Extracting soil water storage pattern using a self-organizing map

Abstract

Soil water is one of the important controlling factors regulating plant growth in a semi-arid environment. It is distributed non-uniformly and exhibits spatial patterns in the landscape. Research has identified different controls of soil water in wet and dry state; however, there are no objective methods for differentiating the dry and wet states. The objective of this study was to extract the temporally persistent spatial patterns of soil water using a Self-Organizing Map (SOM) and identify benchmark sites for different patterns of soil water. Soil water storage along a transect of 576m at 4.5m intervals in a farm field in Saskatchewan, Canada was monitored from 2007 to 2009. A SOM was used to extract the patterns of soil water storage of the surface 0.2m depth compartment. Two patterns were identified: pattern 1 for dry states and pattern 2 for wet states, with the occurrence frequency of 52.9% and 47.1%, respectively. Pattern 1 was often found in summer and fall, and pattern 2 in early spring, and summer in wet years. Soil texture and organic carbon (OC) content showed strong correlation with soil water storage in dry states (pattern 1) and wet states (pattern 2). Based on the aforementioned finding, benchmark sites from each of the two patterns were identified, and they were not adjacent in terms of spatial location. Therefore, soil water states can be identified by using SOM and differentiation of wet and dry states enable identification of more representative benchmark sites.