Event Title

Development of a Stochastic Weather Generator

Presenter Information

Kristen Yeager

Location

Eccles Conference Center

Event Website

htpp://water.usu.edu/

Start Date

3-29-2011 10:25 AM

End Date

3-29-2011 10:30 AM

Description

Although General Circulation Models (GCMs) are used to provide insight into scenario planning, ecological risk assessments, and potential long-term climatic changes, the results of such models do not possess the spatial and temporal resolutions needed to analyze these effects at the watershed scale. In order to obtain spatially and temporally relevant inputs into hydrologic models that are reflective of the trends necessary to evaluate various scenario plans, ecological risk assessments, and climate change trends, stochastic weather generators have been developed to produce synthetic hydrometeorological sequences that are statistically similar to the historical data. The work presented here describes the development of a nonparametric stochastic weather generator that can provide various hydrometeorological data, including precipitation, temperature, pressure, relative humidity, cloud cover, wind speed, direct and total radiation at both the daily and hourly scales. One advantage of using nonparametric models is that the models are adaptive to the historical data used instead of requiring prior knowledge or certain assumptions of the underlying distributions of the historical data. A k-nearest neighbor (k-NN) approach was used to generate daily precipitation data, which then provided input into a disaggregation scheme based on hourly event distributions to create hourly data. The daily temperature model relies on a combination of parametric and nonparametric models, while the hourly temperature data is selected using a k-NN weighted bootstrap. The other parameters were conditioned on precipitation to select the appropriate values based on k-NN resampling. Overall, the models are generally able reproduce the basic statistics of the historical record with regards to the mean, standard deviation and correlation structures.

This document is currently not available here.

Share

COinS
 
Mar 29th, 10:25 AM Mar 29th, 10:30 AM

Development of a Stochastic Weather Generator

Eccles Conference Center

Although General Circulation Models (GCMs) are used to provide insight into scenario planning, ecological risk assessments, and potential long-term climatic changes, the results of such models do not possess the spatial and temporal resolutions needed to analyze these effects at the watershed scale. In order to obtain spatially and temporally relevant inputs into hydrologic models that are reflective of the trends necessary to evaluate various scenario plans, ecological risk assessments, and climate change trends, stochastic weather generators have been developed to produce synthetic hydrometeorological sequences that are statistically similar to the historical data. The work presented here describes the development of a nonparametric stochastic weather generator that can provide various hydrometeorological data, including precipitation, temperature, pressure, relative humidity, cloud cover, wind speed, direct and total radiation at both the daily and hourly scales. One advantage of using nonparametric models is that the models are adaptive to the historical data used instead of requiring prior knowledge or certain assumptions of the underlying distributions of the historical data. A k-nearest neighbor (k-NN) approach was used to generate daily precipitation data, which then provided input into a disaggregation scheme based on hourly event distributions to create hourly data. The daily temperature model relies on a combination of parametric and nonparametric models, while the hourly temperature data is selected using a k-NN weighted bootstrap. The other parameters were conditioned on precipitation to select the appropriate values based on k-NN resampling. Overall, the models are generally able reproduce the basic statistics of the historical record with regards to the mean, standard deviation and correlation structures.

https://digitalcommons.usu.edu/runoff/2011/Posters/11