Incorporating Evapotranspiration Processes in the Rainfall–Runoff–Inundation (RRI) Model and Validating the Model Outputs with the MODIS and GLEAM Evapotranspiration Products

Distributed Hydrological Models (DHMs) have become essential tools for generating information on water resources and water-related disasters under global change (e.g., land-use and climate changes) to support decision making strategies for sustainable development. The Rainfall-Runoff-Inundation (RRI) model, which is a water-budget model, is used widely for event-based simulations for flood hazard and risk assessments. The soil-vegetation Evapotranspiration (ET) is a key component of the water budget and its accurate estimation within DHMs is very important for simulating accurate soil-vegetation water storages, surface runoff, peak river discharges, flood depths, and inundation extents. However, it has not been incorporated within the RRI model explicitly. Therefore, the applications of RRI model are limited to flood hazard and risk assessments of the past flood events with the aid of individual event-based calibrations. The absence of ET estimation is recognized as a major drawback for the applicability of the RRI model to water resources management as well as to water-related disaster management under global change. Accordingly, this research coupled a land surface model (Hydro-Sib2) with the RRI model’s 2-D diffusive wave equations to incorporate land-atmosphere-vegetation interaction and thereby estimating soil-vegetation ET components. The new model, the Water and Energy Budget-based Rainfall-Runoff-Inundation (WEB-RRI) model, was calibrated and validated using a long-term (~20 years) river discharge data in the Kalu River basin, which is located in a wet climatic region, and in the Mundeni River basin, which is located in a dry climatic region, in Sri Lanka. The model-simulated ET outputs were compared with the MODIS (MOD16) and GLEAM ET products during 2002-2009. The results showed that the simulated 8-day trends in wet and dry regions agreed well with both MODIS and GLEAM. Particularly, the wet and dry cycle is well captured by the model and comparable to both validation datasets. The model performance indices showed better results with GLEAM (RMSE: 0.7 mm/day - 0.95 mm/day, correlation: 0.35 – 0.39,) than MODIS (RMSE: 0.96 mm/day - 1.04 mm/day, correlation: 0.14 – 0.25). The biases can be attributed to errors in rainfall amount and its distributions, the missing values in MODIS data during the cloudy period, error generated in the retrieval algorithms of MODIS and GLEAM, MODIS LAI data, meteorological or reanalysis inputs as model forcing data. As the WEB-RRI model overcomes the limitations of the RRI model, it will be used for solving issues in integrated water resources management as well as water-related disaster (i.e., floods and droughts) management under the water cycle variability as well as global change in future studies.