Examining the benefits and sensitivity of vertically heterogeneous hyper resolution land surface model towards simulating a farm scale soil moisture profile in Upper Bhima Basin, India

Soil moisture is an important state variable with high spatiotemporal variability depending on land and climate variables. The importance of various physical controls on soil moisture varies depending on the scale and extent of the study. At a fine scale, soil properties are proven to be critical in defining spatiotemporal variability of soil moisture. In the context of agricultural applications in India, soil moisture estimates at the farm scale (finer spatial resolution) over various root depths are essential.  Traditional Land Surface Models (LSMs) are limited to large spatial scales (in the order of tens of kilometers). They have been designed for synergistic coupling with Earth system models. Besides, they do not account for the vertical heterogeneity of soil. LSMs, including Noah-MP, use a lookup table to obtain soil properties corresponding to soil texture while assuming vertically homogeneous soil texture. Recent studies proved that accounting for vertical heterogeneity in the soil using state-of-art soil maps and pedotransfer functions in LSM can significantly improve the surface soil moisture estimations. However, the effects of incorporating vertical heterogeneity in soil properties on deeper layer soil moisture simulations are yet to be explored. Considering the importance of farm scale root water uptake processes, understanding soil moisture heterogeneity at deeper layers is essential. In this context, the present study hypothesizes that a hyperresolution LSM, which accounts for subgrid spatial heterogeneity while maintaining soil heterogeneity between layers, can improve the characterization of rootzone soil moisture. 

In this work, we used HydroBlocks, a semi-distributed hyper-resolution LSM, which uses Noah-MP at its core, and the concept of Hydrologic Response Units (HRU) to simulate the land surface variables. The analysis is carried out for the first time in India over the Upper Bhima Basin, for the year 2020. Initially, we investigated the benefit of vertical heterogeneity in soil properties to simulate soil moisture at five different layers till one meter deep using HydroBlocks. We used SoilGrids data for different layers to calculate soil hydraulic properties using PTFs and feed them as inputs in the HydroBlocks model. We compare HydroBlocks surface and rootzone soil moisture to existing reanalysis and satellite products, including GLEAM, ERA5-Land, SMAP-L3, and SMAP L4 statistically in terms of bias, ubRMSE and R². Further, an intercomparison of surface and rootzone soil moisture simulations is made between the two cases of Hydroblocks model, first with vertically homogeneous soil properties, and second, with vertically heterogeneous soil properties. The objective of this step is to emphasize the role of vertically heterogeneous soil layers in a hyper-resolution LSM towards addressing the spatiotemporal variability of soil moisture. Finally, a soil parameter sensitivity analysis (using Sobol technique) is carried out using HydroBlocks for five soil layers (up to 1 meter depth), for the first time, to assess the influence of eight soil textural parameters such as wilting point, porosity, pose size distribution, and likewise, on soil moisture simulations. In this process, we also assessed the seasonal variability of parameter sensitivity.