Generating Synthetic Daily Precipitation Realizations for Seasonal Precipitation Forecasts

Abstract

Synthetic weather generation models that depend on statistics of past weather observations are often limited in their applications to issues that depend on historical weather characteristics. Enhancing these models to take advantage of increasingly available and skillful seasonal climate outlook products would broaden applications to include proactive soil and water resources management, better prediction of achieving production targets, and weather-related risk assessment. In this paper, an analytical method was developed that enables generation of daily precipitation time series for seasonal forecasts up to 12 months ahead. The method uses historical weather observations to establish reference precipitation statistics (monthly precipitation amount, number of rainy days per month, and wet–wet and dry–wet day transition probabilities) and subsequently adjusts these statistics to reflect the forecast departures from long-term average monthly precipitation. This reference and forecast departure approach ensures that generated precipitation is consistent and compatible with the forecast and the local climate characteristics as well. The method was tested with precipitation data from the USDA Agricultural Research Service (ARS) weather station at Temple, Texas, and the National Weather Service Cooperative Observer Program (NWS-COOP) data at Tallahassee, Florida, for a hypothetical seasonal precipitation forecast. Several 100-year time series of generated daily precipitation reproduced average monthly precipitation within