Multivariate ordination approach for identification of sub-regional homogeneities in Gujarat, western India

Identification of homogeneous invariant geographical units assumes considerable importance in the context of current attempts at developing environmental indices to monitor changes. Data on the edaphic and climatic parameters for the state of Gujarat—a predominantly arid to semi-arid province in western India—has been used to elaborate an approach that employs principal component analysis (PCA) as the basis for unbiased clustering of similar geographic units at sub-regional scales. PCA identified different sets of edaphic and climatic variables responsible for separating one unit from another. With the exception of a few cases, soil variables exhibited greater influence on the separation of units than climate. Response of some agro-ecological variables (total cultivable area, forestland and normalized difference vegetation index) to both original and ordinated variables in each of the identified edaphic-climatic units (ECU) has been evaluated using step-wise multiple regressions. The edaphic-climatic parameters were good predictors of these variables, although ordination did not necessarily enhance the predictive power of the independent variables. Instead, in many cases original variables showed better correlations. The response of these variables, chosen on the basis of varying levels of dynamism (slow to seasonal fluctuations), showed that independent of the degree of dynamism, the dependence of these variables was distinct in each ECU. This provided a reality check on the agro-ecological and bio-climatic differentiation among the ECUs.