Disaggregation of Daily Rainfall for Continuous Watershed Modeling

Abstract

A daily rainfall disaggregation method, yielding regional hourly rainfall estimates, is presented with application to continuous-simulation watershed models. The disaggregation method estimates the number and volume of rain events each day and selects measured intensity patterns from a nearby station to simulate the hourly rainfall trace for each storm. No partial events are allowed, and start times are taken from a uniform probability distribution. A single calibration parameter, designating the smallest event threshold, must be estimated for each season. Comparing with hourly recorded data, the method reproduces well the variance, lag-1 autocovariance, and probability of zero rainfall. Application of the method in the upper Charles River watershed in eastern Massachusetts demonstrates that the method greatly improves the river flow simulation capability of the hydrologic model over alternate simulations using hourly rainfall outside the watershed. It is concluded that this disaggregation method provides a means of using daily rainfall data for watershed model calibration (for both runoff and total contaminant loads) when an hourly rainfall database is available in the same climatic region.