A nonlinear Granger causality framework to identify the effect of soil moisture on precipitation: A case study of the USA

Soil moisture influences precipitation through its impact on water and energy cycles. Understanding of soil moisture-precipitation (SM-P) coupling is important for improving weather forecasting. However, the sign and pattern of SM-P feedback are still controversial, mainly caused by the difficulty to establish causal relationship and the highly nonlinearity in land-atmosphere process. To address these two problems, we established an innovative nonlinear Granger causality framework which is based on regression modelling by machine learning, time series decomposition techniques, spatial impact modeling, a hybrid feature selection method and nonlinear Granger causality test. As an example, we used this framework to identify sign and hot spot of SM-P feedback over USA at the daily scale. The results highlighted the importance of nonlinear atmosphere response in land-atmosphere dynamics, and identified hot spots of SM-P coupling, which can help quantify the role of land surface states in weather and climate predictability. We envision this innovative framework to improve weather forecast, climate, and Earth system models worldwide, and can offer new perspectives in various feedbacks of Earth systems.