After a short demonstration about the need for operational rainfall forecasts, a literature survey, constraints for model development and definition of criteria of model performance a simple multi-variate autoregressive model is presented that forecasts point rainfall spatially distributed, for short lead times, quantitatively and automatically. For model operation only remotely transmitted standard raingage data is necessary.

The model is expanded to forecast and estimate also areal rainfall from point measurements using the Kaiman filter technique and the theory of regionalized variables.

It is demonstrated that the Kaiman filter model is superior for areal rainfall estimation (compared to the Thiessen procedure) as well as for areal rainfall forecasting compared to trivial forecasts and the point forecasting model mentioned above.

This content is only available as a PDF.