NOAA's 3-month overlapping climate forecasts have been produced operationally since 1995 and have generated interest in the water resources community. While the forecast skill and the uncertainty of forecast outcome remain the focus of much discussion, the assessment of the impact of a forecast on the water resources system must be determined to establish the utility of the forecast for water resources decision making. Thus, the probabilistic characteristics of a seasonal climate forecast must be translated into a corresponding probabilistic hydrologic response. This is a critical step to promote a wider acceptance and utilization of NOAA's forecasts in the water resources user community. For a number of water resources applications this can be achieved by modeling the hydrologic system for the range of forecasted conditions. Most process-based hydrologic models operate on a daily time scale and require daily weather input. While past hydrologic response can be modeled using the actual observed weather data, future potential hydrologic responses must rely on stochastically generated weather data that reflect possible weather realizations of the future. A number of weather generators exist that generate daily weather based on historical weather records. A method to adjust the precipitation parameters of daily weather generators to reflect NOAA's forecasted seasonal precipitation conditions is the object of this paper. For illustration purposes the experimental stochastic weather generator SYNTOR is used, but the method is also applicable to other generators such as CLIGEN, WGEN, US CLIMATE and GEM.

The drivers of SYNTOR are probability measures of occurrence of precipitation on any given day (called transitional probabilities) and distribution parameters of daily precipitation amount. The values of these measures and parameters vary by month and location to capture the local seasonal pattern of precipitation characteristics. To adjust the precipitation parameters of SYNTOR, NOAA's 3-month overlapping forecast departures are first transformed into non-overlapping monthly values at a location of interest following the procedure previously presented at AMS by Schneider and Garbrecht (2002, 2003). Thereafter, the observed relationship between monthly precipitation amount and number of rainy days at the location of interest is used to partition the forecasted precipitation departure into a forecasted departure for daily precipitation amount and number of rainy days. The probability measures of occurrence of precipitation are adjusted for the forecasted number of rainy days based on the observed local and monthly relationship between number of rainy days and transitional probabilities. Once the proper number and sequence of rainy days are generated, the forecasted change in precipitation amount is imposed by adjusting the generated daily precipitation by the percent change in forecasted daily precipitation amount. The method is applied with weather data for Temple, Texas, and probability of exceedance curves for generated normal and forecasted monthly precipitation are shown. The so forecasted daily weather can then be used to establish the hydrologic response to the forecast.