Improving the Computational Effciency of Downscaling GCM data for Use in SWAT
Links to Fileshttps://userpages.umbc.edu/~gobbert/papers/REU2014Team2.pdf
MetadataShow full item record
Type of Work9 pages
RightsThis item is likely protected under Title 17 of the U.S. Copyright Law. Unless on a Creative Commons license, for uses protected by Copyright Law, contact the copyright holder or the author.
This project creates software tools to streamline the computational procedures to generate high-resolution weather parameters from low resolution Global Climate Models (GCM) to be input into the Soil and Water Assessment Tool (SWAT) and to visualize GCM temperature and precipitation data as well as SWAT outputs of crop data. Data from GCMs have relatively low spatial resolution ( 100km x 100km), which needs to be downscaled to a higher resolution to match the resolution of the observed data so that both the data sets can be input into SWAT. We consider several decades of historical simulations from GCMs, and surface-based observations of historical temperature and precipitation data over the Missouri River Basin (MRB). The downscaling method involves two steps: i) bilinear interpolation to ll in values at higher resolution and ii) Linear or Tobit regression between the GCM and the observed data to accurately capture the features of observed data in the GCM temperature and precipitation at high resolution. This downscaled data is then used to generate forecasts, which are used as inputs for SWAT. Based on these statistical downscaling methods, the SWAT outputs are compared. Since the data is large, we focus on maximizing computational e ciency through parallelization of the data generation, model tting and forecasting. A graphical R interface is developed to facilitate the modeling and visualization components at each step. The interface allows climate models and SWAT outputs to be more easily compared for di erent scenarios, thus help assess climate variability and its impact on crop yields over MRB.