Improving Precipitation and Streamflow Forecasts for Amite River Basin in Louisiana

1
Improving Precipitation and Streamflow Forecasts
for Amite River Basin in Louisiana Ashutosh S.
Limaye, Global Hydrology and Climate Center
(GHCC), 320 Sparkman Drive, Huntsville, AL 35805
(Ashutosh.Limaye_at_msfc.nasa.gov) William L.
Crosson, GHCC James F. Cruise, Univ. of Alabama
at Huntsville and Keith Stellman, Lower
Mississippi River Forecast Center, Slidell, LA
Flooding is a major issue in the Amite River
basin (3100 Km2), which covers parts of Baton
Rouge, LA. The watershed has flooded more often
than any other in the region. In this study, we
looked at a major precipitation event resulting
from Tropical Storm Allison (June 2001). This
event produced significant amounts of rainfall in
Amite River basin. Parts of the watershed
received in excess of 25 inches of rain,
resulting in severe flooding and damages.
Forecast hourly precipitation based on storm
movements
Compute errors using Stage III
Use the error analysis in Kalman filter to
improve future forecasts
6-hr forecasts from HPC
Use hourly forecasts to improve LMRFC forecasts.
Combined forecast called Kalman forecast
6-hr LMRFC forecasts
Distributed hydrologic model SHEELS
Distributed hydrologic model SHEELS tracks water
movement in multiple soil layers as well as
overland flow using topography driven network of
grid cells. Soil moisture accounting is driven
by the energy exchange among ground, canopy and
the atmosphere.
Kalman Filter
Motivation Lower Mississippi River
Forecast Center (LMRFC) produces Stage III
precipitation estimates (fused product of NEXRAD
radar estimates with observed data). Rainfall
forecasts from NOAA Hydrometeorological
Prediction Center (HPC) are adjusted by LMRFC to
produce QPF. LMRFC QPF are made for every 6 hour
period. The objective here is to use a
Kalman filter assimilation scheme to improve the
precipitation and streamflow forecasts using 6-hr
LMRFC forecasts and real-time Stage III data. The
bias and RMSE comparisons shown here are similar
to the NWS National Precipitation Verification
Unit (NPVU) statistics. During TS Allison event,
the Kalman precipitation forecasts show about 15
improvement over the LMRFC forecast when compared
to the Stage III estimates. The
distributed hydrologic model SHEELS was run with
Stage III, LMRFC QPF and Kalman QPF. Streamflow
comparisons with the observed are shown for the
TS Allison event.
Streamflow estimates made using Kalman forecast
were more accurate than those made using LMRFC
forecast when compared to the observed. The
improvement was also obvious in the timing and
magnitude of the event.