Investigating the Runoff Disparities in Land Surface Models

Land Surface Models (LSM) provide a vast range of runoff estimates. Although many LSMs applications are not focused on land hydrology, runoff is essential for the water balance closure and it should be better looked after. As LSMs become increasingly complex it becomes harder to identify what causes the disparity between the models. Therefore we propose a study to improve the understanding of the estimated runoff from 6 LSMs. We investigate the influence of some land features on daily runoff time series (hydrographs). To do so, we created a metamodel for each LSM, different land cover (LC) classes, and clay- and sand-predominant soils. The LSMs were taken from the Earth2Observe (Htessel, Jules, Orchidee and Surfex) and GLDAS (Noah and Catchment) projects and have a spatial resolution of 0.25o x 0.25o. The metamodel structure consists of a slightly modified LSTM that includes topography, daily evapotranspiration variables, and rainfall + snowmelt as input. The metamodels were trained with 18 cells and 10 years duration and validated with the same number of cells and years. We defined 20 classes of land cover applying k-means to the original land cover maps of each LSM to obtain a comparable set between the models. Therefore, there will be 240 metamodels (6 LSMs x 20 LC classes x 2 soil textures) from which we can provide hydrographs given controlled meteorological conditions. With this tool available we can observe the runoff response to a rainfall event after changing prior soil conditions. For example: a) a dry soil condition (several days without rain); b) a wet soil condition (several days of rain); c) or how fast the soil becomes completely wet/dry (no changes on the runoff event from that day onwards). Then, we can evaluate the importance of the soil texture compared to the structure of the LSMs for each class of land cover, specifically focusing on indices such as total volume, peak volume, and recession coefficient. So far, we already observed a one-order-of-magnitude difference in the subsurface runoff of the Catchment LSM compared to the others. With this analysis, we expect to obtain insights into what is causing such great disparities in runoff estimates and point towards paths of higher agreement.