83rd Annual

Tuesday, 11 February 2003: 12:00 PM
Operational rainfall and flow forecasting for the Panama Canal Watershed
Konstantine P. Georgakakos, Hydrologic Research Center, San Diego, CA; and J. A. Sperfslage
Poster PDF (398.9 kB)
In 1998, the Hydrologic Research Center designed and implemented at the Panama Canal Authority a prototype coupled hydrometeorological system for real time distributed rainfall and flow forecasting within the 3,200-sqkm mountainous Panama Canal Watershed. The system uses as input ETA forecasts (on an 80-km grid) and observations from surface hydrometeorological stations, upper air radiosondes, and a 10-cm weather radar, to produce mean areal precipitation and flow forecasts on scales of 150-400 sqkm with a maximum 12 hour lead time and with hourly resolution. The system design combines embedded cloud and hydrologic models and state estimators for data assimilation and uncertainty estimation. The Panama Canal Authority has been using the system operationally since October 1998, and the present paper presents the validation that was performed on the rainfall forecasts produced by the PANMAP system component using the operational data.

Analysis of the PANMAP forecasts for the periods July-December 2000 and August-December 2001 was performed with performance measures that are commensurate to the intended use of the rainfall forecasts for hydrologic purposes. The error in forecasting the volume of rainfall is the primary such measure used as a bulk statistic for validation, but individual forecasts with lead times from 1 hr to 12 hrs were also considered during periods with significant storm rainfall. The analysis considered the use of a variety of input data such as the use of ETA forecasts of surface and upper-air fields (temperature, pressure, humidity, winds), and the use of upper air radiosonde observations and surface ALERT data for the short-term forecast of surface mean areal rainfall over 10 sub-catchments of the 3,200-sqkm Panama Canal Watershed. The validation effort was performed in close collaboration with the staff of the Meteorology and Hydrology Section of the Panama Canal Authority.

The primary conclusion from the validation effort is that the coupled system produced useful forecasts during times of heavy rainfall that contributed to producing effective management decisions during these periods. The PANMAP forecasts have lowest bias when radiosonde (RAOB) data are used to produce input to the rainfall prediction component for forecast lead-times out to 3 hours, and when new ETA forecasts are used to produce input to the rainfall prediction component for forecast lead-times from 6 to 12 hours. The use of surface ALERT data exclusively yields reasonable hourly nowcasts, but results in significant underestimation of rainfall volume for longer lead-times. There is an overall overestimation (in some cases significant) of low rainfall amounts that is persistent throughout the validation period, with the best results obtained for the heavier rainfall amounts. Best forecast performance is found for the northeastern mountainous catchments of Rio Gatun, Rio Boqueron, Rio Pequeni and Rio Chagres and for the Lake Gatun area in the central region of the Panama Canal Watershed.

Supplementary URL: