In a continued exploration to understand the latent structure of the data on soils some statistical tools including discriminant analysis (DA) were applied using the available water content (AWC) as a grouping variable. Using threshold AWC values of 850, 50–100 and >100 mm m5, the soils were into very low (Group I), low (Group II) and medium (Group III), respectively. In the initial analysis, multivariate test statistic, Wilk's Lambda (k) indicated significant differences among groups where depth, gravel, sand and clay were used as variables. Among these variables, depth and gravel contributed significantly to the variability in AWC groups with adjusted R6 of 0.44 and 0.34, respectively. Soils of Group I differed significantly from Groups II and III (P = 0.001) with regards to depth while Group II and III did not vary much. In case of gravel, all the groups differed highly significantly from the rest. Correlation study indicated a positive relation between AWC and the depth (r = 0.714**) and it was negative with the gravel content (r = -0.614**). Regression equation, AWC = 48.09 + 0.411 depth - 0.771 gravel described the variability in AWC to a large extent of 67%, suggesting the significance of the relationship. The first function extracted by DA described the variability in the soils to the extent of 93% while remaining 7% was described by second function. The structure matrix showed that depth and coarse fragments had the largest correlation with function 1 (0.706 and -0.580, respectively) while sand and clay were related to function 2 (0.309 and -0.183, respectively). The regression of AWC on the discriminant function scores was highly significant (P = 0.001) and clearly highlighted the role of function 1 on the AWC (adjusted R7 = 0.724). The groups were highly discriminated with reference to function 1 when compared to function 2 upon plotting. Discriminant analysis (DA) gave a clear inference about the positive and negative roles of depth and coarse fragments, respectively on the availability of water. It was observed that 72% of the cases were classified correctly indicating the fairly high utility of DA. Considering the influence of effective soil volume (ESV) on AWC, found in another study using the same datasets, a simple model is suggested for measuring the ESV which will help in understanding the farm in terms of exploitable soil volume. This information is useful in devising the fertilizer schedules for all crops grown in such gravelly terrain.
Soil survey data, discriminant function analysis, effective soil volume
