectral bands (1DD, 300 and 400 nm) were measured, while the spectral techniques provided data at numerous specific wavelengths within the given range. 1. Multispectral Photography (a) Film/Filter combination The usefulness of multispectral photography in terrain analysis has been widely acknowledged in the literature (for a literature review see Appendix II-2). Terrain was analyzed by photographic means and the visible to near-IR wavelength range was explored with the following film/ f i l t e r combination: 1) Tri-X-Aerographic 2403, Black and white, with Idratten f i l t e r 25 2) Tri-X-Aerographic 2403, Black and white, with Uratten f i l t e r s 12 and 59 3) Ektachrome 2448, Color 4) Aerochrome IR 2443, IR-color, with Lilratten f i l t e r 12. These emulsions and f i l t e r s were chosen because they made use of the best spectral sensitivity range for s o i l s and vegetation discrimination on the ground. The f i l t e r properties were calibrated prior to the aerial mission with a laboratory-spectrophotometer, the results of which are given in Figure 14. Transmission % 100-| 50 Uratten F i l t e r 12 (yellow) Uratten F i l t e r 25 (red) Uratten F i l t e r 58 (green) 500 600 700 nm Wavelength Figure 14. Spectral transmission properties of f i l t e r s . 34 Two 70. mm V/inten Aerial cameras were coupled for simultaneous observations with the different f i l m / f i l t e r combinations. The Peace River study area was covered twice (30 minutes between flights) with a Cessna 180 aircraft, and the aerial set-up i s illustrated in Plate V. Ground samples of the test area were collected over a period of 32 hours from the time of the f l i g h t mission and the chemical conditions of the samples were analyzed in the laboratory. When developed, the films were analyzed quantitatively using image enhancement techniques. Plate V. Multispectral photographic set-up. 35 (b) Image enhancement through density s l i c i n g Four s e l e c t e d frames from a l l f i l m / f i l t e r combinations were analyzed separately using a S p a t i a l Data Systems Model 703 densitometer. Ldith t h i s s l i c e r , 64 grey s c a l e s were d i f f e r e n t i a t e d and to enhance s p e c i f i c density l e v e l s , d i f f e r e n t c o l o r s were assigned to these l e v e l s . The o r i g i n a l frames were viewed with a t e l e v i s i o n camera and the enhance-ment was portrayed on a TV screen. By combining s e v e r a l density steps and s e l e c t i n g a s e r i e s of c o l o r combinations a number of c o l o r enhance-ments were obtained. In a d d i t i o n to the s l i c i n g an a r e a l q u a n t i f i c a t i o n of any s e l e c t e d c o l o r w i t h i n each frame was p o s s i b l e by using the b u i l t - i n a r e a l planimeter. The s l i c e r set-up i s i l l u s t r a t e d i n P l a t e VI below. P l a t e V I . Density s l i c i n g set-up. 36 (c) Image enhancement through a d d i t i v e c o l o r viewing Rather than a n a l y z i n g the d i f f e r e n t f i l m / f i l t e r bands s e p a r a t e l y , a c o l o r a d d i t i v e viewing enhancement technique was a l s o evaluated. Each f i l m was p r o j e c t e d i n an assigned c o l o r and a superimposition of these p r o j e c t i o n s produced a f a l s e c o l o r image. By varying the c o l o r s i t i s p o s s i b l e to create a multitude of f a l s e c o l o r imagery and the one g i v i n g the greatest c o l o r c o n t r a s t s f o r any s p e c i f i c a n a l y s i s was chosen on s u b j e c t i v e grounds. ( i ) Diazo f i l m Diazo f i l m s of various c o l o r s were s e l e c t e d and the o r i g i n a l t r a n s p a r e n c i e s were placed on these f i l m s and exposed to u l t r a v i o l e t l i g h t . The UV-rays "burned out" the c o l o r on the diazo sheets according to the amount of l i g h t t r a n s m i t t e d through the o r i g i n a l t r a n s p a r e n c i e s . Dark grey areas on the o r i g i n a l s p r otect the Diazo-sheets and upon de-velopment i n ammonia vapor r e t a i n more c o l o r than l i g h t grey areas. D i r e c t contact t r ansparencies i n the s e l e c t e d c o l o r s were produced f o r each of the fo u r f i l m s , and a superimposition of d i f f e r e n t c o l o r - f i l m / f i l t e r combinations produced f a l s e c o l o r images. In t h i s way the en-hancement of d i f f e r e n t grey shades was p o s s i b l e and the va r i o u s c o l o r u n i t s were then compared w i t h the chemical data from the s o i l samples. (For a d e t a i l e d d i s c u s s i o n on the procedure see Warrington 1975.) Green, blue, red, cyan, yellow and magenta dzachrome Diazo f i l m s were used and the contact transparencies were produced and developed with a CIEL 547 Diazo white p r i n t e r . By combining transparencies from d i f f e r e n t f i l m s a s e r i e s of p i c t u r e s r e p r e s e n t i n g d i f f e r e n t s e n s i t i v i t y ranges were produced, which i n the extreme case r e s u l t e d i n an image of the 4DQ-90D nm wavelength range. ( i i ) I 2 S c o l o r p r o j e c t i o n The same frames from a l l four s p e c t r a l bands were placed i n a I S a d d i t i v e c o l o r viewer. Each s i n g l e frame was p r o j e c t e d on the same screen and with proper r e g i s t r a t i o n a superimposition of a l l bands was p o s s i b l e . To enhance the a d d i t i v e image d i f f e r e n t exchangeable c o l o r f i l t e r s were 37 placed between each single band projection and by altering the light intensity i t was possible to produce an additive false color enhanced image which contains spectral information from a l l bands. This technique i s similar to the Diazo method but has the advantage of greater f l e x i b i l i t y and smaller loss of resolution. The reason for this i s that color f i l t e r s and light intensity can readily be altered during the projection, and the original film transparency i s used rather than a color copy as with the Diazo method. Since the Diazo film i s less sensitive than the color and black and white photographic film, information loss i s inevitable with the Diazo method. In addition information loss also occurs when copying the original film. 2. Spectral Reflectance Measurements in the Laboratory Two sets D f samples were analyzed. The spectral reflectance of 20 s o i l samples was measured at the USGS in Denver, Colorado using a Cary 14 spectrometer with a bi-directional reflectance attachment (Hunt and Ross 1967). Fresh MgD was used as a standard Qver the 400-2000 nm wavelength range. The reflection curves of an additional 38 samples were determined at the USDA Research Laboratory in Ldeslaco, Texas. In this case a Beckman Model DK-2A spectrophotometer with a reflectance attachment was used. BaSO, was used as a standard and measurements were made over the k 500-2500 nm wavelength range. Because different instruments and standards were used the two sets of data were analyzed separately. After removal of the > 2 mm fraction the samples were dried and compacted with a press. The USDA samples were pressed into bottle caps with 11 tons of pressure applied with a Model C -30 press. 38 The effect of particle size and cation extraction on the spectral signatures was assessed by removing various particle size fractions from a number of samples and by analyzing soils before and after neutral ammonium acetate extractions. The resulting spectral values.at the different wavelengths were then compared with the chemical data to examine possible relationships between the chemical and spectral con-ditions. 3. Field Spectral Analysis The f i e l d spectral remote sensing mission was carried out in three parts: (a) airborne spectral reflection measurements, (b) ground spectral reflection measurements of s o i l samples, and (c) data evaluation and com-parison. Both the airborne and ground spectral reflection measurements were obtained by using a si l i c o n diode array spectrometer with 256 continuous wavelength bands in the 400-1050 nm wavelength range. With this system scattered light from different targets was projected onto an array of s i l -icon diodes. The array measurements were read off, digitized and stored onto tape by a data acquisition mini-computer. Spectral reflection curves covering the entire 400-1050 nm wavelength band were constructed by re-trieving the stored data and using reflected / incident light ratios. A comparison between the airborne and ground samples of different fields was attempted and the influence Df chemical variation Dn the measurements was investigated. The spectrometer system used for this analysis was developed by the Remote Sensing section of Ocean and Aquatic Affairs in Victoria and was kindly made available by Dr. Jim Gower. For a detailed description of the system see Walker et a l (1974). 39 (a) Airborne spectral reflection measurements The airborne mission was carried out along a predetermined flight path. In conjunction with the spectral measurements a video film and a 35 mm photographic coverage were obtained. The aerial mission took 45 minutes uith a Beech aircraft and the ground samples were collected over a 7 hour period, starting 3 hours prior to the airborne mission. Given the viewing conditions and sensitivity of the spectrometer, the flying height of 330 m, exposure time (3 seconds) i t was determined that each airborne spectral reflection signature corresponded to a 12 x 7 m ground surface. The measurements obtained in this way represented a continuous record of 12 x 7 m sections along the flig h t l i n e . Spectral reflection curves for each section were retrieved from the tape and plotted on con-tinuous graph paper for visual analysis. The di g i t a l data were used as a basis for the numerical analysis between f i e l d and sample parameters. The accuracy of the flig h t mission with respect to ground sampling was assessed by viewing the video film and the 35 mm photographic coverage. Seventy-five % of the sample locations were positively identified while the remaining 25% were excluded from the detailed analysis. (b) Ground reflection measurements of s o i l samples The collected s a i l samples were divided into two fractions, one far chemical analysis and the other for direct spectral reflection measurements. For the latter operation samples were dried in the laboratory (105°C) and the > 2 mm fraction was removed. Each sample reflection was measured independently and before and after each reading an incident light measure-ment using a white BaSO^ card was made. To determine the effect of grain size and ammonium acetate extraction Qn the reflection, several samples were measured in their original dry condition (no sieving) after sieving and after extraction. The samples were measured outdoors and the readings were directly fed into the computer. The set-up for the measurements i s illustrated in Plate Mil on page 40. Plate VII. Spectrometer set-up (c) Data evaluation and comparisons The data were evaluated and compared in the following ways: (i) Effect of vegetation coverage on spectral signature was examined by comparing similar fields with different types and amounts of vegetation coverage. ( i i ) The relationship between the airborne spectral measure-ments and the photographic coverage was assessed by plotting reflection values from a series of selected wavelengths along the f l i g h t l i n e . In this way the spatial variation across the fields was established. A comparison between this and a density scan on the aerial photograph was made to establish the extent to which these two methods are related. kl ( i i i ) The relationship between the airborne spectral observation and the chemical conditions was attempted in two ways: f i r s t l y mean f i e l d values were computed for both the spectral and chemical conditions and a field-by-field comparison followed. Secondly individual spectra and their corresponding sample site were identified and relationships were established using correlation regression analysis. (iv) The ground spectral measurements were treated in the same way by correlating the spectral with the chemical properties for a l l samples. The effect of certain sample parameters was also investigated. (v) Finally a direct comparison between the airborne and the ground spectral reflection values was made to determine the compatability of the two methods. CHAPTER II HIERARCHICAL TERRAIN UNITS In this study the natural landscape is viewed as a continuum in which changes are generally gradational. This concept was postulated by Uhittaker (1962) who f e l t that the fundamental problem in dealing with nature i s one of identifying basic units within such a continuum. His views are that gradational changes with relative discontinuities are more common in nature than abrupt boundaries and this i s in contrast with most terrain classification schemes which are based an identifying local discontinuities. Some success by the latter method has been achieved at very high hierarchical levels and in specific environments. However, in detailed studies emphasis on relative discontinuities becomes more important. Sites, which can be considered elementary landscape segments, are used as fundamental units in the natural landscape. The site concept as substantiated by Bourne (1931), Linton (1951), Glanzovskaya (1963) and many others - i t s suitability as a basic unit, and i t s practical useful-ness in terrain analysis - has been demonstrated by Wright (1973) and T r o l l (1966). In many terrain analysis schemes the idea of recurring units and recurring patterns i s postulated. In contrast the approach pursued in this study attempts to conform with the idea of relative discontinuities. An effort i s made to determine the degree of similarity among units by using a number of measured parameters. The degree of similarity concept (Berry 1958) i s fundamental to numerical classification and in order to apply i t to the present context the fallowing factors were i n i t i a l l y investigated: A. The internal variability of units at different hierarchical levels; 43 B. The significance of selected parameters in characterizing units; C. The internal homogeneity of the basic units and their s u i t a b i l -ity to numerical treatment. A. THE INTERNAL VARIABILITY DF UNITS AT DIFFERENT HIERARCHICAL LEVELS It is essential that individual units have a sufficient internal homogeneity to be useful in a classification. As a result primary em-phasis i s placed on establishing the type and magnitude of internal chemical variability present in the different units. F i r s t -the primary site units are investigated and this i s followed by a comparison•between different hierarchical units. 1. Internal Site Variability In a pilot study three sites were selected for a detailed analysis of internal v a r i a b i l i t y . Geomorphological c r i t e r i a such as surface con-figuration and origin of parent material were used for selecting two sites in landform units of marine origin while the third was chosen in a unit of outwash origin. The site size varied from 25 to 56 m^ . In each site a number of s o i l pits were dug and samples from the A, B and C horizons were collected for chemical analysis in the laboratory. The sample distribution and identification numbers are illustrated in Figure 15 and w i l l be referred to during the numerical treatment procedure. Site I Site II Site III (Marine unit) (Marine unit) (•utwash"unit) Figure 15. Sample distribution in detailed site analysis. Since relatively small sample sizes were analyzed an attempt was made to use nrjn-parametric stat i s t i c s wherever passible. (a) Variability: F i r s t approximation Following laboratory analysis the internal variability amongst the three sites was examined by using the Mann Whitney nan-parametric sig-nificance test. The aim of the test i s to determine on the basis df the sample populations whether a significant difference exists among the data of the three sites. The Null hypothesis (Ho) states that there i s na significant difference between the sample populations i f a probability of oL - 0.01 i s reached. In contrast the research hypothesis states that two samples are stochastically different i f a probability of oL = 0.01 i s reached. The following available elements were used in this test: Ca, Mg, Na, K, Mn, Zn, Si, CI, Cu, P, and the results are provided in Figure 16. Pb and S values were eliminated from the analysis because of their extremely IDW values which were at the lower limit of instrument detects--a b i l i t y . Accordingly an analysis of variability was not possible. Some general comments can be made from the results shown in Figure 16: (1) The same group of elements can be used to differentiate the two marine sites from the outwash site in the B horizon. The same is true for another set of elements at the C horizon. (2) The most complex conditions seem to occur at the A horizon. (3) Sodium and Magnesium are the most significant parameters in the C horizon. These results indicate that the three sites can be differentiated on the basis of their chemical environments despite their similarity in surface configuration. This i s also substantiated by Wright (1973) who refers to sites as: "areas of small extent, each characterized by a particular environment." In addition, even the sites k5 originating from similar parent material (sites I and II) can tie d i f f e r -entiated on the basis of at least three out of eleven considered chemical parameters. A Horizon Site I Mg Na K Zn PL\ Ca Mg K Mn Si Zn Site III Site II K Mn Si Zn B Horizon Site I K Si Zn PO,, K Mn Si CI PO, Site III' Site II K Si Zn PO, C Horizon Site I Ca Mg Na K Si PO^ Mg Na Zn Site III — Site II Ca Mg Na K Si PO^ Figure 16. Comparison of site variability using significance test. Elements showing a distribution which i s significantly different from site to site at oC - 0.01. (b) Magnitude of variability The internal variability of each site was measured by plotting the range of values for each parameter (see Figure 17). It i s clear that even when dealing with these small units"a substantial internal v a r i a b i l -ity i s observed. This conforms with work reported by Beckett and Webster (1971) who state that: "Even in the natural landscape' as much as half of the coefficient of variance within 1 ha i s already present within a few m2." av.Ca ppm 3000 2000 1000 -E av.Mg ppm 1500 1000 500H I jj A B C Horizons A B C av Na ppm 200 100 av.Cu ppm 0* rr, A B C Horizons B C av.K ppn 600 400 0 EH . M 1000 av. Mn Ppm 750 500 25G A B C A B av.Zn ppm 100 H , £0 60 40 av.S04 ppm 20 av.Si ppm 120C 800 400 av. PO4 ppm 100 80 . 60 40-j 20 A B C A B i l av.C! ppm 3C0 200 100 A B ETUSitel M?rine £Z2 Site 2 Marine S I Site 3 Outwasti Figure 17. Range Df site variability. cn kl The variability observed i s not systematic and the covariation alters with individual elements under consideration. The magnitude i s also ex-pressed in terms of the coefficient of variation % CU (= — x 100) and, depending on parameter, CU values ranging from 7 to 65% were obtained. The relevance of these values in regard to different hierarchical units i s the subject of the next section. 2. Comparison of Variability between Hierarchical Units To contrast the internal variability of sites with those of higher hierarchical orders, data from 37 sites within two types of landform units were used. The sites were selected from units of marine and outwash origin and an overview of the hierarchical subdivision and their associated sample distribution i s provided in Figure 18. Types of Units Landform Units MARINE A a b ' Sites Figure 18. Hierarchical distribution. OUTWASH A i (a) Uariability: First approximation The variability within landform units and unit types waseagain assessed using the Mann Whitney Significance Test previously described. In the case of landfarm units comparisons were made between marine unit a and outwash unit a. Similarly marine unit type was compared' to outwash unit type. Results of the comparative tests are given in Figures 19 and 20. 48 Landform units Outwash b Outwash a Marine b Marine a Marine a Ca (ABC) Mg (ABC) Na (C) Ca (BC) Mg (C) Na (C) no significant difference Marine b Ca (ABC) Mg (ABC) Na (AC) K ( O Ca (BC) Mg (BC) Na (ABC) K(C) Mn (A) Zn(C) P0 4(B) Outwash a no significant difference Note: A, B, and C refer to sample horizon at which distributions were Outwash b significantly different at ot = 0.01. Figure 19. Elements showing a significantly different distribution when compared at landform unit level. Figure 20 shows that thettype of variability present in each unit at the different levels of the hierarchy i s such that a differentiation i s passible in a l l cases for a considerable number of chemical parameters. There i s strong evidence that several parameters rather than a single one should be used for assessing the landform units. Type of Landform Units Marine Unit Type *— Outwash Unit Type Ca(A+C)Mg(A+E)H(C) Si(B+C) Mn (C) CI (B) SO^ (A) Cu (A) (Elements with significantly different distribution at oL = 0.01) Figure 20. Significant difference between type of landform units. 49 (b) Magnitude of variability Using the data of corresponding sites the internal variability at each hierarchical level was compared quantitatively with the site var-i a b i l i t y . The coefficient of variation (% CU) i s used in Figures 21 and 22 in which the internal variation i s presented for each element. Inde-pendence in sample distribution i s provided at the site and landform unit le v e l . However for practical reasons this i s not possible at the landform type level. Also i t should be pointed out that the use of the coefficient of variation i s only partially satisfactory since i t i s dependent on data which are normally distributed. Given the relatively small sets of samples this cannot be assumed. Nevertheless the use of the coefficient i s useful when concentrating on trends rather than absolute values. Several general observations can be made from Figures 21 and 22: (1) The variation within the sites i s , almost without exception, smaller than the variation within the larger hierarchical units (CU range from 5-65% depending on element and site; 59 out of 66 show th i s ) . (2) The coefficient of variation tends to increase with increasing size of units for most elements considered (41 out of 66 cases show th i s ) . (3) The CU values are large even in the smallest units. This indicates that up to 50% of the total variance i s already present within the smallest spatial unit. The integrated average of a l l cation variation within a site i s less than 15%. (4) In the case of site variability the CU values of the C horizon tend to be marginally smaller than in the A and B horizons, which i s not as evident at higher hierarchical levels. 50 OUTWASH UNITS 1S9 > 43 O c o ~ 100 J n > c. £ 60 20 A-HORIZON 180-1 140 > O site landform landform unit unit type 100 60 20 B-HORIZON 1C0 140 > 3? site landform | landform unit j unit type 100 60 20 C-HORIZON site landform landform unit unit type MARINE UNITS > 120H O 100 o e o c E0-| SO 40 20 A-HORIZON 120 too H > <J 80 60 40 20 site landform landform unit unit type B-HOR1ZON 120-100 > O 80 60-1 40 20 C-HORIZON site landform landform unit unit type LEGEND ; AV. Ca AV. K AV.Mg + AV. Na •*•*•» AV.Mn AV.Si site | landform unit landform unit type Figure 21. Coefficient of variation for different hierarchical units. 51 OUTWASH UNITS A-HORIZON / ; ; / / / / . \ / / -If***' — / 4 4 • 't no 140-I 100 > Site landform landform unit unit type 60 20 H B-HORI20N C-HORIZON / / / V x/ r. x/ / >X/ ISO 140 100 eo 20-^ site landform I landform I I unit unit typ6 MARINE UNITS site landform | landform unit unit type •V A-HORIZON •V N V / - T - 7 4 / 180 140 100 60 20-t B-HORIZON C-HORIZON 180 140 100 > o 60 20 7*" site landform landform 1 site landform landform 1 site landform landform unit unit type 1 I unit unit type 1 unit unit type LEGEND — CI - Zn -H- Cu P04 SO4 Figure 22. Coefficient of variation for different hierarchical units. 52 From the above i t can be concluded that the use of the site as a basic individual of the landscape i s ju s t i f i a b l e . B. THE SIGNIFICANCE OF SELECTED PARAMETERS IN CHARACTERIZING UNITS Individual parameters used in this study are of fundamental im-portance and were assessed i n i t i a l l y at the site level and then in units at higher levels. 1. Parameter Significance at the Site Level Using the results of the significance test recorded in Figure 16 i t i s possible to assess the usefulness-of individual parameters in characterizing the site units. A summary of the most important parameters (significantly different at oC= D.01) at specific horizons i s given in Table 3 below. Table 3 Parameters most significant in characterizing site units (A, B, C refer to sample horizon) Available elements Marine site Marine site IIvs II Marine site I vs Outwash site III Marine site II vs Outwash site III Ca A C C Mg AC AC C Na C AC C K AB ABC ABC Si AB BC ABC Mn AB AB A Zn AC AB AB Cu PO^ B ABC BC SO^ CI B 53 From Table 3 a number of conclusions can be reached: (1) The three sites should not be differentiated on the basis of one single element. (2) In the present case Cu, CI, and.SO^ are poor indicators of site differentiation since significant difference could not be detected at any level (B horizon only for CI). (3) Ca, Mg, and Na proved to be most useful in differentiating sites at the C horizon, while Si, Zn, and Mn were indicative of A horizon differences. 2. Parameter Significance at Different Hierarchical Levels Similarly results of significance tests were used to assess par-ameter importance at higher hierarchical levels. Besides the marine and outwash units considered in Figures 19 and 20 one glacio-marine unit (characterized by S sample sites) was also included. A comparison at landform unit and landform unit type levels i s given in Tableso4 and 5. A number of deductions can be made from the results: - No unit could be differentiated on the basis of data from single chemical elements; - Available Ca, Mg, and Na at B and C horizons proved to be the most useful distinguishing characteristics; - SO^, Cu, CI, Mn, and P0^ were only of limited use for character-izing units; - In the present example a differentiation of units on the basis of the chemical data i s not possible i f the units are of common or similar origin. Table k Significant parameters useful for characterizing landform units Comparison between landform units: Dutwash a Marine a Dutwash a Marine b Dutwash a Glacio-marine Outwash b Marine a Outwash b Marine b Outwash b Glacio-marine Outwash Dutwash •• a Marine a Glacia-marine Marine b Glacio-marine Marine a Marine b Available elements: Ca BC BC C ABC ABC C o A • Mg Na K Si C C BC ABC B AB ABC C ABC BC C C C Difference B A AB B AC Difference Mn A Zn C C Cu Cl A P% B A SO^ [Mote: A, B, and G refer to sample horizons.at which specific element distributions was significantly different ( £ = 0.01). 55 When analyzing the parameters within the associated hierarchical levels (Table 5) some general grends become evident: - Available Ca and Mg (at C horizon) remain characteristic dis-tinguishing parameters at a l l levels of the hierarchy; - The usefulness of A horizon samples to distinguish units de-creases somewhat with an increase in the hierarchical level; this i s made evident by the number of elements which show significant differences. - Available Na, K, and Si remain important in differentiating some of the units. Table 5 Characteristic parameters for differentiating outwash from marine units at different levels of the terrain hierarchy Hierarchical Level Site Landform Unit Landform Unit Type Units Compared: Site I vs Site III Site II vs Site III Marine a vs Outwash a Marine b vs Outwash a Marine vs Outwash Available elements: Ca C C BC BC AC Mg AC C C BC AC Na AC C . C ABC K ABC ABC B C Si BC ABC BC Mn A C Zn AB AB C Cu A CI B P ABC . BC B S A Note: A, B, and C refer to sample horizon at which the distribution of the specific elements was significantly different ( at = 0.01). 56 On the basis of the observations made from Tables k and 5 i t can be concluded that: (1) Available Ca, Mg, Na, K, and Si are the most useful parameters by which units can be characterized at a l l levels. The f i r s t three are more useful and consistant at the C horizon level while available Si i s important for A horizon discrimination. (2) Data from available Mn and Zn might be useful when differen-tiating units at the site level. C. INTERNAL HOMOGENEITY OF THE BASIC UNITS AND THEIR SUITABILITY FOR NUMERICAL TREATMENT Having identified units and the parameters by which these units can be characterized i t is essential to determine whether these basic sites are sufficiently homogeneous for numerical treatment. The three sites for which detailed data were available were again used for this purpose. I n i t i a l l y , grouping was attempted with single horizon data and was f o l -lowed by a classification using the complete data observation from a l l horizons. 1. Numerical Grouping using Single Horizon Data A l l samples from A horizons of a l l three sites were grouped numer-ic a l l y using available Ca, Mg, Na, H, Mn, Zn, and Si data. Phosphate was not included in this grouping because i t was f e l t that land use ( f e r t i l i z e r application and grazing) could potentially influence i t s variability greatly. The hierarchical grouping method described on page 28 was used to develop clusters in seven dimensional space. UBC C-group Computer Program (Patterson and Uhitaker 1973) was used for this purpose and the results of this manipulation are provided in Figure 23 on the following page. 57 ITEMS GROUPED Step I 3 Error 1 2 9 . 14 18 5 6 10 11 20 3 7 13 15 17 k 8 12 16 19 II I I I I U I I I I ' M M V I I Site} I 1 2 3 5 6 7 8 9 10 11 12 13 Ik 15 16 17 18 19 9 1 2 3 1 9 9 1 10 5 6 4 12 14 7 8 11 15 16 18 3 133 16 20 9 12 17 19 2 7 11 2 16 17 9 16 0.405 0.524 0.541 0.868 0.964 1.032 1.146 1.543 1.568 2.149 2.517 2.523 3.637 4.386 5.148 10.620 12.180 42.740 65.479 I I I I I I I I •"j-Sit^ II Y Sit f lit Figure 23. Dendrogram resulting from numerical grouping of A Horizon data. Considering that: Samples # 1-8 belong to Site I, Samples # 9-15 belong to Site II, Samples #16-20 belong to Site III, i t i s evident that a reasonable grouping has been achieved. Samples from each site form distinctive groups. The only query i s that Site II joins Site III before i t joins Site I. Since Sites I and II originate from the same parent material i t would have seemed logical for Sites I and II to have joined together f i r s t . The same procedure was followed for samples from B and C horizons, the results of which are given in Figures 24 and 25. 58 ITEMS GROUPED Step I 3 Error 1 3 8 13 17 16 4 9 14 18 2 6 11 ID 19 5 7 15 12 2D ' Y i i i i T T I I I I Y 1 11 15 0.5D5 2 13 14 0.600 3 17 18 0.802 4 2 5 1.648 5 19 20 1.722 6 11 13 2.D83 7 2 3 2.4D7 8 6 7 2.663 9 10 12 3.429 ID 1 16 3.757 11 9 11 4.745 12 2 4 4.947 13 17 19 6.368 14 1 2 7.111 15 6 8" 9.691 16 9 10 10.269, 17 1 6 16.352 18 1 9 34.818 19 1 17 46.D71 ! I I I I I I I ! I I I Y i i i i i r i Uf Site] II U - T - ' I I i ! u Figure 24. Dendrogram for samples from B Horizons. It i s immediately apparent that some mis-grouping occurs in the B horizon. This i s partially understandable since sampling.the B horizon is always a delicate task and matching B horizons from different s o i l pits is d i f f i c u l t . This has been pointed out by Drees and Wilding (1973) who found up to 20% vertical variation for Ca, K, Fe, and T i in t i l l and up to 40% in outwash. As expected the domination of parent material i s clearly evident from Figure 25. A distinction between the two marine units i s not pos-sible and the overall level of grouping i s very low for both the outwash and marine group. This suggests that both groups are relatively homogen-eous while at the same time the between group difference i s very pronounced. ITEMS GROUPED Step I 3 Error 1 9 3 15 20 6 13 4 12 18 7 10 5 14 17 2 8 11 16 19 V?~7 i i i ! V TT I I V 1 2 3 5 6 7 8 9 10 11 12 13 Ik 15 16 17 18 19 11 15 17 19 11 12 1 6 k 3 3 1 3 1 1 1 1 5 13 16 20 11 Ik 10 7 k 2 9 8 3 16 IB 1 11 16 17 1 16 0.203 0.371 0.472 0.486 0.533 0.606 0.924 0.983 0.138 1.158 1.607 2.544 3.596 4.423 8.568 10.085 14.982 17.171 90.138 I I I I Y Y i i i Y V I I I L _ J Y Sijte Figure 25. Dendrogram for samples from C Horizons. 2. Numerical Grouping using Data from A l l Horizons Having considered each horizon separately, i t i s passible to use a l l information at once by describing each s a i l pit as to chemical con-centration in A, B and C horizons. A grouping in 21 dimensions i s attempted and the results are given in Figure 26. It i s evident that the classification has improved to the extent that discrete groups were obtained which essentially correspond to the sampling scheme.(see Figure 26). These findings suggest that there i s sufficient internal homogeneity in each site and a distinct overall chemical difference between the three sites which allows a numerical treatment. 60 ITEMS GROUPED Step I 3 Error 1 11. 15 1.855 2 13 lk 2.549 3 k 5 4.905 4 3 it 5.337 5 6 7 5.986 6 10 12 6.381 7 9 13 6.593 8 1 3 7.572 9 17 19 8.114 10 16 20 9.208 11 2 6 9.690 12 10 11 10.607 13 9 10 13.267 14 16 18 13.993 15 2 8 18.546 16 1 2 27.855 17 16 17 30.821 18 1 9 99.736 19 1 16 196.971 1 2 9 12 20 3 6 13 11 18 4 7 14 15 17 5 8 10 16 19 I I I I I I Site jI I i I I Y I I I I I I I I Y • i I S i t e ' l l ! 4 - 1 t i l l Figure 26. Dendrogram far the combined A, B, and C Horizon samples. Under the assumption that these results are representative of the total landscape situation i t i s implicit that a characterization of sites on the basis of single sample data i s jus t i f i a b l e . 61 CHAPTER III NUMERICAL APPLICATIONS Results presented in the previous chapter indicate that most hier-archical units have sufficiently distinct chemical conditions to justify a numerical treatment. The aim in using numerical grouping procedures i s to establish whether site specific information can be used to quantify larger genetic units. Data from the Peace River and Fraser Ualley study areas were used for this purpose but because Df inherent differences between the areas and modifications in the analysis each data set was treated separately. The results from both studies were compared to assess the r e l i a b i l i t y and suitability of the method. A. NUMERICAL TREATMENT OF FRASER UALLEY DATA A numerical areal quantification of chemical conditions was attempted by using parametric data from Wb sites. Since the inclusion of redundant parameters w i l l bias the grouping two procedures were ex-amined: (1) Direct numerical grouping using parameters selected by significance tests, and (2) numerical grouping after parameter screening through factor analysis. 1. Direct Numerical Grouping using Parameters Selected through Significance Tests In Chapter II available Ca, Mg, K, Na, Mn, S i , and Zn i n the A, B, and C horizons were identified on the basis of significance tests as being most important in characterizing units. These parameters were used to quantify sites and group them together according to similarity of chemical conditions. Five major landform units were identified on aerial photographs on the basis of form, origin and parent material. Two marine units 62 (Ma and Mb), two outwash units (Oa and Ob) and one glacio-rnarine unit (GM) were selected for detailed analysis. Within these units 45 sites were analyzed in detail and a numerical assessment of sites using the above mentioned parameters was made. The hierarchical grouping proce-dure described in Chapter One was used for this purpose and the class-i f i c a t i o n method can be considered an objective one since a l l data were treated simultaneously and indiscriminately D f origin. The results of this numerical procedure are provided in the form of a dendrogram and spatial diagram in Appendix III-2 and III-3 in which sites with similar chemical conditions are grouped together. To show whether the groups are indicative of specific landform units the origin of each group member is listed in Table 6 below. Table 6 Origin of group membership from direct numerical grouping Group Associated Site Members A Mb Ma Oa Ob B Ma Oa Ma C GM GM Ob Oa Ob Ob Oa Ob Oa Ob Ob GM GM GM Oa Ob D GM Mb E Ma Ma Mb F Mb Mb G GM Ma H Oa GM Mb Mb Mb Ma Mb Ma I Mb Ma Note: Groups with less than two members were not considered. (42 sites grouped into 9 classes) From the above table a number of observations are possible: (1) There is l i t t l e evidence that the grouping obtained is indicative of single landform units. 63 (2) While there i s some indication of a discrimination along landform unit types, only a few groups of pure sites were obtained. (3) Sites from the glacio-marine unit do not form distinct groups but rather reflect their mixed origins by being associated with either Df the two types of landform units. Preliminary conclusions can be reached from these observations. There i s some evidence that sites can be classified numerically into groups which reflect genetic type of landform units. However the class-i f i c a t i o n i s not ful l y satisfactory and results from the present study do not indicate that a numerical grouping of hierarchical units i s readily passible on the basis of chemical conditions. The inconclusive results can partially be explained by the f o l -lowing: (a) Heterogeneity of units; (b) distinctness of units; (c) other factors affecting chemical distribution; (d) choice of parameters used. (a) Heterogeneity of units IMDne of the landform units under analysis can be considered pure. Two kinds of inclusions which do not necessarily reflect the genetic conditions of the unit are present: - inclusions which took place during landform genesis; and - subsequent inclusions or genetic alterations. The study area has undergone a complex Quaternary history and during the deposition of marine sediments, for example, a number of interruptions occurred. Evidence of this are the localized gravel deposits found in exposed areas with the otherwise pure marine sediments. A number of modifications have taken place since the original sediment deposition. For example, f l u v i a l modifications were responsible 64 for the deposition, removal and alteration of conditions in a number of places. Also evidence of aealian sand deposition was found in protected areas. These considerations point towards differences in the nature of the originating processes leaving some polygenetic effects which complicate an analysis of the chemical landscape. (b) Distinctness of units Another question is whether, the genetic units are distinctly d i f -ferent. In the case of residual deposits i t has been demonstrated that i t i s possible to have differences on the basis of parent material (Swan 1970, Chesworth 1973). When analyzing deposits which have undergone substantial modification through transport i t is less certain that a distinction in chemical conditions i s readily possible. However, i t is f e l t that the genesis i t s e l f i s important in producing characteristic chemical conditions. This important point is pursued in greater detail later in the chapter. (c) Other factors affecting chemical conditions Factors other than genesis and parent material affect the chemical conditions of each unit. The most important are the s o i l forming factors of topography, climate, biological action, time and land use. The influence of some of these factors on the chemical distribution is the subject of discussion in Chapter Four. (d) Choice of parameters used As mentioned previously the parameters were chosen on the basis of a significance test. Of the seven elements chosen only two were significant for a l l units. A I S D , most of the other parameters are significantly different only at one or two of the three sample horizons. Because a l l parameters for the three horizons were.used indiscriminately a biased classification might have resulted. To overcome this problem 65 an additional parameter screening i s attempted through factor analysis and mil l be examined in the next section. 2. Numerical Grouping after Parameter Screening through Factor Analysis The importance of parameters in characterizing units can be assessed by using factor analysis. Based on the assumption that correlated variables contain redundant information i t i s possible to reduce the 21 variables (7 chemical elements at three sample horizons) to fewer factors. Each factor accounts for a portion of the total variance and those uith the highest values can be considered the best distinguishing properties for subsequent groupings. In the present example 62% of the total variance was condensed into four factors and each subsequent factor accounts for an addition of less than 6%. The Importance of each individual parameter within the factors can best be seen in Table 7 of the factor loadings. The four isolated factors can be identified as follows: Factor one represents Potassium and A horizon Factor two represents C horizon data Factor three represents S i l i c a Factor four represents B horizon data Using these four factors each site was then weighted and the resulting factor scores were used for numerical classification. Again the same grouping procedure was used and the resulting dendrogram i s provided in Appendix IH-k. A much improved grouping i s obtained by the method shown in Table 8. Several observations can be made: (1) Again glacio-marine samples are not sufficiently distinct to form their own group. Instead they join one or two types of landform units depending on similarity in genesis. 66 Table 7 Factor Loadings Factors Variable 1 2 3 4 CA A * 0.711 -0.117 0.279 -0.406 CA B 0.487 0.034 0.338 *-0.566 CA C 0.542 * 0.606 0.350 0.074 MG A * 0.735 0.035 0.094 -0.330 MG B 0.455 0.314 -0.047 -0.440 MG C 0.517 * 0.697 0.269 0.195 K A * 0.601 -0.469 0.144 -0.180 K B * 0.669 -0.564 0.095 0.093 K C * 0.805 -0.224 0.222 0.088 MIM A 0.410 0.090 -0.457 0.014 MM B 0.362 0.043 -0.069 * 0.526 MN C 0.217 -0.116 -0.115 0.316 NA A 0.504 0.272 -0.537 0.173 NA B 0.465 0.379 -0.529 0.003 NA C 0.491 * 0.683 0.033 0.379 SI A -0.273 0.201 * 0.639 0.348 SI B -0.285 0.179 * 0.579 -0.043 SI C -0.315 -0.064 * 0.672 0.176 ZN A 0.587 -0.575 . 0.035 0.319 ZN B 0.428 -0.505 0.039 0.468 ZN C 0.430 0.025 0.426 0.120 Cumulative % of 26.4% 40.4% 53.1% 62.3% total variance * Most significant parameters in each factor. Table 8 Origin of group membership after factor analysis Group Associated Site Members A Mb Ma GM Ma B Mb Ma Ma C Mb Mb Mb Ma Ma D GM Ma E Mb Ma Mb F Ma GM Ma GM Ma Ca Mb Oa Mb G GM Oa Oa Ob Ob Ob GM GM GM Oa H Oa Ob Oa Ob Ob Ob Ob Ob Note: Groups with less than two members were not con-sidered. (41 sites grouped into 8 classes) 67 (2) Again no grouping of sites Df single landform units i s observed. (3) Not considering sites from the glacio-marine units only one of the eight groups includes members from opposite types of units, compared with three before factor analysis. The areal distribution of the sites and resulting groups are represented in Figure 27. This grouping i s obviously not perfect but nevertheless a strong genetic influence can be observed in that similar types of genetic landform units.are quantified by groups of sites. Since both of the marine units as well as the two outwash units are contiguous i t can be argued that the subdivision i s too a r t i f i c i a l and does not reflect the natural conditions. To investigate this hypoth-esis the same numerical technique was used to analyze data from the Peace River study area. B. NUMERICAL TREATMENT OF THE PEACE RIVER AREA DATA A study framework similar to that employed in the Fraser Ualley was used in analyzing the data from the Peace River area. In the latter study 72 sites from four types of landform units were included in the analysis. The distribution of units and associated sites i s revealed in Figure 28. Exchangeable cations (Ca, Mg, Na, and K in the A and C horizons, and % Carbon in the A horizon) were used as parameters in this study. 1. Direct Numerical Treatment of the Peace River Area Data The data collected were similar to those from the Fraser Ualley study despite differences in area, climate, parent material, units, and chemical parameters. The results from a direct grouping of the combined Ma + Mb : MARINE UNITS O a + O b : OUTWASH " A - H NUMERICAL GROUPINGS Scale: 0 1 2 3 4 5 km • Figure 27.' Spatial grouping of sites after factor analysis. 65 64 9 56 63 i; 24,2! 22 2 26 2 3 M0L2O2S29 a 27 2819! 53 SV 4 2 l % 7 S 3 ^ 5 1 " UNITS ABBREVIATIONS USED SAMPLED SITES T i l l T l 55, 56, 58, 59, 61, 62, 63, 64, 65, 66, 67 II T2 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 69, 72 n T3 60 Sandstone SI . 57 II S2 70, 71 II S3 8 • ' Lacustrine L l 9, 13, 14 II L2 5 II L3 51, 52, 53, 54 Lacustro-till LT1 10, 11 II LT2 12, 15, 16, 17 II LT3 1, 2, 3, 4, 6, 7, 31, 32, 33, 34, 35, 36, 37, 38, 39 II LT4 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 68 Figure 28. Units and sample distribution for the Peace River study area. 70 A and C horizon data are provided in the dendrogram in Appendix III-5. The sites associated with each group are list e d in Table 9 below. Table 9 Group members after numerical classification Group Associated Site Members A LT3 LT3 L2 LT2 L l B LT3 L l C LT3 LT3 T l T3 LT3 D LT3 LT2 T2 LT2 LTk LTk E T2 T2 LT3 T2 LT3 LTk F T2 T l T l T l G T2 T2 L3 H L l L3 L3 L3 I LT3 LT3 LT3 LTk LTk LT3 LT3 LT1 LT1 LTk LTk LTk LTk LTk 3 T2 T2 T l T l T l T l K T l T l T l LTk L LT2 LTk M T2 SI S2 S2 T2 LT3 T2 T2 T2 Mote: Groups uith less than two members uere not considered. (70 out of 72 sites uere grouped) The grouping results revealed that: (1) Distinctions between sites from lacustro-jfcill units and sites originating an either t i l l or lacustrine deposits are d i f f i -cult to make. This can be explained by the fact that incnmany cases gradations between pure and mixed deposits were evident in the f i e l d . (2) As observed in the previous study no single site group i s representative of a single landform unit. 71 (3) Considerable evidence for a grouping along landform types is evident however. It appears that in only two out of 13 cases were sites from different genetic units grouped together. This is most apparent in group M where three sites from t i l l units are associated with sites from sandstone units. A partial explanation for this can be found from f i e l d observation which shows most sandstone exposures are masked to some extent with t i l l . 2. Direct Grouping of C Horizon Data To determine whether such a grouping is consistent at depth only C horizon samples were chosen for this analysis. The same values for exchangeable Ca, Mg, Na, and K for C horizon samples were employed in the grouping procedure. As shown in Appendix III-6 and Table 10 a similar grouping was obtained. In this case however only one group, Group G, contains sites from different genetic units. Table 10 Grouping of sites with C horizon data only Group Associated Site Members A LT3 LT3 LT3 L2 LT2 L l L l L l B LT3 LJk LJk LT3 LT3 LT1 LT1 LT3 LJk T2 Tl T l C LT2 LT3 LT4 LTh LTk Llk LJh LTk D LT2 L3 L3 LT3' L3 L3 E T2 T2 T2 F LT2 Tl T2 T2 T l T2 T l T l T l T l T l T2 Tl T l T2 TZ T l G 1 2 T2 S2 T2 SI S2 Mote: Groups with less than two members were not included. 72 The spatial distribution of these sites i s given in Figure 29 which shows that single landform units do not have chemical features distinct enough to be quantified numerically with the site technique. 3. Numerical Grouping after Parameter Screening through Factor Analysis As in the Fraser study, the site information for the Peace River area was f i r s t screened through factor analysis and a subsequent num-erical grouping of factor scores was made. In this way the most s i g n i f i -cant factors, accounting for 75% of the variance, were maintained (see Table 11). The areal distribution Df the site groups i s given in Figure 3D. Table 11 Factor loadings for the Peace River area data Variable Factors 1 2 3 4 1 CARB A 0.107 -0.090 •-D.B21 -0.045 2 CA A 0.301 -0.124 *-0.752 -0.167 3 MG A * 0.810 -0.002 -0.258 0.D17 4 NA A -0.136 * 0.801 -0.1D2 0.246 5 h* A -0.251 0.385 •-0.717 D.238 6 CA C 0.279 * 0.735 0.106 -0.259 7 MG C * 0.831 0.179 -0.002 0.097 8 NA C 0.545 * 0.700 D.102 -0.0D5 9 K C 0.119 D.013 0.064 * 0.931 Cumulative % of total 27.3% 47.7% 63.4% 75.0% variance * Significant parameters The parameters which are most important in each factor are: - exchangeable Mcj (in A and C horizons); - exchangeable _Na (in A and C horizons) and Ca (in C horizon); - A horizon (exchangeable Ca, K, and C) - exchangeable IK (in C horizon) Figure 29 : Grouping of Sites using C-Horizon Data Figure 30 : Grouping of Sites after Factor Analysis 75 Following multiplication of these factors by the original site information a weighted score for,each site was obtained and used for numerical grouping. Results of the grouping are given in Appendix III-7 and in Table 12 below. Table 12 Grouping of sites after factor analysis Group Associated Site Members A LT3 LT3 LT3 B L2 L l L l LT2 LT2 C L T H T2 T2 LT3 T l D T2 T2 E LT3 LT3 LT2 T l T3 LT3 F LT3 LT3 LT3 LT3 LTU L T H L T H LT1 LT1 L T H L T H L T H LTH- LT2 G T2 LT3 LTU L T H L T H L T H H T2 T l T2 T l T l T l T l T2 T l T2 T2 T2 T l T l I T2 SI S2 S2 T l T l T2 J L l L T H L3 L3 The results from this last grouping procedure are encouraging in that only one group (I) contains sites from different genetic units, in comparison with two for the unfactored data. In a l l other cases a sep-aration between genetic types i s possible. Also in at least two cases (groups A and F) distinct l a c u s t r o - t i l l site types were obtained. However on the whole the Peace River factor treatment did not produce s i g n i f i -cantly better results than the direct grouping. C. COMPARISON OF THE TbJO AREAS AND ASSESSMENT OF RESULTS Interpretation of the results of this numerical hierarchical framework is best attempted by f i r s t comparing the two study areas. 76 In a second step the results of the two treatments are compared and an evaluation of the usefulness of the technique i s made. 1. Comparison of the Two Study Areas The two study areas were chosen because they can be characterized by geomorphological units which have undergone a complex Quaternary history. In both cases some units are SD complex that a single genetic classification i s not possible for practical purposes (e.g. lacustro-t i l l and glacio-marine units). The parent material for almost a l l units i s non-residual and has been modified in the process of transport and deposition. Apart from these geomorphological conditions a number of d i f f e r -ences between the two areas could be responsible to some degree for the alteration of chemical conditions. The main differentiating factors have been outlined in Table 13. Another important difference can be found in the choice and use of soluble chemical parameters. Available elements were analyzed for the Fraser study and exchangeable elements were analyzed for the Peace River study. These extraction procedures should produce similar results but upon sample testing i t was evident that the values were not directly comparable. As a result a value comparison between the areas was not made. Instead the results from the two separate grouping analyses were compared. 2. Comparison of Results Since the same hierarchical framework and numerical procedure were used in both studies a comparison of the grouping results was possible despite the inherent differences in the surface conditions and method of analysis. 77 Table 13 Differences between study areas Fraser Valley Peace River Differences in: Study Area Study Area Climatic conditions: - Latitude 49.5°l\l 55.5'IM - Regime Maritime climate i n - Moderate continental fluenced by the coast climate, warm dry mountains summers, cold winters - Mean monthly tem- + 4°C to + 17°C -16.5'C to + 15.5°C perature range - Frost free days 180 days 70 - 90 days - Precipitation 130 - 150 cm 40 - 50 cm (mean annual) - Snowfall 80 cm 170 cm (mean annual) Landforms: - Topography Undulating Undulating with deep canyon type dissection - Elevation 30 - 100 m a.s.l. 700 - 800 m a.s.l. - Slopes 0 - 36° 0 - 4 0 ° Parent Material: - Genesis Marine and Fluvio- Glacial and lacustrine glacial deposits deposits Vegetation: - Type and cover Alder forest 15% Aspen and Black Spruce forest 50% Land Use: - Arable agriculture 5 % 30 % - Pasture land 40 % 5 % - Subdivisions 15 % -- Intensity of land Considerable applica^ L i t t l e application of use tion of f e r t i l i z e r s f e r t i l i z e r s . Burning. - Length of use in 100 years 50 years settled agriculture 78 (a) Hierarchical units Results frc-m both grouping procedures revealed that the proposed hierarchical landform unit framework has validity at the site and landform unit type level. In both cases a numerical classification of sites produced groups which to- a high degree reflect genetic types of landform units. Chemical conditions of individual landform units do not appear to be distinct enough for ready discrimination by numerical grouping techniques. From the grouping of the chemical data i t i s further evidenttthat mixed units of polygenetic origin do not produce sufficiently distinct chemical conditions to yield a separate genetic type classification. Only in the case of the l a c u s t r o - t i l l units in the Peace River area was i t possible to identify two groups of site types which were characteris-t i c of this type of unit. The remaining sites had conditions more similar to associated genetic type units. Although a clear chemical differentiation between types Df land-form units was possible in both study areas, the results from the Fraser Valley are more complex. IMo single site group which reflects polygenetic or mixed units was obtained and one out of eight groups of sites contain members from distinctly different genetic units. The greater complexity in the Fraser Valley i s attributed to a more complicated Quaternary history and to a more intensive alteration of the units by land use. Slightly more contrasting chemical conditions were observed between genetic types of landform units in the Peace River area. IMo group contained sites from more than two types of landform units and in four out of ten cases classes representing single type units were obtained. Even the mixed l a c u s t r o - t i l l units produced two groups of sites which portray distinct chemical conditions. A l l other groups are made up of a core of basic sites to which selected members from mixed units were added. The fact that a l l t i l l type landform units can be described by five classes of sites - two pure, two containing 79 sites from l a c u s t r o - t i l l units and one including sites from sandstone units - suggests gradational changes between landform units or the lacustrine i s reworked t i l l D f similar lithology. (b) Numerical method The hierarchical grouping procedure used in this study made possible a numerical treatment of multiparameter data. The inclusion of the greatest possible number of parameters does not produce a good classification, as shown by the results of the direct classification of both data sets. The use D f parameters which have few differentiating characteristics only tend to bias the classification. The assessment and screening of variables through factor analysis prior to numerical grouping improves the classification considerably for the Fraser Ualley data. This area was inherently more complex and the screening method produced groups which were genetically more homogeneous and more readily interpretable. The i n i t i a l 21 variables (7 chemical parameters in A, B, and C horizons) were reduced to four factors which can be interpreted as: (1) Potassium and A horizon, (2) C horizon, (3) S i l i c a , and (4) B horizon. In this case the variables were success-ful l y reduced without the loss of significant information. The same method was less useful far the Peace River data, where the chemical data were somewhat less complex. In this case a direct grouping of the unscreened data produced better results than the factor analysis. When using the numerical grouping method directly i t should be noted that the choice of parameters is most c r i t i c a l . The omission D f important parameters or the addition of redundant variables greatly affect the grouping results. The differences in results from Tables 9, 10 and 12 show this effect clearly. Despite the use of slightly 80 different chemical parameters results uhich are readily interpreted uere obtained for both study areas. 3. Implications and Conclusions The hierarchical geomorphological framework used in this study proved to be useful at the site and landform unit type level. Sites were classified according to similarity in chemical conditions and the resulting groups of site types are to a large degree indicative of genetic types of landform units. A number of site types could therefore be used to quantify larger units. Individual landform units, however, could not be identified by this method. This implies that the overall genesis and parent material are considerably more important at the site scale than local conditions in climate, position, land use, biota etc. This view i s upheld by Solntzev (1972) who stated that inherent characteristics are far more important in determining chemical conditions than biotic and human factors. The group of sites derived from the Peace River area classification are highly reflective of gradational changes within and between types of landform units. T i l l type units, for example, are quantified by five site types, two of which are unique, while the others include sites from related or polygenetic units. This means that i f the concept of recurring units i s valid i t i s most likely to be applicable within single or associated types of genetic landform units. In no cases are sites from distinctly different genetic units grouped together (e.g. t i l l vs. lacustrine). The numerical procedure used in the present study proved to have several advantages. Because of i t s f l e x i b i l i t y a large amount of multiparameter data can be handled with efficiency and speed. The method is objective since a l l data are treated indiscriminately'and 81 simultaneously. The problem of parameter choice was partially solved by using significance tests and factor analysis. The objectivity of such a procedure i s sometimes questioned (Johnson 1968) because of the subjective decisions which have to be taken during the analysis. The most important factors are: (1) The level at which the groups are extracted from the den-drogram i s a subjective choice. This problem was handled in the following way: Next to the dendrogram the degree of similarity was plotted as a number representing within group variance. The group extraction was made at that level where a distinct break in the distance values occurred. j (2) The number of factors to be extracted is a subjective decision. In the parameter selection procedure factors are produced which account for a certain part of the tatal variance. The f i r s t few factors usually account for a larger portion of the total variance. In the present case thennumber of factors was limited at the point where only a slight increase in total accounted variance occurred withtthe addition of the next factor. (3) Input variables - Chemicals-are selected on the basis of a linear correlation, an assumption which i s not necessarily correct. Data transformation can only partially solve this problem. One otherlimportant problem with numerical hierarchical procedures i s that when used in the absence of distinct natural clusters the method becomes unreliable, producing an a r t i f i c i a l classification. Despite these drawbacks i t i s f e l t that the method Is a useful tool for providing a multi-dimensional analysis helpful in understanding the complexity of chemical terrain conditions. 82 CHAPTER IV FACTDRS AFFECTING PARAMETER VARIABILITY When investigating chemical terrain variability i t i s necessary to identify those factors uhich are responsible for significantly a l -tering the chemical conditions. In this study the factors employed by Jenny (1941) in his model of s o i l development uere used as guidelines. The effect on chemical distribution by such factors as geomorphology, parent material, time, biota, land use and climate uere investigated. A number of arguments for and against the use of s o i l forming factors have been postulated by Bunting (1965), Buol et al (1973), Kline (1973) and Birkeland (1974). No satisfactory quantitative explanation for the processes of s o i l development has yet been obtained for the s o i l factor equation (Kline 1973, Yaalan 1975) because of (a) the complexity of factor interaction, (b) the polygenetic nature of the land surface, and (c) the problem of selecting single parameters by uhich these factors can be quantified. This complexity is further complicated by the dynamic nature of the system in uhich almost a l l elements migrate. Perel'man (1961 arid 1967) for example differentiated betueen biogenic, physical, chemical and mechanical migration. The usefulness and relative importance of individual factors i s scale dependent and w i l l be analyzed in this context in the present chapter. A. MACRO- TP MESD-SCALE ANALYSIS Structure, stage and process are the most significant considerations in the physical landscape at the macro-scale level. The s o i l factor uhich relates best to structure i s parent material; stage i s indicated by time, and processes are related to climate, hydrology, vegetation and land use. 83 1. Parent material Parent material exerts considerable influence in the early stages of s o i l development. Over time however other factors tend to outweigh i t s influence (Chesworth 1973, Birkeland 1974). Solntsev elaborated on this concept stating that "inherent nature" has a far more important impact on the natural environment than a l l biota. He sees the total environment as being shaped by biotic modifications of the basic "geomic" environment. The influence of parent material i s of course best ex-pressed in the C horizon while A and B horizon data can only rarely be used to indicate parent material. To assess the impact of parent material in the Peace River study area i s d i f f i c u l t , since most landform units are depasitional with only a small proportion of residual material present. Glacial deposits tend to reflect the lithology over which the glaciers have passed. The com-plex Quaternary history in which Canadian Shield and Cordillera ice sheets interacted (Mathews 1963) makes such interpretations more d i f f i -cult. The s o i l development is of relatively recent age and despite the variety of deposits a relationship between parent material and chemical distribution i s expected. Landform units intthe Peace River area were chosen according to form and genesis. Three types of units - t i l l , l a c u s t r o - t i l l , and lacustrine units - were identified. The chemical conditions of the units were quantified by using the site specific data for each type of unit. When analyzing the data by a Mann Whitney significance test (see Figure 31) a significant difference was obtained for several parameters in a l l cases. Parent material alone cannot be wholly responsible for producing these different chemical conditions since modification through genesis is likely to have l e f t a strong imprint. Mg(C) !Ma(A + C) Sites from t i l l units Sites from lacustrine units Na (A + C) Mg (C) Sites from lacustro-t i l l units Mg (C) Na (A) K (C) Figure 31. Result of significance test ( cL = 0.01) Uhen assessing the same differences at the meso scale of individual landforms a somewhat different picture emerges. The result from the significance test in Figure 32 reveals that some, but not a l l , units have a significantly different chemical environment. Only two t i l l and two l a c u s t r o - t i l l units had the same chemical conditions while the majority of the other units were significantly different. Not only did some D f these units undergo the same genesis but, as indicated by Mathews (1963), a number of units were made up of the same source material. Exchangeable Ca was found to be a poor discriminator in a l l cases. This could be the result 'of translocation rather than leaching since dry climatic conditions prevail. In addition t i l l units can predominantly be discriminated by C horizon data while lacustrine type deposits tend to be identifiable on A horizon data. This could reflect permeability conditions with indication of degree of leaching. Differences in land use could equally explain this effect, especially since the lacustrine deposits are under cultivation while the t i l l units are predominantly under forest cover. The observation that most of the individual landform units showed significant differences in chemical conditions indicates that other factors, such as environmental conditions under which the deposits were formed, are responsible for the differences. From this i t i s evident 85 that parent material i s a mere useful indicator of chemical conditions at the macro-scale despite local differences in topography, biota and land use. LTk LT3 LT2 L3 L l T2 T l Mg C Na C K C Mg A + Na C K C C Mg C Na C Mg A + C Na A + C K A + C Mg A + C Na A + C K C 12 Mg C Na C Mg A + Na C C Mg C Na C Mg A K A Na C Mg A + C Na C L l Mg A + C Na A Mg C Na A + K A C Mg A + C Na A H A Mg A + C Na C L3 Mg A + C Na A Mg A + Na A C Mg A + C Na A Only parameters showing significant differences at oC = 0.05 are li s t e d . The following chemicals were considered for A and C horizon LT2 Mg A LT3 Mg A Ca A samples: Ca, Mg, Na, K. Figure 32. Results of significance test between chemical parameters of different landform units. 2. Time Factor It i s generally understood that chemical conditions are dynamic and change over time. The chemical imprint l e f t upon the s u r f i c i a l layer by parent material diminishes over time. It i s however d i f f i c u l t to estab-l i s h exact time limits to assess nu l l i f i c a t i o n of parent material. In addition climatic variations (Collins et a l 1970), erosion and deposition (Kleiss 1970), and biotie cycling (Unite 1973) w i l l influence the chemical conditions. "Although no attempt was made to follow up these concepts, an analysis of short term effects was made in the case of the Peace River study area. 86 As mentioned in Chapter One, two sets of samples uere taken over the same f i e l d area uith a three-week time difference. This not only allowed an assessment of time variation hut i t also made passible an evaluation of sampling adequacy. The second set of samples uas collected uithin a 32 hour period and because of the time constraint only about 50% of the previously analyzed sites uere resampled. In this last pro-cedure only A horizon samples uere considered and a comparison between the two duplicate sets uas then made using a T-test. I n i t i a l l y i t uas assumed that particle size of the samples would not change over a three-week period. Therefore wet sieving and hydrometer measurements were made for each duplicate sample. Of the 40 sets two showed grain size distribution which was larger than errors inherent in the method (> 9%). These differences were attributed to sampling error and the two sets were eliminated from subsequent analysis. For a l l other samples a significance test was made between the chemical parameters using a T-test. At a level of 98% exchangeable IMa and K values differed significantly between the two sets. No evidence of short term variation could be found in the data for C, exchangeable Ca, and Mg. This indicates that the sampling procedure was reasonably adequate and that Na and H were the most dynamic elements over the three-week period. The relative mobility of elements in weathering i s , according to Anderson et a l (1958): Ca>Na>Mg>H. This sequence cannot be used to explain the above findings. However, Haljonen and Carlson (1975) pointed out that in an organically rich sedimentary environment the relative mobility i s Na>K>Mn>Ca. Potassium changes can probably be attributed to biotic action but as mentioned by LJhite (1973) Na i s usually excluded by common grasses in a solodized environment at the expense of selectively accumulating K and Mg in the leaf tops. McLean and Carbanell (1972) also found some type of fixing mechanism which can temporarily tie up Mg and Ca for later release. Sodium variations could also have resulted from 87 aerial migration of salts (Tsyguanenko 1968, Smith et a l 1970), or through the actions associated with the hydrological cycle (Crisp 1966). The latter i s more probable since IMa is the most soluble element considered. These short term variations point towards a dynamic environment and the time factor has to be considered important in studies of the variability of soluble chemicals over the land surface. 3. Process Oriented Factors At the macro-scale climate, hydrology, biota and land use are im-portant factors relating to processes. (a) Climate Since the work of Glinka in 1892 climate has been considered the most important single factor in weathering (Oilier 1969) and s o i l forma-tion (Jenny 1941). At the macro-scale the extensive works by Strakhov (1967, 1969 and 1970) are given as an example. In the present study an assessment of the macro climatic effect upon the chemical distribution was attempted by selecting two f i e l d areas in different climatic zones. Because of inherent differences between the two study areas and modification of the method of analysis a direct com-parison was not possible. The meso climatic effects were examined by comparing site aspect values to chemical conditions. In contrast to the findings by Cooper (1960) and Hembree and Rainwater"(1961) no simple relationship was found at this mesb scale. The effect of aspect on weathering and chemical distribution i s of course a function of slope angle, vegetation and r e l i e f and since the majority of the sites were located'on more gently sloping sedimentary landform units the aspect effect was not expected to be dominant. In the absence of site specific climatic data on precipi-88 tation and temperature a more detailed analysis of other climatic effects uas not pursued. (b) Hydrology The only hydrological parameter considered uas s o i l moisture uhich when analyzed over short periods of time i s more dynamic than any of the chemical parameters. Therefore only the data from the 32-hour sampling scheme was suitable for assessing s o i l moisture v a r i a b i l i t y . Using the coefficient of variability as a relative indicator i t was found that for A horizon samples percent s o i l moisture, after percent Carbon and ex-changeable Sodium, was the third most variable parameter for the study area. As such s o i l moisture was more variable than exchangeable Calcium, Magnesium, Potassium, and Phosphorous. Contrary to expectations a sig -nificance test did not reveal any difference in s o i l moisture between types of landform units. Logistical problems of sampling C horizons over short time periods prevented a similar comparison at the C horizon level. In addition, using the same data set no direct correlation between s o i l moisture and chemical conditions was found. Cc) Vegetation and land use The influence of land use and vegetation on the chemical distribution of soluble elements has been the subject of extensive investigations ( c f . Vinogradov 1963, Brazilevich and Rodin 1971, Unite 1971, Gerasimov et a l 1972, Vertraeten 1972). In the present study only a superficial assess-ment of potential effects by these factors could be made. Two land use categories - forested and arable land - and percent vegetation cover were recorded at each si t e . These parameters proved to be of l i t t l e value since no correlation between them and the chemical distribution was obtained. A significance test using the chemical data from forested sites vs arable sites within single landform units did not reveal any indication as to effects of land use. Only the percent s a i l moisture content was s i g n i f i -cantly different between forested and arable land intthree landform units. 89 The choice Df parameters as well as the masking effect of other factors are probably responsible for the lack of positive evidence with regard to chemical conditions. 8. ME50-5CALE INVESTIGATION Investigations related to geomorphology uere examined in greater detail at the mesa and micro scale. In geomorphology emphasis has been placed on relating surface features to s o i l properties (for a review see Appendix I ) . The motivation behind this effort i s that i f a direct relationship can be identified i t would be possible to predict s o i l con-ditions from surface parameters. Surface features are more readily iden-t i f i a b l e and observable by remote means and as such this would simplify the evaluation of s o i l properties considerably. Unfortunately specific relationships between surface form and s o i l properties only occur within discrete geographic areas and are made more complex by their interdepen-dence on other factors such as climate, time, erosion and deposition. Evidence produced in Chapter Three indicated that chemical varia-b i l i t y i s already substantial in meso-and micro-scale units. An assess-ment of factors which influence such variability was nevertheless attempted at the meso scale using form and process parameters for the Peace River study area. 1. Correlation amongst Parameters Numerous studies have identified parameters by which surface forms can be quantified ( c f . Curtis et a l 1965, Gregory and Brown 1966, Daniels et a l 1970b, Gerenchuck et al 1970, King 1970, Evans 1974, Speight 1974, Blong 1975). Most of these studies are concerned with morphometry and as stressed by-Lustig (1969) more emphasis should be directed towards relating form featuresto s o i l process parameters. 90 A direct correlation between slope and s o i l parameters has been reported by TrDeh (1964), Acton (1965), walker and Ruhe (1968), and Young (1972). The principle effects are on s o i l thickness and s o i l drainage, the latter being affected most significantly by slope position and shape. Since drainage affects the soluble chemical conditions i t follows that slope w i l l also affect chemical conditions. Furley (1968) demonstrated such a relationship using Carbon, Nitrogen, pfl and slope values, while Wilcox et al (1957) used chemical data from a mountain stream to show the elevation effect on chemical movement. Using form related parameters such as slope angle, r e l i e f , and slope position a multiple correlation analysis was performed with the Peace River site data to see whether predictions of chemical properties can be made from these parameters. The site data were treated indis-criminately and chemical versus form relationships were investigated. No direct correlation was obtained for slope angle, slope position and soluble chemicals but the previously mentioned relationship of slope angle with s o i l thickness and drainage was confirmed. The significant correlations are lis t e d in Figure 33 below: r = 0.49 r = -0.48 Soil color % sand % s o i l moisture in (value and chrome) A horizon r = 0.45 Soil depth Slope angle (°) (cm) r = -0.68 Figure 33. Correlations with slope values. Correlation coefficients are significant at JL = 0.005 D.F. = 71 The only significant correlations between geomorphological and chemical parameters are given in Figure 34 an the following page. 91 ex. Mg in A horizon (ppm) r = -0.47, Elevation r = -0.57 r = -0.38\ ex. Ca in C horizon (ppm) r = +0.36 ex. Mg in C horizon (ppm) Figure 34. Correlations related to elevation. Correlation coefficients are significant at oi = 0.005 D.F. = 71 These findings suggest that a possible relationship exists between elevation and exchangeable Ca and Mg. It can be argued that elevation i s not a purely geomorphological attribute in that i t also affects chemical conditions by Influencing climate. To see whether such a relationship has general application, data from single landform units were tested in a similar manner. In only four out of seven landform units could a sim-i l a r correlation be confirmed, and this indicates that a more complex relationship underlies this connection. To assess the elevation-chemical relationship in greater detail central tendency stati s t i c s were then used. 2. Mean Values UJhen testing the correlations within single landform units i t was shown that units associated with lacustrine environment tended to have higher values for exchangeable Mg and IMa (this was best seen in the C horizon data). Because of their genesis these units are generally located in the lower section of the study area. More interesting, however, is the fact that on the basis of the chemical data two types of lacustrine deposits were identified. This is evident from Table 14 where the mean values of each unit are l i s t e d . 92 The associated l a c u s t r o - t i l l units show similar trends (see Table 15). In a l l cases type I has very high Mg values, while type II has s i g n i f i -cantly lower values. Table 14 Mean values for lacustrine units Chemicals Type I Units Type II Units (in ppm) L l L2 L3 Mg (A Horizon) 234 276 29 Na " 34 7 40 Mg (C Horizon) 663 830 74 Na " 77 114 . 37 Table 15 Mean values for lacustro - t i l l units Chemicals Type I Units Type II Units (in ppm) LT2 LT3 LT4 Mg (A Horizon) 152 257 114 Na " 9 4 7 Mg (C Horizon) 390 414 33 Na " 46 30 38 Referring to Figure 28 i t i s evident that Type I (LT2 and LT3, L l and L2) are contiguous units separated from Type II (L3 and LT4) by the major "canyon of the Beatton River. This suggests"that the two"type units east and west of the canyon have been formed under different genetic environments. This can be confirmed by an analysis of the position of the deposit with respect to elevation. The lacustrine and associated deposits east of the Beatton Canyon are 30 to 45 meters higher in elevation than similar deposits to the west suggesting that the eastern deposits 93 are older and were possibly formed from different source material or under different environmental conditions. This demonstrates that in the case Df the Peace River area the relationship between elevation and chemicals i s the result of genesis of the deposit, and the form and position of the units can be used as an indicator of genesis. An eval-uation of single parameters does not necessarily lead to a simple solu-tion and as already shown other factors might obscure simple relationships. To cope with this complexity a multiparameter approach was assessed and i s discussed in the next section. C. ME50- TO MICRO-SCALE INVESTIGATIONS In Chapter Three i t was demonstrated that landform units are only partially unique and factors other than form affect the chemical d i s t r i -bution. The form-chemical relationship can however be analyzed in other ways: (1) by using multiparameter data to assess single landform units, and (2) by additions of new parameters to the numerical grouping. 1. The Use of Multiparameter Data in Assessing Single Landform Units Using data from single genetic landform units has the advantage of minimizing the influence of such factors as parent material, regional climate, and time. Within such a framework the form-chemical relationship was assessed by analyzing the data from 15 sites within the t i l l 2 land-farm unit (T2). The same clustering, procedure as described in Chapter One was used to classify these sites according to their similarity with respect to soluble chemical conditions. Values for Carbon, exchangeable Ca, Mg, Na, and Y\ in the A and C horizons were considered and the resulting grouping is given in Figure 35 on the following page. Each group (A, B, C, and D) can be referred to as a distinct chemical site-type. 94 ITEMS GROUPED Step I J Error 1 9 13 0.917 2 2 6 1.044 3 9 15 1.497 k 5 14 1.707 5 10 11 2.747 6 8 9 3.390 7 k 5 3.661 8 2 12 4.391 9 2 10 8.335 10 k 8 8.758 11 1 3 12.157 12 7 13.900 13 1 2 19.891 Ik 1 4 37.596 1 12 5 13 3 10 14 15 2 11 8 7 6 4 9 I I I I I Y J I Figure 35. T i l l unit 2, chemical site types resulting from numerical grouping. If we l i s t the physical properties of the respective sites within each group the following data (Table 16) are obtained: Table 16 Physical properties associated with sites Parameters A B C D Particle size 12 - 27% (35)* 22 - 28% 20 - 30% 8 - 19% (sand fraction) Slope at site (°) 1.5 - 4 (18)* 0.5 - 1.5 5 - 11 5 - 7 Slope below site 1 - 3 (13)* 0.5 - 3 4.5- 12 4 - 9 ( ° ) Penetrometer 100 - 219 209 - 238 145- 209 80 - 260 readings Elevation (m) 789 - 784(703)* 736 - 808 756- -793; 705 - 775 pH 5.5 - 7.2 5.1 - 5.9 5.1- 5.7 4.9 - 6.3 * One of the five members of group A showed extreme values. 95 The values in this table suggest that besides.the slope information no single parameter can be used as a differentiating criterion. Never-theless each group of sites seems to portray a distinct combination of physical properties. This can be seen when the site locations are plotted on aerial photographs (see Plate UIII). Despite differences in land use, groups A, B, and C are contiguous and they f a l l within a dis-tinct spatial pattern with regard to the position within the landform unit. A schematic block diagram representation of the landform unit i s given in Figure 36. From this i t i s possible to obtain a better spatial impression as to the location of sites and subsequent groups. This presentation would suggest that a relationship exists between the chemical distribution and position on the landform unit. Group B seems to be representative of the gently sloping upper section of the unit, while group A refers to sites in the lower section on the eastern side of the unit. Group C on the western extremity of the unit i s at a slightly lower elevation and more steeply sloping. Finally group D joins two sites from difEerent areas. As evident from the tree graph (Figure 35) these sites are not similar and they join each other at a late stage in the classification. scale 0.5 N 1 km FIGURE 36: TILL UNIT T2 CHEMICAL DISTRIBUTION Plate VIII. Part of T2 landform unit 97 The lack of consistancy in group D can be attributed to inherent shortcomings in the classification method uhich assumes that ultimately a l l sites have to join groups to make up the total landscape. Similarity can be expressed as being inversely proportional to distance from origin. The sooner two sites join on the tree graph the more similar they are. Results from the above described grouping procedure suggest that a numerical processing of site data on the basis of chemical information shows promise at the lowest level of the hierarchy, and that position within the landform unit seems to influence the chemical conditions. 2. The Addition of New Parameters to the Numerical Grouping As mentioned previously only some chemical parameters were found to be useful as distinguishing properties. It i s of course possible to add any number of additional parameters to the grouping procedure in the hope of improving the output. Since the units being analyzed are landform units some distinguishing properties which are expressive of form could be added to the procedure. This was pursued with data from another single landform unit, the l a c u s t r o - t i l l unit LT3. In addition to the chemical parameters (Ca, Mg, K, Na) slope angle at the site, elevation and Carbon content in the A horizon were included in the grouping. The dendrogram from such a grouping i s given in Figure 37 an the following page. Again four main groups can be identified. Before analyzing the individual groups i t should be stated that unit LT3 i s heterogeneous, having a low r e l i e f hummocky surface with internal, poorly drained depressions. Site 14 does not join any group and i s located at the edge of the Beatton Canyon to the west. The exact loca-tions of these groups are represented in Figure 38. The range of values of some of the more important properties for each group was compiled and. is presented in Figure 39 on page 99. 98 ITEMS GROUPED Step I 3 Error 1 7 6 9 2 10 8 lk 3 5 13 k 12 11 T—T I I V I I 1 8 13 1.764 2 7 10 2.39it 3 8 11» 2.438 k 1 2 3.667 5 5 12 it.112 6 8 9 5.558 7 5 6 8.762 8 7 11.770 9 5 8 12.173 10 1 3 16.796 11 it 5 17.752 12 it lit- 22.289 13 1 k , itit.513 T Y AA 4i Figure 37. Grouping of sites in landform unit LT3. Group A consists of three sites originating on the upper sections of a moderately sloping rise in the central part of the unit. Group B consists of sites from depressional areas, where stagnant water persists for part of the year. Group C is a combination of sites originating from sections of two more extensive slopes (2-5° in angle and several hundred meters long). Group D consists of sites from gentle slopes in the upper sections of hummocks. B FIGURE 38.SITE LOCATION AND GROUPING D ppm 2000 1500 1000 500 Ca(A) D D ppm 1500 1000 500 A B C 0 C a (C) D • D % 12 8 4 A B C D Carbon D D • Mg (A) ppm 600 400 200 • • ppm 1300 800 400 A B C D Mg(C) D • N s (A) (o) 6 • D ppm 12 8 4 n • A B C D ppm 80 60 40 20 D Na (C) A B C D Slope Angle m 700 H 670 640 A B C D A B C D A B C D Elevation ° 0 K (A) ppm 300 200 100 D D ppm 120 80 40 A B C D K (C ) • DDO A B C D Color Value • • • A B C D A B C 0 FIGURE 39. RANGE AND VARIABILITY OF DIFFERENT GROUPS 100 D. CONCLUSION Three basic problems uere encountered uhen attempting to identify factors uhich significantly affect chemical conditions: (1) the scale of investigation, (2) the choice of parameters, and (3) factor interac-tions. The scale of investigation i s most important since some factors such as climate and, to a lesser degree, vegetation and land use are more readily assessed at the macro- to mesa-scale uhile geomorphological par-ameters are more useful at the meso- to micro-scale. The hierarchical framework i s advantageous in this context since the most detailed informa-tion can be integrated for use at the mesa- and macro-scale. The coice of parameters by uhich factors can be quantified i s of primary importance and, as evidenced by the inconclusive assessment of the climate, biota and land use effect, simple parameters most often do not permit an adequate assessment. A multiparameter approach gives better results although an assessment of the functioning and importance of individual variables becomes more complex. Parameters should be indica-tive of processes and genesis in order to allow a more r i g i d analysis. This was demonstrated by the example of elevation and slope position. Factor interactions are largely responsible for obscuring relation-s h i p s . This problem can best be overcome by a selective treatment of factors in uhich some parameters are held constant uhile the variability of others i s assessed. The genetic-geomorphDlogical unit frameuork used in this study is quite useful in this respect since i t provides a medium in uhich variation in parent material i s minimized. The relative impor-tance of such factors as biota, land use and time could not be assessed from the collected data and there i s evidence that no simple relationship amongst parameters can properly account for the individual factor behavior. 101 CHAPTER M REMOTE SENSING: MULTISPECTRAL PHOTOGRAPHY A. INTRODUCTION 1. Remote Sensing Applications Chemical terrain variability studies cannot be undertaken without gathering a large amount of quantitative parametric data. Such a process has inherent time and cost problems and to see whether i t can be improved a number of remote sensing techniques have been evaluated. The aim of this chapter i s to examine the potential of predicting chemical conditions by multispectral techniques. As mentioned in Chapter One the physical properties and structure of an object greatly influence the emission and reflection of electromagnetic energy. To determine whether the spectral signatures are suitable for the quantification of selected chemicals, spectral reflection measurements were obtained via photographic means and also through direct d i g i t a l measurements. The photographic approach is examined in this chapter while the spectral measurements are the subject of Chapter Six. A detailed literature review on the application of multispectral sensing in detection of s o i l chemical conditions i s provided in Appendix II. 2. Multispectral Photography The introduction of color photography has greatly improved the interpretation of complex ground scenes (Colvocolesses 1975, Aldrich 1966 etc.). The reason for this improvement i s generally attributed to obser-vations by physiologists who claim thattthe eye-brain combination can identify same 20,000 colors while differentiating only 200 shades of grey tones (Evans 1948). To take advantage of this potential, multi-spectral photography and color enhancement techniques have been developed. Their usefulness in quantifying chemical surface conditions w i l l be examined in this study. 102 Rocks, soils and vegetation components are the main constituents of surface composition. Inference with regard to chemical conditions can be made directly by using geological evidence. Vegetation dependent methods, according to Gilbertson and Langshaw (1975) can be either geo-botanical or biogeochemical. In the geobotanical approach plant species indicative of certain s o i l and mineral conditions are used (Cole et a l 1974) while the biogeochemical approach involves the uptake of s o i l chemicals by vegetation. This uptake w i l l affect the spectral properties of the vegetation which can then be detected by remote spectral measure-ment (Yost and Uenderoth 1971b, Howard et a l 1971). The spectral response of different vegetation surfaces and bare soils and rocks i s therefore of great potential value in chemical terrain analysis. To emphasize certain spectral bands which are more useful in contrasting such different sur-faces multispectral photography was used. Emphasis was placed on the green, red, infrared and the f u l l color spectral bands t Q contrast d i f -ferent vegetation and s o i l conditions. The four band f i l m / f i l t e r combination described in Chapter One was obtained for the Peace River study area while simultaneous ground sampling took place. Two sample frames (# 40 and 45) covering the l a c u s t r o - t i l l unit LT3 were selected for detailed analysis. The f i r s t represents a grass covered surface; a considerable amount of bare s o i l i s exposed in the second frame. An assemblage of the four-emulsion photographic cover for both study areas is given in Plates IX and X. A visual analysis of the multispectral images reveals that a greater differentiation of terrain conditions i s possible from the color film. This i s in agreement with findings by Piech and Idalker (1974), Anson (1970) and Parry et al (1969). The false color IR-film on the other hand seems more suited for a separation of vegetation type (Gates 1970). Uith this multi-emulsion photography a stepwise re-creation of sur-face reflection i s possible by relating different tonal and textural con-ditions to the collected ground data. These tonal variations were 1D3 C D Plate IX. Four band multispectral photography. Frame 45, grass covered surface. A = IR color, B = f u l l color, C = green band, D = red band. 105 quantified by density slicing and additive color viewing enhancement techniques. IMAGE'QUANTIFICATION THROUGH DENSITY SLICING A spatial Data System Density Slicer was used to quantify density differences for each single image. The machine allowed a separation of 64 density classes and by assigning a number of different color combina-tions to individual density levels an enhanced image was created. The' color combination was chosen subjectively so that the best experimental contrast could be obtained. An example of the enhanced density image i s provided in Plate XI. It should be noted that each color represents an equal density area and with a built-in planimeter i t was possible to determine the percent cover of each color category. The ground sample conditions were then related to the colored density levels in an effort to determine possible relation-ships. 1. Vegetated Surface Frame No. 45 was chosen because of i t s interesting vegetation pattern (Plate IX) which in fact reflects the s o i l moisture and drainage conditions. The vegetation pattern is used as an indicator of drainage and chemical conditions. The locations of the sample reference sites were transferred to the aerial cover and to each of the sliced images (see Plate XII for sample location). Each of the four-band images was analyzed separately by the density sli c i n g method and because vegetation was a chief dis-criminator the best results were obtained from the IR-cDlor film (see Plate XIII and Table 17). 1D.6 P l a t e X I . Color enhanced density l e v e l s . P l a t e X I I . Ground sample l o c a t i o n s , (vegetated s u r f a c e ) . 1D7 Plate XIII. Color enhanced equal density area from color-IR film. Table 17 Color categories related to ground conditions Color Sample Soil % ex. ex. ex. ex. ex. Percent Category No. Moisture C Ca Mg Na h P Aerial ppm ppm ppm ppm Coverage Light blue 218 7.6 2.7 1200 170 3.9 74 3 43% 220 7.5 2.9 2780 68 3.9 66 3 Yellow 213 4.3 2.9 840 161 3.2 66 4 7% 221 2.7 2.1 840 114 2.5 179 3 Dark blue 217 48.3 10.4 1980 134 9.9 238 3 41% and 222 13.8 8.9 1500 229 3.9 203 5 purple 216 10.5 4.3 740 164 7.6 203 3 108 From Table 17 i t is evident that the following three moisture categories could readily be differentiated: (1) low moisture conditions < 4.5 % H;?0 (yellow) 7 % of total area (2) medium moisture conditions ro 7 % H20 (light blue) 43 % of total area (3) high moisture conditions > 10 % H2O (blue,purple)41 % of total area A partial differentiation between Carbon content and exchangeable IMa and K was also possible with the same density subdivision. The results of the remaining three images were somewhat less promising and are included in Appendix IV-1. The density classes viewed in this example are repre-sentative of vegetation cover, amount of s o i l exposed through the vegeta-tion cover, and drainage conditions. Despite the low absolute number of samples these results are encouraging. 2. Bare Soil Surface Frame IMo. 40 was chosen because of a minimum vegetation cover at the time of exposure and the presence of a distinct s o i l color pattern. Again drainage conditions are greatly responsible for the s o i l variation, which i s visible by the tonal pattern on the aerial image. The chemical conditions were again related to the tonal differences. To f a c i l i t a t e the analysis the ground reference site location i s indicated in Plate XIV/ on the following page. The closest relationship between tonal categories and chemical ground conditions was obtained from the color and color IR frames (Plates XU and XV/I, and Tables 18 and 19). More complex findings resulted from an analysis of the black and white 500-600 and 600-700 nm wavelength band, the results of which are given in Appendix IV/. The color categories l i s t e d in Tables 18 and 19 subdivide the sites into discrete groups of high, medium, and low organic matter and s o i l moisture content. A distinct grouping with regard to exchangeable Ca and 109 to a lesser degree of Mg and K is also possible with the same color groups. Plate XIV/. Exposed s o i l surface, indicating reference site location. The color s l i c i n g provided a slightly better separation between the different parameters than the color IR but in both cases a clear distinc-tion between s o i l moisture conditions and Carbon content was shown. As mentioned by Buckman and Brady (1969) cation exchange capacity is usually related to texture, type of clay, and organic matter content (Curtain and Smillie 1976). It i s therefore conceivable that the separa-tion of exchangeable Ca and Mg was only passible because the Carbon -exchangeable cation relationship exists, while the tonal variations are not directly influenced by soluble salts. This was partially confirmed by a correlation analysis of the ground samples in which percent Carbon and exchangeable Ca were significantly correlated (r=- 0.51, sig. at JL = 0.005). Lilhen looking at the individual wavelength bands i t i s apparent that tonal contrasts in a l l four bands (500-600, 600-700, 400-700, 500-900 nm) largely reflect organic matter content. The films covering larger wavelength bands (400-700 and 500-900 nm) did however produce better dis-criminations. 110 These f i n d i n g s should houever be considered merely e x p l o r a t o r y s i n c e the data a v a i l a b l e uere i n s u f f i c i e n t f o r a more ri g o r o u s s t a t i s t i c a l a n a l -y s i s . P l a t e XV. Density enhancement from c o l o r photography. Table 16 Density c l a s s e s and a s s o c i a t e d ground c o n d i t i o n s ( c o l o r frame) Color Category Sample IMo. % S o i l Moisture 0/ ZD 3 C ex. Ca ppm ex. Mg ppm ex. IMa ppm ex. H ppm ex. P ppm Green 200 15.7 5.3 1140 342 14 293 3 201 8.2 4.9 1060 366 17 105 4 L i g h t blus >202 6.2 2.7 660 272 18 90 5 & grey 203 3.0 2.5 840 132 7 74 3 204 7.3 1.2 380 118 17 62 3 206 5.5 1.5 700 78 8 144 3 209 5.6 1.1 380 204 21 51 4 Yel l o u j - 205 11.9 2.6 640 260 13 62 3 brouin Purple 208 14.5 8.3 4440 235 17 59 1 Black 207 54.9 40.9 2640 401 27 94 6 Plate XVI. Density enhancement from infrared color photography. Table 19 Density classes and associated ground conditions (IR-color frame) Color Category Sample No. % So i l Moistun 0/ A> 2 C ex. Ca ppm ex. Mg ppm ex. Na ppm ex. K ppm ex. P ppm Color Blue 200 15.7 5.3 1140 342 14 292 3 3/2 201 8.2 4.9 1060 366 17 105 4 4/2 2D2 6.2 2.7 660 277 18 89 5 5/2 205 11.9 2.6 640 260 13 62 3 4/2 Light blue 203 3.9 2.5 840 132 7 74 4 5/2 White 204 7.3 1.2 380 118 17 62 3 6/2 206 5.5 1.5 700 78 8 144 3 6/1 209 5.6 1.1 380 204 21 51 4 6/3 Pink 207 54.9 40.9 2640 400 27 93 6 2/1 208 14.5 8.3 4440 235 17 59 1 6/1 j 112 3. Discussion on Results and Procedure Although adequate results uere obtained a number of shortcomings inherent in this method should be discussed. (a) Technical problems (i) Photographic factors such as type D f emulsion, d i f f e r -ences in processing and exposure, variation in atmospheric conditions and sun angle, a l l influence density measurements. (Ii) The slicing procedure i s subject to distortion at the edges. This results in disturbances along the border of the images. ( i i i ) The same color combination mil l give different contrasts and color balance when used with different frames and films; this makes a comparative frame-by-frame analysis d i f f i c u l t . (iv)' The transfer of reference sample points onto the sliced image i s d i f f i c u l t , especially when a complex tonal pattern i s present. (v) The density sli c i n g system used has a sensitivity range which i s too refined, giving the operator too many density categories to which color combinations have to be assigned. This makes a repeated analysis tedious and complicated. (vi) The photographic separation of different wavelength bands i s limited by the f i l t e r and film properties; the resulting densities are not readily comparable. (b) Application problems A number of factors contribute to making up photographic tone and a l l of these have to be considered in such an analysis. The di f f i c u l t y of separating interactions has been pointed out by Evans et al (1976) 113 who found that the photo tone did not correlate well uith s o i l properties in the case of sa t e l l i t e photography. Lithologic patterns, particle size, chalk content, organic matter, s o i l moisture, a l l contribute to making up photographic tone, but no consistant relationship has yet been devel-oped. The density measurements on the color and color IR frames should be made at a specific wavelength by extracting different f i l t e r densities, •nly then can the color potential revealed on the film be utilized to i t s f u l l capacity. In the present example the slicing was merely a measure of transmission which was independent of wavelength. This resulted of course in a significant loss of information. This problem can however be solved by using a scanning microscope photometer or density scanner in which the measurements are made through several f i l t e r s . This latter method was successfully applied by Cihlar and Protz (1972) and Maurer (1974), but because the equipment was not available i t could not be pursued in this study. A l l the problems mentioned above make the routine comparative anal-ysis questionable. Nevertheless, by carefully controlling the photographic factors (Yost and Ldenderoth 1967, and wenderoth and Yost 1974) and by comparing tonal properties only within single images a quantification of gross tonal variations i s possible. This was successfully demonstrated for crop inventories by Learner et a l (1975) and Stoner et a l (1976) and for land use classification by Byrne and Munday (1972). (c) Summary of findings In view of the exploratory nature of this analysis conclusions should be drawn with caution. In this study a distinction between units of different s o i l moisture and organic matter content was possible by density s l i c i n g . Other chemical conditions such as exchangeable Ca, Mg, Na, and H content could partially be differentiated because they are associated with moisture and organic matter. Their distinction i s most 114 probably the result of direct correlation between Carbon and exchangeable cations rather than from the individual cations themselves. This i s especially true in the case of low ion concentrations. Ca, 'IMa, and Mg behave in the same manner as percent Carbon in Table 19 while averages of Ca, Mg, and IMa behave the same way in Table 18. Such a relationship is also found for the vegetated surface (Table 19) but the expression i s much weaker. This was confirmed by ranking tables of the color categor-ies (see Appendix IV-6). Although areas of different s o i l moisture and organic matter conditions could be identified with a l l four spectral bands the slicing was most successful from the color and color IR frame (400-700 nm and 5DQ-9D0 nm wavelength band). The color band proved to be more useful for the study area where a large amount of bare soils was exposed, while the color IR was mare useful for the vegetated study area, confirming prior experience. C. IMAGE ENHANCEMENT THROUGH ADDITIVE COLOR VIEWING 1. Introduction From the previous analysis i t i s evident that a considerably greater discrimination of surface conditions was possible on images covering a wider wavelength range and by using different colors for enhancement. Rather than analyzing single images an addition of a l l multispectral bands should therefore produce better results. The addition of different multispectral images can be accomplished in two ways, (a) by producing diaza-film, and (b) by optical projection through different color f i l t e r s . Poor results were obtained from the diazo-film method. This can be attributed to a reduced sensitivity of the diazo-film and the limited f l e x i b i l i t y in adjusting color combinations and brightness. As a result this method was not pursued beyond the stage of f i r s t t r i a l and was abandoned in favor of additive color projections. 115 Image enhancement by optical projections as described in Chapter One showed greater promise. A l l four individual film emulsions were super-imposed by projections with the I S system. A different color was assigned to each individual projection by exchangeable f i l t e r s , while the brightness was controlled by selectively changing the light intensity of the projec-tion lamps. Because overlapping spectral bands were used the enhancement technique could not be employed to i t s fullest advantage. Nevertheless, adequate enhancements were obtained for the same study areas described in the density slicing analysis. 2. V/eqetated Surface Because the four emulsions were obtained from two consecutive over-flights, registration was somewhat d i f f i c u l t . A slight shift in f l i g h t exposure did not allow a perfect superimposition of a l l four bands for this study area. A two-band superimposition (color and color IR) was possible and is reproduced in Plate XVII on the following page. The ground reference sites were once again transferred to the new image and color categories with their associated ground information are lis t e d in Table 20. Although adequate separation of s o i l moisture and organic matter was possible i t i s evident that such an image is not necessarily more useful than the color or color IR band alone. 3. Bare Soil Surface The study area with limited vegetation cover used for the density sl i c i n g analysis was used once again. A four-emulsion projection using a l l multispectral images was produced according to the specifications l i s t e d in Table 21 on page 117. Plate XVII. Color additive image (color and color IR band). Table 20 Color categories and associated ground conditions Color Category Sample Mo. % HrjD 0/ /o c ex. Ca ppm ex. Mg ppm ex. Na ppm ex. K ppm P ppm Very dark 217 48.3 10.4 1980 134 9.9 238 3 red Dark red 214 28.8 2.7 760 86 4.8 183 3 222 13.8 8.9 1500 229 3.9 203 5 215 6.4 3.6 1160 120 3.5 215 3 216 10.5 4.3 740 164 7.6 203 3 Light 218 7.6 2.7 1200 170 3.9 74 3 purple 219 4.3 2.9 840 161 3.2 66 4 220 7.5 2.9 2780 68 3.9 66 3 White- 221 2.7 2.1 840 114 2.5 179 3 purple 117 Table 21 Specifications used to create additive color image Film Band Projection F i l t e r Relative Illumination IR (50TJ-9DD nm wavelength) white 4.0 Black & white 500-600 nm green 6.5 Black & white 600-700 nm red 5.0 Color 400-700 nm blue 6.5 The resulting image and the associated ground conditions are given in Plate XV/III and Table 22.Dn the following page. Again equal density areas correspond with discrete organic matter and s o i l moisture values. In addition exchangeable Ca, IMa, and Mg values seem to be closely related to the s o i l moisture and organic matter con-ditions since equally distinct exchangeable cation categories were ob-tained. 4. Discussion of Results and Procedure By additive color viewing i t i s possible to produce a terrain image which contains information from a wider wavelength range than would have been possible from currently available aerial film (400-900 nm vs 500-900 nm wavelength range). In addition the color potential was easily utilized by altering both the color f i l t e r s in the projector and the light intensity. Image registration proved to be a serious problem. This i s to a great extent caused by the fact that the multispectral photography was obtained from two consecutive overflights. A slight shift in fli g h t exposure caused a distortion on the superimposed image, which resulted in a loss of resolution. P l a t e XVIII. Color a d d i t i v e image ( a l l bands) Table 22 Color c a t e g o r i e s and associated ground c o n d i t i o n s Color Sample % % ex. ex. ex. ex. Category No. H 20 C Ca Mg IMa K ppm ppm ppm ppm Brounish 200 15.8 5.3 1140 342 17 292 v i o l e t 201 8.2 4.9 1060 366 17 105 L i g h t v i o l e t 202 6.2 2.7 660 277 18 89 203 3.9 2.5 840 132 7 74 Very l i g h t 204 7.3 1.2 380 118 17 62 v i o l e t 206 5.5 1.5 700 78 a 144 209 5.6 1.1 380 204 21 51 Red orange 205 11.8 2.6 640 260 13 62 Orange 208 14.5 8.4 4440 235 17 59 207 54.9 40.9 2640 400 27 93 119 Rather than classifying the tonal variation in the step-like manner used in the slicing method a single color representation i s made for each band. By altering the colors for each image i t is passible to produce a picture in which the original tonal variations are s t i l l present but are now portrayed in different color combinations. The contrasts are there-fore more gradational than in the sliced images making a transfer of ref-erence points simpler. The results of the additive color enhancements are adequate but not necessarily superior to, far example, the color or color IR band. This i s illustrated in Plate XIX where the three images are contrasted. Ranked color categories (see Appendix IV/-6) show no great difference and confirm the pattern described. There are two reasons why the enhancement did not produce superior results: (1) loss of resolution due to registration d i f f i c u l t i e s ; and (2) superimposition of overlapping spectral bands. The former clearly caused a loss in resolution while the latter limited color contrasts. A B C Plate XIX. Image comparison, A = IR color, B = color enhancement; C = f u l l color. 120 D. CONCLUSIONS Both the density slicing and additive color viewing techniques are useful in the attempt to quantify chemical terrain conditions. However both techniques, as summarized in Table 23, have advantages and disadvan-tages. Table 23 Comparison of density sli c i n g and color additive techniques Method Advantages Disadvantages Density slicing The tonal and spatial element Only single images can of each equal density area be analyzed. can readily be quantified. By assigning color combina- A large number of photo-tions different density levels graphic and atmospheric can be contrasted for visual conditions affect the observations. density values only some of which can readily be controlled. Additive color A colored image containing A quantification of viewing information from a l l spectral color variation i s bands can be produced. subjective. A control aver color balance Registration of a l l and color contrast i s spectral images can readily possible. result in a loss of resolution. Choice of spectral bands i s c r i t i c a l . No evidence was found that a separation of chemical conditions i s more readily passible from selected narrow-band images (e.g. 500-600 nm and 600-700 nm wavelength). However the color band (400-700 nm) proved to be more useful for discriminating bare s o i l conditions while the color IR (500-900 nm) was superior for a similar analysis from a grass covered surface. An addition of a l l bands could prove to be mare useful but because of loss in resolution caused by image registration problems 121 and using images of overlapping wavelength ranges, no improved enhance-ment was obtained. Since a quantification of equal color areas i s not readily possible on the additive color image a density sli c i n g after addition could be more useful provided the resolution loss can be held to a minimum. A quantitative measurement of color rather than densities seems more effective but problems in controlling photographic factors as indicated by Thomson (1975) w i l l remain a serious handicap making universal appli-cation of this technique questionable. 122 CHAPTER Ul DIGITAL MULTISPECTRAL SENSING A. INTRODUCTION Direct spectral measurements of terrain surfaces and s o i l samples has become of increasing interest to scientists. This i s based Dn the understanding that every object emits and reflects electromagnetic energy in a characteristic manner. The more important factors i n f l u -encing overall spectral reflectance are: (1) the spectral distribution Df the solar energy at the time of analysis, (2) the spectral scattering of atmospheric particles, and (3) the physical properties of the object. The identification of properties and the analysis of their effect on the total spectral signature Df the object i s of great interest as i t enables the observer to quantitatively predict selected properties through remote means. Investigations of this nature have been undertaken by airborne analysis, in situ f i e l d measurements, and laboratory observa-tions. The laboratory method is the most promising since i t makes It possible to have a closer control over the otherwise dynamic illumination and atmospheric conditions. Unfortunately, site and vegetation conditions cannot readily be reproduced in the laboratory and in these cases a direct f i e l d or airborne approach i s more justif i a b l e . The most important multispectral analyses of soils and minerals are reviewed in Appendix IL From this literature review i t i s evident that a number of s o i l properties have been identified as influencing spectral signatures. Pettry et a l (1974), Myers (1975) and Reeves et al (1975) listed mineral content, particle size, s o i l texture, s o i l color, iron oxide, surface roughness, s o i l moisture, and organic matter content as important parameters. The aim of the Fraser Ualley mission was to examine the extent to which soluble chemicals can be identified and predicted in a quantitative 123 manner from d i g i t a l multispectral measurements. The method of analysis has been described in Chapter One and the data analysis and results are reported in this chapter. There are four parts to the data evaluation: (1) Airborne spectral analysis, (2) f i e l d measurements of samples, (3) laboratory observations of samples, and (H) a comparison of the results obtained by the three methods. B. AIRBDRIME SPECTRAL ANALYSIS The airborne data uere examined under the following three topics: (1) assessment of ground control and spectral values, (2) contrasting average per f i e l d data, (3) analysis of site specific data. 1. Assessment of Ground Control and Spectral Values With the help of a video film and 35 mm black and white photographic coverage i t was possible to reconstruct the exact fli g h t passage and assess the accuracy of the sampling scheme. Cut of 40 s o i l sample sites 29 could positively be identified on the films. Small diversion from the pre-deter-mined flight line was unavoidable and resulted in 11 ground sample sites not being totally covered. Thus the chemical data for these remaining samples were not used in the detailed analysis. (a) Identification of spectra vs corresponding site location The task of selecting the individual spectra which correspond to the sampled ground locations was accomplished as follows: In a l l cases the f i e l d borders were covered with vegetative matter either along ditches or in the form of hedges. This proved to be an excellent i n d i -cator for matching up individual spectra with their corresponding ground 124 Figure 40: Influence of Vegetation on Reflection (mean field values) —>— • • • » »— 450 SSO 650 750 850 950 Wavelength in nm 125 sample site since vegetation causes a distinct increase in reflection in the near IR wavelength range ( > 750 nm) (see Figure 40). In this way the f i r s t and last spectra of each f i e l d could easily be identified. The number of spectra for each f i e l d were counted and the aerial coverage represented on each spectrum was calculated. Once the spectra and ground sample positions were plotted on the same aerial photography i t was pos-sible to find corresponding pairs. (b) Assessment of spectral values As mentioned previously the spectral reflection values were produced as ratios of reflected vs incident light (R/I). To show that these spec-t r a l observations are not the result of systematic changes in the incident light conditions during the f l i g h t , spectral curves from target G were 456 550 650 750 850 250 Wavelength in nm Figure 41. Spectral reflection range of f i e l d G]_ to G^. 126 analyzed in greater detail. Target G consisted of four adjacent fields each having similar physical and chemical conditions, and parent material of common origin. The spectral range from f i e l d Gj_ to G^ platted in Figure 41 (page 125) shows that a l l four contiguous fields-have over-lapping spectra with no visible evidence that the spectral changes are a result of systematic changes in illumination conditions during the f l i g h t . 2. Contrasting Average per Field Ualues To avoid i n i t i a l l y the problem of selecting spectral curves with their corresponding ground sample site, average f i e l d reflection values were computed. As shown in Figure 42 unique spectral reflection curves were obtained for each f i e l d . T a r g e t G, which consists of four contig-uous fields with similar parent material and similar chemical conditions, could not be separated on the basis of spectral values and was for this analysis treated as one target. To further emphasize the distinctiveness of the 400-1000 nm wave-length spectra, the spectral range for each f i e l d was plotted in Figure 43. From these charts i t i s evident that the spectral reflection values for each f i e l d were expressive of their distinct'characteristics since almost no overlap exists between the eight targets considered. (a) Visual comparison with chemical data To show the range of chemical conditions present between the d i f f -erent targets, average f i e l d values were compiled in Table 24. It i s evident that a clear distinction in chemical conditions exists for a l l fields except targets A and B which both developed on outwash parent material. 127 450 550 650 50 50 50 ' 0 5 0 WAVELENGTH In nm Figure h2. . Mean target reflection curves. 128 — . 1 1 • . 1 • 1 1 1 < > 450 550 650 750 850 950 Wavelength in nm ' I I > I I u ' l t l I ' « 450 550 650 750 850 950 Wavelength in nm Figure 43. Spectral reflection curves for soils from different parent materials. Table 24 Average chemical conditions of fields Average Chemical Condition of Fields Grain Size Target # of # of ex. ex. ex. ex. Fe Fe Fe airborne s o i l % H20 % C Ca Mg Na K Dith. Pyr. Ox. % sand % s i l t % clay spectra samples ppm ppm ppm ppm % % % A 26 3 5.9 1.2 80 3 4 32 1.5 0.2 0.7 70 21 9 B 46 6 6.0 1.8 91 5 3 51 not measui -ed 74 21 5 C IDS n o s a m p l e s c o l 1 e c t e d D 16 6 8.5 6.1 980 101 58 292 1.6 0.5 0.6 25 52 23 E 20 1 30.4 46.3 1470 420 90 305 1.1 0.8 0.9 0 66 34 F 20 3 25.0 32.3 1870 676 39 277 1.1 0.5 0.6 0 61 39 G 131 14 4.5 2.6 1027 81 18 63 1.2 0.3 0.9 8 62 30 H 23 6 6.3 6.5 753 220 31 107 0.7 0.4 0.5 14 60 26 130 (b) Correlations between average values Correlations between mean spectral and mean chemical values were determined for a l l parameters at 12 different wavelengths. The results of this can be seen in Figure 44 in which a number of chemical param-eters show significant correlation with spectral reflection at different wavelengths. For example exchangeable Na and % C correlate best with the spectral data at 450-550 nm wavelength. Best values for exchangeable Mg are at 600-750 nm, exchangeable Ca at 750-1000 nm and particle size at 900-1000 nm. It should be stressed that this analysis i s only considered as a f i r s t approximation since the average chemical values per f i e l d were compiled from very small sample populations (between 3 and 6 samples per f i e l d ) . 3. Analysis of Site Specific Data (a) Correlations The ground sample data for which corresponding airborne spectra could be identified were used for correlation and regression analyses. Again correlation coefficients between spectral and chemical data were determined for 12 wavelengths at 50 nm intervals covering the 450-1000 -nm range. The results of this analysis are provided in Figure 45 and a comparison between these and the results obtained from the average .data analysis revealed a similar pattern. Iron (Pyrophosphate extraction) which was not used in the average data analysis proved to be the only new correlated parameter. Particle size (%), organic matter (% C), and iron (% in the form of organic complexes) correlated most strongly with spectral reflection measurements. These results are in agreement with laboratory and a i r -borne observations made by Bowers and Hanks (1965), Baumgartner et a l WAVELENGTH nm 450 550 650 750 850 950 1050 Figure kk. Correlations of average target values. 132 Figure 45. Correlations amongst parametersr.using site specific data. 133 (1970), and Kristof and Zachary (1971). The large number Df additional correlations i s somewhat suspect since they could be the result of auto-correlations and correlations amongst variables. The latter has been referred to in the literature by Buckman and Brady (1969) and Curtain and Smillie (1976). They noted that organic matter and texture are usually highly correlated with CEC over a wide range of parent material. Some of the- correlations amongst variables were confirmed when analyzing the correlation structure (see Figures 46 and 47). To eleminate some of these problems site specific information from targets G and H were analyzed separately. These targets have similar chemical and physical surface conditions, and since they originate from the same s o i l parent material a much smaller chemical variation i s ex-pected. This would allow a more rigorous analysis and accordingly i n -formation from 15 sites within these targets was used. In contrast to the previous situation only Mg and Na, and Ca and K are correlated amongst each other. Ldhen comparing the spectral data with the chemical information i t i s evident that only Mg and Na are significantly correlated (see Figure 48). Besides being more significantly correlated Mg shows a similar peak value (at 750 nm) as in the previous analysis. It can therefore be suggested that in the present example exchangeable Mg influences spectral reflection. Figure 46. Significant correlations amongst site specific parameters ( U = 0.005) Figure 47. Significant correlations amongst parameters using average target data ( at = 0.D05) 13H WAVELENGTH nm 4S0 550 650 750 S50 950 % sand us reflect. % Carbon vs reflection - l - i - ex. K " " ex. Na 11 " *++ ex. Ca " " +••.•+"• ex. Mg " " Figure 48. Spectral and chemical correlations of targets G and H. 135 (b) Predictions Using significantly correlated parameters, regression lines were calculated to predict chemical concentrations from the spectral data for a l l sample sites. Adequate predictions uere obtained for: % Sand from reflection values at 900 nm wavelength % Carbon " " " 6QD nm " % Iron " " '• 700 nm " Exchangeable Mg 11 " " 750 nm " Results from the regression analysis are included in Appendix V/-1. Ldhen using the spectral data from targets G and H alone exchangeable Mg can be predicted as follows: Y = 592 - 32D6X where Y = Mg concentration in ppm, and X = % spectral reflection at 650 nm. This regression line was f i t t e d to the data in Figure 49, and observed and predicted values are reproduced in Table 25 on the following page. 400 i l _ , , , , , S 10 15 20 25 Reflection in % at 650nm Figure 49. Regression of exchangeable Mg with reflection values at 650 nm wavelength. 136 Table 25 Results Df prediction o f exchangeable Mg from spectral reflection values at 650 nm wavelength Observed Predicted Residual (ppm) (ppm) 164 233 -69 267 217 70 277 279 - 2 126 107 19 76 61 15 85 100 -15 Sk 76 -12 65 60 5 78 100 -22 66 68 - 2 103 91 11 185 123 62 78 68 10 69 99 30 55 k5 10 The standard error is substantial (36 ppm) but can be considered adequate for prediction in the present example. Also upon testing no significant difference occurred between the predicted and the observed distribution. The problem with this latter screening procedure is that (1) the sample observations were once again reduced to a questionable level, and (2) the effect of Carbon on the spectral reflection was minimized. This latter factor i s of considerable importance. «+. Summary of Airborne Spectral Analysis The airborne results have a number of seriaus limitations which should be noted: (1) Relatively few sites (29) were used in this study, 137 (2) no attempt could be made to confirm the results by successive missions, and (3) parameters other than those measured might be responsible for the spectral variations. Nevertheless the results can be summarized as follows: (1) Characteristic spectral curves were obtained for each s o i l surface of different geomorphological origins. The d i f f e r -ences were such that the within variances were considerably smaller than the between variance and a separation according to curve shape is readily possible from the a i r . (2) The spectral reflection curves are influenced by chemical parameters. The most important ones seem to be % Carbon, grain size (% sand) and iron (% in the form of Fe-organic complexes). Several exchangeable cations seem also to be related to spectral conditions, the effect of individual cations could however not be separated from Carbon content since a correlation between Carbon and CEC seems to be present in a l l cases. (3) A quantitative prediction of same of these parameters was possible, but i t should be noted that they are only applicable to the study area and were not tested elsewhere. (4) The relationship amongst parameters and the proportional influence of each factor on the spectral variation could not properly be assessed from the airborne data.. Possible obliteration of spectral changes by other parameters and intercorrelations amongst variables make such an analysis complex. 138 C. GROUND SPECTRAL MEASUREMENTS The method described in Chapter One uas used for the outdoor spectral analysis of the samples. The aims of these measurements uere to determine the accuracy of the instrument, to assess the effect of selected parameter variation on the spectral signature, and to compare the airborne values uith those obtained on the ground. 1. Instrument Assessment The instrument arrangement used for the airborne mission uas utilized for the sample measurements. The reproducibility of the spectral signatures uas assessed by spli t t i n g several samples and measuring each fraction at up to one hour time intervals. This time interval corresponds approximately uith that during uhich the airborne measurements uere made. Figures 50a and 50b shou that the results are adequate but not excellent. In the uorst case an error of up to 7% of the total reflection uas en-countered. The overall shape of the spectral curve uas houever reasonably consistant. The reason for the error in measurements can be attributed tb (a) the instrument accuracy, (b) variations in incident light conditions, and (c) possible variation in surface of samples. Although each sample uas spread out evenly slight surface variation could have resulted in small spectral alterations. 2. The Effect of Parameter Alteration on Spectral Reflectance Myers and Allan (1968) and Pettry et a l (1974) revieued some of the more important factors uhich affect spectral reflection. Particle size, surface roughness, s o i l moisture, organic matter content, s o i l color, and iron oxide uere found to be the most important parameters. Some of these uere i n i t i a l l y assessed uith the f i e l d spectrometer. (a) So i l moisture The spectral response resulting from s o i l moisture changes has been discussed by Bouers and Hanks (1965), Shields et a l (1968), Hoffer and 139 50 S50 650 750 850 950 1050 Wavelength in nm Figure 50a. Reproducibility of measurements. I , , , , T 1 • ' i 1 ' 450 550 650 750 850 950 1050 Wavelength in nm Figure 50b. Reproducibility of measurements. 140 Johannsen (1969), Condit (1970), Planet (1970), and Pettry et a l (1974). They noted that the overall spectral reflectance decreases with an i n -crease in s o i l moisture, and that the best spectral contrasts are ob-tained at low moisture levels. In addition, Hoffer and Johannsen (1969) found that when altering s o i l moisture content the shape of the spectral curve i s only maintained in clayey soils but not in sandy s o i l s . Pettry et a l (1974) pointed out that the moisture-spectral relationship of a single s o i l i s curvilinear in that the reflection increases more rapidly at low moisture levels than at higher levels. To eliminate this variable in the present analysis a l l s o i l samples were dried in the laboratory prior to the spectral measurements. (b) Organic matter content The.spectral response resulting from organic matter alterations has been examined by Bowers and Hanks (1965), Shields et al (1968) and Mathews et al (1973b). They found that a removal of Carbon by oxidation resulted in an increase in the overall reflection of the samples. These latter results should however be accepted with reservation since the oxidation procedure used to remove the Carbon also alters other chemical conditions, this has been demonstrated by Lavkulich and Ldiens (1970) who found that a number of s o i l constituents were affected by the oxidation procedure. These in turn could also contribute to some of the spectral variation. Pettry et a l (1974) illustrated the organic matter effect on soils from a single s o i l series. They noted that at low organic matter content, small organic matter variations caused a rapid change in spectral reflec-tion while at high levels only small spectral changes were observed on soils with different organic matter content. The data from the present study showed a similar relationship between organic matter and reflection although soils from markedly different s o i l series were used. The relationship i s illustrated in . Figure 51 on the following page. 141 40 30 § 20 n TO o 10 H X x V x 10 20 30 % Reflectance at 900nm 40 5G Figure 51. Relationship between % Carbon and spectral reflectance. From Figure 51 i t is evident that a linear relationship is only useful at low Carbon content. (c) Particle size effect The effect of particle size on spectral reflectance has been investigated by Bowers and Hanks (1965) and Shockley et a l (1962). To minimize this variable in the present study the ) 2 mm fraction of each sample was removed by sieving. The effect of different particle size was examined on the basis of four samples in which both the < 2 mm fraction and the <C 0.15 mm fraction were measured with the spectrometer. A com-parison between the two fraction spectra, 'revealed that an overall increase in spectral reflection was obtained by removing the coarser fractions (Figure 52). The increase was evident in a l l four test samples and was proportionally higher at longer wavelength ranges. <0.15mm fraction 450 550 650 750 350 950 Wavelength in nm Figure 52a. Effect of particle size on spectral reflection. 40 n 30 o 20 1 n o 0> «j cc >p 10 H < 0.15mm fraction / • 429, <2mm fraction / <015mm fraction ^ - 436 <2mm fraction 450 850 950 550 650 750 Wavelength in nm Figure 52b. Effect of particle size on spectral reflection. 143 (d) Extraction of exchangeable cations Myers and Allen (1968), uJiegand et a l (1975), and Richardson et a l (1976) a l l noted that s o i l salinity m i l l affect spectral signatures. The exchangeable cations effect on spectral measurements uere investigated on three samples by measuring the original and the neutral ammonium acetate extracted sample (see Figure 53 and Table 26). It uas found that pos-it i v e l y higher spectral values uere obtained from the extracted samples. Table 26 Chemical conditions of samples used in Figure 53 Sample IMo. C % Ca (ppm) Mg (ppm) IMa (ppm) K (ppm) Fe D Fe p % Fe 0 % 405 5.5 740 277 77 92 9 4 5 427 1.0 85 3 4 25 17 2 8 411 1.5 930 65 16 52 7 2 a 40 i 30 c (0 450 550 650 750 S50 950 1050 Wavelength in nm Figure 53. Reflectance before and after extraction. 144 The results in Figure 53 are not conclusive since soluble organic substances uere also removed by the extraction procedure. The amount of organic matter removed cannot be readily analyzed and merits a more ex-tensive investigation. Nevertheless, judging from the light yellou color of the extract i t i s evident that dissolved organic matter could also be responsible for the spectral alterations. A similar investigation uas made in the laboratory and i s discussed later in this chapter. (e) Parent material Condit (1970) classified soils according to the shape of the spectral reflectance curves. This uas also pursued in the present study where the spectral signatures of a l l samples were compared. It became evident that samples originating from similar parent material produce characteristic spectral groups (see Figure 54 a and b). The within group range was such that a distinction on the basis of parent material type was readily pos-sible. 3. Comparison between Chemical and Spectral Data The potential of predicting chemical conditions from spectral measurements was once again investigated. (a) Correlations A correlation analysis was made between the spectral values at 13 wavelengths and the chemical values of the samples. The results of this procedure are given in Figure 55. Significant correlations were obtained for five parameters. Exchangeable Mg and % Carbon were best correlated with spectral reflectance values at 900-1000 nm wavelength. The other relationships were: exchangeable K at 550-650 nm, exchangeable Ca at 1000-1050 nm, and Fe at 700-800 nm. The significant relationships amongst variables i s best illustrated by the correlation structure in FigureA56i,n It should be noted that a number of significant correlations were present amongst the chemical parameters and these in turn influence 145 40i 3<H 20-1 o> o c (0 <J 1 10 Outwash 'Field A B Fluvial/ Marine. " F i e l d D Organic Field E F 4S0 550 650 750 850 Wavelength in nm Figure 54a. Parent mater i a l e f f e c t (range of values). 950 1050 40 30 8 c n u o 20 10 Deltaic ( low Na.Mg.C) Field G _ • -mj —- ' _ ''"beTtaic (higher Na,Mg,C) Field H 450 550 650 • 750 Wavelength in nm 850 950 1050 Figure 54b. Parent mater i a l e f f e c t . 450 Wavelength in nm 550 CEO 750 850 950 050 Na vs Spectral Reflectance * * C a vs •+• •+- Mg vs i—i—i K vs C vs Fe vs s i g n , l e v e l <^=0.005 V - V -Figure 55. Significant correlations between chemicals and spectral reflectance. -Figure 56. Significant correlation amongst chemical parameters ( 06 = 0.D05). 147 the overall spectral relationships. To eliminate some of these inter-relationships a selected sample treatment could be attempted. In view of the relatively small sample papulation (n = 29) such a procedure was not pursued; nevertheless the importance of Carbon in this analysis i s evident. (b) Regressions Table 27 below l i s t s the regression equations which can be used for predicting chemical parameters from spectral measurements. Table 27 Regression equations for selected chemical parameters Elements Wavelength Regression Equation Standard Error Df Y % Fe % C ex. Mg (ppm) ex. H (ppm) 7 DO nm 900 nm 1000 nm 600 nm Y = 0.73 - 0.015 X Y = 59.0 - 1.85 X Y = 1075 - 31.3 X Y = 373 - 13.5 X 0.122 5.98 117.40 63.40 4. Summary of Ground Measurements A number of parameters which influence spectral reflection have been investigated. S o i l moisture and particle size effect were minimized by drying and sieving procedures in the laboratory. Organic matter content seems to be the most significant factor affecting spectral reflectance in samples with low Carbon content. The extraction of exchangeable cations also affects the spectral values. However, at this stage i t i s not clear whether the removal of cations i s totally responsible for such changes or whether soluble organic matter which i s also removed by the extraction contributes to such changes. Soils could be differentiated according to parent material by an analysis of the spectral curves. Characteristic curve shapes were 148 obtained for each s o i l group. Finally, correlations and regression analysis revealed that a prediction of Carbon, exchangeable Mg and Fe is passible but that the standard errors in such an investigation remain relatively high. D. LABORATORY MEASUREMENTS Two independent sets of samples uere used for the spectral reflec-tance measurements in the laboratory. The f i r s t set consisting of 20 samples was analyzed at the U.S. Geological Survey laboratory in Denver, Colorado, under the direction Df Dr. H. Hunt. The second set of 38 samples was analyzed at the U.S. Department of Agriculture laboratory in Ueslaco, Texas, under the supervision of Dr. C. Wiegand. To test the compatability of the two types of measurements three duplicate samples were analyzed in both laboratories. The results of this comparison are given in Figure 57. It i s apparent that the absolute values as well as the overall spectral curves are different. This can be attributed to differences in instrument response, differences in sample preparations, and differences in reference standards used (BaSQ^ vs MgO). The proportional differences seem to be consistant but since only three samples were tested i t was not passible to determine a cor-rection function. Instead the two sets of samples were analyzed separ-ately and the end results were then compared. 1. Extraction of Exchangeable Cations The spectral reflection before and after neutral ammonium acetate extraction was measured for 5 samples (2 USGS, 3 USDA). As can be seen from Figure 58 an increase in spectral reflectance after extraction was observed in four out of five samples. The increase i s proportionally greater at longer wavelength ranges and for samples with higher cation content removed. Unfortunately the results are not conclusive since a negative spectral response was obtained from one sample and Carbon content 60 • — . After Extraction 50 A 40 30 J 411 /405 Before Extraction o | 20 10 / // Nr. c% Ca Mg Na K Fe % 405 5.5 740 277 77 92 .4 411 1.5 930 65 16 52 .8 -» 1 1 1 > I 350 450 550 650 750 850 950 1050 T 1 1 1 1 I I I « 1500 2000 Wavelengh in nm Figure 58a. Spectral response before and after extraction (measurements at • USGS laboratory). WAVELENGH in nm Figure 58b. Spectral response before and after extraction (measurements at USDA laboratory). 152 uas also higher for the samples uith the most apparent spectral increase. The spectral changes can therefore not be attributed totally to cation variations since soluble Carbon i s also removed by the extraction proce-dure. 2. Parent Material Effect When the spectral curves uere platted far a l l measured laboratory samples distinct spectral groups uere once again obtained on the basis of s o i l parent material (see Figures 59a and b). The distinction betueen each spectral group is such that characteristic curves can easily be recognized in most cases. 3. Comparison betueen Spectral and Chemical Data (a) Correlations In accordance uith the previous studies a correlation analysis of a l l chemicals uas performed uith a l l spectral measurements. The correla-tion coefficients for the tuo sets of data i s given in Figure 60. In both cases exchangeable Mg uas significantly correlated uith spectral values at 750-1000 nm uavelength. Other significant correlations uere found for exchangeable hi at 500-600 nm and % Fe at 600-800 nm. Percent Carbon uas significantly correlated uith reflection values but the uave-length at uhich the best correlations uere obtained differed betueen the tuD laboratory studies. Best results for the USGS studies uere obtained at 800-1000 nm uhile the USDA samples uere most useful at 550-650 nm. Aside from this discrepancy the tuo sets of data produced similar corre-lation results. Significant correlations amongst variables uere present in both data sets, in the case of the USDA measurements significant correlations uere limited to Ca-Carbon-Mg. 20 -c SOIL PARENT MATERIAL CZI ORGANIC (muck) I 1 Sample(s) ESJ ORGANIC (p»or ) 1 CZJ OUTWASH 3 CZJ DELTAIC 8 C D LACUSTRO TILL 9 CT] FLUVIAL /MARINE 4 CD LACUSTRO TILL (depreuional) 3 500 lfiOC 1500 2004 2500 WAVELENGTH in nm o UJ 40 -20 • 10 A S«> 1000 1500 20C0 2500 nm U l WAVELENGTH in nm ^ Figure 59b. Parent material effect (USDA measurements) -0.1--0.2 H -0.3--0.4--0.5 0.6 400 600 800 1000 1S00 Wavelength in nm 2000 2500 USDA MEASUREMENTS + it- f " / . — t — 1 — ( — I — t — 1000 1500 2000 Wavelencith in nm LEGEND: Spectral Reflection Fe vs. » •+-+-+- Mg vs. if »* •H- Ca vs. •i >* Na vs. *t ji — i — i - K vs. n »» U-.005) Figure 60. Correlation between chemicals and laboratory spectral measurements. Ul 156 (b) Regressions Chemical parameter predictions uere once again attempted from the spectral values. The regression equations uhich produced the best results are list e d in Table 28 belou. Table 28 Comparison of regression equations from different laboratory measurements USGS Measurements Elements Wavelength Regression equation Standard error of Y n % Fe 700 nm Y = 0.68 - 0.0105 X 0.064 % C 850 nm Y = 50 - 1.21 X 5.51 15 ex. Mg 850 nm Y = 889 - 20.5 X 124.70 ex. H 550 nm Y = 367 - 11.8 X 69.70 USDA Measurements Elements wavelength Regression equation Standard error of Y n % Fe 800 nm Y = 0.75 - 0.015 X 0.12 % C 600 nm Y = 24.2 - 0.93 X 6.4 29 ex. Mg 750 nm Y = 617 - 18.1 X 126.0 ex. H 600 nm Y = 313 - 10.4 X 96.2 Again i t should be noted that the standard errors are quite large making the predictive value only partially satisfactory. 4. Summary of Laboratory Measurements Despite the different measuring procedures the tuo sets of samples produced remarkably similar results. Characteristic groups of spectral 157 curves uere obtained in both cases for soils originating from similar parent material. There i s however a considerable difference in spectral curve shape and absolute spectral values. This i s attributed to d i f f e r -ences in instrument response and measuring procedure since on the basis of three duplicate samples i t i s evident that a similar proportional difference was obtained between the three samples. For both sample sets significant correlations between the spectral values and % Fe, % C, exchangeable Mg, and exchangeable K were obtained. Regression lines were similar for both sample.sets producing somewhat better predictions in the case of % Fe and exchangeable K from the USGS measurements. The wavelength at which best prediction results were ob-tained differed somewhat between the two methods (Fe 7D0 vs 800 nm, % Carbon 850 vs GOO nm, exchangeable Mg 850 vs 75D nm, exchangeable H 550 vs 600 nm). Predictions of Fe, C, Mg, and K from spectral values were partially successful, but standard errors remained substantial. There i s strong evidence that cation concentration affects spectral values but measurements before and after extraction did not produce ab-solutely conclusive evidence. Significant correlations between Carbon and exchangeable Mg were found in both cases and this partially influences the overall relationship. E. COMPARISON AMONGST AIRBORNE, GROUND AND LABORATORY SPECTRAL MEASUREMENTS A number of factors differed for the three types of measurements: reflection standards, atmospheric conditions, number of measurements, surface area and instrument response. In addition, s o i l moisture and particle size variability were minimized for the ground and laboratory . measurements. As a result a direct comparison i s somewhat d i f f i c u l t . Nevertheless a relative comparison of spectral curves, correlations and predictions i s made in this section.. 158 1. Spectral Curves When comparing the three types of measurements for corresponding site-samples the fallowing observations can be made from Figure 61: (1) The laboratory measurements of the samples produced generally higher reflectance values than their counterparts on the ground. This uas more pronounced in the case of the USGS measurements. The airborne observations in turn produced the louest overall values. For soils uith high organic matter content, however, the reflectance increase is not pronounced. (2) The spectral increase is proportionally larger at longer wavelengths in the case of the USGS measurements. (3) A distinct atmospheric absorption band at 950 nm i s only observed on the airborne measurements. (4) Characteristic spectral curves were obtained for soils from different parent materials. Such a distinction i s readily possible from a l l three measurements. (5) Except for the samples with high Carbon content a greater spectral discrimination is possible from the laboratory measurements. 2. Correlations Because of the differences in sample observations and sample treat-ments the correlation trends rather than the absolute correlations were compared. From Figure 62 i t is apparent that similar correlations were ob-served for the laboratory and ground measurements. The airborne corre-lations between spectral reflectance and Fe, C, IMa, and H tend to decrease drastically in the longer wavelength ranges. This can in part 159 Figure 61b. 160 450 550 6 5 0 750 850 950 1050 Figure 61c. Comparison betueen different types of measurements. 404 —» i 1 1 1 1 1 1 1 ! ' » • • 450 550 650 750 850 950 1050 WAVELENGTH in nm Figure 61d. 161 c o a o o 400 wavelength in nm 600 gOO 1000 400 wavelength in nm 600 SCO 1000 Fe vs. Spectral Reflectance C vs. Spectra! Reflectance wavelength in nm 600 800 400 wavelength in nm 600 800 1000 ex. Ca vs. Spectral Reflectance ex. Mg vs. Spectral 'Reflectance c t> i) o o c o aj w o o wavelength in nm 400 600 800 1000 wavelength in nm 600 8 0 0 1000 o o o u ex. Na vs. Spectral Reflectance ex. K vs. Spectral Reflectance Legend: Airborne Measurements Laboratory Measurements USDA i-*—•-• Ground Measurements Laboratory Measurements USGS Figure 62. Correlation trends-(chemical us. spectral values). 162 be attributed to the atmospheric absorption band which i s most manifest at 950 nm wavelength. 3. Regressions The regression lines used to predict the % Fe, % C, exchangeable Mg, and exchangeable K were compared in Table 29. The regression lines from the laboratory and ground measurements' are very closely related. The airborne relationships are similar but have a slightly steeper angle. This can be attributed to the reduced overall spectral values obtained from the a i r . Straight linear relationships were used for the predictions although there i s evidence, especially in the case of Carbon, that a curvilinear relationship i s more applicable at higher concentrations. The predictions are therefore only applicable for restricted s a i l conditions and the same equation should not be applied for predicting extreme chemi-cal conditions. The standard error for % C and % Fe are slightly reduced in the case of the ground and laboratory observations (see Figure 63) but vary only slightly for exchangeable Mg and n. From this i t can be con-cluded that the measured relationships from the three methods are similar and that the discrimination was improved from the airborne to the labora-tory measurements. F. SUMMARY 1. Predictions: Multispectral airborne, ground, and laboratory measure-ments were found to be useful in predicting chemical parameters such as % C, % Fe, exchangeable Mg, and exchangeable K. 2. Relative importance of individual_parameters: The cause and effect of individual parameters on the spectral reflection values could not be isolated in a satisfactory manner. There is strong evidence that % Carbon is the most important parameter. Carbon was however correlated with several cation concentrations and since no selective, non-destructive extraction procedure was available i t was impossible 163 Table 29 Comparison of regression lines betueen different measurements Element Type of Measurements Wavelength Regression Equation Standard Error n % Fe Airborne 700 nm Y = 0.80 - 0.032 X 0.165 % 29 Ground 700 nm Y = 0.73 - 0.015 X 0.122 % 29 Laboratory (USGS) 700 nm Y = 0.68 - 0.0105X 0.064 %• 15 Laboratory (USDA) SOO nm Y = 0.75 - 0.015 X 0.12 % 29 % Carbon Airborne 600 nm Y = 39.0 - 2.41 X 9.30 % 29 Ground 900 nm Y = 59.0 - 1.85 X 5.98 % 29 Laboratory (USGS) 850 nm Y = 50.0 - 1.21 X 5.51 % 15 Laboratory (USDA) 600 nm Y = 24.2 - 0.93 X 6.40 % 29 ex. Mg Airborne 750 nm Y =937.7 -449.6 X 127.1 ppm 29 ppm Ground 1000 nm Y = 107.5 - 31.6 X 117,.4 ppm 29 Laboratory (USGS) 850 nm Y = 889 - 20.5 X 124.7 ppm 15 Laboratory (USDA) 750 nm Y = 617 - 18.1 X 126.0 ppm 29 ex. K Airborne 550 nm Y = 530 _ 33.3 X 76.2 ppm 29 ppm Ground 600 nm Y = 373 - 13.5 X 64.8 ppm 29 Laboratory (USGS) 550 nm Y = 367 - 11.8 X 69.7 ppm 15 Laboratory (USDA) 600 nm Y = 313 10.4 X 96.2 ppm 29 164 Standard Error for Fe Standard Error for C S E % .25 .20 .15 .10 .05 S E % 5 SI 3 3 o o o ea L_ o S3 to >> o o J3 10 8 6 4-01 o 3 O o 3 3 o o X) (0 < Q I CO Z> o "S k_ o Xi 3 Standard Error for Mg SE ppm Standard Error for K SE ppm 125 100-801 75 60-40-25-^ 0) c o 73 C 3 O o CO o CO I ZD Ss o ro b. O X! (0 < Q CO O 20 c o c 3 O o CO o CO ZD >. o CO o xi < Q CO ZD >. t— O ra o J3 Figure 63. Standard errors. 165 to demonstrate the spectral response of soluble Carbon or cation removal in isolation. The extraction which removes both types of parameter generally produced an increase in spectral reflectance, but from the data i t was not clear whether the response resulted from both parameters or merely from soluble Carbon extraction. Variations in s o i l moisture and particle size present in the a i r -borne observations did not seem tD obstruct the chemical predictions since the relationships were similar to the laboratory and ground measurements where the effect of these parameters was limited. 3. Spectral curves: There was no selective wavelength which was most useful for the analysis. Rather an analysis D f the total spectral curve was most productive. Soils from similar parent material pro-duced characteristic spectral curves which were clearly identifiable from a l l three measurements. The 500-1100 nm wavelength range was found to be most useful for the present analysis. k. Type of measurements: Similar relationships were obtained from a l l three types of measurements. The airborne analysis produced the lowest overall values while the spectral discrimination was highest in the laboratory. Laboratory measurements Df soils with low organic matter content ( < 7%) revealed considerably higher spectral values than airborne measurement. In the case of organic soils such an increase was very small however. Although the absolute values were different between the ground and the laboratory measurements the overall relationships were closely related. Finally, an atmospheric absorption band slightly reduced the usefulness of the airborne ob-servations in the 950 nm wavelength range. 166 CONCLUSION L i t t l e research emphasis has been placed on the study of chemical landscape va r i a b i l i t y . This can be attributed to: the large number of factors uhich control the chemical distribution, the dynamic nature and complexity in uhich these factors interact, and the inherently large number of samples required to establish a proper data base. In the present study the variability of soluble chemicals over tuo Quaternary landscapes uas examined and a method uas developed in uhich fundamental problems such as scale of investigation, definition of spatial units and choice of parameters by uhich these units can be quantified uere the main considerations. A hierarchy of landform units uas used to allou the ready transfer of information from one scale to the next. A genetic geomorphological framework uas chosen because of the ease of recognition and i t s relationship to parent material and genesis. Numeri-cal techniques were applied to select and analyze parameters and to allow a rapid multiparameter treatment. Finally remote sensing techniques were examined to determine whether the chemical data base for such studies can be generated in a more efficient way. The conclusions of this research can best be summarized under the following headings: A. The suitability of hierarchical geomorphological units in studying chemical terrain variability; B. The use of numerical grouping techniques in processing multi-parameter data; C. Prediction of chemical ground conditions by multispectral remote sensing methods. 167 A. THE SUITABILITY DF HIERARCHICAL GEOMORPHOLOGICAL UNITS IN STUDYING CHEMICAL TERRAIN VARIABILITY An ascending hierarchy of genetic gee-morphological units ranging from "site" to "site type", "landform units", and "landform unit types" uas found to be useful in studying the variability of soluble chemicals over the landscape. The units uere selected on the basis of form, parent material, particle size, and inferred genesis, and their internal chemical variability uas tested on the basis of a pilot study in a marine and outuash sedimentary environment uhich led to the following conclusions: (1) The within unit variability proved to be smallest at the site level (up to 65% CV) and increased to over 200% CV for a number of parameters at the highest hierarchical level. (2) The best chemical differentiation was possible at the "landform unit type" level where significant differences in chemical conditions were found in a l l cases. At the lower "landform unit" level such a distinction was not possible in the case of units of similar origin. This suggests that the type of genesis and parent material even in this Quaternary environ-ment is more important than local environmental conditions and land use. (3) The importance of genetic geomorphological units was also evident from the observation that in a l l units the variability of the C horizon was marginally smaller than their A and B horizon counterparts. (4) Of the twelve chemical parameters analyzed (av. Ca, Mg, K, Na, Zn, S i , CI, Pb, Cu, PO^, SO^, CI) two (SO^ and Pb) were eliminated from the analysis because of their low overall values, while significance tests showed that Ca, Mg, Na, K and, to a lesser degree, Si and PO^ were the most useful parameters by which the units could be differentiated. 168 (5) In no case uas i t possible to differentiate units on the basis of a single chemical parameter and this provided the incentive for a multiparameter numerical analysis. B. THE USE DF NUMERICAL GROUPING TECHNIQUES IN PROCESSING MULTIPARAMETER DATA To speedily classify terrain units on the basis of multiparameter chemical data an average distance cluster program uas used in uhich a large number of sites uere grouped together according to similarity in chemical conditions in a l l horizons. When comparing the numerically derived groups uith the hierarchical units the follouing conclusions uere draun: (1) The groups of sites uhich uere derived numerically from the chemical data alone corresponded very closely to "landform unit types" uhich uere selected on the basisjof form, parent material, particle size, and inferred genesis. This suggests that a close relationship exists betueen type of geomorpholog-i c a l units and chemical conditions, and that sites can be used to describe landform unit types. (2) The same relationship could not be confirmed at the louer hierarchical level uhere only very feu site groups uere representative of individual landform units. This uas the case in both Quaternary f i e l d areas and confirmed the findings in the previous section uhere the type of genesis and parent material uere found to be more important than local conditions. (3) Despite the previously mentioned screening of parameters by significance tests i t uas found that the Fraser Ualley area uas more complex chemically than the Peace River area. The results of the grouping uere therefore slightly less accurate than in the more undisturbed Peace River area, thus confirming the suggestion by Webster and Beckett (1970) that variability 169 i n intensively cultivated areas i s greater than i n more undisturbed areas. TD improve the grouping procedure a factor analysis was performed prior to the numerical c l a s s i f i c a t i o n i n which a l l considered parameters were reduced to four factors without a substantial loss i n variance. Such treatment produced results which i n the case of the Fraser Valley were more closely related to "types of landform units". (4) Few physical parameters were d i r e c t l y correlated with chem-i c a l variables and the numerical grouping procedure proved to be of additional use i n analyzing factors which affect the chemical d i s t r i b u t i o n i n single landform units. Numer-i c a l l y derived groups of chemical s i t e s were found to be related to d i s t i n c t combinations of physical properties suggesting that a multiparameter relationship seems to e x i s t . (5) A numerical analysis of s i t e s on the basis of chemical and form related physical parameters was also found to be useful. The s i t e groups obtained by t h i s means were largely represent-ative of types of drainage at d i s t i n c t positions within the landform unit. C. PREDICTION DF CHEMICAL GROUND CONDITIONS BY MULTISPECTRAL REMOTE SENSING METHODS In studies of chemical t e r r a i n v a r i a b i l i t y elaborate sampling and intensive laboratory analysis are necessary to generate an adequate data base. The present study attempted to lessen these problems by assessing the potential of multispectral remote sensing techniques for quantita-t i v e l y predicting chemical ground conditions. To determine whether spectral signatures are suitable for the quantification of selected chemicals spectral measurements were made via photographic means and through direct d i g i t a l measurements with a spectrometer. The conclusions 170 of these investigations are in tuo parts: (1) multispectral photographic applications and (2) di g i t a l multispectral measurements. 1. Multispectral Photographic Applications (1) Using f i l m / f i l t e r combinations i t uas possible to record spectral information on the photograph in discrete spectral bands. The information in each band uas slightly different and in most cases complementary. (2) Density slicing and color enhancement of equal density areas uere found useful uhen related.to the ground chemical condi-tions. By identifying ground sample points on the sliced images i t uas possible to determine the range of chemical concentrations for different colored density levels. In this uay. a distinction betueen areas of different s a i l moisture and Carbon content uas possible. An areal plani-meter built into the density sli c i n g system alloued a spatial quantification of each density level and thus the chemical distribution over space. (3) Exchangeable Ca, Mg, andMa could be differentiated partially by this method; this i s probably the result of a direct correlation betueen Carbon and exchangeable cations rather than from the individual cations themselves. (4) The same relationship could be found on both a vegetated and a non-vegetated surface. In the former the vegetation surface uas used as an indicator of ground conditions and although similar trends uere observed the overall expression uas ueaker. (5) The uider bands (color 400-700 nm and color IR 500-900 nm) uere more useful uhile the 500-600 and 600-700 nm band produced i n -conclusive results. The slicing of color IR proved to be 171 slightly superior for an analysis of the vegetated surface, while the color sli c i n g gave best results for the exposed s o i l surface. (6) In response to the greater u t i l i t y of the wider wavelength band an additive color technique was used to produce a com-posite image of a l l four original film bands (400-9QD nm). Similar results were obtained with regard to chemical quanti-fication but the resulting image was not superior to the normal color or color IR band. The lack of improvement can probably be attributed to technical problems in the generation of the additive image. 2. Digital Multispectral Measurements (1) Multispectral measurements were made with a spectrometer from the air and from samples on the ground and in the laboratory. A l l three types of measurements proved to be useful in pre-dicting chemical parameters such as % C, % Fe, exchangeable Mg and exchangeable K. (2) Despite differences in measuring technique a l l three types of measurements produced similar regression trends. Linear relationships were used for the prediction but there i s evidence that such functions are only applicable at inter-mediate and low concentration levels. (3) The cause and effect of individual parameters on the spectral reflectance values could not be isolated in a satisfactory manner, but strong evidence was produced that Carbon i s the most important parameter. Carbon was however correlated with several cation concentrations and since no selective, non-destructive extraction procedure was available i t was impos-sible to demonstrate the spectral response of soluble Carbon of cation removal in isolation. The extraction which removes 172 both types of parameter generally produced an increase in spectral reflectance, but from the data i t uas not clear whether the response resulted from both parameters or merely from soluble Carbon extraction. Variations in s o i l moisture and particle size present in the airborne observations did not seem to obstruct the chemical predictions since the relationships uere similar to the laboratory and ground measurements uhere the effect of these parameters uas limited. (4) There uas no selective uavelength which uas most useful for the analysis. Rather an analysis of the total spectral curve was most productive. Soils from similar parent material pro-duced characteristic spectral curves which were clearly identifiable from a l l three measurements. The 500-1100 nm wavelength range was found to be most useful for the present analysis. (5) Similar relationships were obtained from a l l three types of measurements. The airborne analysis produced the lowest overall values while the spectral discrimination uas highest in the laboratory. Laboratory measurements of soils uith low organic matter content ( < 7%) revealed considerably higher spectral values than airborne measurement. In the case of organic soils such an increase was very small however. A l -though the absolute values were different between the ground and the laboratory measurements the overall relationships were closely related. Finally, an atmospheric absorption band slightly reduced the usefulness of the airborne observations in the 950 nm wavelength range. The techniques evaluated in this research w i l l not replace the more conventional terrain analysis methods, but because of their flex-i b i l i t y , ease of handling, and speed of data generation and evaluation they hold considerable potential which can be used to complement terrain analysis investigations. 173 APPENDIX I LITERATURE REVIEU - SOIL CHEMICAL VARIABILITY AS RELATED TO GEOMORPHOLOGY A. INTRODUCTION The relationship between soils and geomorphology and i t s role in s o i l landscape variability i s the topic of this review. Soil variability can be assessed in four dimensions. The f i r s t three are the spatial dimensions, while the fourth i s concerned with variability over time. Although a l l four dimensions are considered, emphasis i s placed on s o i l landscape variab i l i t y . Huggett (1975) noted that the progression of a science as envisaged by Sylvester-Bradley (1967) passes through three phases: taxonomic, historic, and functional. The f i r s t i s concerned with identifying basic parameters and units by which an understanding of genesis can be developed. In the historic phase the progression of genesis over time is studied. Finally, in the functional phase, an understanding of control factors i s attempted, so that the functioning of the system as a whole can be under-stood and predicted. Using such a concept i t can be stated that until recently pedologists have been preoccupied with questions of taxonomy. A number of micro-scale studies have however been undertaken to gain an understanding of the functional changes of specific parameters over short time periods. Despite these efforts and as argued by Huggett (1975) a functional understanding of the system as a whole is s t i l l in i t s infancy stage, the problem being that a comprehensive macro-scale theory cannot readily be developed from an understanding of specific microscale components alone. There are a number of reasons why progress'in understanding the s o i l system has been slaw. Some of the important factors are (1) the complexity 174 of the system as a whole, and (2) the absence of a readily definable functional s o i l unit. 1. Complexity of So i l System The complexity of the system i s best illustrated by the example D f the factor function of s o i l development by Jenny (1941). Although he demonstrated how individual factors such as climate, parent material, r e l i e f , biota and time can influence s o i l development (Jenny 1961) no solution of the total function has yet been accomplished. A number of authors ( c f . Kline 1973, Runge 1973, Yaalon 1975) have elaborated on the complexity of such an equation and have concluded that a practical resolution i s improbable i f not impossible. 2. Absence of a Readily Definable Functional Soil Unit Vertical movement has long been considered D f primary importance in profile development and far this reason l i t t l e attention has been given to the identification of three-dimensional soil.units. Only for mapping purposes did i t become necessary to use spatial units. A r t i f i c i a l taxonomic mapping units such as the s o i l series were developed for the purpose of minimizing the within unit variance. Attempts to look at functional s o i l units have ranged from the catenary concept to toposequence, site, facet, watershed, and geomorphological units. A l l of these are closely related to geamorphology and are discussed in this context. B. SOIL GEOMORPHOLOGICAL RELATIONSHIP When studying the s a i l landscape i t is evident that pedology and geamorphology overlap in many areas. Jenny (1941) list e d r e l i e f as a primary factor D f s o i l formation. It should however be noted that r e l i e f exerts i t s e l f mostly in an indirect way by affecting microclimatic, hydro-logical and geomorphological processes, which in turn are responsible for 175 changes in s o i l development. Nevertheless surface shape and composition are closely related to the distribution of soils and this relationship forms the basis of a s o i l landscape framework. Generally three types of landscape components can be recognized: paint, linear, and spatial com-ponents. 1. Point Component of the Landscape Altitude, aspect, and position can be referred to as point specific, although spatial zones are often related to them. A l l these components affect climate and drainage conditions which in turn are responsible for altering s o i l development. (a) Altitude Altitude i s primarily responsible far influencing the temperature and precipitation regime and i t i s for that reason that different soils occur at different altitudes. The altitude effect on s o i l development and chemical movement have been investigated by c f . Wilcox et al (1957), Floate (1965), Cortes and Franzmeier (1972). (b) Aspect Microclimatic conditions are mostly affected by aspect. Analyses as to the effect of aspect on s o i l chemistry have been reported by c f . Cooper (196D), Hembree and Rainwater (1961), Birkeland (1974). (c) Position A number of investigators have noted that topographic position i s most important in differentiating drainage, sedimentary and erosional processes. Spratt and Mclver (1972) for example demonstrated the effect of position on s a i l f e r t i l i t y . Since the effect of position can however only be assessed on a comparative basis i t w i l l be discussed in greater detail under the topic of linear landscape components. 176 2. Linear Landscape Components Hillslopes generally consist of a sequence of continuous slope segments which are referred to as slopes, toposequences, and catenas. They are linear landscape components and can be quantified by gradient, length and form. A toposequence i s usually referred to as a series of slope segments over similar parent material. It differs from a catena in that the latter includes drainage considerations and uas originally restricted to residual conditions. (a) Slope gradient Numerous attempts have been made to relate slopes or toposequences to s o i l properties. In i t s simplest form slope gradient uas found to be inversely correlated uith s o i l thickness and drainage (Young 1972). The relationship i s made more complex however by slope evolution and processes of weathering, erosion, transportation and deposition. It is for this reason that models of slope retreat have been used in order to establish soil-slope relationships (see for example Beckett 1968, Furley 1968, etc.). It i s recognized that some processes are more dominant on different sections of the slope and this in turn affects the s o i l development con-siderably. This observation was used by Ruhe and bJalker (1968), and walker and Ruhe (1968) in the development of a soil-slope model. They attributed the variations in relationships at different positions within the slope to sedimentary and erosional processes and differences in parent material age. The overall position within the slope i s therefore of basic importance. This was enforced by Bunting (1961), Protz et a l (1973), and Furley (1974) who noted that position was more important than gradient in influencing s o i l development. Slope gradient i s however the main determinant of pro-cesses. In addition, when confined to individual slope segments, a strong correlation between slope gradient and a number of chemical parameters was found to exist in the upper erosional slope segment (Whitfield and Furley 1971, Furley 1971). Slope gradient was positively correlated with pH, CO3, 177 IMa, K and Ca. While C and IM decreased uith an increase in slope gradient for the same segment. These relationships did not appear to be valid in the louer depositional slope. In studies involving catenas and toposequences, an assessment of s o i l development and chemical movements are made by anal-yzing s o i l conditions along the slope sequence. Without making specific references to slope gradient, changes in s o i l conditions are reported in the dounslope direction. Yaalon-' et al (1972) for example found a 50-80% increase in Mn dounslope in three catenas of different parent materials. Dan et al (1968), Young (1969), and McKeague et al (1973) described s o i l variations, Adams and Walker (1975) PO^ movements, IMyhan et a l (1972) s o i l nutrient changes, and Dodd et a l (1964) salt movement over catenary se-quences. Finally Walmsely and Lavkulich (1975) shoued hou uater chemistry can be used to establish soil-landform relationships. Although only some of the above studies specifically noted the effect of slope gradient i t i s evident that i t forms an important part in these investigations. (b) Slope length Bunting (1964), Walker et a l (1968b), and Furley (1971) noted that slope length is related to s o i l development. A clear separation of slope length and gradient is however not readily passible. Slope length in conjunction with gradient exerts i t s e l f most strongly on the erosional and depositional processes on the slope. (c) Slope form The slope form, such as rate of change and radius of curvature, has also been noted as a factor in influencing s o i l development (Curtis et a l 1965). The relation of lithology and slope form was elaborated by Swan (1970, 1971). Troeh (1964) found good correlations between shape of slope and drainage classes, while Dan et a l (1964) and Walker et a l (1968c) mentioned slope form as a factor in horizon development and loess accumula-tion. 178 3. Spatial Components of Topography Two directions have been followed in using spatial components to establish geomorphology-soil relationships: (a) to relate point and linear features to space, and (b) by a direct analysis of spatial units. (a) Catena vs s o i l association Most slope investigations are concerned with linear processes and as noted by Reynolds (1975) the majority of toposequences and catena studies use linear sampling schemes. On this basis spatial implications are not readily passible. With regard to catena Webster (1965) noted that in his study area a representative catena could be related to the characteristic s o i l pattern which covers large areas. In this case a catena becomes an indicator of three-dimensional units of s o i l associations. Ruhe and Walker (1968) used the spatial dimension to f i t their slope-soil model to the loess landscape. Some Russian landscape classifications are based on geochemical migration^ (Perel'man 1961, 1967). Glazovskaya (1963, 1968) related such migrations to catenary sequences and developed a typology of geochemical catenas which are representative of the major climatic zones (Glazovskaya 1970). The use of the catena as a functional unit has been suggested by Dan and Yaalon (1964). Huggett (1975), however, argued that such units might only have validity an surfaces where straight flowline conditions are dominant. Specific slope types (components of catenas) have been related to s o i l patterns in steepland areas by Campbell (1973). Others, such as Dan et a l (1968), Hleiss (1970), and Huddleston et al (1975), distinguished between slope sections of depositianal and erosional nature, in order to establish chemical s o i l landscape relationships. Similar examples were reviewed by Daniels et a l (1971). 179 (b) Geomorphological units Geomorphological surfaces have long been used as a basis for s o i l mapping. Such units should be identified on the basis of form and parent material and should preferably be genetic units. Because of the poly-genetic nature of the landscape the relationship betueen geomorphological units and soils i s however not totally satisfactory. Webster and Beckett (1964) noted that on a reconnaissance scale the variation of a number of s o i l chemicals within such mapping units i s far too great to have pre-dictive value. In more detailed analyses where emphasis i s placed on processes of genesis and composition of parent material the geomorphology-s o i l relationships are often more useful (Bettenay 1968). Crofts (1974) noted that detailed geomorphological mapping with emphasis on genesis can be of great use in physical terrain analysis. Similarly Cooke and Doornkamp (1974) reviewed geomorphological mapping based on morphology, s u r f i c i a l material and processes. Ruhe (1956, 1974), and Ruhe et a l (1971) used stratigraphy and chronology in describing the nature of soils on geomorphic surfaces. Similarly, coastal plains (Daniels et a l 1970a, Gamble and Daniels 1974), and lake sediments (Spilsbury and Fletcher 1973, Kaljonen and Carson 1975) proved to be useful in studying pedologic conditions. Finally, Huggett (1975) suggested that on theoretical considerations the watershed should be used as a functional unit of the s a i l landscape in mass balance studies of chemical components. The s o i l landscape re l a -tionship in such units has been investigated by Mala et a l (1974). Up to a certain scale the mapping of uniform process areas within such units however would be more promising since linear hydrological relationships da not necessarily give a good reflection of spatial v a r i a b i l i t y . Most geomorphological units have great internal v a r i a b i l i t y . This i s largely related to the polygenetic nature of the landscape, where during the evolution of the units a number of different processes l e f t their 180 imprint. The relative importance of these processes and their role in the genesis of the unit are therefore c r i t i c a l . This uas demonstrated by Huddleston et a l (1973, 1975). C. SOIL LANDSCAPE VARIABILITY The lateral variability of s o i l chemicals has been extensively revieued by Beckett and Webster (1971). The present section i s therefore limited to an evaluation of the potential usefulness of different spatial units in predicting chemical conditions. The most desirable units are functional ones in uhich the internal variability i s minimal and the betueen unit variance great. In nature such units are rare since, as noted by Beckett and Webster (1971), 50% of the total variance uithin a f i e l d may already be present uithin the f i r s t feu meters 2. It i s generally accepted that the variability increases uith the size of the units (Beckett 1967},, Mclntyre 1967) but the magnitude and form of such a relationship are not u e l l understood (Beckett and Webster 1971). When dealing uith the landscape i t i s variability rather than uniformity uhich dominates. Most of the factors responsible for altering chemical conditions cannot readily be observed. It i s therefore unlikely that a comprehensive landscape classification system can be developed in uhich a l l factors are considered. Instead units are usually chosen on the basis of readily observable s u r f i c i a l expressions such as geology, landform, vegetation cover and land use. In this context a distinction should be made betueen a r t i f i c i a l units and those related to the natural landscape. 1. A r t i f i c i a l Units A number of taxonomic s a i l units such as the pedon, s o i l type and s a i l series, as uel l as most of the agricultural fields, can be considered a r t i f i c i a l . 181 (a) Pedon - s p i l series The pedon i s the smallest three-dimensional s o i l unit of the land-scape. It consists of a few meters-3 of s o i l and i s largely of academic and taxonomic interest. Drees and Wilding (1973) found that when confined to individual horizons, pedon variability i s greatly affected by parent material. They found up to k5% variability of pedons on outwash while pedon on loess and t i l l varied up to 20%. The care taken in the sampling procedure i s most important in such investigations. Mapping units based on the s o i l series are a r t i f i c i a l since they are only partially related to topography and geology. Their main purpose is to minimize the within unit variance and they are often delineated by analogy of sampling analysis. In complex landscapes i t i s often impossible to map pure s o i l units. In such cases the variability i s of course greatly increased. There i s however no universally accepted level of variability for s o i l series mapping units. It i s evident that the greater the level of uniformity the more useful the s o i l maps become for management purposes. (b) Agricultural fields Agricultural fields are rarely delimited on purely physical grounds. Social and economic factors often play some role in the selection. For agricultural production and management a knowledge of the within f i e l d variation i s of great importance. The results of a number of such studies are given by Beckett and Webster (1971) who noted that far cultivated fields the s o i l chemical variability i s often greater than for similar areas under natural conditions. This i s especially the case for such elements as C, Ca, K, and P. 2. Unitss related to the Natural Landscape Geological and geomorphological c r i t e r i a are most often used as the basis for landscape unit delimitation. The chemical variability within such units has been assessed at different scales. 182 (a) Site-polypedon-land facets A l l these units are delineated on the basis of parent material and surface form. The importance of such units in ecological studies has been stressed by Wright (1973). The site variation for several chemical parameters has been investigated by Mader (1963). He noted that the variability uithin a site uas smaller than that betueen sites, thus sat-isfying the basic principle of classification. But he further stressed that the relative variability of different chemicals uas great. Land facet variations uere investigated by Mitchell and Perrin (1967), Webster and Beckett (1970), Mitchell (1971), and Areola (1974). Mitchell and Perrin (1967) concluded that uithin a limited geographical area land facets, uhich are recurrent terrain surfaces uith uniform appear-ance, are sufficiently homogeneous and mutually different to be used for land inventory studies. Areola (1974) confirmed these findings by con-trasting chemical uith physical s o i l parameters. Some of the chemical parameters did houever shou great variability, making the usefulness of such units in predicting chemical conditions less valid. (See also Webster and Beckett 1970.) A l l these authors classified facets on the basis of analogy and found that drainage conditions and lithology are largely responsible for the differentiation of groups of facets. (b) Watershed The uatershed as a functional unit of the s o i l landscape has been suggested by Schumm (1956). In mass balance studies of nutrients (see for example Crisp 1966, Bormann and Likens 1969, Cleaves et a l 1970) such units have proven to be useful. An analysis of the spatial s o i l variation uithin such units i s more complex. The basic problem of relating linear processes to space, rather than delineating uniform process areas remains. V/reeken (1968, 1973) assessed s o i l variability uithin a uatershed and concluded that differential 183 and spatial-age transgressions of soils and geomorphological parameters complicated s o i l interpretations. (c) Geomorphological mapping units Most s o i l mapping units are delineated on aerial photographs and as such are based on geomorphological surface expressions. Depending on the complexity of the landscape, the detail of analysis, the scale, and aim of mapping, emphasis can be placed on morphology, geology, topography and genesis. A reduction in variability within the mapping units i s usually achieved when a l l of these factors are considered. This was confirmed by Webster and Beckett (1964) who noted a decrease in variability for units chosen on a geological and physiographic basis. The same authors did however note that the variability of some of the chemical parameters remained high, making a prediction of chemical conditions unsatisfactory. Areola (1974) considered geology, r e l i e f , slope, ground drainage, climate and vegetation in his investigation of land facets mapping units. He found a high degree of homogeneity for those factors which are genetic-al l y and morphologically relevant for s o i l description. Chemical properties such as exchangeable cations and organic matter did however vary considerably. McCormack and Wilding (1969) found that the within unit variation in a complex landscape iscoften greater than the between unit variation. They suggested that more detailed geomorphic factors such as slope, position and aspect could be responsible for a part of this variance. Crosson and Protz (1973) had more success in predicting organic matter, pH, and clay content for similar mapping units. In such an analysis the complexity of the landscape and the scale of mapping i s of c r i t i c a l importance. In general i t can be stated that the smaller the scale the greater the chance of including impurities in the mapping units. At a very large scale where emphasis can be placed on slight differences in slope, aspect and position, an improvement in variability i s possible. 184 These generalities are however only applicable to some s o i l parameters and i t i s evident that a number of chemical properties are too variable t D be usefully classified within such a framework. For those variables an assessment of the cause of the variation must be explained by a number of additional factors related to climate, hydrology, biota, and land use. Finally, Mitchell (1971) found that the predictability of some mapping units i s only valid within a limited geographic area and cannot readily be applied to unit prediction over wide expanses of terrain. D. CONCLUSION The complexity of the s o i l system i s evident by the absence D f a readily acceptable functional s o i l landscape unit. No simple factor relationship can adequately account for the s o i l development and variability but a close relationship between the shape and composition of the geomor-phological surface and s a i l development does exist. In s a i l landscape studies such relationships have formed the basis of s o i l mapping. Unfortunately the majority of investigations have em-phasized linear slope relationships which cannot always be related to space. Geomorpholagical surfaces, especially when analyzed in terms of age, compo-si t i o n , stratigraphy and genesis, have proven to be most useful. The poly-genetic nature of the landscape does however introduce a complication factor which cannot be salved quickly. When the above c r i t e r i a were used to delineate the s o i l landscape, units could be identified which showed reasonable internal homogeneity while exerting distinct differences amongst them. Most of the parameters used in s o i l description were found to have smaller variance within the units than between them. This i s however not necessarily the case with chemical parameters, which tend to be considerably more variable than their physical counterparts. 185 In a detailed analysis of simple landscape where emphasis can be placed on slope, position, aspect and drainage the variability for such parameters i s reduced, but i t s t i l l remains at a disturbingly high le v e l . This suggests that factors other than the considered geomorphological factors contribute considerably to the chemical va r i a b i l i t y . As a result the value of small scale mapping units i s therefore only partially satis-factory in predicting s o i l conditions. 186 APPENDIX II A REVIEW DF MULTISPECTRAL SENSING TECHNIQUES AS APPLIED IN CHEMICAL TERRAIN ANALYSIS A. INTRODUCTION Remote sensing applications in soils, vegetation and terrain studies have been reviewed by Myers and Allan (1968), the Remote Sensing Committee (1970), Carroll (1973), Pettry et a l (1974), Myers (1975) and Reeves et al (1975). The present review i s therefore confined to multispectral applica-tions in the visible and near IR wavelength range to detect chemical s o i l conditions. Certain vegetation, soils, and geological parameters are often indicative of chemical ground conditions and constitute an important part of this review. As mentioned by Calwell (1966) "Transmission, reflection, absorption, emission and scattering of electromagnetic energy by a particular kind of matter are selective with regard to wavelength and are specific for that particular kind of matter, depending primarily upon i t s atomic and molecu-lar structure." To make f u l l use of this concept multispectral techniques have been developed in which target reflections are measured and compared in a number of discrete wavelength bands. Reflection in the visible and near IR wavelength can be measured photographically by different film/ f i l t e r combinations, or by direct measurements with a spectrometer. The two methods are treated separately in this review. When analyzing chemical surface conditions with multispectral tech-niques, one i s inevitably confronted with the question of what is sensed. The sensor usually produces an integrated record of the vegetation cover, s o i l background and shadow effect, and a separation of the total reflected energy into individual components is d i f f i c u l t . Comparative methods and causative factors have been used to identify ground conditions. Chemical surface conditions have been investigated by analyzing bare surfaces and via geobotanical evidence. Both methods are reviewed here. 187 B. MULTISPECTRAL PHOTOGRAPHY The sensitivity range for a direct recording of reflected energy i s betueen 400 and 900 nm uavelength for conventional f i l m / f i l t e r combina-tions. Reflections from other uavelength bands can only be rendered visible on photographic film by indirectly transforming the originally sensed data by a photosensitive c e l l (Baker and ScDtt 1975). A considerable information loss usually occurs through this transformation, but as mentioned by Reeves et al (1975) the spatial information i s often as important as the spectral data, especially uhen dealing uith geological and soils interpretations. The traditional use of black and uhite film i s being challenged by the availability of color. The eye-brain combination can identify far more colors than grey shades (Evans 1948, Slater 1975) and as a result better interpretations are obtained by the use of color. Black and uhite film i s preferably being replaced by different color films (Colvolesses 1975), or used uith different f i l t e r combinations to provide the basis for additive color vieuing (Anderson 1969, Yost and Uenderoth 1971a, Ross 1973, and Ldenderoth and Yost 1974). In either case the addition of color has introduced a neu dimension to photo interpretation. The aim of multi-spectral photography i s to produce refined photographic images of a target in uhich certain spectral elements are emphasized or contrasted. 1. Subtractive Color The color produced by conventional color films i s the result of a subtractive process in uhich the three dye layers regulate the transmission of radiation of specific uavelengths. In terrain studies tuo such films, the f u l l color and the color IR, have been used uith increasing success. (a) Soils and geological applications Most soils and geological applications have been confined to mapping applications and assessments of drainage conditions. Such studies have some relevance uith reference to chemical conditions since parent material 188 i s an important factor in influencing s o i l chemistry. Anson (1970) found a greater accuracy in s o i l identification with color and color IR films. This was confirmed by Parry et a l (1969), Gerbermann et a l (1971), Valentine et a l (1971), and Acton and Stonehouse (1972). They a l l found that s o i l boundaries could be interpreted with greater ease on color film. Improved analysis of s o i l drainage with color and color IR was reported by Kuhl (1970), and Parry and Turner (1971). Similarly floodplain analysis and mapping were fac i l i t a t e d with the two color films (Parker^et a l 1970, Orr and Quick 1971, Anderson and Uobber 1973, Conway and Holz 1973, and Piech and Walker 1974). (b) Vegetation applications Color IR film proved to be of significant use in vegetation studies, since plants reflect considerably more radiation in the near IR wavelength range (Gates 1970, Knipling 1970, and Gausman 1974). Because of this, species and forest type identification i s improved, as indicated by Heller et a l (1964), Aldrich (1971) for forest type, Driscoll and Coleman (1974) for shrubland, and Philpotts and Wallen (1969), and Murtha (1972) for vegetation damage. Vegetation can be used in the form of geobotanical or biogeochemical evidence (Gilbertsan and Langshaw 1975). The former refers to the use of indicator species inferring chemical conditions while the latter emphasizes the uptake of selected minerals by the plants. Such an uptake w i l l often affect the spectral properties of the vegetation. Some success in the identification of copper bearing strata using geobotanical evidence and multispectral photography has been reported by Cole et a l (1974). 2. Additive Color Simultaneous exposure of black and white film with different f i l t e r combinations has become the principle behind multispectral photography. Such photographic systems have been described by Marlar et a l (1967), Yost and Uenderoth (1967), Hawarth (1972), and Slater (1972, 1975). The different black and white bands can then be superimposed in projecting the individual 189 bands through different exchangeable f i l t e r s . Uith the proper selection a host of different color and false color enhanced images can be produced (Yost 1969). (a) Soils and geological applications Direct applications in agriculture and s o i l science have been reported by c f . Uiegand et a l (1971), Yost and Uenderoth (1971a), Behe (1972), Croun and Pauluk (1972), Tarnocai (1972), and Evans (1975). Most of these are concerned uith s o i l type identification and s o i l mapping. A large number of ERTS applications are based on the same principle, uith the exception that the original sensor i s a non-photographic multi-channel scanner uith an extended uavelength range. Digital data from four uavelength bands can be transformed into black and uhite photographic images and add-it i v e color images can then be produced by superimposition. Such ERTS imagery applications have been described by Anuta et a l (1971), Lathram (1973), Cipra (1973), Parks et a l (1973), and Steinhardt et a l (1974) for soils and geological mapping applications. (b) Vegetation applications Land use, forest and vegetation type mapping have been done success-ful l y from ERTS imagery. Feu of them are houever related to chemical ground conditions and are therefore omitted in this review. Most of the above mentioned photographic applications are based on interpretive analysis and subjective comparisons, emphasizing the spatial rather than the spectral element. 3. Quantitative Analysis of Photography A number D f studies have been undertaken to quantify the photographic imagery by measuring film densities. In this uay an automatic interpreta-tion of terrain conditions can be attempted (Rib and Miles 1969). The basic problem i s hou to identify those factors uhich make up the density values. 190 Unfortunately this is a complex problem since a number of factors are involved, some of uhich cannot be controlled very readily (Hunter and Bird 1970). The most important components of density are tone and tex-ture (Hoffer et al 1972) and these are dependent on. (1) spectral compo-sition of the incident radiation at the time of exposure, (2) photographic factors such as film type, lenses, aperture, exposure, f i l t e r s , develop-ment, and (3) physical and chemical conditions of the ground. Some measures to limit the photographic factors have been described by Malan (1974) and Wenderoth and Yost (1974), but l i t t l e control i s possible over atmospheric and ground conditions. The effect of vegetation, soils, and photographic factors on tone has been described by Hoffer et a l (1966), and Richardson et al (1975). Given the above mentioned problems i t i s evident that density applications should be treated uith caution. (a) Soil and geological applications Evans (1976) found that predictions of lithology, particle size, chalk content, organic matter, and s o i l moisture conditions from tonal measurements uere only partially successful. Sharma et a l (1972) produced similar results for geological unit interpretation. The method i s someuhat more useful uith regard to the spatial element uhere equal density areas can readily be enhanced, mapped and quantified uith an areal planimeter. Mapping applications for soils (Tanguay 1969, Frazee et a l 1972, and Odenyo and Rust 1975) and s o i l limitations (Frazee et a l 1971) have been most useful. Measurements of color densities have also been pursued. Cihlar and Protz (1972), Symeonakis and Pratz (1974) had only limited success in s o i l mapping using color density values. Density measurements on multispectral space photography uere made for s o i l salinity determinations by Ldiegand et a l (1975). A partial dis-crimination betueen high and lou saline conditions uas passible from the 600-700 nm uavelength band despite the vegetative cover. Color and color IR film proved to be less useful. 191 To make better use of the multispectral concept Vincent (1973) proposed a ratio method in uhich spectral ratio values from the most contrasting bands are used. Such an approach i s most useful uith direct spectral values rather than uith density measurements, since small var-iation in film and processing conditions can alter the density values readily. A ratio method using color film has been discussed by Piech and Walker (1974). They used the blue and red density ratio to delineate s o i l moisture and s o i l textural conditions. Cb) Vegetation application Density measurement on color IR film has been used by Brooner and Simonette (1971), and Learner et a l (1975) for crop identifications. More recently Maurer (1974) classified agricultural crops by their tex-tural patterns using the color density measurements and discriminant analysis. Stoner et a l (1976) used color and IR color densities for estimating canopy density, yield and biomass. The vegetation conditions uere houever not related to chemical ground conditions. The same tech-nique uas used by Thomas et a l (1966) to relate density values to s o i l s a l i n i t y . Similarly Wiegand et a l (1975) used density values to classify s o i l salinity conditions in several sites in Texas. The latter found that density differences betueen vegetation and bare soils uere most useful for discriminating betueen different salinity classes. Finally, a method producing spectral reflection curves from density measurements of color photographs has been described by McDouell and Specht (1974) uho used this method for crop identification.„ Controls over exposure and processing conditions are essential, houever, and limit such applications considerably. 192 MULTISPECTRAL SENSING In order to emphasize the spectral rather than the spatial element, direct spectral measurements of the reflected and emitted energy of the ground surface are desirable. Spectrometers uith f i l t e r combinations and grading devices are generally used for this purpose (Hovis et al 1967, Hunt and Ross 1967, Holmes 1970, and Raines and Lee 1974). A number of factors have been identified uhich influence the reflec-tion characteristics, Watson (1972) list e d ueathering, surface conditions and lichen caver as factors uhich complicate spectral identification of rocks. Gates (1970) mentioned s a i l texture, s o i l moisture, surface rough-ness, mineral and chemical conditions, and s a i l color as important param-eters affecting s o i l reflection values. Similarly Carlson et al (1971) and Pettry et a l (1974) listed species composition, leaf and plant structure surface roughness, moisture conditions and plant stress as important element in spectral analysis of vegetation. Spectral measurements can be obtained in the laboratory from in situ f i e l d measurements and from airborne analysis. Applications u i l l be dis-cussed in terms of these three methods. 1. Soils Applications (a) Laboratory measurements In laboratory analysis a rigorous control aver illumination conditions is passible and i t i s for this reason that laboratory spectral reflection measurements have been most successful. Steiner and Guterman (1966) and Talchel'nikov (196B) published a series of reflection curves for Russian soils, uhile Condit (1970) produced a number of American s o i l spectra. These studies did not relate s o i l parameters to spectral values but uere concerned uith determining different types of spectral curves. 193 Bouers and Hanks (1965) in a more detailed analysis found that s a i l moisture, particle size and organic matter content a l l affect the reflection of radiant energy. They noted a decrease in total reflection uith an increase in s o i l moisture, especially at the longer uavelength bands. These observations uere confirmed by Hoffer and Johannsen (1969) uho also suggested that the crust formation uhich results from surface drying com-plicates the reflection measurements. The reflectance difference betueen uet and dry soils and the factors influencing the accuracy of such measurements uere discussed by Planet (1970). Pettry et a l (1974) noted that although both s o i l moisture and organic matter affect the reflection measurements, variations in the latter result in a considerably smaller reflection change at lou organic matter values. Soil moisture assessments uere also attempted in the infrared range. Skidmore et a l (1975) developed a method based on uater absorption characteristics in the IR range. Laboratory measurements of s o i l color have been described by Shields et a l (1966) and Karmanov (197LT). The relationship betueen s o i l color and moisture-organic matter content have been investigated by Shields et al (1968). They found that differences in the nature of organic matter such as humic to fulvic acid ratio and percent NaOH extractable humic acid are responsible for reflection differences. Furthermore the removal of organic matter by oxidation results in a considerable increase in reflection (Bouers and Hanks 1965, Hatheus et a l 1973b, and Pettry et al 1974). Free iron oxide and clay type uere found to alter reflectance values from soils (Matheus et a l 1973b). Some of these results should houever be treated uith caution since removal of organic matter by oxidation (H202 or muffler furnace treatment) and extraction of iron by the sodium-dithionite method are not necessarily non-destructive and can alter other s a i l proper-tie s . 1 % (b) In situ f i e l d measurements It i s not readily possible to relate laboratory measurements to f i e l d observations since solar radiation i s variable and cannot be con-trolled. In situ analysis has mostly been related to dynamic elements such as s o i l moisture. Cipra et al (1971) for example measured in situ s o i l reflectance in the visible uavelength range. They confirmed the laboratory findings by observing that the same type of s o i l under dry conditions produced higher reflectance values than under uet conditions. They also found that surface roughness, crust formation and shadou sig-nificantly influence the measurements. Skidmore et al (1975) developed a potential f i e l d method by u t i l i z i n g uater's property to absorb certain infrared uavelength. Blanchards et al (1974) confirmed the reduction effect in reflection by an increase in moisture. In addition they des-cribed a method of relating s o i l temperature differences to s o i l moisture conditions in the thermal IR uavelength range. A decreasing difference in day to night s o i l temperature uas related to an increase in s o i l moisture. Similar temperature related methods uere described by Myers and Heilman (1969), and Idso et al (1975). In none of these studies did the reflection values relate to chemical conditions. (c) Airborne applications May and Petersen (1975) related laboratory spectral curves to those obtained from an airborne scanner. When applying correction factors for solar radiation and atmospheric attenuation adequate comparisons could be made. Multispectral scanner data from both aircraft and sat e l l i t e sensors have been applied to s o i l mapping (Hristof and Zachary 1971, and Kristof 1971). They noted that the use o f spectral s o i l classes in automatic terrain analysis has potential but that such factors as surface roughness, texture, color and variations in illumination conditions caused significant problems in mapping. The use of computer implemented pattern and c l a s s i f i -195 catiDn analysis to analyze multispectral data has been suggested by Steiner et a l (1969), Fu (1976), and Kettig and Landgrebe (1976). Applications by Cipra et al (1972) and Hristof and Zachary (1974) uere only partially successful for s o i l mapping. Drainage pattern and organic matter content uere readily observed but these authors f e l t that due to several complicating factors such multispectral techniques u i l l only supplement but not replace the traditional survey and mapping procedures. More success uas reported by Baumgardner et a l (1970), and Kristof et a l (1973, 1974) uhen mapping s o i l organic matter uithin a confined geographical area. The potential of mapping other types of s o i l chemicals has been assessed by Montgomery and Baumgardner (1974). Soil parent material uas mapped uith reasonable accuracy by Matheus et al (1973a) using similar techniques. Finally multispectral techniques uere used in s o i l salinity detection from aircraft (Prentice 1972) and Skylab (Richardson et a l 1976) data. The latter found that the 1090-1190 nm uavelength bandouas most useful for bare s o i l s . The use of reflection differences betueen bare soi l s and associated vegetation proved to be an even better indicator over the 690-1700 nm uave-length range. 2. Geological Applications Spectral analysis of rocks and minerals is made more complicated by the existance of isomorphous minerals, uhere certain elements are readily substituted for others in selected positions in the geometrical frameuork of atoms (Hunt and Salisbury 1970). Also impurities and contaminations resulting from ueathering and sample preparation have been listed as c r i t i c a l in spectral analysis (Hunt and Salisbury 1970, Vincent et a l 1975). Electronic and vibrational processes uithin the molecular structure produce spectral features uhich can be related to chemical composition. Multispectral techniques should therefore at least on theoretical grounds be suited for rock and mineral discrimination. 196 (a) Laboratory measurements Despite extensive spectral analysis of minerals and rocks in the laboratory by Hovis (1966), Adams and F i l i c e (1967), Hunt and Salisbury (1970-1974), Salisbury and Hunt (1974), and Vincent et a l (1975), the spectral behaviour of minerals and rocks is s t i l l not well understood. The primary aims behind these spectral studies are (1) to determine spectral curves for different rocks and minerals in building up a reference library, (2) to select uavelength bands by uhich rocks and minerals can best be contrasted, and (3) to relate chemical conditions to spectral values for prediction purposes. Vincent et al (1975) found a number of significant uavelengths by uhich certain chemical elements can be detected (see also White and Keester 1966, and Hunt and Salisbury 1970). While such applications uere i n i t i a l l y aimed at space exploration (Adams 1968) an increasing interest has developed to apply such spectral techniques to geological explorations on Earth. (b) In situ measurements Only a limited number of in situ spectral measurements have been reported. Raines and Lee (1975) used a multifilter spectrometer to measure in situ reflectance of sedimentary rocks. They concluded that total spectral curves rather than specific uavelength reflection values should be used. Also large numbers of measurements are required to f u l l y account for a l l surface conditions. Watson (1972) reported that surface conditions, ueathering, and lichen complicate such measurement endeavors. (c) Airborne applications A comparative evaluation of sand deposits by spectral reflection uas made by Romanova (1968). Brennan and Lintz (1971) uere able to discriminate betueen rock types, but mentioned that reflection"differences resulting from moisture variations, s o i l and vegetation cover may exceed lithological differences. Vincent and Thomson (1972) shoued that a discrimination of 197 high s i l i c a content rocks, carbonate rock and vegetated targets i s possible in some cases. Infrared spectral emittance values uere used for geological mapping in volcanic terrain by Lyon (1972) and Pohn (1974).. Finally, Vincent and Thomson (1972) and Vincent (1973) shoued that uith the use of ratio images greater control over atmospheric conditions and better dis-crimination betueen rock types uere obtained. A similar approach uas suggested by Vincent (1972) to differentiate betueen iron compounds. 3. Vegetation Aspects The vegetation factors uhich alter spectral reflection have been revieued elseuhere ( c f . Myers 1970, 1975). This section i s merely devoted to applications of vegetation as an indicator of chemical ground conditions. (a) Laboratory and in situ studies The potential of the early detection of mineral deficiency in plants by multispectral techniques has been evaluated in a number of laboratory and greenhouse studies (Gausman et a l 1969a, Richardson et a l 1969, Ward 1969, and Younes et al 1974). In some of these studies nutrient deficiency and high salt content could not readily be detected from spectral measure-ments of the leaves. Weber and Olson (1967) and Gausman et a l (1969b) noted that shade type of c e l l s produced under greenhouse conditions, s o i l moisture conditions in early leaf development, leaf age (Brack and Mack 1972), disease, and s o i l temperature conditions a l l complicate such an analysis. Due to the structural change induced by saline conditions i t is houever possible to use some plants as indicators of salinity conditions (Gausman 1969). Similarly nutrient deficiency in several agricultural crops could be detected by Myefsb(1970), Al Abbas et a l (1974), and Younes et a l (1974). Most encouraging results uere reported by Gausman et a l (1973) uho detected iron, magnesium, phosphate, sulfur and nitrogen deficiencies in Cucurbita pepo test plants using multispectral methods. 198 In situ spectral signatures of forest trees uere produced by Kalensky and Wilson (1975). They suggest that site conditions are important in tree differentiation but do not refer to specific chemical conditions. (b) Airborne observations The interpretation of airborne spectral observations is made more complex by the fact that the spectral signature i s the result of the integrated effect of vegetation and s o i l factors. Reflection differences between bare soils and crop residue have been investigated by Gausman et al (1975). Attempts to extract underlying s o i l spectra from grasslands have been reported by Tucker and Miller (1974), uhile changes in s a i l patterns uith increasing crap canopy uere investigated by Hristaf and Baumgardner (1975). Another attempt at partitioning uas made by Mansel and Oohannsen (1973) far agricultural crops. Finally the spectral change induced by crop grouth uas investigated by de Boer et a l (1974) over the grouing season. The most promising chemical applications have resulted from studies related to species identification. Spectral signatures of specific plant species uere used to delineate uetland conditions by Carter and Anderson (1972). Mineral prospecting'using indicator plants uas reported by Press and Norman (1972). The identification of areas uith high copper content by spectral methods has been suggested by Cole et a l (1974). Houard et a l (1971) used ponderosa pine reflection signatures to delineate high and lou copper content areas. Finally Yost and Uenderoth (1971b) uere able to correlate soil-geochemical information uith tree reflection data in the case of mineralized balsam f i r s . CONCLUSION Dn theoretical grounds multispectral remote sensing methods seem uell suited for analysis of surface chemical conditions. Unfortunately as evidenced in this revieu a very large number of factors complicate 199 such analysis. None of the methods discussed above produce outstanding results, and i t should be stressed that multispectral techniques should be used only in conjunction uith traditional survey methods, since at best they produce complementary information. The spatial element is favored in using the multispectral photographic method. The f a c i l i t y to enhance and contrast the aerial images i s of use for mapping and inventory purposes and can be considered as an improvement over the traditional air photo interpretation method. A quantitative analysis of photo density i s useful in same instances but cannot readily be applied on a continuous or sequential basis. Direct multispectral measurements hold more promise but because of the dynamic nature of solar energy and atmospheric conditions i t i s d i f f i c u l t to produce consistant and comparative results. Uith this latter method the spectral rather than the spatial element i s stressed. Target and ground conditions can be contrasted by comparing the spectral curves. Both methods f a c i l i t a t e but dD not eliminate f i e l d and sample analysis. Uith the multispectral approach a partitioning af the effect of different factors is passible in same instances but in general obliteration of one factor by another is more common. Controlled laboratory experiments hold the key in future application. To relate the laboratory results to those from the air i s houever a d i f f i c u l t task. Finally, in analysis of chemical ground conditions from the air the indirect method of using selected plant species as indicators of chemical conditions holds the greatest promise. APPENDIX III SUPPLEMENTARY INFORMATION ON NUMERICAL GROUPING APPENDIX I I I - l : The location of the sites which were used in the Fraser Valley project are indicated in Figure 64 below: LEGEND • G M : GLACIO-MARINE UNIT 6 a *b : OUTWASH UNITS Mcn-b : MARINE UNITS SAMPLED SITE L O C A T I O N S - « Scale 1 2 3 4 5 km N Figure 64. Locations of sites in Fraser Valley study area. 201 APPENDIX III-2 I ~ . u -« » » » • » » * * « « * * : J — H e g c o [ co IS eg i n l CM ! i CM i y « » • # # I ! i-i i ! I b i -' i — i — " T i -o •i. if * | W * t * * fcl it H ir t CM vO in CT» vT — — - r r -o o o co cn OD, crj m ^ CO "O o - r i s |— — — « — m o> CD f j i f > o - t <r U l co Ln (M o i CO VA vO * -K ! -K * * I ! i c o i n c W h - c n — ' f - ' O O - H o — r o i n c n ^ o — \ * ' t o m Cfi , i <x> o*> !*- cj> M> — - a cn *o eft <N; t o ; — r - r - W i n c o i f l N O i ' u i - O r - c n o ^ ' O c J .4- CM, m m ^ H - r r t ( Y ) c s J r ^ ! ^ c o ^ " < » ^ r - < T , o « i ) r ^ ( > o - — h - r - o <r o-> ^ - o j N N N ( N m n S ' 4 - ' T i n , * - a ^ 00 c o <-* CM I n i n ' O O O O O O O C O O O O O O O O O O O O j CM CO CM to 4 " pr\ Pl *Tj >s- 00 en i n — r - J co — — CM m (*i <^ _ m — CM CN ro CM. eg <o i n vtf r - c o o» O CM CM CO M3 ^> —J — ro —. ro —. vri ro -* uV o r- od r - i n — ft O M Mj O CO v*- CO CO U \ v£> -O-LO (T> - 0 t o o — r - H cn —'! CM CO IA' ITJ CM CMi h i n K r - — c r CM cM r - r~ -4* CM M> r- c o <r CM - • t- — *o r- M> >r r~ co — M l 4 " — O N - O CD CO M i CM m M c n <r O co: rv) -st co m i r— o CM;:- i n 0 0 0 0 0 o ! 0 0 — • — ^- ~-r-.04-o-3-o'cMincol*-tnrr» csj^ ,^ cn<rcMJ——cM co <j-r-a»tocn«o!^ so <o m CP CM CM CM <r I — CM o*\ 0 .— CM ro 4-] m so r-' 00 a> O) ^ CM CM CM CM CMl'CM CM CM CM CM CO 1 C^ 'C w C M C M i M m f n ' m M D c o c r , t — — CO VO - T MJ] — cM <T, CM I <J» —• CO CO (O CM CM CO cTij CM CM H> —V CM — CM] — N m ' 4 - i n ^ r - c o o v o —* CM] m M CO t o CO c n r o r n rrv >T 4* Mr m MT s j - 4> Figure 65. Dendrogram of direct grouping of sites (Fraser Ualley). 202 APPENDIX III-3 Figure 66. Spatial grouping of sites in the Fraser Ualley area (direct grouping). ID C ra cn o ^ o ~n ro H D CO D. ID h) ra D m 4 •c CO CO 3 V-1 l - J D ro -u •< CD D. H-CO c+ c+ ra DJ • h) D C "D H-D ID 03 -1) r+ ro •-j CO n c+ • CO D CO 1—1 < cn cn ITEMS GROUPED STEP I J ERROR 1 3 44 14 -.22 18 7 38 4 10 40 16 31 43 39 32 19 24 26 42 _3.0_ 13_ 33 37 JI I 12 2l_ -20 25. 28 IS 27 _3A 29 23 45 17 41 34 35 14 14 . 38 17 1 _3L7_ 7 8 9 10 11 _12_ 13 14 15 16 17 _L8_ 27 24 _ 11-23 7 L l _ 30 33 .40 32 22 „45_ 34 25 .12.. 43 10 5760527 0037479 , 5828600.. .6394844 .7284555 7.38332.7_ ,8122625 .8275537 , 8832550-.8984499 ,4703150 .0376.4.72-. 24 14 37.. 7 4 _8_ 19 20 21. 22 23 _2A_ 13 19 15 29 7 _2.4_ 25 26 27_ 28 29 _3.Q_ 9 24 _. 1-4 5 _L5_ 27 23 39 31 16 _20-37 21 41 35 14 _26_ 28 36 _36 44 6 _2A 1 3 3 3 3 ..3. 3 3 3 3 4 _5_ 5.0382042 5. 4033375 6. 1403084.. 6. 9492912 7.1512632 ...7.56 8494.8 8.0268612 9.2469940 . 9.4410400-9. 5774994 9.8781538 ?. 9.090 ?67. I 31 7 32 7 33 2 34 15 35 4 _3J6. 3 13 11 -19 29 7 -1.8.. 37 2 38 4 . 39 3_ 40 1 41 2 _42 2_ 5 9 _4 3 8 _15_ I I _L4..A8.9.4.L0 *_ 18.544327 * 23.153275 * 23. 869644 «_ 26.498489 * 26.947433 * -2 9... 6.9.9 Q.6 6 *_ 33.369141 * 38.981918 * _40.755966 *_ 46.477249 * 53.586624 * 78.065842 « l_l 1 I I l_l 3= "tl Tl m t—I X l _ J _ L I 1 43 44 42 2 116.85938 175.38330 * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * IN3 • -n H-ID C ro cn CD ^ CT -o ra ra u 0) D-n hi ro D ID za 4 ra < 3 ro H Q - I ) ra CL ro H-CO h) ro # n c+ ID H • C in H' 3 ID D tn c+ ra tn ITEMS ««oufro 1 3 ERROR 1 S 7 71 ' 0.0*7*547~ • » 41 42 0.1010231 • 3 4 5 47 0.195(754 4 57 70 0.2137412 • I 61 t* 0.213317* t 10 1 1 0.2171/857 • 7 2 2 41 O.2AO9 0I3 • J 3 7 45 0.3434701 • 9 I d 40 0.3170712 10 it 43 0.4409142 11 27 31 0.44SOI99 a 12 3 3 3 5 0.4S41359 • 13 26 30 0.4S5863D 14 5 5 38 0.5250948 • 15 3 7 •)> 0.35(9434 I t 25 5 7 0.5JJ34I 1 * 17 17 2 0 0.551(130 • IS 15 13 0.4130095 • 19 il " ts 0.4196513 • 1 0 7 92 0-4801829 • 21 It 72 o-tSioosn • 22. 59 4 7 0-4848943 23 7 3 7 0 . tl05359 * 2V 51 54 0.9041.57 • 25 14 22 0.92287*4 a t 10 45 0.93S37JO • 27 * it 1.I«3](*l 28 5 9 it 1.1940470 • 23 27 28 1.2BI9739 3 0 » 17 1.335751* 31 25 4 9 1.4155445 32 18 n I.4979848 33 59 *i 1.5442443 • 3* 44 <,t 1. 5 C 5 I 9 Z 3 5 5 12 1.4028381) it 27 11 1. 801 2 532 • 37 10 41 1.1409485 34 35 40 2. HOI 170 39 7 5 3 2.2222402 4 bo 13 44 2.3202 8.3 • 41 1* 45 2.47?3539 • W it 39 2.8223214 • 43 5 14 9.432(1(0 44 29 55 3 . 89*5581 45 27 50 3.5304992 * Hi 3 5 J.8573508 • 4/7 25 JS 3.9222440 • te 1 to 4.2077248 • 43 i 15 4.2440329 50 1 2 4.4200931 * 51 21 27 4.4043310 » 51 4 34 5.4130373 • 33 J 18 6.0815926 • 54 19 51 7.9598834 « S S 25 26 8.3127928 54 4 I g.5076471 • 57 7 59 8.3J77047 • 5 3 1 5 9.4383698 • 55 13 S 3 J0.92Z546 t o 21 29 13.392756 • 61 1 3 15.777283 • (I 4 21 14.120602 - » tl 7 I t 19.417419 • f 7 15 24.757843 • tt 4 19 24.690710 ii I S 52 31.8)5960 • tl 7 24 47.3UZ72 • it 1 4 50.797454 • 5 9 a 13 41.471634 • 7 0 i 9 88.344SC2 • 71 i 7 44.23X4] » 11 •~rr I T 34 Sg 54 4 3 T T I IBI T I.I i r T T 10 T 4 5 41 41 kl I * t 2 2 i r i i i.i I I i.i i.i T i I . . i •hzzz I «9 16 62 il, 4 M if ts M 2 6 4S 30 38 7J> 2 » I T T I I I I I I I.I I I I I K I l — i I . . . . - I I \~rr-i i i i i-.i i i i i J3 TJ m CT 1—1 X M I . I I ro • •F-"n • H* ID C hi ro • cn U3 o ro ZT D a D . hi hi H - D N ID D hi 01 3 D-01 D ct- -t) 01 D . f N H -"0 hi ro ra OJ n n c+ ro lO hi D < C ro "D H -•a • ID D ra H ' ra 01 c CD H -ID i t e h s enoupro ISTiP I J EMOR 41 40 42 31 45 52 14 35 I " 2 5 26 O.OIU853 • 2 64 65 0.0169061 • 3 2S 71 0.0205024 • 4 30 57 0.03604-39 • 5 41 42 0.0381461 • 6 _. .» ._ 40 0 . 0 4 1 2 3 2 2 _ . « T 1 2 0.0451813 • S 27 62 0.0522618 • 7 43 0.0712652 • 10 37 40 0.0771865 • 11 52 54 0.0648584 • 12. ._ 44 . 4 8 . . O.0905BI8 • 13 55 64 0.0959529 • 14 58 59 0.0970454. » 16 7 £8 O.I0212SS • 16 45 46 0.1204700 • 17 27 63 0.1349703 < 18 22 . 72. . 0.1350827. __• 19 SI 67 0-1383812 • 20 18 58 0.145*158 < 21 28 29 0.1493723 » 22 11 37 0.1500206 < 23 19 23 0.1606283 • 24 4 . 3 5 . 0.1647170 « 25 30 70 0.1856396 < 26 36 39 0.1899055 > 27 4* 47 0.1899241 ' 28 20 £1 0.2704757 29 31 44 0.2747397 > 30 . 1 0 . _ 11 0.2752930 31 51 53 0. 2834067 32 9 14 0.2862301 < 33 17 16 0.3143137 34 18 55 0.3734803 35 22 49 0.3919274 26 25 . . 30 . 0.4796256 37 45 50 0.4894177 38 17 19 0.6150209 39 4 41 0. 4203402 40 5 12 0.668Z4IO 41 10 27 0.7123370 42 7 36 0.72697Z7 _ 43 34 51 0.8750196 44 20 22 0.9120477 45 ( 33 1.0340042 46 5 13 1.0593872 47 • 20 56 1.I6270O7 48 31. 45 1.1629181 49 1 3 1. 201 1423 50 7 10 1.2984734 31 21 28 1.3254509 52 17 18 1.6717978 53 S 9 1.8880987 54 16 52 2.0587416 SS 4 6 2.1150913 54 15 31 2.6441669 57 1 6 4.1914072 59 17 20 4.3033183 59 4 7 4.3608999 60 16 34 .4 452 0 61 25 38 5.5783936 62 14 21 6.1749907 53 4 15 7.9061717 «« 4 16 11.283966 65 17 25 12.671737 ft 1 49 13.219637 61 4 32 18.626450 6t 4 24 41.743103 63 * 17 47.835602 70 1 8 57.41J434 71 1 4 89.875487 I I I I I I I I I I I I I I I I I I I I I JJ..LJ. I I I I I I I I I I I I I I I I I I I I .L.I.J-1. I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I I l_l I. I I I I I I I I I I I I I I I I I I I I II II. I I I I I I I I I I I I I I I I I .1.1. I i I I I I I I I I I I.I I I I I I I I I. I . I—I i U..I. i _ "i i" T i i i I.* i i I I I. . I I I I L I J l_ I. I 48 54 S3 I I I I l_ l I I F! 23 55 56 30 25 57 26 70 3 71 36 I I I I ..I . J . I I I I. I ° I I I l_l I l _ i_.l_. I_-I I . I I I I I I C I I 1. I.I I II I I I I I I .1—1 I I I I l-l I I I I L l._l I I I I I "1 — I 33 m 11—* i • 206 Figure 70,. Dendrogram analysis. of numerical grouping after factor 207 APPENDIX IV SUPPLEMENTARY DATA ON DENSITY SLICING AND ADDITIVE COLOR VIEWING APPENDIX IV-1: R e s u l t s Df density s l i c i n g Df the 500-600 nm band black and u h i t e image (bare s o i l s u r f a c e ) : P l a t e XX. Enhanced density l e v e l s (500-600 nm band). Table 30 Density c l a s s e s and associated ground c o n d i t i o n s Color category tt Df S i t e % S o i l Moisture % C ex. Ca ppm ex. Mg ppm ex. Na ppm ex. K ppm ex. P ppm Green 203 3.9 2.5 840 132 74 67 4 Blue 204 7.3 1.2 380 118 17 63 3 205 11.9 2.6 640 260 13 62 3 206 5.5 1.5 700 78 8 144 3 Uhite 207 54.9 40.9 2640 400 26 93 6 208 14.9 8.3 4440 235 27 60 1 Pink 209 5.6 1.1 380 204 219 51 4 208 APPENDIX IU-2: R e s u l t s of density s l i c i n g of the 600-700 nm band black and white image (bare s o i l s u r f a c e ) : I P l a t e XXI. Enhanced density l e v e l s (600-700 nm band). Table 31 Density c l a s s e s and associated ground c o n d i t i o n s (black and white 600-700 nm wavelength banoV) Color # of % S o i l % ex. Ca ex. Mg ex. Na ex. K ex. P category S i t e Moisture C ppm ppm ppm ppm ppm Purple 204 7.4 1.2 380 118 17 62 3 206 5.5 1.5 700 78 8 164 3 209 5.6 1.1 380 204 21 51 4 Brown 205 11.9 2.6 640 260 13 62 3 Green 208 14.5 8.3 4440 235 17 59 1 Dark 207 54.9 40.9 2640 401 27 94 6 blue 209 APPENDIX IV/-3: R e s u l t s of density s l i c i n g Df the 600-700 nm band black and white image (vegetated s u r f a c e ) . P l a t e XXII. Enhanced density l e v e l s . Table 32 Density c l a s s e s and associated ground c o n d i t i o n s C o l o r # of % S o i l % ex. Ca ex. Mg ex. Na ex. K ex. P category S i t e Moisture C ppm ppm ppm ppm ppm Dark 214 28.8 2.7 760 86 4.8 183 3 green 216 10.5 4.3 740 164 7.6 203 3 219 4.3 2.9 840 161 3.2 66 4 221 2.7 2.1 840 114 2.5 179 3 L i g h t 215 6.4 3.6 1160 120 3.5 215 3 green 220 7.5 2.9 2780 68 3.9 66 3 Grey 218 7.6 2.7 1200 170 3.9 74 3 Black 217 48.3 10.4 1980 134 9.9 238 3 222 13.8 8.9 1500 229 3.9 203 5 1 210 APPENDIX IV-4: Results of density slicing of the 500-600 nm band black and white image (vegetated surface): Plate XXIII. Enhanced density levels. Table 33 Density classes and associated ground conditions Color # of % S o i l % ex. Ca ex. Mg ex. Na ex. K ex. P category Site Moisture C ppm ppm ppm ppm ppm Dark 217 48.3 10.4 1980 134 9.9 238 3 blue 222 13.8 8.9 1500 229 3.9 203 5 Light 220 7.5 2.9 2780 68 3.9 66 3 blue 219 4.3 2.4 840 161 3.2 66 4 Yellow 221 2.7 2.1 840 114 2.5 179 3 216 10.5 4.3 740 164 7.6 203 3 214 28.8 2.7 760 86 4.8 183 3 Green 215 6.4 3.6 1160 120 3.5 215 3 218 7.6 2.7 1200 170 3.9 74 3 211 APPENDIX IU-5: Results of density slicing Df the color image (vegetated surface): Plate XXIV/. Enhanced density level. Table 34 Density classes and associated ground conditions Color category # of Site % S o i l Moisture 0/ /a c ex. Ca ppm ex. Mg ppm ex. Na ppm ex. K ppm ex. P ppm White 219 4.3 2.4 840 161 3.2 66 4 215 6.4 3.6 1160 120 3.5 215 3 220 7.5 2.9 2780 68 3.9 66 3 Light 218 7.6 2.7 1200 170 3.9 74 3 blue- 221 2.7 2.1 840 114 2.5 179 3 green Pink 216 10.5 4.3 740 164 7.6 203 3 214 28.8 2.7 760 86 4.8 183 3 Black 217 48.3 10.4 1980 134 9.9 238 3 222 13.8 8.9 1500 229 3.9 203 5 212 APPENDIX IU-6: Ranking tables for density slicing and COICT additive images discussed in Chapter Five. Tables 18-20 and 22 (p. 110-118) mere ranked to determine the trends and relationships amongst parameters. Rank 1 refers to highest chemical concentration, while rank k refers to lowest levels. Table 35 Chemical ranks for density levels from color image (vegetated surface) Color category % Boil Moisture % C ex. Ca ppm ex. Mg ppm ex. Na ppm ex. K ppm Light blue 2 2 1 3 2 3 Yellow 3 3 3 2 3 2 Dark blue and 1 1 2 1 1 1 purple Table 36 Chemical ranks from additive color image (vegetated surface) Color % Soil % ex. Ca ex. Mg ex. Na ex. K category Moisture C ppm ppm ppm ppm Very dark red 1 1 1 2 1 1 Dark red 2 2 3 1 2 2 Purple 3 3 2 3 3 White 4 k k k 4 3 213 APPENDIX 11/-6 (cont'd) Table 37 Chemical ranks for density levels from color image (non-vegetated surface) Color % Soil % ex. C ex. Mg ex. Na ex. H Category Moisture C ppm ppm ppm ppm Green 2 2 2 1 2 1 Light blue- if 3 it 3 2 green Yellou-broun 3 3 3 3 Purple 1 11 1 2 1 if Table 36 Chemical ranks for density levels from color IR-image (non-vegetated surface) Color % S o i l % ex. C ex. Mg ex. Na ex. K Category Moisture C ppm ppm ppm ppm Blue 2 2 2 2 3 1 Light blue if 3 3 3 if if White 3 if if if 2 2 Purple 1 1 1 1 1 3 Table 39 Chemical ranks from additive color image (non-vegetated surface) Color Categories % Soil Moisture % C ex. C ppm ex. Mg ppm ex. Na ppm ex. H ppm Violet 2 2 2 2 2 1 Light violet it 3 3 3 if 3 Red orange 3 it if if 3 2 •range 1 1 ' 1 1 1 it 214 Appendix V : Supplementary Data on Chemical Predictions from , Spectral Reflection Measurements. Figure 71 a + b : Predictions from Airborne Measurements.. 0.8000 1 *„Iron_£ in fom of orgetilc cow^Lxeo ) ve. Reflection pt 700 nr Al - 0.8059 . -1.181 . A* 0.6400 FRATlolCOEFF.I- u.9| FPR08CC0EFF.)» 0.0019 STD.ERR.CONST.= 0.1410 STO.ERR.COEFF0.9219 0.4800 1 > i i STD.ERR. Al - 0.(648 " R50 ' 0.306 1 BURblN-KATJOrj STA.* 1.904 AUTOCORRELATION COEFf. » 0.4751E-01 r i i« n • u i SS 0.3200 i . i i 2 I I 1 ".1600 _ i i 1 • -0.0 1 1 1 -1 — 0.8409E-01 o. 1400 0. 1120 OBSERVED PREDICTED RESI0UAL OBSERVED PRE01CTE0 RE5IDUAL 1. 0.4000 0.42 54 -0.2537E-01 11- a. icoo 0.21157 -0. 1857 21. 2. 0.4000 0.4101 -0.101OE-01 12. 0.5C00 0.3169 0.183! 22. I. 0.4000 0.4721 -0.7213E-01 13. 0.2000 0.2625 -0.624aE-01 23. * . 0.4000 0.5089 0.9|;7E-01 14. 0.4000 0.3013 0.9371E-01 24. 0.3009 £.2625 0.3752E-01 15. 0.0 0.278 1 -0.2781 25. *. 0.5000 0.2937 C.2C63 16. 0.0 0.2393 -0.2393 26. T. 0.'000 0.2701 0.1299 17. CO 0.3996 -0.3 996 27. f . O.J 000 0.2625 -0.6248E-OI 1>. 0.2000 0.1241 0.7591E-01 28. 0.3000 0.2937 0.6344E-02 19. 0.2C00 0.1447 O.5332E-01 29. IS. 0.3000 0.2701 0.2989E-01 20. 0.2000 0.9418E-01 0.1058 0, 1680 OBSERVED 0.8000 0.5000 0.5000 0.6000 0.0 0.5000 0.5000 0.4O00 0.4000 0.1960 PREO1CTED-0.4737 0.3264 0.4489 0.4966 _ 0.3o<.9 0.3719" 0.4333 0.3 789 0.3856 | %fiP=L. _0.2240 _ •_ RESIOUAL 0.3263 -0.2876E-01 0.5109E-01 0.103» _;0.3649 0.1281 0.4668E-01 0.2106E-01 0.1440E-01 % Cftrton va.^ prctraljRtf^ tlon^ C^fX^ rw^ avelength^ _A.l- -240.9 « AT FR»T10(C0EFF.I-FPROBICOEFF.I" STO.ERR.CONST.-STO.ERR.COCFF.-_itP.tRR.-Al _-_ RS9 19.80 0.0002 7.234 54.14 9.266 .4231 »«RBJN-W»T30N STA.- 1.220 AUTOCORRELATION COGFF. * 0.3591 II 1 t JI • Z 2 1* _0B5eRVEP_ 4.400 3.830 5.500 2.400 3.200 2.800 1. 3.IOO I . 1.500 %. 2.000 10. 2.400 -«1 _PREDICTED-1J.97 12.78 16.90 4.545 1.606 5,96 7 ~ 2.'7a7~ 1 .606 3.9(.7 1.606 RCS10UAI ~ "-V.565" -8.985 -11.40 " -2.145 1.594 -I.147_ 0.4133" -0. 1062 -1.967 0.7938 13. 13. 14. 15. 16. 17; 18. 19. 20. OBSERVED ~" i.'fOO" 2.600 1.6 00 2.100 4.800 4.100 —6\7oo"~ i.ooo 1.500 1.000 1—• 0 O.TIOOC-Ol 0.1170 0.94O0C-01 0.1400 PRE01CTEO RESIDUAL 085£R<C0 — "3.967 -2.167 46.30 4.3«5 -1.945 22. 21.«0 1.028 C.5720 25. 4.545 -2.445 24. 2.787 2.Oil 25. 0.4498 3.650 26. 14.71 -8.012 27. -3.887 4.887 28. -2.176 3.676 29. -5.573 6.5 73 0.1630 \%Wl-215 Appendix V cont. : Figure 71 c + d: Predictions from Airborne Measurements. gxcharigttibiQ ffg vs. Rafjeetlun fit 750 W*v»len(|tl> ; _ .»»..«. 9)37.7 .t.zW- .* .*!" FR«T10(COEFf.1" 42.35 fpuoeccotff.l' o.oooo STD.ERR.CONST." 120.6 STO.ERR.COEFF.* 761.7 STP.ERR. A3 "_ 127.1 r s e « 0.6107 OURBIN-UATSON STA." AUTOCORUELi!I ON COEFF 470.0 1 I j - >. 1 . 190 0. -^ "sio.o i 11. I — 0.92 DOE OBSERVE p_ 1. 166.0 2. 207.0 3. 277.0 " ». 126.0 S. 76.00 «. 85.00 7. 64.00 6. 65.00 ». 78.00 10. 66.00 PRED1 CTEO_ 308.7 28*. 4 393.4 151.5 . 103.0 127.2 "42.48 91.06 127.2 103.0 RESIOUU - - - - 1 6 4 - 7 -2.614 -116.4 -23.54 -26.96 -42.25 21.52' -26.06 1 3 1 4 15. 1 6 oesesvEo PREDICTED 103.0 — 127.2 _ 185.0 163.4 78.00 41 .06 69.00 139.6 27.00 103.0 55.00 54.87 ; 116.0 Z40.3 3.000 -136.5 4.000 -101 .3 2.000 -183.1 01 0.1190 RESIDUAL — -24.25 ~ 21.57 -13.06 -70.64 -75.96 0.1274 - 1 2 4 . 3 -139.5 105.3 185.1 0.1460 0.1730 OBSERVED 420.0 630.'0 610.0 790.0 10.00 —21\ 72. 23. 2 4 . 25. 26. ~'?.T.~ ie. 29. 1 1 l„, \%eefL 0.2000 0.227O PREDICTED RESIDUAL 'UTit 148.8 245.3 367.3 -144.5 199.0 76.00 176.0 118.0 340.9" '481.2 364-7 422.7 154.5 186.2 - '282."3" 197.1 208.0 77.24-21.14 -90.05 6".300 Appendix V : Figure !2 a + b Predictions from Ground Measurements. 216 - nn>— ]M6 c a v s ' r f c . r - ' f FRATIO r p n o e - 7 0 EKK STD ERR STD t«R R'SQ VIA VAH A e ( 8 ) i B . l / i l < B ) ( 7 ) K 3 2 0 . 0 6 0 0 3 7 2 . 6 - 1 3 . 4 9 4 4 . 4 2 0 . 0 0 0 0 1 40. 'ii 2 . 0 2 3 _ _ 6 4 . 66 0 . 6 2 2 0 / N O . 0 8 3 E K V E O CALCUL AT.O R E S I D U A L / 1 1 . 7 5 . 0 0 0 1 6 4 . 6 7 - 8 9 . 8 7 1 / 2 . 7 7 . 0 0 0 1 3 7 . 9 0 - 6 0 . 8 9 8 / 1 I 3. 9 2 . 0 0 0 1 7 2 . 9 6 - 8 0 . 9 6 3 / . 4. 6 2 . 0 3 0 44.8 4 3 1 7 . 1 5 7 / . 1 I s . 6 5 . 0 0 0 6 9 . 1 1 9 - 4 . 1 184" 1 6 . 6 1 . 0 0 0 6 6 . 4 2 1 - 5 . 4 1 1 1 / . V . 6 3 . 0 0 0 4 f i . 8 8 9 1 4 . 1 1 1 / e . 5 2 . 0 0 0 6 . 4 2 9 9 4 3 . 5 7 0 2 6 0.0 9 . 4 3 . 0 0 0 - 9 . 1 1 0 - 2 6 . 1 1 8 / 1 0 . 6 0 . 0 0 0 <-.4_ l - 6 . 4 2 1 1 / L 1 1 ' . 4 7 . 0 0 0 ' " " 4 2 . 3 7 5 - 1 5 . 3 7 5 " / 12. 3 0 . 0 0 0 - 4 5 . 5 1 5 7 5 . 5 1 5 / 1 3 . 4 7 . 0 0 0 2 1 . 9 1 6 2 5 . 0 8 4 / 14. 5 4 . 0 0 0 1 0 2 . 8 3 - 4 8 . 8 3 4 / 15. 6 S . 0 0 0 7 0 . 4 6 7 - 1 . 4 6 7 0 / 16 . 1 1 2 . 0 0 5 6 . 9 8 1 5 5 . 0 1 9 / I T ; A | . 0 0 0 I 2 5 . T 6 - 6 4 . 7 6 1 / t e . 3 7 . 0 0 0 9 6 . 0 9 1 - 5 9 . 0 9 1 2 0 0 . 0 1 9 . 2 5 . 0 0 0 7 9 . 9 0 7 - 5 4 . 9 0 7 / 20. 4 6 .COO 9 6 . 0 9 1 - 5 0 . 0 9 1 / 21 . 2 5 . 0 0 0 8 6 . 3 0 2 - 6 0 . 3 0 2 / 22 . 9 7 . 0 0 0 1 4 3 . 2 9 - 4 6 . 2 9 3 / 2 3 . " 2 8 3 . 0 0 ' 1 5 9 . 4 8 123 . 5 2 / 2 4 . 2 9 3 . 0 0 1 4 3 . 2 9 1 4 9 . 7 1 / 2 3 . 5 2 0 . 0 0 2 0 3 . 9 8 1 1 6 . 0 2 / 2 6 . 2 8 5 . 0 0 2 2 4 . 2 1 *" 6 0 . 7 9 0 / 2 7 . 2 5 0 . 0 0 288 . 9 4 - 3 8 . 9 4 5 / 2 8 . 2 5 8 . C O 2 5 7.93 4 0 . 0 7 4 i * d T B r ~ \ .... "29. " ~ _ 0 S 7 C 0 1 1 9 1.64 137358" 8 0 . 0 0 ', i i i t • ', 2 t . 11 1 1 1 1 ', i 1 1 I 1 1 1 2 0 . 0 0 -//I t l t t U U t \itltttttt 1 / / / / / / / / / 1 / / / / / / / / / 1 / / / / / / / / / / / / / / / / / / 1 / / / / / / / / y | / / / / / / / / / i / / / / / / / / / . 6.000 11 . 0 0 10.00 2 1 . 0 0 26.00 3 1 . % R£FLECTIOP4 AT keo,~ DER VAR M9 T s o r o IND VAR 1000 T O M ST COCFF 8 -31.34 sro EAR .A| STD ERR 181 S ' 10 £ W » i T l 1 3 3 . 7 4 . 1 6 0 1 1 7 . 4 0 . 6 6 9 6 NO. 0 5 S E R V I D CALCULATED R I S I D - A L 1 . 1 6 6 . 0 0 2 3 8 . 2 2 - 7 4 . 2 1 5 2. 2 « 7 . 0 0 1 5 3 . ( 0 1 3 3 . 4 0 3. 1 7 7 . 0 0 2 4 4 . 4 8 3 2.518 4. 126.CO 56.45 5 6 9 . 5 4 5 5. 7 6 . 0 0 0 " 112.e» - 3 6 . 8 6 4 " " " «. 85.CCO I S A . 6 0 - 4 9 . 6 0 0 7 . 6 4 . 6 0 0 4 0 . 7 6 6 2 3 . 2 1 4 ». 6 5 . 0 0 0 21 . 9 5 3 4 3 . 5 1 7 1. 7 8 . C O Q 1 1 2 . 8 6 - 3 4 . 8 6 4 10. 6 6 . C C O 1 2 8 . 5 3 - e ; . 5 ! 2 11 . 1 0 3 . 0 0 ~ Hi. 74 •Si.TH 1 2 . 185.00 - 6 e .5*. : s 3 . J O 11 . 78.CCO *:.;.'! 1 5 . 1 T 7 14. 6 8 . 0 0 0 1 9 7 . ' « - 1 2 8 . 4 8 15. 27.COO 14! . 0 7 - 1 : 4.07 14. sr.coo 6 5 . 9.', / - I S . . 5 7 17. 5 .0000 9 7 . 1 9 5 ' - 9 . . 1 9 5 11. 6.C0O0 25.1 1 7 - 1 9 . 117 19. J.occo - 6 . ' . 6. ' 3 4 5 . 6 2 9 3 0 . 4 .COCO I E . 8 5 0 - 1 4 . 8 5 0 2 1 . 2 . 0 000 5 9 . 5 8 9 - 3 7 . 5 6 9 2 2 . 1 0 .CCO 1 6 3.00 - 1 5 3 . 0 0 2 3 . 7 6 . 0 0 0 '25. I i 7 5 0 . F - 3 ' 2 4 . 118.00 3 1 . 3 1 5 8 6 . 6 1 5 t i . 1 1 6 . 0 0 2 0 3 . 7 4 - 8 7 . 7 4 3 I t . 6 3 0 . 0 0 4 7 » . 5 2 1 5 0 . 4 6 2 7 . 8 1 0 . 0 0 5 6 7 . 2 6 4 2 . 7 3 7 2 8 . 7 9 0 . 0 0 4 6 J . 6 V 3 2 6.15 2 9 . """' 4 2 0 . 0 0 - • 4 1 6.92 - 6 8 . 9 1 8 , 7 - ' ° 2 2 . 0 0 2 7 . 0 3 J I . 0 0 3 7 . 0 S L A S H E S CA IME i l A « I S IS U . . 5 - v _ _ Z _ _ _ _ _ _ '% TZFUCTICH AT I O O O « « Appendix V : Figure 72 c + d: Predictions from Ground Measurements. D E C 2ND VAR VAR F E 7 0 0 TME *.u AND C O N S T COEFF FRAT 10 A B (B) 0 . 7 2 3 1 - 0 . 1 4 6 7 E - 0 1 1 6 . 3 5 AF i U S t L TC H . U THE R E C i R E S i l G N L l . \ t O.8000 -f FR06 (31 0.0006 NO. 2. 3. 4. 5. ' 14. 19. 20. "21 23. 24. 26. 27. 25. S T D ERR IA| 0 . 8 8 4 9 E - 0 1 OBSERVED 0 . 4 0 0 0 0 0 . 4 0 0 0 0 0 . 4 C O 0 0 . 0 . 4 0 0 O O 0 . 3 C O C 0 0 . 5 C 0 0 0 0 . 4 0 0 0 0 0 . 2 0 0 0 0 0 . 3 0 0 0 0 0 . 3 0 0 0 0 0. I0O0O 0 . 5 0 0 0 0 0 . 2 0 0 0 0 0 . 4 0 0 0 0 0 . 2 0 O 0 G 0. 2 0 0 0 0 " 0 . 2 0 0 0 0 0 . 5 0 0 0 0 0 . 4 0 0 0 0 0 . 5 0 0 0 0 0 . 5 0 0 0 0 0 . 6 0 0 0 0 " 0 . 6 0 0 0 0 " STD ERR ( B l 0 . 3 6 5 6 E - 0 2 5TO ERR 1 VI 0 . 1 2 2 0 C A L C U I A T E O 0 . 4 4 0 4 7 0 . 4 0 4 7 7 0 . 4 5 0 8 8 0 . 3 0 6 6 1 " 0 . 3 3 3 3 8 0 . 3 2 5 9 4 0 . 3 1 1 0 7 0 . 2 6 6 4 5 0.3 2 4 4 5 0 . 3 2 6 9 2 " 0 . 3 1 8 5 1 0 . 2 0 6 9 5 0 . 2 8 2 8 1 0 . 3 6 3 1 3 0 . 3 0 2 1 4 0 . 3 3 9 3 3 " 0 . 3 3 0 4 0 ' 0 . 3 9 8 B 2 0 . 3 8 0 9 6 0 . 5 2 9 7 1 0 . 6 0 2 6 0 0 . 5 5 6 4 1 0 . 3 9 5 1 6 ~ 1 R E S I O U A L - 0 . 4 0 4 7 I E - 0 1 - 0 . 4 7 7 J 6 E - 0 2 - 0 . 5 0 8 8 2 E - 0 1 0 . 9 3 3 9 4 E - 0 1 """-0 . 3 3 3 7 9 E - O l " 0 . 1 7 4 0 6 0 . 8 8 9 3 2 E - C 1 - 0 . 6 6 4 4 7 E - U 1 - 0 . 2 4 4 5 4 E - 0 1 _ - 0 . ? 8 3 l 7 E - O I _ - 0 . 2 1 8 5 1 0 . 2 9 3 0 5 - O . 6 2 8 0 8 c - 0 l 0 . 3 6 6 7 3 E - 0 1 - 0 . 1 0 2 1 4 _ - 0 . 1 3 9 3 3 - 0 . 1 3 0 4 0 0 . 1 0 1 1 8 0 . I 9 0 2 S E - 0 1 - 0 . 2 9 7 1 4 E - 0 1 - 0 . 1 0 2 6 0 0 . 4 3 5 1 3 1 : - 0 1 " " 0 . 2 0 4 8 4 / / 1 u II i m i Milium I II 11 in II i II II in/i 11 mn un i u 11111 m 11111 u i1immii \u m mm m i mt\ 5.000 11.00 17.00 23.00 21.00 25. %tCFLtCTIOM Ar 700 » m 0 ERR S 1 0 OR STO ERR f Sw IAI 181 I 1 1 1 . 6 * 4 C . 1964 6 . 484 0 . 7 » 3 3 0 6 S I P V E D CAICI'IATED RES ICM4L 4 . 4 0 : 0 1 > . 9 7 0 - 9 . 5 4 9 4 3.0000 9 . 1 4 . - C - 5 . 5 4 2 0 5 . J C 6 0 1 4 . 7 1 0 - 9 . 1 1 0 0 1 . 4 C 0 O 1 . 0 1 2 ) 1 . 1 8 7 7 " ~ " J . 2 C 0 O ~ ' ' 6 - . J 3 9 S ' • ! . « « 2 . 8 P 7 0 • . 5 2 7 ) - 1 . 7 2 5 5 1 . 1 0 0 0 - O . 9 4 3 S 6 E - 0 1 3 . 1 8 6 4 1 . 5 0 0 0 - 2 . 5 0 4 7 4 . 0 0 4 7 2.OCCO 3 . < ^ 3 7 -i.«c»; 2 . 4 0 00 4 . 8 S 5 5 - 2 . 4 J J S " " 1 . 1 0 0 0 '- ' f . 2 4 5 7 "' - 3 . 4 6 1 7 " 2 . 4 0 C O - « . 5 8 3 4 t i . 5 8 3 t . 6 0 0 0 0.S27I6 0. 7 7 2 8 4 2 . 1 0 7 4 t.te-i - 4 . 5 0 1 S 4 . S C 0 0 4 . 7 1 » 1 0 . 6 ) 4 4 : - -4 . 1 C C O 1 . 1 * 7 4 I . 9 C 2 6 1 . 4 0 C O " 5 . 4 S i S " - 2 . 0 5 4 S O . 7 C C 0 O 1 . 3 8 2 3 - 0 . 4 8 2 4 6 1 . 0 0 0 0 - 3 . 9 8 5 6 4 . 4 8 5 6 1 . 5 0 0 0 0 . 2 7 1 8 4 1 . 2 2 i ? I . 0 0 0 0 1 . 5 4 7 6 - -O . S 4 7 J 7 4 . 4 C 0 0 9 . 7 1 2 2 - 5 . 3 1 2 2 4 . ( 0 0 0 3 . 6 0 3 7 1 . 2 9 4 3 6 . 5 0 O O 3 . C . 4 8 4 3 . 4 5 1 6 6 . 7 0 0 0 1 4 . 1 5 5 - 7 . 4 5 4 7 1 1 . 4 0 0 2 7 . ( 5 2 - 6 . 4 5 2 4 40 . 1 0 0 3 4 . 5 1 6 5 . 5 8 3 8 55.400 2 7 . 6 6 7 7 . 7 3 2 7 4 4 i l 0 0 " 5 0 . 2 5 9 "14. 0 4 r OEP 1*0 VAR VAR CARBON 9 0 0 COEFF 8 -1.851 FRAT10 181 67.07 F R R 0 8 181 —*7T6o~ "I7T00"" I 1 111 t 11 1 i l l 1 t l . t 1 IU 11111: ini 1111111 ni 111111 II\I 111111 im 1111111 m 1 in in 1 ii 11111111\ 111 m 111 \ imimiMiimim 12.00 17.00 22.00 27.00 32.00 37.oj % R f F i r c 7 1 0 M AT 300 ~ . 218 BIBLIOGRAPHY ABLER, R., ADAMS, J . and GOULD P. 1971 S p a t i a l o r i e n t a t i o n . P rent i ce H a l l , Inc. p. 149-189. . ACTON, D.F. 1965 The re l a t i on sh ip of pattern and gradient of s lopes to s o i l type. Canadian Journa l of S o i l Sc ience, V o l . 45, p. 96-101. ACTON, D.F. and STONEHDUSE, H. 1972 Experimental program on the a p p l i c a -t i o n of remote sensing f o r s o i l survey i n Saskatchewan. 1st Canadian Svmp. on Remote Sensing, Ottawa. Dept. of Energy, Mines and Resources, p. 295-303. ADAMS, J .A . 1968 Lunar and Martian surfaces: Pe t ro log i c s i gn i f i c ance of absorpt ion bands i n the n e a r - i n f r a r e d . Science, V o l . 159. p. 1453-1455. ADAMS, J . S . and FILICE, A. 1967 Spect ra l re f l ec tance 0.4 to 2.0 microns D f s i l i c a t e rock powder. Journal of Geoph. Research, VDI . 72, p. 5705-5715. ADAMS, J .A . and UJALKER, T.W. 1975 Some proper t ies of a chrono-toposequence of s o i l s from granite i n New Zealand. Forms and amount of phosph. Geoderma 13, p. 41-51. AL ABBAS, A .H. , BARR, R., HALL, J . D . , CRANE, F . L . , BAUMGARDNER, M.F. 1974 Spectra of normal and nut r ient d e f i c i e n t maize l eaves . Agronomy Journa l , V o l . 66, p. 16-20. ALDRICH, R.C. 1971 Space photo f o r land use and f o r e s t r y . Photoqr. Eng., V o l . 37 No. 4, p. 389-401. ALDRICH, R.C. 1966 Forest app l i ca t i ons of 70 nm c o l o r . Photogr. Eng., V o l . 32, p. 802-810. ALLAN, U. and SEUELL, J . 1973 Remote sensing of fa l low s o i l moisture by photography and i n f r a red f i n e scanner. Tran. Am. Soc. of Agr. Eng., V o l . 16, p. 700-706. ALLISON, L .E . , BOLLEN, Id.B. and MOODIE, C D . 1965 T o t a l Carbon, i n : Methods of S o i l Ana ly s i s , Part II (B lack, C A . ed . ) , Agronomy 9, p. 1346-1366. American Soc. Agron., Madison Wise. ANDERSON, D.H. and HAUKES, H.E. 1958 Re la t i ve mob i l i t y of the common elements i n weathering of sch i s t and grani te areas. Geochimica et Cosmochimica Acta 14, p. 204-210. ANDERSON, R. 1969 The parameters o f co l o r reproduct ion i n add i t i ve co lo r a e r i a l photography. Seminar on New Horizons i n Color A e r i a l Photo- graphy , The Coliseum, New York. A.S.P. p. 107-114. 219 ANDERSON, R. and U10BBER, F. 1973 Wetland mapping i n Neu Jersey . Photoqr. Eng., V o l . 39 No. 4, p. 353-358. ANSON, A. 1970 Color a e r i a l photography i n the reconnaissance of s o i l and rocks . Photoqr. Eng., V o l . 36 No. 4, p. 343-354. ANUTA, P., KRISTOF, S., LEVANDOWSKI, D., PHILLIPS, T . , MACDONALD, R. 1971 Crop, s o i l and geo log ica l mapping from d i g i t i z e d mu l t i s pec t r a l s a t -e l l i t e photography. 7th In tern . Sym. on Remote Sensing of Env i r on - ment, Ann Arbor, Michigan, p. 1983-2016. AREOLA, 0. 1974 Photo - in te rpre ta t i on of land facet s as a s o i l mapping technique. Geoforum 20, p. 25-38. ARMSTRONG, J . E . 1957 S u r f i c i a l geology of Neu Westminster area, B.C. Geo l . Surv. Can., Paper 59-9, p. 1-27. BAKER, R., and SCOTT, M. 1975 E l e c t r o - o p t i c a l remote sensors u i t h r e l a t e d o p t i c a l sensors, p. 325-366, i n : Manual of Remote Sensing (Reeves, R., ed . ) , The Am. Soc. of Photogrammetry, F a l l s Church, V i r g i n i a , 2144 pp. BAUMGARDNER, M.F., KRISTOF, S . J . , JOHANNSEN, C . J . , ZACHARY, A .L . 1970 E f f e c t s of organic matter on the mu l t i s pec t r a l p roper t ie s of s o i l s . Proc . of the Indiana Acad, of Science 1969, V o l . 79, p. 413-422. BAZILEVICH, N. and RODIN, L. 1971 Geographical r e g u l a r i t i e s i n p roduc t i v i t y and the c i r c u l a t i o n of chem. elements i n the E a r t h ' s main vegetat ion types. Sov iet Geoqr., V o l . 12, p. 24-53. BECKETT, P.H.T. 1967 L a t e r a l changes i n s o i l v a r i a b i l i t y . Journ. Aust . Ins t . A g r i c . S c i . , V o l . 33, p. 172-179. BECKETT, P.H.T. 1968 S o i l formation and slope development. Z e i t s c h . Geomorph., VDI . 12, p. 1-24. BECKETT, P.H.T. and WEBSTER, R. 1965 A c l a s s i f i c a t i o n system for t e r r a i n . M.E.X.E. Report # 872, p. 247; # 873, p. 142; # 874, p. 39. BECKETT, P. and WEBSTER, R. 1969 A rev ieu of s tudies on t e r r a i n eva luat ion by the Oxford-M.E.X.E.-Cambridge Group 1960-69. M i l i t a r y Engineering Experimental Establishment Chr i s tchurch, Report #1123, Dept. of A g r i c . Sc ience, Oxford. BECKETT, P.H.T. and WEBSTER, R. 1971 S o i l v a r i a b i l i t y : a r e v i e u . S o i l s and F e r t i l i z e r s , V o l . 34, no. 1, p. 1-14. BECKETT, P.H.T., WEBSTER, R., MCNEIL, G., MITCHELL, C. 1972 T e r r a i n eva luat ion by means of a data.bank. Geoqr. Jou rn . , V o l . 138, p. 430-456. 220 BEHE, G. 1972 Evaluation of multiband data f o r resources and land use inve n t o r i e s . 1st Canadian Svmp. on Remote Sensing, Ottawa. Dept. of Energy Mines and Resources, p. 305-316. BERRY, B. 1958 A note concerning methods of c l a s s i f i c a t i o n s . Ann. Ass. Am. Geoq., V o l . 48, p. 300-303. BETTENAY, E. 1968 Geom. and s o i l d i s t r i b u t i o n . Transact. 9th Intern. Conqr. S o i l S c i . , Adelaide, p. 615-621. BIRKELAIMD, P. 1974 Pedology, Weathering and Geomorphological Research. Oxford Univ. Press, London, 285 pp. BLACK, C A . (ed.) 1965 Method of S o i l A n a l y s i s , V o l . 2, p. 891-896. Am. Soc. of Agronomy Inc., Madison, Wisconsin. BLANCHARD, M., GREELEY, R., G0ETTELMAN, R. 1974 Use of v i s i b l e near IR and thermal IR remote sensing to study s o i l moisture. Proc. of the 9th Intern. Svmp. on R.S. of the Environment, p. 693-700. BL0NG, R.J. 1975 H i l l s l o p e morphometry and c l a s s i f i c a t i o n : a Neu Zealand example. Z e i t s . Geomorphology N.F. 19 # 4, p. 405-429. de BOER, T. f BUNNIK, IM., VAN KASTERNEN, H., UENK, D., VERHDEF, W., L0DR, G. 1974 Investigation i n t o s p e c t r a l signature of a g r i c u l t u r a l crops during state of growth. Proc. of the 9th Intern. Symp. on R.S. of the Environment, p. 1441-1455. B0RMANN, F.H. and LIKENS, G.E. 1969 The watershed-ecosystem concept and studies of nutrient cycles, in: The Ecosystem Concept i n Natural Resources Management (Van Dyne, G., ed.), Academic Press, N.Y., p. 44-76. BOURNE, R. 1931 Regional survey and i t s r e l a t i o n to stock taking of the a g r i c u l t u r a l resources of the B r i t i s h Empire. Oxford Memoirs 13. BOWERS, S.A. and HANKS, R.J. 1965 R e f l e c t i o n of radient energy from s o i l s . S o i l S c i . , V o l . 100 #2, p. 130-138. BRACH, E. and MACK, A. 1972 Differences i n reflectance properties of diseased plants grown under d i f f e r e n t environments. 1st Canadian Symp. on Remote Sensing, Ottawa, Dept. of Energy, Mines and Resources, p. 103-110. BRENNAN, P. and LINTZ, J . 1971 Remote sensing of some sedimentary rocks. Proc. of the 7th Intern. Symp. on R.S. of the Environment, p. 253-267. BROONER, W.C and SIMONETTE, D.S. 1971 Crop di s c r i m i n a t i o n with c o l o r IR photography. Remote Sensing of the Environment, V o l . 2, p. 21-35. BUCKMAN, H. and BRADY, W. 1969 The Nature and Properties of S o i l s . MacMillan, London, p. 96-104. 221 BUNTING, B.T. 1961 The role of seepage moisture in s o i l formation, slope development and stream i n i t i a t i o n . Am. Journ. of Sci., Vol. 259, p. 503-518. BUNTING, 8.T. 1964 Slope development and s o i l formation on some Bri t i s h sandstones. Geog. Journ., Vol. 130, p. 73-79. BUNTING, B.T. 1965 The Geography of Soils. Hutchinson Ltd., London, 213 pp. BUOL, S., HOLE, F., McCRACKEN, R. 1973 S o i l Genesis and Classification. The Iowa State Univ. Press, Ames, 360 pp. BYRNE, R. and MUNDAY, J. 1972 Photo density and the impact of shifting agriculture on subtropical vegetation: A case study in the Bahamas. Proc. of the 8th Intern. Symp. on Remote Sensing of the Environment, p. 1311-1326. CAMPBELL, I.B. 1973 Pattern of variation in steepland sails: variation on a single slope. Neu Zealand Journ. of Science, 16, p. 413-434. CARLSON, R., YORGER, D., SHAW, R. 1971 Factors affecting the spectral properties of leaves uiith special emphasis on leaf uater status. Agron. Journ., VDI. 63, p. 486-489. CARROLL, D. 1973 Remote sensing techniques and their application to s o i l science. Soils and F e r t i l i z e r s , Vol. 36, No. 7, p. 259-266. CATTEL, B. 1965 Factor analysis; introduction to essentials. Biometrics, Vol. 21, p. 190-215 and p. 405-435. CHESWORTH, W. 1973 The parent rock effect in the genesis of s a i l . Geoderma 10, p. 215-225. CHRISTIAN, CS. and STEWART, CA. 1968 Methodology of integrated Survey, in: Aerial Surveys and Integrated Studies. Proc. Toulouse Conf. 1964; UNESCO, Paris, p. 233-280. CIHLAR, J. and PROTZ, R. 1972 Color film densities for soi l s P.I. Photoqr. Eng., Vol. 38, No. 12, p. 1091-1098. CIPRA, J.E. 1973 Mapping s o i l associations, using ERTS MSS data. LARS Info. Note 101773. Laboratory for Applications of Remote Sensing, Purdue Univ., West Lafayette, p. 3A 1-10. CIPRA, J., BAUMGARDNER, M., ST0NER, E., MACDONALD, R. 1971 Measuring radiance characteristics of s o i l uith a f i e l d spectroradiometer. Soil Sci. Soc. Am. P r o c , Vol. 35, p. 1014-1017. 222 CIPRA, J . , SWAIN, P., GILL, 3., BAUMGARDNER, M., HRISTOF, S.J. 1972 Definition of spectrally separable classes for s o i l survey research. Proc. of 8th Intern. Symp. on R.S. of the Environment, p. 765-769. CLEAVES, E.T., GODFREY, A.E., BRICKER, O.P. 1970 Geochemical balance of a small watershed and i t s geomorphic implications. B u l l e t i n of the Geological Soc. of America 81, p. 3015-3032. CLINE, M.G. 1949 Basic p r i n c i p l e s of s o i l c l a s s i f i c a t i o n . S o i l S c i . , Vol. 67, p. 81-91. COLE, M., JONES, E., CUSTANCE, N. 1974 Remote sensing i n mineral explor-ation, i n : Environmental Remote Sensing Appl. and Achievements (Barret and Cu r t i s , eds.), E. Arnold, p. 49-66. COLLINS, J.B., WHITESIDE, E.P., CRESS, C E . 1970 Seasonal v a r i a b i l i t y of pH and lime requirements i n several Southern Michigan s o i l s when measured i n different ways. S o i l S c i . Soc. Am. P r o c , Vol. 34, p. 56 COLVOCOLESSES, A. 1975 Information through color imagery. Journal of Research, U.S.G.S., Vol. 3, No. 2, p. 127-129. COLWELL, R.N. 1966 Uses and l i m i t a t i o n s of multispectral remote sensing. Proc. 4th Symp. on R. S. of the Environment, Ann Arbor, Mich., p. 71-160. GONDII", H. 1970 The spectral reflectance of American s o i l s . Photoqr. Eng., Vol. 36, No. 9, p. 955-966. CONWAY, D. and HOLZ, R. 1973 The use of near infrared photography i n the analysis of surface morphology of an Argentine a l l u v i a l floodplain. Remote Sensing of Environment, V o l . 2, p. 235-242. COOKE, R.U. and DOORNKAMP, J . 1974 Geomorpholoqy i n Environmental Management. Claredon Press, Oxford, 413 pp. COOPER, A. 1960 An example of the role of microclimate i n s o i l genesis. SDil S c i . , V o l . 90, p. 109-120. CORTES, A., FRANZMEIER, D.P. 1972 Climosequence of ash-derived s o i l s i n the Central C o r d i l l e r a of Colombia. S o i l S c i . Soc. Am. P r o c , Vol. 36, p. 653-659. CRISP, D.T. 1966 Input and output of minerals for an area of Pennine moorland: the importance of p r e c i p i t a t i o n , drainage, peat erosion and animals. Journal of Applied Ecology 3, p. 327-348. CROFTS, R. 1974 Detailed geomorphological mapping and land evaluation i n Highland Scotland, _in: Process i n Geomorpholoqy (Brown and Waters), Inst, of B r i t i s h Geogr. Trans., Special Publ. #7, p. 231-249. 223 CROSSON, L.S. and PROTZ, R. 1973 P r e d i c t i o n of s o i l properties from stereophotD measurements of landform properties. Can. Journ. of S o i l Science 53, p. 259-262. CROWN, P. and PAWLUK, S. 1972 Remote sensing f o r s o i l resources inventories; uses of multiband photography. 1st Canad. Symp. on Remote Sensing, Dept. of Energy, Mines and Resources, p. 317-322. CURTIN, D. and SMILLIE, G.LJ. 1976 Estimation of components of s o i l c ation exchange capacity from measurements at s p e c i f i c surface and organic matter. S o i l S c i . Soc. Am. P r o c , V o l . 40, p. 461-462. CURTIS, L.F., DOORNKAMP, J.C., GREGORY, J.H. 1965 The de s c r i p t i o n of r e l i e f i n f i e l d studies of s o i l s . Journ. of S o i l S c i . , V o l . 16, p. 16-30. DAN, J.,and YAALDN, D.H. 1964 The a p p l i c a t i o n D f the catena concept i n studies of pedogenesis i n Mediterranean and desert f r i n g e regions. Trans. 8th Int. Congr., S o i l S c i . 5, p. 751-758. DAM, J . , YAALON, D.H., KOYUMOJISHY, H. 1968 Catenary s o i l r e l a t i o n s h i p i n I s r a e l , The Netanya Catena on Coastal Dunes of the Sheron. Geoderma, V o l . 2, p. 95-120. DANIELS, R.B., GAMBLE, E.E., CADY, J.G. 1970a Some r e l a t i o n s among coas t a l p l a i n s o i l s and geomorphic surfaces i n North C a r o l i n a . S o i l S c i . Soc. Am. P r o c , V o l . 34, p. 648-653. DANIELS, R., NELSON, L., GAMBLE, E. 1970b A method of cha r a c t e r i z i n g nearly l e v e l surfaces. Z e i t s . Geomorphology, N.F. 14, p. 175-185. DANIELS, R.B., GAMBLE, E.E., CADY, J.G. 1971 The r e l a t i o n s h i p betueen geomorphology and s o i l morphology and genesis. Advances i n Agronomy, V o l . 23, p. 51-68. DAY, P.R. 1950 Phy s i c a l basis of p a r t i c l e s i z e a nalysis by the hydrometer method. S o i l S c i . , V o l . 70, p. 363-374. DODD, J.D., RENNIE, D.A., CDUPLAND, R.T. 1964 The nature and d i s t r i b u t i o n of s a l t s i n uncultivated.saline s a i l s i n Saskatchewan. Canad. Journ. S o i l S c i . , V o l . 44, p. 165-175. DREES, L.R. and WILDING, L.P. 1973 Elemental v a r i a b i l i t y within a sample u n i t . S o i l S c i . Soc. Am. P r o c , V o l . 37, p. 82-87. DRISCOLL, R. and COLEMAN, M. 1974 Color f o r shrubs. Photoqr. Eng., V o l . 40, p. 451-459. EVANS, I.S. 1972 General geomarphometry, deviations of a l t i t u d e and descr i p t i o n s t a t i s t i c s , p. 17-90, _in:~ S p a t i a l Analysis i n Geomorphology (Chorley, R., ed.), Harper and Row Publ., N.Y., 393 pp. 224 EVANS, R. 1948 An Introduction to Color. John LUiley, N.Y., 370 pp. EVANS, R. 1975 Multiband photography for soilvisurv/ey i n Breckland East Anglia. Photoqrammetric Record, Vol. 8, p. 297-308. EVANS, R., HEAD, J. , DIRKZWAGER, M. 1976 A i r photo tones and s o i l properties: implications for interpreting s a t e l l i t e imagery. Remote Sensing of Environment, Vol. 4, p. 265-280. FARSTAD, L., LORD, T.M., GREEN, A.J., HORTIE, H.J. 1965 S o i l survey of the Peace River area i n B r i t i s h Columbia. Report # 8 of the B r i t i s h Columbia S o i l Survey, UBC and Research Branch, Canada Dept. FLOATE, M.J.S. 1965 Dis t r i b u t i o n of organic matter and phosphorus fractions i n a topographic sequence of s o i l s i n southern B r i t i s h Columbia. Can. Journ. S o i l S c i . , V o l . 45, p. 323-329. FRAZEE, C, MYERS, V., ItlESTIN, F. 1971 Remote sensing for detection of s o i l l i m i t a t i o n s i n a g r i c u l t u r a l areas. Proc. of the 7th Intern. Symp. on R.S. of the Environment, p. 327-343. FRAZEE, C , MYERS, V., WESTIN, F. 1972 Density s l i c i n g techniques for s o i l survey. S o i l S c i . Soc. Am. P r o c , Vol. 36, p. 693-695. FU, H. 1976 Pattern recognition i n remote sensing of the Earth's resources. IEEE - Trans, on Geoscience Electronics, Vol. GE 14, No. 1, p. 10-18. FURLEY, P.A. 1968 S o i l formation and slope development. Zeitsch. Geomorph., Vol. 12, p. 25-42. FURLEY, P.A. 1971 Relation betueen slope form and s o i l properties developed over chalk parent materials, i n : Slopes, Form, and Process (Brunsden, D.), I n s t i t u t e of B r i t . Geogr. Spec. Publ. No. 3, p. 141-162. FURLEY, P.A. 1974 Soil-slope-plant relationships i n the Northern Maya Mountains, B e l i z e , Central America. Journ. of Bioqeoqraphy, V o l . 1, p. 171-186. GAMBLE, E.E. and DANIELS, R.B. 1974 Parent material of upper-and middle coastal- p l a i n s o i l s i n North Carolina. S o i l S c i . Soc. Am. P r o c , Vol. 38, p. 633-637. GATES, D.M. 1970 Physical and physiological properties of plants. Remote Sensing i n Agriculture and Forestry, National Academy of Science, Washington, D.C., p. 224-252. GAUSMAN, H.W. 1974 Leaf reflectance of near I R. Photogr. Eng., Vol. 40, No. 2, p. 183-191. GAUSMAN, H.W., ALLEN, W., CARDENAS, R. 1969 Reflect, of cotton leaves and the i r structure. Remote Sensing of Environment, Vol. 1, p. 19-22. 225 GAUSMAN, H.W., ALLEN, Id., MYERS, V., CARDENAS, R., RICHARDSON, A. 1969 Relation of light reflectance to cotton leaf maturity. Proc. of the 6th Intern. Symp. on R. S. of the Environment, Vol. 2, p. 1123-1143. GAUSMAN, H.U., ESCOBAR, D.E., RODRIGUEZ, R.R. 1973 Discriminating among plant nutrient deficiencies with reflectance measurements. Proc. 4th Biennal Workshop on Aerial Color Photography in the Plant Sciences, Univ. of Maine, Orono, July 10-12, 1973, p. 13-27. GAUSMAN, H.W., GERBERMANN, A., WIEGAND, C , LEAMER, R., RODRIGUEZ, R., NORIEGA, J. 1975 Reflectance differences between crop residues and bare s o i l s . Soil Sci. Soc. Am. P r o c, Vol. 39, p. 752-755. GERASIMOV, I., ISAHOV, Y., PANFILOV, D. 1972 The internal circulation of matter in the principal types of natural ecosystems of the USSR. Soviet Geoqr. Review and Transl., Vol. 13, p. 441-447. GERBERMANN, A., GAUSMAN, H., WIEGAND, C. 1971 Color and color IR films for s o i l identification. Photoqr. Eng., Vol. 37, No. 4, p. 359-364. GERENCHUK, K.I., GORADH, I.K., TOPCHIYEV, A.G. 197D A method for estab-lishing some parameters of the morphology structure of landscapes. Soviet Geogr. Review and Transl., Vol. 9, No. 4, p. 262-271. GILBERTSON, B. and LONGSHAW, T. 1975 Multispectral aerial photography as exploration tool, I: concepts, techniques and instrumentation. Remote Sensing of Environment, Vol. 4, p. 129-146. GLAZOVSKAYA, M.A. 1963 On geochemical principles of the classification of natural landscapes. Int. Geol. Rev., Vol. 5, p. 1403-1431. GLAZOVSKAYA, M.A. 1968 Geochemical landscapes and types of geochemical s o i l sequences. 9th Trans. Int. Conqr. Soil Sci., Adelaide, Vol. 4, p. 303-312. GLAZOVSKAYA, MlAi1 1970 Types of geochemical s o i l catenas. Soviet Geoqr. Review and Transl., Vol. 9, p. 235-245. GOLD, A., ASHER, B. 1976 Soil reflectance measurement using a photographic method. Soil Sci. Soc. of Am. Journ., Vol. 40, p. 337-341. GREGORY, K., and BROWN, E. 1966 Data processing and the study of land form. Zeit. fur Geom., N.F. 10, p. 237-263. GREWELING, T. and PEECH, M. 1965 Chemical Soil Tests. N.Y. State College of Agriculture. Bull. 960. HELLER, R., DOVERSPIKE, G., ALDRICH, R. 1964 Identification of tree species on large scale panchromatic and color aerial photographs. Agric. Handbook # 261, Forest Service, USDA. 226 HEMBREE, CH. and RAINWATER, F.H. 1961 Chemical degradation on opposite flanks of the Wind River range. USGS Water Supply Paper 1535 E, p. 9. HILLS, A.G. 1966 The classification and evaluation of land for multiple uses. Forest Chronicle, p. 1-25. HILLS, A.G. and PORTELANCE, R. 1960 The Glackmeyer Report of Multiple Land Use Planning. Ont. Dept. of Land and Forests, 210 pp. H0BS0IM, R.D. 1972 Surface roughness in topography, a quantitative approach. in: Spatial Analysis in Geomorpholoqy (Chorley, R.J., ed.), Methuen Ltd., London, p. 221-246. HOFFER, R., HOLMES, R., SHAY, J. 1966 Vegetative, s o i l and photogr. factors affecting tone in agricultural remote multispectral sensing. Proc. 4th Symp. on Remote Sensing of the Environment, Univ. of Michigan, p. 115-134. HOFFER, R. and JOHANNSEN, C 1969 Ecological potential in spectral signa-ture analysis, jin: Remote Sensing in Ecology (Johnson, P., ed.), Univ. of Georgia Press, Athens, Georgia, p. 1-16. HOFFER, R., ANUTA, P., PHILLIPS, T. 1972 ADP, Multiband and multi-emulsion digitized photos. Photogr. Eng., Vol. 38, IMo. 38, p. 989-1001. HOLMES, R. 1970 Field spectroscopy: in remote sensing uith reference to agriculture and forestry. Nat. Acad, of Science, p. 298-323. HORTON, R.E. 1945 Erosional development of streams and their drainage basins - hydro physical approach to quantitative morphology. Geol. Soc. Am. Bull., Vol. 56, p. 275-370. H0VIS, Id. 1966 IR-spectral reflectance of some common minerals. Applied Optics, Vol. 5, p. 245-248. HOVIS, W., KLEY, Ld., STRANGE, M. 1967 F i l t e r wedge spectrometer for f i e l d use. Applied optics, Vol. 6, p. 1057-1058. HOWARD, J., WATSON, R., HESSIN, T. 1971 Spectral reflectance properties of Pinus Ponderosa in relation to copper content in the s o i l -Malachite Mine, Jefferson County, Colorado. Proc. of the 7th Intern. Symp. on Remote Sensing of the Environment, p. 285-297. HOWARTH, P. 1972 Film and f i l t e r combinations for the study of the Peri-glacial landscape of the High Arctic. Proc. 1st Can. Symp. Remote Sensing, p. 379-385. HUDDLESTON, J.H. and RIECKEN, F.F. 1973 Local soil-landscape relationship . in Western Iowa,.I: Distribution of selected chemical-physical properties. S o i l S c i . Soc. Am. P r o c , Vol. 37, No. 2, p. 264-270. 227 HUDDLESTOIM, J.H.. WALSH, J.A., JOWETT, D., RIECKEN, F.F. 1975 Local s o i l -landscape relationships in Western Iowa: Quantification of pedologic and geologic effects. S o i l Sci. Soc. Am. Proc, Vol. 39, p. 908-913. HUGGETT, R.J. 1975 Soil landscape systems: a model of s o i l genesis. Geoderma, Vol. 13, p. 1-22. HUNT, G. and ROSS, H. 1967 A bi-directional reflectance accessory for spectroscopic measurements. Applied Optics, Vol. 6, No. 10, p. 1687-1690. HUNT, G. and SALISBURY, J. 1970 Visible and near IR spectra of minerals and rocks: I Silicates. Modern Geology, Vol. 1, p. 283-300. HUNT, G. and SALISBURY, J. 1971 Visible and near IR spectra of minerals and rocks: II Carbonates. Modern Geology, Vol. 2, p. 23-30. HUNT, G. and SALISBURY, J. 1971 Visible and near IR spectra of minerals and rocks: IV Sulfides and Sulfates. Modern Geology, Vol. 3, p. 1-14. HUNT, G. and SALISBURY, J. 1973 Visible and near IR spectra of minerals and rocks: VII Acid igneous rocks. Modern Geology, Vol. 4, p. 217-224. HUNT, G., SALISBURY, J., LENHDFF, . 1974 Visible and near IR spectra of minerals and rocks: IX Basic and ultrabasic igneous rocks. Modern Geology, Vol. 5, p. 15-22. HUNTER, G. and BIRD, G. 1970 C r i t i c a l terrain analysis. Photogr. Eng., Vol. 36, No. 9, p. 939-952. IDS0, S., SCHMUGGE, T., JACKSON, R., REGINAT0, R. 1975 The u t i l i t y of surface temperature measurements for the remote sensing of surface s o i l mater status. Journ. of Geophysical Research, Vol. 80, No. 21, p. 3044-3049. JACKSON, M.L. 1956 Soil Chemical Analysis. Univ. of Wisconsin, Madison, p. 160-163. JENNY, H. 1941 Factors of Soil Formation. McGraw-Hill, N.Y., 281 pp. JENNY, H. 1961 Derivation of state factor equations of soils and eco-systems. Soil S c i . Soc. Am. Proc, Vol. 25, p. 385-388. JOHNSON, R. 1968 Choice in classification, the subjectivity of objective methods. Ann. Ass. Am. Geoqr., Vol. 58, p. 575-589. JUNKINS, J.L. and JANCAITIS, J.R. 1971 Mathematical terrain analysis. Am. Congr. on Surv.. and-Mapping. 31st Ann. Meeting, Wash., D.C, p. 658-678. KALENSKY, Z. and WILSON, D. 1975 Spectral signature of forest trees. Proc. 3rd Can. Symp. Remote Sensing, p. 155-172. 228 KARMANOV, I. 1970. Study D f s o i l s from spectral composition D f reflected radiation. Soviet S o i l S c i . , Vol. 2, p. 226-237. KETTIG, R. and LAIMDGREBE, D. 1976 C l a s s i f i c a t i o n of multispectral image data f o r extraction and c l a s s i f i c a t i o n of homogeneous objects. IEEE - Trans, on Geoscience Electronics, Vol. GE 14, No. 1, p. 19-26. KING, B. 1970 A parametric approach to land system c l a s s i f i c a t i o n . Geoderma, V o l . k, p. 37-46. KLEISS, H. 1970 H i l l s l o p e sedimentation and s o i l formation i n Northeastern Ioma. S o i l S c i . Soc. Am. P r o c , Vol. 34, p. 287-290. KLINE, J.R. 1973 Mathematical simulation of so i l - p l a n t relationships and s o i l genesis. S o i l Science 115, p. 240-249. KNIPLING, E.B. 1970 Physical and physiological basis for the reflectance of v i s i b l e and near infrared radiation from vegetation. Remote Sensing of Environment, V o l . 1, p. 155-159. K0LJ0NEN, T. and CARLSON, L. 1975 Behaviour of the major elements and minerals i n sediments of four humic lakes i n Southwestern Finland. Fennia 137, p. 1-47. KRISTOF, S.J. 1971 Preliminary multispectral studies of s o i l s . Journ. of S o i l and Water Conservation, V o l . 26, p. 15-18. KRISTOF, S.J. and ZACHARY, A.L. 1971 Mapping s o i l types from multispectral scanner data. Proc. of the 7th Intern. Symp. on Remote Sensing of the Environment, p. 2095-2097. KRISTOF, S.J. and ZACHARY, A.L. 1974 Mapping s o i l features from m u l t i -spectral scanner data. Photoqr. Eng., V o l . 40, No. 12, p. 1427-2434. KRISTOF, S.J., BAUMGARDNER, M., JOHANNSEN, C. 1973 Spectral mapping of s o i l organic matter. LARS Info. Note 030773, Laboratory for Appl. of Remote Sensing, Perdue Univ., West Lafayette, p. 1-15. KRISTOF, S.J., BAUMGARDNER, M.F., JOHANNSEN, C.J. 1974 Spectral mapping of s o i l organic matter. ITC Journ. 1974-4, p. 479-499. KRISTOF, S.J. and BAUMGARDNER, M. 1975 Changes of multispectral s o i l patterns with increasing crop canopy. Agronomy Journal, Vol. 67, p. 317-321. KUHL, A.D. 1970 Color and IR photo, for s o i l s . Photoqr. Eng., Vol. 36, p. 475-482. LACATE, D.S. 1961 A review of land type c l a s s i f i c a t i o n and mapping. Land Economics, Vol. 37(3), p. 271-278. LATHRAM, E.H. 1973 Geological application of ERTS imagery i n Alaska. 3rd ERTS Symp. Abstract. NASA/Goddard Space Centre, 39 pp. 229 LAVKULICH, L.M. 1974 Method of Soil Analysis. Ped. Lab, UBC, Dept. of Sai l Sci., 221 pp. LAVKULICH, L.M. and WIENS, J.H. 1970 Comparison of organic matter des-truction by Hydrogen Peroxide and Sodium Hypochlorite and i t s effect on selected mineral constituents. Soil Sci. Soc. Am. P r o c , Vol. 34, p. 755-758. LEAMER, R., WEBER, D., WIEGAND, C. 1975 Pattern recognition of soi l s and crops from Space. Photogr. Eng., Vol. 41,p. 471-478. LINTON, D.L. 1951 The delimitation of morphological regions, in: Essays in Geography (Stamp, D. and lilooldridge, S.U.), Longmans, London, p. 199-218. LUSTIG, L.K. 1969 Quantitative analysis of desert topography, _in: Arid Land in Perspective (McGinnies and Goldman, B.), The Am. Ass. for the Advancement of Science and Univ. of Ariz. Press, p. 47-58. LUTTMERDING, H.A. and SPROUT, P.N. 1966 Soil Survey of Lanqley Municipality and Barnston Island. B.C. Dept. of A g r i c , Kelouinai B.C., p. 1-161. LYON, R. 1972 Infrared spectral emittance in geological mapping: airborne spectrometer data from Pisgal Crater, California. Science. Vol. 175, p. 983-985. MABBUTT, J.A. 1968 Review of concepts of land classification, in: Land Evaluation (Stewart, G.A.), MacMillan, Melbourne, p. 11-27. MABBUTT, J.A., LICHFIELD, W.H., SPECK, N.H., SOFOULIS, J., WILCOX, D.G., ARNOLD, J., BROOKFIELD, M., WRIGHT, R.L. 1963 General report on lands of the Wiluna-Mackatharra area. CSIRO, Austr. Land Res. Ser., Vol. 7, p. 1-214. MABBUTT, J.A. and STEWART, G.A. 1965 Application of geomorphology in integrated resource surveys in Australia. Revue Geomorph. Dynamique, Vol. 6, p. 1-13. MADER, D. 1963 So i l variability - a serious problem in s o i l site studies in the Northeast. Soil S c i . Soc. Am. P r o c , Vol. 27, p. 707-709. MALAN, 0. 1974 Color balance of color IR film. Photogr. Eng., Vol. 40, p. 311-316. MAL0, D.D., WORCESTER, B.K., CASSEL, D.K., MATZD0RF, K.D. 1974 Soil landscape relationships in a closed drainage system. Soil S c i . Soc. Amer. Proc, Vol. 38, p. 813-818. MANSEL, P. and JOHANNSEN, C. 1973 An application of remotely sensed data to- agricultural land use distribution analysis. The Prof. Geoqr., Vol. 25, No. 3, p. 242-248. 230 MARLAR, T. and RINKER, J. 1967 A small four camera system for multi-emulsion studies. ' Photoqr. Eng., Vol. 33, No. 4, p. 1252-1257. MATHEWS, H.L., CUNNINGHAM, R.L., CIPRA, J.E., WEST, T.R. 1973a Application of multispectral remote sensing to s o i l survey research in southeastern Pennsylvania. Soil Sci. Soc. Am. P r o c, Vol. 37, No. 1, p. 88-93. MATHEWS, H.L., CUNNINGHAM, R.L., PETERSEN, G.W. 1973b Spectral reflectance of selected Pennsylvania s o i l s . S o i l S c i . Soc. Am. Proc, Vol. 37, p. 421-424. MATHEWS, W.H. 1963 Quaternary stratigraphy and geomorphology of the Fort St. John area, Northeastern B.C. Dept. of Mines & Pet. Res., Victoria, 22 pp. MAURER, H. 1974 Quantification of textures - textural parameters and their significance for classifying agricultural crop types from colour aerial photographs. Photogrammetria, Vol. 30, p. 21-40. MAY, G., and PETERSEN, G.W. 1975 Spectral signature selection for mapping unvegetated s o i l s . Remote Sensing of Environment, Vol. 4, p. 211-220. McCORMACK, D. and WILDING, li . 1969 Variation of s o i l properties within mapping units of soils with contrasting substrata in Northwestern Ohio. Soil S c i . Soc. Am. Proc, Vol. 33, p. 587-593. McDOWELL, D. and SPECHT, M. 1974 Determination of spectral reflectance using aerial photography. Photogr. Eng., Vol. 40, p. 559-568. McINTYRE, G.A. 1967 Soil sampling for s o i l testing. Journ. Aust. Inst. Agric. Sci., Vol. 33, p. 308-320. McKEAGUE, J.A. and DAY, J.H. 1966 Dithionite and Oxalate extractable iron and aluminum as aids in differentiating various classes of s o i l s . Can. Journ. So i l Sci., Vol. 46, p. 13-22. McKEAGUE, J.A., MACD0UGALL, J.I., MILES, N.M. 1973 Micromorphological, physical, chemical and mineralogical properties of a catena of soils from Prince Edward Island in relation to their classification and genesis. Can. Journ. S o i l Sci., Vol. 53-3, p. 281-295. McLEAN, E. and CARBONELL, M. 1972 Calcium, Magnesium and Potassium satur-ation ratios in two soils and their effect upon yields and nutrient contents of German millet and a l f a l f a . Soil Sci. Soc. Am. P r o c , Vol. 36, p. 927-936. MEHRA, O.P. and JACKSON, M.L. 1960 Iron Oxide removal from soils and clays by Dithionite-Citrate system buffered with Sodium Bicarbonate. Clay and Clay Minerals, Vol. 5, p. 317-327. MELTON, M.A. 1957 An analysis of the relations amongst elements of climate surface properties and geomorphology. Techn. Rep. # 11, Office of Naval Research, Dept. of Geology, Columbia Univ., p. 1-37. MITCHELL, C.W. 1971 An appraisal of a hierarchy of desert land units. Geoforum, Vol. 7, p. 69-75. 231 MITCHELL, C.W. 1974 Terrain Evaluation. Longman, London, 221 pp. MITCHELL, C.W. and PERRIN, R. 1967 The subdivision of hot deserts of the uorld into physiographic units. l i e Symposium international de Photo-interpretation, Paris 1966 (Trans.), Vol. 4, p. 89-106. MONTGOMERY, 0. and BAUMGARDNER, M. 1974 The effects of the physical and chemical properties of soils on the spectral reflectance of s o i l s . LARS Info. Note 112674, Laboratory for Appl. of Remote Sensing, Purdue Univ., West Lafayette, p. 1-119. i MURTHA, P. 1972 A guide to air photo interpretation of forest damage in Canada. Canadian Forest Service, Publ. 1292, Qttaiua, 62 pp. MYERS, V. 1970 Soi l , Water and Plant Relations: remote sensing uiith , special references to agriculture and forestry. Nat. Acad, of Science, Wash., D.C., p. 253-295. MYERS, V. 1975 Crops and Soils, in: Manual of Remote Sensing (Reeves, R.), American Photogr. Society, Falls Church, Virg., Vol. II, p. 1715-1813. MYERS, V. and ALLEN, W. 1968 Electrooptical remote sensing methods as non-destructive testing and measuring techniques in agriculture. Applied Optics, Vol. 7, No. 9,/p. 1819-1838. MYERS, V. and HEILMAN, M. 1969 Termal infrared for s o i l temperature studies. Photogr. Eng., Vol. 35, No. 10, p. 1024-1032. NYHAN, J.W., FREDERICK, L.R., NORTON, D.C. Ecology of nematodes in Clarion-Webster toposequences associated uith Glycine max (L.) M e r r i l l . S o i l S c i . Soc. Am. P r o c , Vol. 36, p. 74-78. 0DENY0, V. and RUST, R. 1975 Application of density s l i c i n g techniques to s o i l survey. Soil S c i . Soc. Am. Pr o c , Vol. 39, p. 311-315. 0LLIER, C. 1969 Weathering. Oliver and Boyd, Edinburgh, p. 103-119. ORR, D. and GJUICK, J. 1971 Construction materials in delta areas. Photogr. Eng., Vol. 37, No. 4, p. 337-351. PARKER, D., LEE, G., MILFRED, C. 1970 Flood plain delineation uith pan and color. Photogr. Eng., Vol. 36, No. 10, p. 1059-1063. PARKS, W.L. and B0DENHEIMER, R.E. 1973 Delineation of major s o i l asso-ciations using ERTS-1 imagery, in: Symposium on Significant Results obtained from Earth Resources Technology Satellite-1, Vol. 1, p. 121-126. PARRY, J., COWAN, W., HEGINB0TT0M, 0. 1969 Soils studies using color photos. Photogr. Eng., Vol. 35, No. 1, p. 44-56. 232 PARRY, J. and TURNER, H. 1971 Infrared photos for drainage analysis. Photoqr. Eng., Vol. 37, No. ID, p. 1Q31-103B. PATTERSDN, J., WHITACKER, R. 1973 UBC - C-Grouping Computer Programme. Univ. of British Columbia, Vancouver, 20 pp. PEREL'MAN, A.I. 1961 Geochemical principles of landscape classification. Soviet Geography, Vol. 2, p. 63-73. PEREL'MAN, A.I. 1967 Geochemistry of Epiqenesis. New York, Plenum Press, 266 pp. PETTRY, D., POWELL, M., NEWHOUSE, M. 1974 Use of Remote Sensing in Agriculture. NASA, Wallops Station, Virginia, and NTIS, U.S. Dept. of Commerce, p. 1-139. PHILPOTTS, L. and WALLEN, V. 1969 Infrared color for crop disease identification. Photoqr. Eng., Vol. 35, p. 1116-1123. PIECH, K., WALKER, J. 1974 Interpretation of s o i l s . Photoqr. Eng., Vol. 40, p. 87-94. PLANET, W. 1970 Some comments on reflectance measurements of wet s o i l s . Remote Sensing of Environment, Vol. 1, p. 127-129. POHN, H. 1974 Near infrared reflectance anomalies of Andesite and Basalt in Southern California and Nevada. Geology, Vol. 2, p. 547-550. PRENTICE, V.C. 1972 Multispectral Remote Sensing Techniques applied to Salinity and Drainage Problems in the Columbia Basin, Washington. Ph.D. Thesis, Dept. of Geography, Univ. of Michigan, Ann Arbor, 224 pp. PRESS, N.P. and NORMAN, J.W. 1972 Detection of ore bodies by remote sensing the effects of metals as vegetation. Trans. Inst. Mining and Metall., Vol. 81, p. 166-168. PROKAIEV, V.I. 1962 The facies as a basic and smallest unit in landscape studies. Soviet Geoqr. Review and Transl., Vol. 3, No. 6, p. 21-29. PROTZ, R., BREWER, R., McKEAGUE, J. 1973 Quantification of s o i l v a r i a b i l i t y . 1972/73 Progress Report of Dept. of Land Resources Science, Univ. of Guelph, p. 76-77. RAINES, G.,,and LEE, K. 1974 Spectral reflectance measurements. Photoqr. Eng., Vol. 40, p. 547-555. RAINES, G. and LEE, K. 1975 In situ rock reflectance. Photoqr. Eng., Vol. 41, No. 2, p. 189-198. REEVES, R., KOVER, A., LYON, R., SMITH, H. 1975 Terrain and minerals, assessment and evaluation, in: Manual of Remote Sensing (Reeves, R., ed.), Vol. II, p. 1107-1351, Amer. Soc. of Photogr., Falls Church, Virg. 233 REMOTE SENSING COMMITTEE 1970 Remote Sensing with Special Reference to Agriculture and Forestry. Nat. Acad, of Science, Wash., D.C, 422 pp. REYNOLDS, S.G. 1975 Soil property variability in slope studies: suggested sampling scheme and typical required sample,sizes. Zeits. Geomorph., N.F. 19, p. 191-208. RIB, H. and MILES, R. 1969 Automatic interpretation of terrain features. Photogr. Eng., Vol. 35, No. 2, p. 153-164. RICHARDSON, A.J., ItllEGAND, C.L., GAUSMAN, H.lil., CUELLAR, J.A., GERBERMANN, A.H. 1975 Plant, s o i l and shadow reflectance components of row crops. Photogr. Eng., Vol. 41, p. 1401-1407. RICHARDSON, A., GERBERMANN, A., GAUSMAN, H., CUELLAR, J. 1976 Detection of saline so i l s with Skylab multispectral scanner data. Photogr. Eng. and Rem. Sens., Vol. 42, No. 5, p. 679-684. RIQUIER, 3. 1974 A summary of parametric methods of s o i l and land evalua-tion, _in: Approaches to Land Classification, FAO S o i l Bulletin, No. 22, FAO, Rome, p. 47-53. R0MAN0VA, M. 1968 Spectral luminance of sand deposits as a tool in land evaluation, in: Land Evaluation (Stewart, G.A.), MacMillan, p. 342-348. ROSS, D. 1973 Simple multispectral photography and additive color viewing. Photogr. Eng., Vol. 39, No. 6, p. 583-591. RUHE, R. 1956 Geomorphic surfaces and the nature of s o i l s . Soil Science, Vol. 82, p. 444-455. RUHE, R. 1974 Holocene environments and s o i l geomorphology in Midwest United States. Quat. Research, Vol. 4, p. 487-495. RUHE, R.V. and WALKER, P.H. 1968 Hillslope models and s o i l formation, open system. 9th Intern. Congr. S o i l Sci., Adelaide, p. 551-560. RUHE, R.V., MILLER, G.A., VREEKEN, W.J. 1971 Paleosols, loess sedimentation and s a i l stratigraphy, in: Paleopedoloqy (Yaalon, D.), Intern. Soc. Soi l Sci., Jerusalem, p. 41-60. RUNGE, E.A. 1973 So i l development sequences and energy models. Soil Science, Vol; 115, p. 183-193. RYDER, J.M. 1972 Pleistocene chronology and glacial geomorphology: studies in Southwestern British Columbia, in: Mountain Geomorpholoqy (Slaymaker, 0. and McPhearson, H.), Tantallis Research Ltd., p. 63-72. SALISBURY, J. and HUNT, G. 1974 Remote sensing of rock type in the visible and near-infrared. Proc. of the 9th Intern. Symp. on Remote Sensing of the Environment, Vol. 3, p. 1953-1958. 234 SCHUMM, S.A. 1956 Evolution of drainage systems and slopes in badlands at Perth Amboy, N.J. Geol. Soc. Am. Bull., Vol. 67, p. 597-646. SCOTT, R.M. and AUSTIN, M.P. 1971 Numerical classification of land systems using geomorphological attributes. Austr. Geogr. Studies, Vol. 9, p. 33-40. SHARMA, S.B., ANANTHANARAYANAN, P.V., VAN DER MEER MOHR 1972 The use of density s l i c i n g for discrimination and identification of geological units. ITC Publ. Series B, No. 67, p. 45-63. SHIELDS, J., ARNAUD, R., PAUL, E., CLAYTON, J. 1966 Measurements of s o i l color. Can. Journ. S o i l Sci., Vol. 46, p. 83-89. SHIELDS, J., PAUL, E., ARNAUD, R., HEAD, W. 1968 Spectrometric measure-ments of s o i l color and i t s relationship to moisture and organic matter. Can. Journ. So i l Sci., Vol. 48, p. 271-280. SHOCKLEY, W.G., KNIGHT, S.J., LIPSCOMB, E.B. 1962 Identifying s o i l para-meters uith an infrared spectrometer. Proc. of the 2nd Symp. on Remote Sensing of the Environment, Univ. of Michigan Press, Ann Arbor, p. 267-288. SKIDMDRE, E.L., DICKERSON, J.D., SCHIMMELPFENNIG, H. 1975 Evaluating surface-soil uater content by measuring reflectance. Soil S c i . Soc. Am. Proc, Vol. 39, p. 238-242. SLATER, P. 1972 Multiband camera. Photoqr. Eng., Vol. 38, No. 6, p. 543-555. SLATER, P. 1975 Photographic systems for remote sensing, i n : Manual of Remote Sensing (Reeves, R., ed.) p. 235-324, Am. Soc. of Photog., Falls Church, Virginia, 2144 pp. SMITH, R., TWISS, P., KRAUSS, R., BROUN, M. 1970 Dust deposition in relation to site, season and climatic variables. Soil Sci. Soc. Am. Proc, Vol. 34, p. 112-117. SOKAL, R.,and SNEATH, P. 1963 Principles of Numerical Taxonomy. John Wiley, N.Y., p. 169-215 and p. 305-312. . SOLNTSEV, N.A. 1962 Basic problems of Soviet landscape science. Soviet Geogr. Revieu, Vol. 3 (6), p. 3-15. SOLNTSEV, N.A. 1973 On biotic and geomic factors in the formation of the natural environment. Soviet Geoqr. Revieu, Vol. 14 (6), p. 347-355. SPEIGHT, J.G. 1968 Parametric description of land forms, in: Land Evaluation (Stewart, G.A., ed.), MacMillan, Austr., p. 239-260. SPEIGHT, J.G. 1974 A parametric approach to landform regions, in: Process in Geomorphology (Broun and Waters), Inst, of British Geogr. Trans. Spec. Publ. No. 7, p. 213-229. 235 SPILSBURY, U. and FLETCHER, K. 1973 Application of regression analysis to interpretation of geochemical data from lake sediments in Central British Columbia. Can. Journ. of Earth Sci., Vol. 11, p. 345-347. SPRATT, E.D. and McIVER, R.IM. 1972 Effects of topographical positions, s o i l test values, and f e r t i l i z e r use on yields of uiheat in a complex of black chernozemic and gleysolic s o i l s . Can. Journ. of Soil Sci., Vol. 52, p. 53-58. STEINER, D. and GUTERMANN, T. 1966 Russian Data on Spectral Reflectance of Vegetation, Soil and Rock Types. Dept. of Geogr., Univ. of Zurich, and U.S. Army European Res. Dffice, Juris Druck and Verlag, Zurich, p. 1-225. STEINER, D., BAMBERGER, H., MAURER, H. 1969 Computer processing and classification of multi-variate information from remote sensing imagery. Proc. 6th Symp. on Remote Sensing of the Environment, Vol. 2, p. 895-908. STEINHARDT, G., FRAIMZMEIER, D., CIPRA, J. 1974 Indiana s o i l associations compared to Earth resources technology s a t e l l i t e imagery. Proc. of the Indiana Academy of Science, Vol. 84, p. 463-468. STOIMER, E., BAUMGARDNER, M., SWAIN, P. 1976 Determining density of maize canopy from digitized photographic data. Agron. Journ., Vol. 68, p. 55-59. STRAHKOW, N.M. 1967 Principles of Lithogenesis. Oliver and Boyd, Edinburgh, Vol. 1, 245 pp. STRAHKOW, N.M. 1969 Principles of Lithogenesis. Oliver and Boyd, Edinburgh, VDI. 2, 609 pp. STRAHKOW, N.M. 1970 Principles of Lithogenesis. Oliver and Boyd, Edinburgh, Vol. 3, 577 pp. STRAHLER, A.N. 1957 Quantitative analysis of uatershed geomorphology. Am. Geophys. Union Trans., Vol. 38, p. 913-920. SWAN, S.B. 1970 Analysis of residual terrain: Johor, Malaya. Ass. Am. Geogr. Annals, Vol. 60, p. 124-133. SWAN, S.B. 1971 Relationships between regolith, lithology and slope in a humid tropical region: Johor, Malaya. Trans. Inst. B r i t . Geogr., Vol. 51, p. 189-200. SYLVESTER-BRADLEY 1967 Evolution versus entropy. Proc. Geol. Soc. London, Vol. 78, p. 137-147. SYMEONAHIS, A. and PR0TZ, R. 1974 Multispectral image enhancement tech-niques for mapping s o i l s . 1973 Progress Report of Dept. of Land Resources Science, Univ. of Guelph, p. 76-77. 236 TAIMGUAY, M., HOFFER, R. MILES, R. 1969 Multispectral imagery and automatic classification of spectral response for detailed engineering soi l s mapping. Proc. 6th Symp. on Remote Sensing of the Environment, p. 33-63. TARNOCAI, C. 1972 The use of remote sensing techniques to study peatland and vegetation types, organic soils and permafrost in the Boreal Region of Manitoba. 1st Can. Symp. on Remote Sensing, Dept. of Energy Mines and Resources, p. 323-335. THOMAS, 0., MYERS, V., HEILMAN, M., WIEGAND, C. 1966 Factors affecting light reflectance of cotton. Proc. of 4th Symp. on Remote Sensing of the Environment, Univ. of Michigan, p. 305-312. SfiHOMSQN, G.H. 1975 The use of photographic materials in remote sensing, in: Remote Sensing Data Processing (van Genderen, J., Collins, W., eds.), The Remote Sensing Society, Univ. of Sheffield, p. 1-18. TOBLER, W.R. and DAVIS, CM. 1968 A d i g i t a l terrain library. Univ. of Mich. Dept. of Geoqr. Techn. Report, ORA Proj. 08055, Ann Arbor, Mich. TOLCHEL'NIKDV, Y. 1968 Characteristic reflection curves and the interpre-tation of desert soils from aerial survey photography. Soviet Soil Science, Vol. 11, p. 1493-1504. TRICART, J. 1971 Cartographie geomorphologique et classement des terres pour l a conservation. Mem, et Doc, Centre de Doc. Cartoqr., CNRS, Vol. 12, p. 215-222. TR0EH, F.R. 1964 Landform parameters correlated to s o i l drainage. So i l S c i . Soc. Am. Proc, Vol. 28, p. 808-812. TROLL, C. 1966 Landscape Ecology. Paper S4 of the Proc. of the 1st Int. Seminar on Integr. Surveys of the Natural Environment, ITC/UNESC0 Centre on Integrated Surveys. TSYGANENK0, A.F. 1968 Aeolian migration of salts and i t s possible geochemical and s o i l forming importance. Soviet Soil Science, Vol. 3, p. 482-487. TUCKER, C.,and MILLER 1974 Extraction of the underlying s o i l spectra from canopy spectroreflectance measurements of the short grass prairie, in: Remote Sensing of Earth Resources (Shahrokhi, F., ed.), Univ. of Tennessee, Tullahoma, Vol. 3, p. 73-83. TURNER, A.K. and MILES, CR. 1967 Terrain analysis by computer. Indiana Acad. Sc i . P r o c , Vol. 77, p. 256-270. VALENTINE, K., LORD, T., WATT, W., BEDWANY, A. 1971 S o i l mapping accuracy from black and white color and infrared aerial photography. Can. Journ. S o i l Sci., Vol. 51, p. 461-469. 237 VAN LOPIK, J.R. and KOLB, CR. 1959 A technique for preparing desert terrain analogs. U.S. Army Engin. Waterway Exp. Station, Techn. Report 3-506, V/icksburg, Miss. VERSTAPPEN, H. Th. and VAN ZUIDAM, R.A. 1975 The ITC System of Geomor- phological Survey. Vol. 7, 3rd ed., ITC, Enschede, Netherlands, p. 1-52. UERSTRAETEN, L,, V/LASSAT, K., LIVENS, J. 1972 Water soluble Carbon and Nitrogen in s o i l and their relation to land use and texture. So i l Science, Vol. 113, No. 3, p. 200-203. VINCENT, R. 1972 An ERTS multispectral scanner experiment for mapping iron compounds. Proc. of the Bth Intern. Symp. on Remote Sensing of the Environment, Vol. 2, p. 1239-2344. VINCENT, R. 1973 Spectral ratio imaging method for geological remote sensing from aircraft and s a t e l l i t e s . Symp. Proceedings: Management and Utilization of Remote Sensing Data, Am. Sac. of Photogr., Fa l l s Church, Virg., p. 377-397. VINCENT, R. and THOMSON, F. 1972 Rock type discrimination from ratioed infrared scanner images of Pisgah Crater, California. Science, Vol. 175, p. 986-988. VINCENT, R., ROWAN, L., GILLESPIE, R., KNAPP, C. 1975 Thermal-infrared spectra and chemical analyses of tuienty-six igneous rock samples. Remote Sensing of Environment, Vol. 4, p. 199-209. VINOGRADOV, B.V. 1963 Biochemical provinces and their role in organic evolution. Geochemistry, p. 214-228. VINOGRADOV, B.V., GERENCHUK, K.I., ISACHENHiO, A.G., RAMAN, K.G., TSCEL'CHUK, Y.N. 1962 Basic principles of landscape mapping. Soviet Geogr. Review and Transl., Vol. 3 (6), p. 15-20. VREEKEN, W.J. 1968 Stratigraphy, sedimentology, and moisture contents in a small loess watershed in Tama County, Iowa. Iowa Acad. Sci. P r o c , Vol. 75, p. 225-233. VREEKEN, W.J. 1973 Soil variability in small loess watersheds; clay and organic carbon content. Catena, Vol. 1, p. 181-196. WALKER, G.A.H., BUCHH0LZ, V.L., CAMP, D., ISHERW00D, B., GOWER, J.F.R. 1975 A silicon diode array spectrometer for ocean colour monitoring. The Canadian Journal of Remote Sensing, Vol. 1, p. 26-30. WALKER, P.H., HALL, G., PROTZ, R. 1968a Relation between landform parameters and soil, properties. S o i l Sci. Soc. Am. Proc, Vol. 32, p. 101-104. WALKER, P.H., HALL, G.F., PROTZ, R. 1968b So i l trends and variability across selected, landscapes in Iowa. So i l S c i . Soc. Am. Pr o c , Vol. 32, p. 97-100. 238 WALKER, P.H. and RUHE, R.V. 1968 Hillslope models and s o i l formation, open and closed systems. Trans. 9th Intern. Conqr. S o i l Sci., Adelaide, p. 551-568. WALMSLEY, M. and LAVKULICH, L. 1975 Landform-soil-uegetatiDn-uiater chemistry relationship, Wrigley area, N.W.T. I Morphology, classification and site description II Chemical, physical, and mineralogical determinations and relationships. Soil S c i . Soc. of Am. Proc, Vol. 39, p. 84-93. WARD, J. 1963 Hierarchical grouping to optimize an objective function. Am. Stat. Assoc., Vol. 58, p. 236-244. WARD, J. 1969 The significance of changes in IR reflectance in sugar maple, induced by s o i l conditions of drought and sali n i t y . Proc. of the 6th Symp. on Remote Sensing of the Environment, p. 120.5-1226. WARRINGTON, P. and RYERSON, R. 1974 Using diazo colour composites to extract information from ERTS-1 multispectral data. 2nd Canadian Symp. on Remote Sensing, Univ. of Guelph, Ontario, p. 289-294. WATSON, R. 1972 Spectral reflectance and photDmorphic properties of selected rocks. Remote Sensing of Environment, Vol. 2, p. 95-100. WEAVER, R., SYERS, 0., JACKSON, M. 1968 Determination of s i l i c a in citrate bicarbDnate-dithionite extracts of s o i l s . S o i l Sci. Soc. of Am. Proc, Vol. 32, p. 497-501. WEBER, F. and 0LSDN, C. 1967 Remote sensing implications of changes in physiologic structure and function of tree seedlings under moisture stress. Am. Prog. Report of Remote Sensing Laboratory, Pacific SW Forest and Range Exp. Station, Nat. Resource Prog. NASA, p. 1-60. WEBSTER, R. 1965 A catena Df soils on the Northern Rhodesia plateau. Journ. Soil S c i . 16, p. 31-43. WEBSTER, R. and BECKETT, P.H.T. 1964 The study of the agronomic value of s o i l maps interpreted from air photographs. 8th Intern. Congr. of Soil Sci., Bucharest, Vol. 88, p. 795-803. WEBSTER, R. and BECKETT, P.H.T. 1970 Terrain classification and evaluation using aerial photography: a review of recent work at Oxford. Photogrammetria, Vol. 26, p. 51-75. WENDEROTH, S. and YOST, E. 1974 Multispectral Photography for Earth Resources. Remote Sensing Infor. Center, Greenvale, N.Y., p. 1-257. WHITE, E.M. 1971 Grass cycling of Ca, Mg, K and Na in solodization. Soil S c i . Soc. of Am. P r o c , Vol. 35, p. 309-311. 239 WHITE, W. and KEESTER, K. 1966 Optical absorption spectra of iron in the rock forming s i l i c a t e s . The American Minerologist, Vol. 5, p. 774-791. WHITFIELD, W. and FURLEY, P. 1971 The relationship betueen s o i l patterns and slope form in the Ettrick Association, Southeast Scotland, in: Slopes, Form and Process (Brunsden, D., ed.), Inst, of B r i t . Geogr., Spec. Publ., IMa - 3, p. 165-174. WHITTAHER, R.H. 1962 Classification of natural communities. Botan. Rev., Vol. 28, p. 2-239. WIEGAND, C, LEAMER, R., WEBER, D., GERBERMANN, A. 1971 Multibase, multi-emulsion space photogr. for crops and s o i l s . Photogr. Eng., Vol. 37, N D . 3, p. 147-156. WIEGAND, C , RICHARDSON, A., GAUSMAN, H., LEAMER, R., GERBERMANN, A., EVERITT, J., CUELLAR, J. 1975 Soil salinity detection. Final Report for L.B. Johnson Space Centre, Houston, Texas, Agric. Research Service, USDA., Weslaco, p. 1-41. WILCOX, J.C., HOLLAND, W.D.,„McDOUGALL, J.M. 1957 Relation of elevation of a mountain stream to reaction and salt content of uater and s o i l . Can. Journ. Soil Sci., Vol. 37, p. 11-20. WOOD, W.F. and SNELL, J.B. 1960 A quantitative system for classifying landforms. H.Q. Quartermaster Research and Engineering Command, Techn. Report EP-124, Natick, Mass. WRIGHT, R. 1972 Principles in a geomorphological approach to land c l a s s i -fication. Zeitschrift fur Geomorphologie, N.F. 16, No. 4, p. 351-373. WRIGHT, R. 1973 An evaluation of the value of site analysis in f i e l d studies in tropical Australia. Zeitschrift fur Geomorphologie, N-E. 17, No. 2, YAALON, D.H. 1975 Conceptual models in pedogenesis: can s o i l forming functions be salved. Geoderma, Vol. 14(3), p. 189-206. YAALON, D.H., JUNGREIS, C., KOYUMDJISKY, H. Distribution and reorganization of manganese in three catenas of Mediterranean s a i l s . Geoderma, Vol. 7(1/2), p. 71-78. YOST, E. 1969 The principles and applications of multispectral color aerial photography, in: Neu Horizons in Color Aerial Photography, p. 367-384. YOST, E. and WENDEROTH, S. 1971a Multispectral color for agriculture and forestry. Photogr. Eng., Vol. 37, No. 6, p. 590-604. YOST, E. and WENDEROTH, S. 1971b The reflectance spectra of mineralized trees. Proc. of the 7th Intern. Symp. on Remote Sensing af the Environment, p. 269-284. 240 YOST, E. and UENDEROTH, S. 1967 Multispectral color aerial photography. Photoqr. Eng., Vol. 33, No. 9, p. 1020-1040. YOUNES, H., ABDEL-AAL, R., HHODAIR, M., ABDEZ-SAMIE, A. 1974 Spectral reflectance studies on mineral deficiency in corn plant. Proc. of the 9th Intern. Symp. on Remote Sensing of the Environment, p. 1105-1125. YOUNG, A. 1969 Slope form and the s o i l catena in savanna and rainforest environment, j un : Geomorphology in a Tropical Environment, Occ. Pap. Br i t . Geomorph. Res. Group No. 5, p. 3-12. YOUNG, A. 1972 Slopes. Oliver and Boyd, Edinburgh, p. 252-257.
UBC Theses and Dissertations
Chemical terrain variability : a geomorphological approach using numerical and remote sensing techniques Schreier, Hanspeter 1976
If you are having trouble viewing or searching the PDF with Google Chrome, please download it here instead.
- 831-UBC_1977_A1 S33.pdf [ 22.12MB ]
- JSON: 831-1.0094324.json
- JSON-LD: 831-1.0094324-ld.json
- RDF/XML (Pretty): 831-1.0094324-rdf.xml
- RDF/JSON: 831-1.0094324-rdf.json
- Turtle: 831-1.0094324-turtle.txt
- N-Triples: 831-1.0094324-rdf-ntriples.txt
- Original Record: 831-1.0094324-source.json
- Full Text
Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url: