School of Civil and Environmental Engineering

Ph.D. Thesis Defense Announcement

Use of Satellite Data for Improved Wetland Modeling and Management

By

Courtney Di Vittorio

Advisor:

Dr. Aris Georgakakos (CEE)

Committee Members:

Dr. Rafael Bras (CEE/EAS), Dr. Jingfeng Wang (CEE), Dr. Yao Xie (ISYE), Dr. Augusto Getirana (University of Maryland)

Date & Time: Monday, December 2nd, 2019, at 4:00 PM

Location: Carnegie Building, Room 122

Complete announcement, with abstract, is attached

Decision support tools used in water resources management allow stakeholders to make informed decisions and assess the trade-offs of alternative management options. However, these tools are typically driven by the regulation of lakes and reservoirs and often do not fully incorporate the dynamics of wetlands. Wetlands provide important socio-economic and environmental services that should be protected and sustained. However, the cost associated with collecting enough in-situ data to accurately characterize the hydrology of many wetlands impedes their integration into hydrologic models. This study investigates how remotely sensed information (from satellites) can be leveraged to develop and calibrate hydrologic models of wetlands in data scarce areas. The value of this research is demonstrated for the Sudd Wetland, a vast seasonal wetland located in South Sudan that is a critical component of the Nile River hydrology. 
 
The Sudd hydrologic model was developed through an iterative approach where the complexity of the model was increased incrementally, following holistic comparisons between the model fluxes and states and the relationships between them. The wetland model was calibrated to the Sudd outflows and flooded area extents, using an objective function that measures how well the model can simulate these fluxes in terms of both magnitude and timing. The Sudd inflows and outflows were estimated from a combination of in-situ river flow data, satellite altimetry measurements, a hydrologic rainfall-runoff model, and multiple autoregressive river routing models. Monthly estimates of the Sudd flooded area extents were derived primarily from MODIS (MODerate Resolution Imaging Spectrometer) optical satellite imagery, using a new wetland land cover classification and inundation mapping procedure that was developed in this research. The Sudd model also relied on gridded precipitation and evapotranspiration estimates. Multiple satellite-based data sources were leveraged to estimate these fluxes, and accuracy was assessed by how consistent the precipitation and evapotranspiration rates were with the overall Sudd water balance. Uncertainties ingrained in the model structure, calibration parameters, and satellite-derived hydrologic data were considered jointly when assessing model performance and introducing new processes and more complex relationships in subsequent model formulations. The final model can simulate the estimated outflows with sufficient accuracy (a Nash-Sutcliffe Efficiency coefficient of 0.75 was achieved); however, discrepancies remain between the simulated and MODIS-derived flooded area extents. Following a thorough analysis of the advancing and receding patterns of the inundation maps, these discrepancies are believed to be related to the connectivity of the full flooded area extents to the main Sudd water body that regulates the outflows. Recommendations for future research are made that might improve the satellite-based wetland inundation maps and hydrologic flux estimates, and the underlying structure of the Sudd hydrologic model.