This project seeks to quantify the inherent predictability of observed precipitation variations over the U.S. on seasonal to multi-decadal timescales. The questions to be addressed include: 1) What amount of observed seasonal-mean precipitation variability is uniquely related to variations in the background state of the observed system during a given time-period? 2) When and where do these inherently predictable variations in seasonal-mean precipitation occur within the U.S.? 3) What is the inherent predictability of short-term (<12-months) and long-term (>12-month) drought occurrences? 4) Do numerical climate models adequately reproduce the inherent predictability of precipitation variations over the U.S.?
To answer these questions, the investigators will (i) construct stochastic weather-generation models based upon observed station-based daily precipitation, which can be used to establish the envelope of variability that arises solely from the random behavior of precipitation events, (ii) identify observed interannual to multi-decadal scale variations in seasonal-mean precipitation across the U.S. that lie outside this envelope produced by chance, (iii) use the observed and stochastically-generated daily-precipitation time-series to detect inherently-predictable variations in the severity of drought, as represented by the Standard Precipitation Index, and (vi) evaluate numerical model capability in reproducing the observed inherent predictability of seasonal-mean precipitation and drought over the U.S., using daily precipitation estimates from numerical model representations of the climate of the 20th century.
The broader impacts of this project will include improving our confidence in climate forecasts for many regions of the U.S by identifying "hot-spot" regions of historical predictability that are strongly influenced by climate-change processes. Further it will establish the groundwork for a better understanding of the role that regional- and large-scale atmospheric circulations and ocean- and land-surface forcings play in modifying regional precipitation on interannual to multi-decadal time-scales. It will also guide future investigations into numerical model systems' portrayal of physical mechanisms that give rise to observed predictability. Finally, it will help determine the applicability of using stochastic weather-generation models for identifying and analyzing inherently predictable precipitation variations in other regions of the world.
This project sought to quantify the inherent predictability of observed precipitation variations over the United States on seasonal to multi-decadal timescales. The questions we were interested in answering included: 1) What amount of observed seasonal-mean precipitation variability is uniquely related to variations in the background state of the observed system during a given time-period? 2) When and where do these inherently predictable variations in seasonal-mean precipitation occur within the United States? 3) What is the inherent predictability of short-term (<12-months) and long-term (>12-month) drought occurrences? 4) Do numerical climate models adequately reproduce the inherent predictability of precipitation variations over the United States? To answer these questions, we constructed stochastic weather-generation models based upon observed station-based daily precipitation, which can be used to establish the envelop of variability that arises solely from the random behavior of precipitation events. We then identified observed interannual to multi-decadal scale variations in seasonal-mean precipitation across the U.S. that lie outside this envelope produced by chance. Further, we used the observed and stochastically-generated daily-precipitation time-series to detect inherently-predictable variations in precipitation extremes, including extreme dry spells leading to droughts or heavy precipitation events leading to flooding. Finally, we are in the process of developing a "unified" stochastic weather-generation model framework that is equally applicable for grid-point based precipitation as well as station-based precipitation. Using daily precipitation estimates from numerical model representations of the climate of the 20th century in combination with the new unified stochastic weather-generation model, we will continue to evaluate numerical model capability in reproducing the observed inherent predictability of seasonal-mean precipitation and drought over the U.S. The broader impacts of the project include improving our confidence in climate forecasts for many regions of the U.S by identifying "hot-spot" regions of historical predictability that are strongly influenced by climate-change processes. Further it establishes the groundwork for a better understanding of the role that regional- and large-scale atmospheric circulations and ocean- and land-surface forcings play in modifying regional precipitation on interannual to multi-decadal time-scales. It also guides future investigations into numerical model systems’ portrayal of physical mechanisms that give rise to observed predictability. Finally, it helps determine the applicability of using stochastic weather-generation models for identifying and analyzing inherently predictable precipitation variations in other regions of the world. The outcomes of the research include the isolation of "hot-spots" and "flare-ups" of inherently predictable precipitation variations across the U.S., which can be used to guide regional climate forecasting and uncover processes driving this variability. The outcomes also include a database of daily values of potentially predictable variance of seasonal precipitation characteristics for all 774 station locations across the contiguous U.S. so that researchers across a broad range of disciplines can determine where the greatest opportunities lie to predict precipitation changes of interest to them. Finally, they will include the development of a "unified stationary stochastic weather model" (U-SSWM) framework for systematically determining the precipitation frequency and intensity parameters for both station-based and numerically-simulated precipitation data, which can be used by other researchers to study the predictability of seasonal precipitation variations in other regions of the world, of climate processes/variables other than precipitation, and of processes/variables outside the climate sciences altogether.
