Improving the catchment scale wetland modeling using remotely sensed data

Authors: LEE, SANGCHUL - University Of Maryland; YEO, IN - University Of Newcastle; LANG, M.W. - Department Of Fish And Wildlife; McCarty, Gregory; Sadeghi, Ali; SHARIFI, AMIR - University Of Maryland; JIN, H. - University Of Maryland; LIU, Y. - University Of Maryland

Submitted to: Journal of Environmental Modeling and Software
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 6/15/2017
Publication Date: 7/1/2017
Citation: Lee, S., Yeo, I., Lang, M., McCarty, G.W., Sadeghi, A.M., Sharifi, A., Jin, H., Liu, Y. 2017. Improving the catchment scale wetland modeling using remotely sensed data. Journal of Environmental Modeling and Software. https://doi.org/10.1016/j.envsoft.2017.11.001.
DOI: https://doi.org/10.1016/j.envsoft.2017.11.001

Interpretive Summary: The potential to routinely map inundated of wetlands at a 30-m resolution with high accuracy (> 90%) was recently demonstrated using Light Detection and Ranging (LiDAR) and time series Landsat data. The mapping algorithm was applied to detect inundated area for small seasonal forested wetlands, which are densely distributed in the Coastal Plain of the Mid-Atlantic Region of the US. Forested wetlands are difficult to map, due to the presence of the forest canopy and the fact that inundation and saturation in these wetland is often ephemeral. Inundation maps were produced annually using Landsat images and LiDAR intensity data acquired in early spring (Jin et al., 2016), when most standing water within a wetland was viewable due to the leafoff status of forest canopy. When applied to watersheds in the Coastal Plain Physiographic Province of the Chesapeake Bay Watershed (CBW), spatial patterns of inundation reflected different weather conditions and followed the change trend of streamflow. In this paper, we presented an integrated wetland-watershed modeling framework that capitalizes on time series inundation maps, with an aim to improve prediction on wetland inundation. We outlined problems commonly arising from input data preparation and parameterization to represent spatially aggregated wetlands within a distributed watershed model. We demonstrated how intra-watershed processes can be better captured by setting spatialized wetland parameters developed from remotely sensed data, as it reduced the degree of model overfitting. We then placed particular emphasis on assessing model prediction using spatial maps of inundation under different weather conditions. This spatial data-model integrated framework was tested with SWAT and Riparian Wetland Module (RWM), an improved SWAT extension for riparian wetlands (RWs). The effects of improved parameterization and process representation on predicted inundation are demonstrated using a case study conducted in the upper river basin of Choptank Watershed, located in the Coastal Plain of the CBW.

Technical Abstract: This study presents an integrated wetland-watershed modeling framework that capitalizes on inundation maps and other geospatial data to improve spatial prediction of wetland inundation and assess prediction uncertainty. We outline problems commonly arising from data preparation and parameterization used to simulate wetlands within a distributed watershed model. We demonstrate how intra-watershed processes can be better captured by spatialized wetland parameters developed from remotely sensed data, because their use reduced the degree of equifinality. We then emphasize assessing model prediction using inundation maps derived from remotely sensed data. This spatial data-model integrated framework is tested using the Soil and Water Assessment Tool (SWAT) with an improved riparian wetlands (RWs) extension, for an agricultural watershed in the Mid-Atlantic Coastal Plain, US. This study illustrates how spatially distributed information is necessary to predict inundation of wetlands and hydrologic function at the local landscape scale, where monitoring and conservation decision making take place.