Files
Abstract
An understanding of the land surface heterogeneity has been deemed important for modeling earth system processes and developing a mechanistic understanding of soil hydrology. In this study, we have developed a Heterogeneity Index (H-index) that quantifies the soil, topography, and seasonal vegetation based heterogeneity at a global scale. Our study follows the scaling nomograph which was introduced to incorporate the scale and site-specific dependence of soil moisture on geophysical heterogeneity and antecedent wetness conditions at the regional scale. The H-index is based on eigenvalue decomposition of sand, leaf area index (LAI), and flow accumulation, hence quantifies the sub-grid scale variability and co-variability of land-surface heterogeneity for large-scale hydrology. The H-index has been modified for global-scale analysis. The H-index has been found to respond to the changes in vegetation cover and can be used to assess the change in land-surface heterogeneity over time. We introduce a modular and efficient algorithm for index generation that processes partially available data-sets such as region-specific elevation from regional LIDAR surveys and improvements in global or regional soil texture maps. Additionally, the formulation of the H-index has the potential to be incorporated with earth system models. In this work, we present the computed H-index for the period of 2002-2020 and provide a comparison of the H-index with available classification measures such as the Major Land Resource Areas (MLRA) and Common Resource Areas (CRA). To demonstrate hydrologically similar regions we cluster Soil Water Retention Parameters (SWRPs) using unsupervised machine learning techniques including K-means clustering, spectral clustering, Gaussian Mixture Model, and Hierarchical clustering. Results show adequate cluster separation that implies a successful classification of near-surface soil moisture dynamics by H-index. Moreover, it demonstrates the potential of replacing representative elementary volume (REV) driven soil hydrology classifiers such as porosity to describe soil hydrology at the remote sensing scale.