Calibration of probabilistic quantitative precipitation forecasts with an artificial neural network

Huiling Yuan, Xiaogang Gao, Steven Mullen, Soroosh Sorooshian, Jun Du, Hann Ming Henry Juang

Research output: Contribution to journalArticle

28 Citations (Scopus)

Abstract

A feed-forward neural network is configured to calibrate the bias of a high-resolution probabilistic quantitative precipitation forecast (PQPF) produced by a 12-km version of the NCEP Regional Spectral Model (RSM) ensemble forecast system. Twice-daily forecasts during the 2002-2003 cool season (1 November-31 March, inclusive) are run over four U.S. Geological Survey (USGS) hydrologic unit regions of the southwest United States. Calibration is performed via a cross-validation procedure, where four months are used for training and the excluded month is used for testing. The PQPFs before and after the calibration over a hydrological unit region are evaluated by comparing the joint probability distribution of forecasts and observations. Verification is performed on the 4-km stage IV grid, which is used as "truth." The calibration procedure improves the Brier score (BrS), conditional bias (reliability) and forecast skill, such as the Brier skill score (BrSS) and the ranked probability skill score (RPSS), relative to the sample frequency for all geographic regions and most precipitation thresholds. However, the procedure degrades the resolution of the PQPFs by systematically producing more forecasts with low nonzero forecast probabilities that drive the forecast distribution closer to the climatology of the training sample. The problem of degrading the resolution is most severe over the Colorado River basin and the Great Basin for relatively high precipitation thresholds where the sample of observed events is relatively small.

Original languageEnglish (US)
Pages (from-to)1287-1303
Number of pages17
JournalWeather and Forecasting
Volume22
Issue number6
DOIs
StatePublished - Dec 2007

Fingerprint

artificial neural network
calibration
forecast
geological survey
climatology
river basin
basin

ASJC Scopus subject areas

  • Atmospheric Science

Cite this

Calibration of probabilistic quantitative precipitation forecasts with an artificial neural network. / Yuan, Huiling; Gao, Xiaogang; Mullen, Steven; Sorooshian, Soroosh; Du, Jun; Juang, Hann Ming Henry.

In: Weather and Forecasting, Vol. 22, No. 6, 12.2007, p. 1287-1303.

Research output: Contribution to journalArticle

Yuan, Huiling ; Gao, Xiaogang ; Mullen, Steven ; Sorooshian, Soroosh ; Du, Jun ; Juang, Hann Ming Henry. / Calibration of probabilistic quantitative precipitation forecasts with an artificial neural network. In: Weather and Forecasting. 2007 ; Vol. 22, No. 6. pp. 1287-1303.
@article{7e104479e6f949ba86c5dfe34317ea5c,
title = "Calibration of probabilistic quantitative precipitation forecasts with an artificial neural network",
abstract = "A feed-forward neural network is configured to calibrate the bias of a high-resolution probabilistic quantitative precipitation forecast (PQPF) produced by a 12-km version of the NCEP Regional Spectral Model (RSM) ensemble forecast system. Twice-daily forecasts during the 2002-2003 cool season (1 November-31 March, inclusive) are run over four U.S. Geological Survey (USGS) hydrologic unit regions of the southwest United States. Calibration is performed via a cross-validation procedure, where four months are used for training and the excluded month is used for testing. The PQPFs before and after the calibration over a hydrological unit region are evaluated by comparing the joint probability distribution of forecasts and observations. Verification is performed on the 4-km stage IV grid, which is used as {"}truth.{"} The calibration procedure improves the Brier score (BrS), conditional bias (reliability) and forecast skill, such as the Brier skill score (BrSS) and the ranked probability skill score (RPSS), relative to the sample frequency for all geographic regions and most precipitation thresholds. However, the procedure degrades the resolution of the PQPFs by systematically producing more forecasts with low nonzero forecast probabilities that drive the forecast distribution closer to the climatology of the training sample. The problem of degrading the resolution is most severe over the Colorado River basin and the Great Basin for relatively high precipitation thresholds where the sample of observed events is relatively small.",
author = "Huiling Yuan and Xiaogang Gao and Steven Mullen and Soroosh Sorooshian and Jun Du and Juang, {Hann Ming Henry}",
year = "2007",
month = "12",
doi = "10.1175/2007WAF2006114.1",
language = "English (US)",
volume = "22",
pages = "1287--1303",
journal = "Weather and Forecasting",
issn = "0882-8156",
publisher = "American Meteorological Society",
number = "6",

}

TY - JOUR

T1 - Calibration of probabilistic quantitative precipitation forecasts with an artificial neural network

AU - Yuan, Huiling

AU - Gao, Xiaogang

AU - Mullen, Steven

AU - Sorooshian, Soroosh

AU - Du, Jun

AU - Juang, Hann Ming Henry

PY - 2007/12

Y1 - 2007/12

N2 - A feed-forward neural network is configured to calibrate the bias of a high-resolution probabilistic quantitative precipitation forecast (PQPF) produced by a 12-km version of the NCEP Regional Spectral Model (RSM) ensemble forecast system. Twice-daily forecasts during the 2002-2003 cool season (1 November-31 March, inclusive) are run over four U.S. Geological Survey (USGS) hydrologic unit regions of the southwest United States. Calibration is performed via a cross-validation procedure, where four months are used for training and the excluded month is used for testing. The PQPFs before and after the calibration over a hydrological unit region are evaluated by comparing the joint probability distribution of forecasts and observations. Verification is performed on the 4-km stage IV grid, which is used as "truth." The calibration procedure improves the Brier score (BrS), conditional bias (reliability) and forecast skill, such as the Brier skill score (BrSS) and the ranked probability skill score (RPSS), relative to the sample frequency for all geographic regions and most precipitation thresholds. However, the procedure degrades the resolution of the PQPFs by systematically producing more forecasts with low nonzero forecast probabilities that drive the forecast distribution closer to the climatology of the training sample. The problem of degrading the resolution is most severe over the Colorado River basin and the Great Basin for relatively high precipitation thresholds where the sample of observed events is relatively small.

AB - A feed-forward neural network is configured to calibrate the bias of a high-resolution probabilistic quantitative precipitation forecast (PQPF) produced by a 12-km version of the NCEP Regional Spectral Model (RSM) ensemble forecast system. Twice-daily forecasts during the 2002-2003 cool season (1 November-31 March, inclusive) are run over four U.S. Geological Survey (USGS) hydrologic unit regions of the southwest United States. Calibration is performed via a cross-validation procedure, where four months are used for training and the excluded month is used for testing. The PQPFs before and after the calibration over a hydrological unit region are evaluated by comparing the joint probability distribution of forecasts and observations. Verification is performed on the 4-km stage IV grid, which is used as "truth." The calibration procedure improves the Brier score (BrS), conditional bias (reliability) and forecast skill, such as the Brier skill score (BrSS) and the ranked probability skill score (RPSS), relative to the sample frequency for all geographic regions and most precipitation thresholds. However, the procedure degrades the resolution of the PQPFs by systematically producing more forecasts with low nonzero forecast probabilities that drive the forecast distribution closer to the climatology of the training sample. The problem of degrading the resolution is most severe over the Colorado River basin and the Great Basin for relatively high precipitation thresholds where the sample of observed events is relatively small.

UR - http://www.scopus.com/inward/record.url?scp=38849171706&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=38849171706&partnerID=8YFLogxK

U2 - 10.1175/2007WAF2006114.1

DO - 10.1175/2007WAF2006114.1

M3 - Article

AN - SCOPUS:38849171706

VL - 22

SP - 1287

EP - 1303

JO - Weather and Forecasting

JF - Weather and Forecasting

SN - 0882-8156

IS - 6

ER -