Application of earth observation and related technology in agro-hydrological modeling

Freshwater is vital for life on Earth, and as the human population continues to grow so does the demand for this limited resource. However, anthropogenic activities and climate change will continue to alter freshwater systems. Therefore, there is a need to understand how the hydrological cycle is changing across the landscape. Traditionally, this has been done by single point monitoring stations; however, these stations do not have the spatial variability to capture different aspects of the hydrologic cycle required for detailed analysis. Therefore, hydrological models are traditionally calibrated and validated against a single or a few monitoring stations. One solution to this issue is the incorporation of remote sensing data. However, the proper use of these products has not been well documented in hydrological models. Furthermore, with a wide variety of different remote sensing datasets, it is challenging to know which datasets/products should be used when.To address these knowledge gaps, three studies were conducted. The first study was performed to examine whether the incorporation of remotely sensed and spatially distributed datasets can improve the overall model performance. In this study, the applicability of two remote sensing actual evapotranspiration (ETa) products (the Simplified Surface Energy Balance (SSEBop) and the Atmosphere-Land Exchange Inverse (ALEXI)) were examined to improve the performance of a hydrologic model using two different calibration techniques (genetic algorithm and multi-variable). Results from this study showed that the inclusion of ETa remote sensing data along with the multi-variable calibration technique could improve the overall performance of a hydrological model.The second study evaluated the spatial and temporal performance of eight ETa remote sensing products in a region that lacks observed data. The remotely sensed datasets were further compared with ETa results from a physically-based hydrologic model to examine the differences and describe discrepancy among them. All of these datasets were compared through the use of the Generalized Least-Square estimation with Autoregressive models that compared the ETa datasets on temporal (i.e., monthly and seasonal basis) and spatial (i.e., landuse) scales at both watershed and subbasin levels. Results showed a lack of patterns among the datasets when evaluating the monthly ETa variations; however, the seasonal aggregated data presented a better pattern and fewer variances, and statistical difference at the 0.05 level during spring and summer compared to fall and winter months. Meanwhile, spatial analysis of the datasets showed that the MOD16A2 500 m ETa product was the most versatile of the tested datasets, being able to differentiate between landuses during all seasons. Finally, the ETa output of the model was found to be similar to several of the ETa products (MOD16A2 1 km, NLDAS-2: Noah, and NLDAS-2: VIC).The third study built upon the first study by expanding the use of remotely sensed ETa products from two to eight while examining a new calibration technique, which was the many-objective optimization. The results of this analysis show that the multi-objective calibration still resulted in better performing models compared to the many-objective calibration. Furthermore, the ensemble of all of the ETa products produced the best performing model considering both streamflow and evapotranspiration.