You are here:
Maps of the Future: Multi-scale Precipitation Modeling and ForecastingEPA Grant Number: U915612
Title: Maps of the Future: Multi-scale Precipitation Modeling and Forecasting
Investigators: Funk, Chris C.
Institution: University of California - Santa Barbara
EPA Project Officer: Boddie, Georgette
Project Period: September 1, 1999 through September 1, 2002
Project Amount: $77,943
RFA: STAR Graduate Fellowships (1999) RFA Text | Recipients Lists
Research Category: Academic Fellowships , Ecological Indicators/Assessment/Restoration , Fellowship - Earth Sciences
The objective of this research project is to build combined physical and statistical models of precipitation to generate high-resolution forecasts, given global climate model input. These high-resolution precipitation fields then may be used to develop historical time-series on continental scales, anticipate potential drought and flooding in the upcoming year, or explore the impacts of climate change. Models are developed for a region in Africa, where there is a pressing need for better information, and the effects of extreme events can exert a heavy toll.
The research begins with a set of daily precipitation values from 1,094 weather stations from continental Africa, and a 33-year times-series of global climate model data (temperature, velocity, specific humidity, etc.). The global climate model data will be used in two ways. First, the precipitation rate will be extracted for each day, for each grid cell situated over Africa. Second, wind velocity, humidity and temperature fields in conjunction with local topography obtained from a digital elevation model will be used to drive a diagnostic model of orographic precipitation. These two steps will result in two estimated precipitation fields: the final step transforms these fields into a conditional probability distribution. The distribution of daily rainfall amounts varies widely with locations in Africa. Different regions have different climate dynamics and different distributions of daily rainfall. To account for this spatial heterogeneity, unique sets of statistical models will be developed for each weather station. These logistic regression models will link the global climate and orographic rainfall amounts to the observed frequency of rainfall events at a range of intensities. Given climate and orographic rainfall amounts, a conditional distribution function can be built, from which probabilities of given events or the maximum likelihood estimate of precipitation can be inferred. The coefficients of these models will be interpolated to a continental grid and used to derive a historic time-series of daily rainfall.
Given climate and orographic rainfall amounts, a conditional distribution function can be built, from which probabilities of given events or the maximum likelihood estimate of precipitation can be inferred.