Short and Medium Range Forecast of Soil Moisture for the Different Climatic Regions of India Using Temporal Networks

Spatio-temporal evolution of soil moisture is a complex process and controlled by several factors including hydro-meteorological forcings. This study borrows a recently developed concept of temporal networks to capture the time-varying association between hydro-meteorological forcings and spatio-temporal evolution of soil moisture. Climate change and dynamic terrestrial environment cause slow but continuous change in the characteristics of hydro-meteorological forcings leading to variation in spatio-temporal distribution of soil moisture. Keeping this in the focus of the study, temporal networks based time-varying modelling framework is adapted for one-month to one-season (three-months) in advance prediction of monthly soil moisture for entire Indian mainland. Results indicate that the association among the hydro-meteorological forcings varies with both space and time. With the increase in prediction lead-time, the strength of association with the variables, such as pressure, wind and temperature, decreases and that with the variables like leaf area index remains informative. Among different seasons, the model shows superior performance for the monsoon and post monsoon periods. Next, the soil moisture based extremes are assessed by utilizing two deficit indices and two wetness indices. The model performance is highly satisfactory but varies over space and seasons, with a marginally better performance for the wetness indices. Overall, given the vast spatial extent of the Indian mainland, the proposed model performs robustly for almost all the climatic regions and may be promising for other parts of the world as well.