A Forecast of the February-March 1999 Anomalous Discharge of the Caroní River, Venezuela

contributed by Stefan Hastenrath¹, Larry Greischar¹, Esperanza Colón² and Alfred Gil²

¹Department of Atmospheric and Oceanic Sciences, University of Wisconsin, Madison, Wisconsin

²Centro de Pronóstico Hidrometeorológico C.V.G. Electrificación del Caroní, CA., Caracas, Venezuela

Following on work done a decade ago (Hastenrath, 1990) which explored the predictability of anomalous discharge from selected rivers in northern South America, an empirical method has been developed to forecast the February-March anomalous discharge of the Caroní river in eastern Venezuela. The present work is documented in Hastenrath et al. (1999). The practical importance of seasonal forecasts of water resources is now acutely recognized by organizations devoted to the generation of hydroelectric power. In particular, the Guri power plant fed by the Caroní rivershed in eastern Venezuela (Fig. 1) with a capacity of 10,000 MW is the second largest in the world. It provides some 70 percent of the total electrical power supply for the country and allows for export abroad. Increasing energy demands and uncertainties in reliable water supply underline the need for advanced planning. Accordingly, there is considerable interest in the development of pertinent forecast methods.

Predictors used were the July-August values of Tahiti minus Darwin sea level pressure difference, an index of the Southern Oscillation denoted by SOI(JA), and the preceding July-August discharge of the Caroní river CAR(JA). The predictand CAR(FM) is the February-March discharge of Caroní river. With reference to 1950-79 the mean of CAR(FM) is 1549 m³s^-1 and the standard deviation is 884 m³s^-1.

Two linear regression models were constructed using the period 1950-79 for training and reserving 1980-98 as an independent verification period for testing the skill of the models. A neural network model was developed using the interval 1950-75 for training proper, the interval 1976-85 served to successively check skill during training, while the interval 1986-98 was reserved as truly independent verification period. A simple regression model using only SOI(JA) as predictor captures 41% of the variance over the verification period. Adding the preceding July-August river discharge as an additional predictor increases the variance captured to 44%. A neural networking model using both SOI(JA) and CAR(JA) as predictors captured 67% of the variance over the 1986-98 verification period. The two linear regression models are:

Model RR CAR(FM) = 357*SOI(JA) + 1000

Model R CAR(FM) = 0.115*CAR(JA) + 357*SOI(JA) - 35

indicating that the Caroní February-March discharge is positively correlated with the preceding July-August discharge and Southern Oscillation Index.

The 1998 value of the SOI(JA) was 3.2 mb, compared with mean and standard deviation values for the training period of 1.49 and 1.47 mb, respectively. This will tend to enhance discharge. Compared with mean and standard deviation values of 8710 and 1245 m³s^-1, the 1998 value of CAR(JA) was 8848 m³s^-1 which will also tend to enhance discharge.

The resulting forecasts for the 1999 February-March Caroní river discharge follow:

RR	2142 m3s-1 (+06.7 SD)
R	2125 m3s-1 (+0.65 SD)
Neural Networking	2142 m3s-1 (+0.67 SD)

This would be comparable to February-March of 1981, 1989, 1990, 1993, and 1996.

Acknowledgments

Real-time input information is gratefully acknowledged as follows: SOI from CPC-NOAA and Caroní discharge from CVG Electrificacion de Caroni (EDELCA).

References

Hastenrath, S., 1990, Diagnostics and prediction of anomalous river discharge in northern South America, J. Climate, 3, 1080-1096.

Hastenrath, S., L. Greischar, E. Colon, and A. Gil, 1999, Forecasting the anomalous discharge of the Caroní River, Venezuela, J. Climate (submitted)

Fig. 1. Orientation map showing Caroní watershed and Venezuela within South America. Solid square indicates discharge gauging station Guri and dot raingauge station Urimán. Heavy solid and dashed lines are isohyets of annual precipitation totals in mm. Including the Guri and Urimán sites, the basin contains a total of 63 hydrological, 59 rainguage, and 11 climatological stations.