Random forest for identification and characterization of groundwater dependent ecosystems

Abstract

Anthropogenic actions such as groundwater pumping, agricultural practices, industrialization, and waste disposal can greatly affect groundwater resources which would eventually drive changes in vulnerable ecosystems. Therefore, it is clear that there is a need to identify the locations of groundwater dependent ecosystems (GDEs) to enable the development of policies that adequately address their protection. The purpose of this study is to propose a method based on geospatial data sets and random forest algorithm to map the distribution of GDEs in the United States at 1 km spatial resolution. This paper presents the results in Nevada. The method is based on the principle that ecosystems will use water in proportion to its availability and the dependence on that resource will be expected to increase with higher aridity of the environment. Results show that random forest is a promising technique for the identification and characterization of GDEs using geospatial data sets as predictor variables.