Characterizing soil moisture regimes, linear and nonlinear soil moisture-latent heat flux dependencies in observations and models

Evapotranspiration or latent heat flux (LE) controlled by variations in soil moisture (SM) is a crucial process affecting the moisture and energy balance at the land-atmosphere interface. With such an important role in the climate system, SM-LE coupling is conventionally examined in the study of land-atmosphere interactions. Studies commonly identify SM-LE relationships by metrics involving one of two statistical approaches: (1) Correlation analysis: fitting a straight line onto the SM-LE space to evaluate the dependency; (2) Breakpoint analysis: determining the critical SM values that separate SM-LE space into different regimes in which land-atmosphere coupling bears fundamentally different behaviors. Beyond these, recent studies have also drawn attention to nonlinear SM-LE dependencies using mutual information analysis. However, at a global view, there is a lack of research that has examined SM-LE relationships merging these three aspects: linear dependency, nonlinear dependency, and SM-LE breakpoints. In this study, we combine these concepts to diagnose more fully SM-LE coupling using data sets from climate models, reanalyses, and observationally constrained data. The global pattern of critical soil moisture values is determined by breakpoint analysis. Then, mutual information analysis is applied on days when SM is within transitional regimes, a range wherein LE is positively sensitive to SM variations, to quantify the SM-LE dependency. This relationship is further decomposed into linear and nonlinear components. Our results show discrepancies among data sets for the global pattern of existing SM regimes, but consistencies in the distributions of linear and nonlinear SM-LE coupling. This implies that although models may have climate biases, the inherent behavior of how LE interacts with SM is rather well-described. The study provides new insights into targets for coupled land-atmosphere model development.