2nd UF Water Institute Symposium Abstract

   
Submitter's Name Syewoon Hwang
Session Name Poster Session: Managing Water and Energy in a Transitioning Environment 1
Category Managing water and energy in a transitioning environment
Poster Number 309
 
Author(s) Syewoon Hwang,  Univerisyt of Florida
  Wendy Graham,  Water Institute, Univeristy of Florida
  Jose Hernandez, University of Florida
  Christopher Martinez, University of Florida
  James Jones, University of Florida
   
  Assessment of Mesoscale Dynamical Downscaling Model (MM5) for Regional Climate Simulation in the Tampa Bay region
   
  This research analyzes the temporal and spatial variability of historic precipitation in Tampa Bay region and evaluates the ability of the mesoscale downscaling model (MM5, Grell et al., 1994), to reproduce this variability. The long term goal of this effort is to evaluate the utility of using MM5 to downscale GCM forecasts and climate change scenarios for improving water management decisions in the Tampa Bay region. Cumulative probability distributions were constructed using observed daily and monthly rainfall at each station, and the spatial correlations between the 53 stations were analyzed for each month using covariance and variogram analysis for both observed data and MM5 predictions. MM5 was run to predict precipitation at 9x9 and 27x27 km2 spatial resolutions and 6-hour temporal resolution over the 23 year period from 1986 to 2008 using the NCEP/NCAR reanalysis data set as initial and boundary conditions. The raw precipitation predictions were then bias-corrected at each observation station using the cumulative probability distribution mapping approach (Wood et al., 2002). Daily and monthly precipitation totals were estimated over the Alafia and Hillsborough River watersheds using the bias-corrected point precipitation and observed variogram functions. MM5 performance was assessed by cross-validating predicted daily and monthly point and total watershed precipitation for each month. Variograms from the bias-corrected daily precipitation predictions in general indicated that MM5 overestimates the strength of the spatial correlation and underestimates the variance of precipitation compared to the observed data, especially in the summer months when convective storms dominate. The simulations for each month reproduced the daily mean point precipitation values with an average error of -0.0641 in (Jul.) to 0.0214 in (Oct.) with an average RMSE of 0.6834 in (Mar.) to 0.9449 in (Sep.) over the 53 rain stations. Monthly mean point precipitation values were reproduced with an average error of -0.7110 in (Jun.) to 0.2732 in (Aug.) with an average RMSE of 1.8154 in (Mar.) to 5.1194 in (Sep.) over the 53 stations stations. Point kriging the bias-corrected daily precipitation fields over the watersheds reproduced the observed rainfall with an average RMSE of 0.3550 in (Jan.) to 0.6977 in (Sep.) over the 53 stations. Block kriging the bias-corrected daily precipitation fields over the watersheds reproduced the observed block kriged rainfall with an average RMSE of 0.3104 in (Feb.) to 0.5932 in (Sep.) over the 23 years. In all cases actual kriging errors were well predicted by the kriging standard deviation estimate. In the next phase of this research the methodology developed here will be used to produce spatially distributed bias-corrected precipitation estimates from downscaled MM5 predictions that use GCM forecasts and IPCC scenarios as boundary conditions. These precipitation fields will be subsequently be used in a hydrologic model to predict streamflow response to climate fluctuations and climate change scenarios in order to improve the operation of water supply reservoirs in the region.