Statistical post processing of extreme weather forecasts

J.J. Velthoen

doi:10.4233/uuid:f6a6096d-9eb1-4a77-b376-34e01b817011

Statistical post processing of extreme weather forecasts

J.J. Velthoen

Statistics

Research output: Thesis › Dissertation (TU Delft)

50 Downloads (Pure)

Abstract

In this thesis we develop several statistical methods to estimate high conditional quantiles to use for statistical post-processing of weather forecasts. We propose methodologies that combine theory from extreme value statistics and machine learning algorithms in order to estimate high conditional quantiles in large covariate spaces. In applications of weather forecasting we show improved predictive skill for precipitation forecasts.

Original language	English
Qualification	Doctor of Philosophy
Awarding Institution	Delft University of Technology
Supervisors/Advisors	Jongbloed, G., Supervisor Cai, J., Advisor
Thesis sponsors	Netherlands Organisation for Science Research (NWO)
Award date	14 Sept 2022
Place of Publication	Ede
Edition	1
Print ISBNs	9789083272726
Electronic ISBNs	9789083272726
DOIs	https://doi.org/10.4233/uuid:f6a6096d-9eb1-4a77-b376-34e01b817011
Publication status	Published - 2022

Funding

This work is part of the research project \Probabilistic forecasts of extreme weather
utilizing advanced methods from extreme value theory" with project number 14612
which is nanced by the Netherlands Organisation for Scienti c Research (NWO).

Keywords

Extreme quantile regression
Statistical post-processing
Extreme value theory
Extreme conditional quantile
Variable selection
Random Forest
Gradient boosting

Access to Document

10.4233/uuid:f6a6096d-9eb1-4a77-b376-34e01b817011

dissertationFinal published version, 11.9 MB

Cite this

@phdthesis{f6a6096d9eb14a77b37634e01b817011,

title = "Statistical post processing of extreme weather forecasts",

abstract = "In this thesis we develop several statistical methods to estimate high conditional quantiles to use for statistical post-processing of weather forecasts. We propose methodologies that combine theory from extreme value statistics and machine learning algorithms in order to estimate high conditional quantiles in large covariate spaces. In applications of weather forecasting we show improved predictive skill for precipitation forecasts. ",

keywords = "Extreme quantile regression, Statistical post-processing, Extreme value theory, Extreme conditional quantile, Variable selection, Random Forest, Gradient boosting",

author = "J.J. Velthoen",

year = "2022",

doi = "10.4233/uuid:f6a6096d-9eb1-4a77-b376-34e01b817011",

language = "English",

isbn = "9789083272726",

volume = "1",

edition = "1",

type = "Dissertation (TU Delft)",

school = "Delft University of Technology",

}

TY - THES

T1 - Statistical post processing of extreme weather forecasts

AU - Velthoen, J.J.

PY - 2022

Y1 - 2022

N2 - In this thesis we develop several statistical methods to estimate high conditional quantiles to use for statistical post-processing of weather forecasts. We propose methodologies that combine theory from extreme value statistics and machine learning algorithms in order to estimate high conditional quantiles in large covariate spaces. In applications of weather forecasting we show improved predictive skill for precipitation forecasts.

AB - In this thesis we develop several statistical methods to estimate high conditional quantiles to use for statistical post-processing of weather forecasts. We propose methodologies that combine theory from extreme value statistics and machine learning algorithms in order to estimate high conditional quantiles in large covariate spaces. In applications of weather forecasting we show improved predictive skill for precipitation forecasts.

KW - Extreme quantile regression

KW - Statistical post-processing

KW - Extreme value theory

KW - Extreme conditional quantile

KW - Variable selection

KW - Random Forest

KW - Gradient boosting

U2 - 10.4233/uuid:f6a6096d-9eb1-4a77-b376-34e01b817011

DO - 10.4233/uuid:f6a6096d-9eb1-4a77-b376-34e01b817011

M3 - Dissertation (TU Delft)

SN - 9789083272726

VL - 1

CY - Ede

ER -

Statistical post processing of extreme weather forecasts

Abstract

Funding

Keywords

Access to Document

Fingerprint

Cite this