Quantitative Real-Time Rainfall Forecasting Using Remote Sensing.
A physically-based rainfall forecasting model for real-time hydrologic applications is developed with emphasis on utilization of remote sensing observations. Radar measured reflectivities, satellite measured brightness temperatures, and surface meteorologic observations are the primary model input quantities. Temporal and spatial scales of interest are lead-times of one-hour and regions on the order of 10-10000 km^2. The model is derived from the principles of conservation of mass for vertically integrated liquid water content in a cloud. Model dynamics are governed by a nonlinear relationship between meteorological input quantities and the model state. To develop a means of accounting for uncertainty in model formulation and measurements, a Kalman filter is incorporated to couple model forecasts with observed quantities. Because implementation of the filter requires a linear system, a statistical-linearization technique is applied to the model dynamics. The Kalman filter state estimator allows real-time updating of the model state and estimation of forecast uncertainty. A calibration study is performed to define guidelines for parameter estimation, evaluate the working range for model parameter values, and select model parameter values for model verification. The model is calibrated and verified using observed data. The observed data includes: radar data collected at the National Weather Service radar site in Oklahoma City; satellite data obtained from the GOES -7 infrared band; and ground level meteorological data obtained from weather stations in the region of the radar site. Results indicate that one-hour lead-time rainfall rate forecasts produced using the calibrated model are of comparable or higher accuracy than forecasts obtained using traditional methods such as persistence or advection -based nowcasting. The performance measures used to evaluate forecast quality include: mean rainfall rate error; root mean square error of rainfall rate; percent of correctly forecasted rainfall coverage; and cross-correlation between forecasted and observed rainfall fields.
- Pub Date:
- January 1992
- Engineering: Civil; Physics: Atmospheric Science; Remote Sensing