Simulating Inundated Land Area in the Red River Basin using the VIC model within the NASA Land Information System

It is a challenge to estimate the volume of water detained on the land surface during ponding events for river flow forecasting on the high plains of the US upper Midwest. Affected by frozen soil physics and mechanisms of drainage tiles and culverts, the process of surface water detention and delayed infiltration is not represented in the operational river forecasting system at the National Weather Service (NWS). Current and prior operational efforts to predict flood crests on the Red River have at times exhibited large errors, thus it is important to improve our understanding of ponding dynamics through modeling approaches, with an ultimate goal of improving the process modeling basis for river forecasting.    

Our study approaches this objective through the application of a macro-scale hydrology model as constrained by remotely sensed imagery of inundated land area (ILA).  The Variable Infiltration Capacity (VIC) model has integrated physics for inland lake and surface storage. Isolated lakes and wetlands within each VIC grid cell are represented as a land cover class with an associated physical parameterization that simulates the water and energy balances of permanent water bodies and seasonally flooded ground.  Here we implement VIC within the NASA Land Information System (LIS), a flexible land surface modeling and data assimilation framework designed to integrate satellite- and ground-based observations, land surface and hydrological models, and advanced data assimilation techniques to produce optimal fields of land surface states and fluxes. For this study, the VIC 4.1.2.k has been implemented into the NASA LIS version 7.1 to take advantage of its modeling infrastructures, such as a variety of forcing data readers, data assimilation routines, a flow routing algorithm and parallel computation.

We present a case study of simulating ILA in the Red River basin. The study area has an extremely low slope and land use is extensively agricultural with almost no trees and other dense vegetation. The parameters of VIC ponding algorithms, such are the depth-area relationship, are derived based on high-resolution DEM data and topographic indices. Inundation related state variables, such as ponding area and water depth, are simulated using forcing data from the North American Land Data Assimilation System (NLDAS) project. The applicability and performance of the large-scale VIC model on simulating inundated land area are evaluated against MODIS ILA observations as well as observed streamflow data.  This study helps to establish a foundation for improving model simulation of ILA in support of improved streamflow prediction.