Analysis of variability of some forest soils in southwestern Oregon

Graduation date: 1981The variability of selected physical, chemical, and morphological\ud soil properties in two landtype mapping units on'the Rogue River National\ud Forest in southwestern Oregon was studied.\ud The objectives of the study were (i) to quantify soil variability\ud in several soil resource inventory mapping units, (ii) to explore methods\ud of describing soil variability in order to make soil map information more\ud useful to the map user, and (iii) to attempt to identify sources of\ud variability.\ud Two mapping units, in the Siskyou and Cascade Mountains, were\ud selected to represent extremes of internal variability. Eight delineations\ud of each map unit were sampled with randomly located transects for\ud a total of 40 sites per map unit.\ud Soil properties exhibited various types of frequency distributions.\ud Normal, skewed, and multi-modal distributions were observed. Nearly half\ud the property-horizon combinations measured in both map units had normal\ud distributions. Chemical properties, such as extractable bases, were\ud consistently positively skewed or approximately log-normal. Square root\ud and logarithmic transformations of the data normalized these distributions\ud and stabilized the variance. These results suggested that for\ud multi-modal and badly skewed populations, assumption of a normal\ud distribution may lead to considerable error if the arithmetic mean and\ud standard deviation are used for predictive purposes.\ud Map unit 74 in the Siskyou Mountains was considered to be more\ud variable, over most properties measured, than map unit 33 from the\ud Cascades. For most properties coefficients of variation (CV) were higher,\ud the sample requirements to estimate population means were greater, and\ud the ranges were wider in map unit 74, as compared to map unit 33.\ud In both map units, chemical properties were more variable than\ud physical or morphological properties, which were about equal in their\ud variability.\ud The number of samples required to estimate the means of properties\ud varied widely and were often prohibitively large (264 for organic matter\ud in the surface of map unit 74). This number could be reduced if the\ud sampling scheme was stratified using estimates of within and between\ud delineation variance.\ud Between 50 and 75 percent of the total variation in most properties\ud of both map units occurred within delineations. This result is desireable\ud from a management and broad planning perspective and tends to support\ud the validity of the map units as designed and mapped.\ud When tested by analysis of variance, most properties in both map\ud units had significantly different delineation means. These differences\ud could often be traced to one particular delineation and were many times\ud not of practical significance.\ud Large or small changes in the values of most properties were found\ud to be as likely to occur at 660 foot separation distances as at 15 foot\ud distances. This tends to indicate a random distribution of variation,\ud when distance alone is considered, which could have important consequences when attempting to characterize soil properties for management interpretations\ud or site evaluations.\ud Chemical properties expressed volumetrically were more variable\ud than the same properties expressed on a weight basis. Volumetric chemical\ud concentrations were greater in map unit 74 than in map unit 33, the\ud reverse of the relationship found with those properties on a weight basis.\ud The proportions of within delineation variance in map unit 74 increased\ud as a result of conversion to volumetric concentrations.\ud Use of principal components analysis confirmed the chemical properties\ud as contributing greater amounts of variation than the physical or\ud morphological properties. Ordination of sites along axes of selected\ud factors also confirmed the general uniformity of the map units as well as\ud the greater variability of map unit 74