Category: Research

Sponsors: National Science Foundation

Students: Courtney Di Vittorio

Description:

Hydrologic models of wetlands enable hydrologists and water resources managers to appreciate the environmental and societal roles of wetlands and manage them in ways that preserve their integrity and sustain their valuable services. However, wetland model reliability and accuracy are often unsatisfactory due to the complexity of the underlying processes and the lack of adequate in-situ data. In this research, we demonstrate how MODIS satellite imagery can be used to characterize wetland flooding over time and to support the development of more reliable wetland models. We apply this method to the Sudd, a seasonal wetland in South Sudan that is part of the Nile River Basin.

The database consists of 16 years of 8-day composite ground surface reflectance data with a 500-m spatial resolution. After masking poor quality pixels, monthly distributions of wetness and vegetation indices were extracted. Based on literature and personal accounts describing the Sudd as well as Google Earth imagery, a set of ground truth locations were identified for nine land classes. Using the ground truth locations, a novel classification procedure was developed that uses the empirical monthly distributions of each individual pixel. This procedure allows pixels to be classified as mixed pixels (or mixels) if their distributions share properties with two different classes. Once the full area of interest was classified, each pixel was evaluated on a monthly scale to determine if, when, and how long it was flooded using a procedure that incorporates spatial information and monthly precipitation data. The result is a set of monthly inundation maps for the full period of interest (2000–2015). An independent set of ground truth locations were selected to validate the land cover classification procedure, which demonstrated a high level of accuracy. The derived monthly inundation series agrees well with existing literature, limited ground observations, and estimated water fluxes into the wetland. This information is currently being used to develop a wetland model as part of a comprehensive modeling system for the Nile River Basin. This novel procedure is general and has many advantages over those in existing research for applications in data scare areas.

Publications:

C. A. Di Vittorio & A. P. Georgakakos. Land cover classification and wetland inundation mapping using MODIS. Remote Sensing of Environment 204, 1-17 (2018). https://doi.org/10.1016/j.rse.2017.11.001 

Acknowledgements:

We are grateful to the Ministries of Water and Irrigation in Uganda, Sudan, and Egypt for the provision of in-situ river flow and other hydrologic data.  Moreover, we are grateful to Dr. Georg Petersen for sharing his ground measurements on the Sudd Wetland. This research was supported in part by the National Science Foundation Graduate Research Fellowship Program under Grant No. DGE-1650044. Any opinions, findings, and conclusions or recommendations expressed in this article are those of the authors and do not necessarily reflect the views of the National Science Foundation or other organizations.

Effects of Water Use and Regulation

In watersheds with appreciable water use and regulation (including storage reservoirs, in-stream withdrawals, and/or inter-basin water transfers), the development of reliable ensemble streamflow predictions (ESP) at downstream locations requires characterization and incorporation of the expected streamflow alterations from natural conditions and their associated uncertainty.  Streamflow alterations can be incorporated if, as part of the ESP forecast generation process, water use and regulation activities are represented with sufficient accuracy.  This approach can be effective in watersheds where flow alterations occur due to large, main stem river projects and well documented water use activities, but it becomes impractical where flow alterations result from many small and/or medium scale storage projects and water use activities distributed throughout the watershed.  In the latter cases, comprehensive information on reservoir filling and depletion, water withdrawals and returns, and/or water transfers is both not readily available and subject to change from year to year, adding bias and uncertainty to the flow forecasts. This research project develops and demonstrates procedures to characterize the aggregate flow alteration biases and uncertainty in watersheds in the latter category and incorporate them in ensemble streamflow predictions at downstream points.

Project Scope and Activities

This research project develops and demonstrates a new method to characterize the aggregate flow alteration biases and associated uncertainty in watersheds with important but largely undocumented water use and regulation activities.

The approach includes procedures to (a) detect the presence of significant upstream regulation and water use influences; (b) correct the ensemble streamflow predictions and associated uncertainty for any biases during periods when upstream regulation and water use influences are detected; and (c) assess the forecast reliability improvements. Validation results are reported for three California watersheds. The forecast adjustment approach has been developed for operational use in routine forecast operations of the U.S. National Weather Service River Forecast Centers.  

Related Publications

1.    Georgakakos, A.P., H. Yao, and K.P. Georgakakos, “Upstream Regulation Effects on Ensemble Streamflow Prediction,” Journal of Hydrology, in press, 2014.

Acknowledgements

This project was funded by the Hydrologic Research Laboratory of the National Weather Service and was carried out jointly by the Hydrologic Research Center (HRC; California) and the Georgia Water Resources Institute (GWRI; Georgia) in collaboration with the NWS River Forecast Centers.