A new approach to extended‐range multimodel forecasting: Sequential learning algorithms
Multimodel combinations are a well‐established methodology in weather and climate prediction and their benefits have been widely discussed in the literature. Typical approaches involve combining the output of different numerical weather prediction (NWP) models using constant weighting factors, eithe...
Gespeichert in:
Veröffentlicht in: | Quarterly journal of the Royal Meteorological Society 2021-10, Vol.147 (741), p.4269-4282 |
---|---|
Hauptverfasser: | , , |
Format: | Artikel |
Sprache: | eng |
Schlagworte: | |
Online-Zugang: | Volltext |
Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
container_end_page | 4282 |
---|---|
container_issue | 741 |
container_start_page | 4269 |
container_title | Quarterly journal of the Royal Meteorological Society |
container_volume | 147 |
creator | Gonzalez, Paula L. M. Brayshaw, David J. Ziel, Florian |
description | Multimodel combinations are a well‐established methodology in weather and climate prediction and their benefits have been widely discussed in the literature. Typical approaches involve combining the output of different numerical weather prediction (NWP) models using constant weighting factors, either uniformly distributed or determined through a prior skill assessment. This strategy, however, can lead to suboptimal levels of skill, as the performance of NWP models can vary with time (e.g., seasonally varying skill, changes in the forecasting system). Moreover, standard combination methods are not designed to incorporate predictions derived from sources other than NWP systems (e.g., climatological or time‐series forecasts). New algorithms developed within the machine learning community provide the opportunity for “online prediction” (also referred to as “sequential learning”). These methods consider a set of weighted predictors or “experts” to produce subsequent predictions in which the combination or “mixture” is updated at each step to optimize a loss or skill function. The predictors are highly flexible and can combine both NWP and statistically derived forecasts transparently. A set of these online prediction methods is tested and compared with standard multimodel combination techniques to assess their usefulness. The methods are general and can be applied to any model‐derived predictand. A set of weather‐sensitive European country‐aggregate energy variables (electricity demand and wind power) is selected for demonstration purposes. Results show that these innovative methods exhibit significant skill improvements (i.e., between 5 and 15% improvement in the probabilistic skill) with respect to standard multimodel combination techniques for lead weeks up to 5. The incorporation of statistically derived predictors (based on historical climate data) alongside NWP forecasts is also shown to contribute significant skill improvements in many cases.
Errors in numerical weather prediction (NWP) systems limit their skill. It is therefore desirable to combine multiple NWP systems to enhance skill, particularly at extended range. Here, a new set of schemes—sequential learning algorithms—is shown to offer significant benefits over existing combination schemes. The schemes are well‐suited to operational use and, in contrast to more traditional schemes, allow weights to evolve dynamically and the transparent inclusion of statistical forecasts. Skill improvements of up |
doi_str_mv | 10.1002/qj.4177 |
format | Article |
fullrecord | <record><control><sourceid>proquest_cross</sourceid><recordid>TN_cdi_proquest_journals_2609674642</recordid><sourceformat>XML</sourceformat><sourcesystem>PC</sourcesystem><sourcerecordid>2609674642</sourcerecordid><originalsourceid>FETCH-LOGICAL-c3227-36a0ce28db35c495318860ce03d5b426567deafc0580c8c64ce6653c9d1a3a413</originalsourceid><addsrcrecordid>eNp10MtKw0AUBuBBFKxVfIUBFy4k9cw9cVeKVwoiKugqTCcnbUKatDMptTsfwWf0SUytW1cHfj7OjZBTBgMGwC-X5UAyY_ZIj0ljotjA2z7pAQgVJQDJITkKoQQAZbjpkfchrXFN7WLhG-tmtG0ofrRYZ5h9f355W0-RzldVW8ybDCuaNx6dDW1RT6_oMy5XWLeFrWiF1tddSG01bXzRzubhmBzktgp48lf75PXm-mV0F40fb-9Hw3HkBOcmEtqCQx5nE6GcTJRgcay7BESmJpJrpU2GNnegYnCx09Kh1kq4JGNWWMlEn5zt-nYXdPuENi2bla-7kSnXkGgjteSdOt8p55sQPObpwhdz6zcpg3T7t3RZptu_dfJiJ9dFhZv_WPr08Kt_AA4Sbu4</addsrcrecordid><sourcetype>Aggregation Database</sourcetype><iscdi>true</iscdi><recordtype>article</recordtype><pqid>2609674642</pqid></control><display><type>article</type><title>A new approach to extended‐range multimodel forecasting: Sequential learning algorithms</title><source>Access via Wiley Online Library</source><creator>Gonzalez, Paula L. M. ; Brayshaw, David J. ; Ziel, Florian</creator><creatorcontrib>Gonzalez, Paula L. M. ; Brayshaw, David J. ; Ziel, Florian</creatorcontrib><description>Multimodel combinations are a well‐established methodology in weather and climate prediction and their benefits have been widely discussed in the literature. Typical approaches involve combining the output of different numerical weather prediction (NWP) models using constant weighting factors, either uniformly distributed or determined through a prior skill assessment. This strategy, however, can lead to suboptimal levels of skill, as the performance of NWP models can vary with time (e.g., seasonally varying skill, changes in the forecasting system). Moreover, standard combination methods are not designed to incorporate predictions derived from sources other than NWP systems (e.g., climatological or time‐series forecasts). New algorithms developed within the machine learning community provide the opportunity for “online prediction” (also referred to as “sequential learning”). These methods consider a set of weighted predictors or “experts” to produce subsequent predictions in which the combination or “mixture” is updated at each step to optimize a loss or skill function. The predictors are highly flexible and can combine both NWP and statistically derived forecasts transparently. A set of these online prediction methods is tested and compared with standard multimodel combination techniques to assess their usefulness. The methods are general and can be applied to any model‐derived predictand. A set of weather‐sensitive European country‐aggregate energy variables (electricity demand and wind power) is selected for demonstration purposes. Results show that these innovative methods exhibit significant skill improvements (i.e., between 5 and 15% improvement in the probabilistic skill) with respect to standard multimodel combination techniques for lead weeks up to 5. The incorporation of statistically derived predictors (based on historical climate data) alongside NWP forecasts is also shown to contribute significant skill improvements in many cases.
Errors in numerical weather prediction (NWP) systems limit their skill. It is therefore desirable to combine multiple NWP systems to enhance skill, particularly at extended range. Here, a new set of schemes—sequential learning algorithms—is shown to offer significant benefits over existing combination schemes. The schemes are well‐suited to operational use and, in contrast to more traditional schemes, allow weights to evolve dynamically and the transparent inclusion of statistical forecasts. Skill improvements of up to 5–15% are found.</description><identifier>ISSN: 0035-9009</identifier><identifier>EISSN: 1477-870X</identifier><identifier>DOI: 10.1002/qj.4177</identifier><language>eng</language><publisher>Chichester, UK: John Wiley & Sons, Ltd</publisher><subject>Algorithms ; Climate prediction ; Climatic data ; expert advice ; extended range ; Learning algorithms ; Machine learning ; Methods ; multimodel ; online prediction ; s2s forecasting ; sequential learning ; subseasonal ; Weather ; Weather forecasting ; Wind power</subject><ispartof>Quarterly journal of the Royal Meteorological Society, 2021-10, Vol.147 (741), p.4269-4282</ispartof><rights>2021 The Authors. published by John Wiley & Sons Ltd on behalf of the Royal Meteorological Society.</rights><rights>2021. This article is published under http://creativecommons.org/licenses/by/4.0/ (the “License”). Notwithstanding the ProQuest Terms and Conditions, you may use this content in accordance with the terms of the License.</rights><lds50>peer_reviewed</lds50><oa>free_for_read</oa><woscitedreferencessubscribed>false</woscitedreferencessubscribed><citedby>FETCH-LOGICAL-c3227-36a0ce28db35c495318860ce03d5b426567deafc0580c8c64ce6653c9d1a3a413</citedby><cites>FETCH-LOGICAL-c3227-36a0ce28db35c495318860ce03d5b426567deafc0580c8c64ce6653c9d1a3a413</cites><orcidid>0000-0003-0154-0087</orcidid></display><links><openurl>$$Topenurl_article</openurl><openurlfulltext>$$Topenurlfull_article</openurlfulltext><thumbnail>$$Tsyndetics_thumb_exl</thumbnail><linktopdf>$$Uhttps://onlinelibrary.wiley.com/doi/pdf/10.1002%2Fqj.4177$$EPDF$$P50$$Gwiley$$Hfree_for_read</linktopdf><linktohtml>$$Uhttps://onlinelibrary.wiley.com/doi/full/10.1002%2Fqj.4177$$EHTML$$P50$$Gwiley$$Hfree_for_read</linktohtml><link.rule.ids>314,780,784,1417,27924,27925,45574,45575</link.rule.ids></links><search><creatorcontrib>Gonzalez, Paula L. M.</creatorcontrib><creatorcontrib>Brayshaw, David J.</creatorcontrib><creatorcontrib>Ziel, Florian</creatorcontrib><title>A new approach to extended‐range multimodel forecasting: Sequential learning algorithms</title><title>Quarterly journal of the Royal Meteorological Society</title><description>Multimodel combinations are a well‐established methodology in weather and climate prediction and their benefits have been widely discussed in the literature. Typical approaches involve combining the output of different numerical weather prediction (NWP) models using constant weighting factors, either uniformly distributed or determined through a prior skill assessment. This strategy, however, can lead to suboptimal levels of skill, as the performance of NWP models can vary with time (e.g., seasonally varying skill, changes in the forecasting system). Moreover, standard combination methods are not designed to incorporate predictions derived from sources other than NWP systems (e.g., climatological or time‐series forecasts). New algorithms developed within the machine learning community provide the opportunity for “online prediction” (also referred to as “sequential learning”). These methods consider a set of weighted predictors or “experts” to produce subsequent predictions in which the combination or “mixture” is updated at each step to optimize a loss or skill function. The predictors are highly flexible and can combine both NWP and statistically derived forecasts transparently. A set of these online prediction methods is tested and compared with standard multimodel combination techniques to assess their usefulness. The methods are general and can be applied to any model‐derived predictand. A set of weather‐sensitive European country‐aggregate energy variables (electricity demand and wind power) is selected for demonstration purposes. Results show that these innovative methods exhibit significant skill improvements (i.e., between 5 and 15% improvement in the probabilistic skill) with respect to standard multimodel combination techniques for lead weeks up to 5. The incorporation of statistically derived predictors (based on historical climate data) alongside NWP forecasts is also shown to contribute significant skill improvements in many cases.
Errors in numerical weather prediction (NWP) systems limit their skill. It is therefore desirable to combine multiple NWP systems to enhance skill, particularly at extended range. Here, a new set of schemes—sequential learning algorithms—is shown to offer significant benefits over existing combination schemes. The schemes are well‐suited to operational use and, in contrast to more traditional schemes, allow weights to evolve dynamically and the transparent inclusion of statistical forecasts. Skill improvements of up to 5–15% are found.</description><subject>Algorithms</subject><subject>Climate prediction</subject><subject>Climatic data</subject><subject>expert advice</subject><subject>extended range</subject><subject>Learning algorithms</subject><subject>Machine learning</subject><subject>Methods</subject><subject>multimodel</subject><subject>online prediction</subject><subject>s2s forecasting</subject><subject>sequential learning</subject><subject>subseasonal</subject><subject>Weather</subject><subject>Weather forecasting</subject><subject>Wind power</subject><issn>0035-9009</issn><issn>1477-870X</issn><fulltext>true</fulltext><rsrctype>article</rsrctype><creationdate>2021</creationdate><recordtype>article</recordtype><sourceid>24P</sourceid><sourceid>WIN</sourceid><recordid>eNp10MtKw0AUBuBBFKxVfIUBFy4k9cw9cVeKVwoiKugqTCcnbUKatDMptTsfwWf0SUytW1cHfj7OjZBTBgMGwC-X5UAyY_ZIj0ljotjA2z7pAQgVJQDJITkKoQQAZbjpkfchrXFN7WLhG-tmtG0ofrRYZ5h9f355W0-RzldVW8ybDCuaNx6dDW1RT6_oMy5XWLeFrWiF1tddSG01bXzRzubhmBzktgp48lf75PXm-mV0F40fb-9Hw3HkBOcmEtqCQx5nE6GcTJRgcay7BESmJpJrpU2GNnegYnCx09Kh1kq4JGNWWMlEn5zt-nYXdPuENi2bla-7kSnXkGgjteSdOt8p55sQPObpwhdz6zcpg3T7t3RZptu_dfJiJ9dFhZv_WPr08Kt_AA4Sbu4</recordid><startdate>202110</startdate><enddate>202110</enddate><creator>Gonzalez, Paula L. M.</creator><creator>Brayshaw, David J.</creator><creator>Ziel, Florian</creator><general>John Wiley & Sons, Ltd</general><general>Wiley Subscription Services, Inc</general><scope>24P</scope><scope>WIN</scope><scope>AAYXX</scope><scope>CITATION</scope><scope>7TG</scope><scope>7TN</scope><scope>F1W</scope><scope>H96</scope><scope>KL.</scope><scope>L.G</scope><orcidid>https://orcid.org/0000-0003-0154-0087</orcidid></search><sort><creationdate>202110</creationdate><title>A new approach to extended‐range multimodel forecasting: Sequential learning algorithms</title><author>Gonzalez, Paula L. M. ; Brayshaw, David J. ; Ziel, Florian</author></sort><facets><frbrtype>5</frbrtype><frbrgroupid>cdi_FETCH-LOGICAL-c3227-36a0ce28db35c495318860ce03d5b426567deafc0580c8c64ce6653c9d1a3a413</frbrgroupid><rsrctype>articles</rsrctype><prefilter>articles</prefilter><language>eng</language><creationdate>2021</creationdate><topic>Algorithms</topic><topic>Climate prediction</topic><topic>Climatic data</topic><topic>expert advice</topic><topic>extended range</topic><topic>Learning algorithms</topic><topic>Machine learning</topic><topic>Methods</topic><topic>multimodel</topic><topic>online prediction</topic><topic>s2s forecasting</topic><topic>sequential learning</topic><topic>subseasonal</topic><topic>Weather</topic><topic>Weather forecasting</topic><topic>Wind power</topic><toplevel>peer_reviewed</toplevel><toplevel>online_resources</toplevel><creatorcontrib>Gonzalez, Paula L. M.</creatorcontrib><creatorcontrib>Brayshaw, David J.</creatorcontrib><creatorcontrib>Ziel, Florian</creatorcontrib><collection>Wiley Online Library Open Access</collection><collection>Wiley Online Library (Open Access Collection)</collection><collection>CrossRef</collection><collection>Meteorological & Geoastrophysical Abstracts</collection><collection>Oceanic Abstracts</collection><collection>ASFA: Aquatic Sciences and Fisheries Abstracts</collection><collection>Aquatic Science & Fisheries Abstracts (ASFA) 2: Ocean Technology, Policy & Non-Living Resources</collection><collection>Meteorological & Geoastrophysical Abstracts - Academic</collection><collection>Aquatic Science & Fisheries Abstracts (ASFA) Professional</collection><jtitle>Quarterly journal of the Royal Meteorological Society</jtitle></facets><delivery><delcategory>Remote Search Resource</delcategory><fulltext>fulltext</fulltext></delivery><addata><au>Gonzalez, Paula L. M.</au><au>Brayshaw, David J.</au><au>Ziel, Florian</au><format>journal</format><genre>article</genre><ristype>JOUR</ristype><atitle>A new approach to extended‐range multimodel forecasting: Sequential learning algorithms</atitle><jtitle>Quarterly journal of the Royal Meteorological Society</jtitle><date>2021-10</date><risdate>2021</risdate><volume>147</volume><issue>741</issue><spage>4269</spage><epage>4282</epage><pages>4269-4282</pages><issn>0035-9009</issn><eissn>1477-870X</eissn><abstract>Multimodel combinations are a well‐established methodology in weather and climate prediction and their benefits have been widely discussed in the literature. Typical approaches involve combining the output of different numerical weather prediction (NWP) models using constant weighting factors, either uniformly distributed or determined through a prior skill assessment. This strategy, however, can lead to suboptimal levels of skill, as the performance of NWP models can vary with time (e.g., seasonally varying skill, changes in the forecasting system). Moreover, standard combination methods are not designed to incorporate predictions derived from sources other than NWP systems (e.g., climatological or time‐series forecasts). New algorithms developed within the machine learning community provide the opportunity for “online prediction” (also referred to as “sequential learning”). These methods consider a set of weighted predictors or “experts” to produce subsequent predictions in which the combination or “mixture” is updated at each step to optimize a loss or skill function. The predictors are highly flexible and can combine both NWP and statistically derived forecasts transparently. A set of these online prediction methods is tested and compared with standard multimodel combination techniques to assess their usefulness. The methods are general and can be applied to any model‐derived predictand. A set of weather‐sensitive European country‐aggregate energy variables (electricity demand and wind power) is selected for demonstration purposes. Results show that these innovative methods exhibit significant skill improvements (i.e., between 5 and 15% improvement in the probabilistic skill) with respect to standard multimodel combination techniques for lead weeks up to 5. The incorporation of statistically derived predictors (based on historical climate data) alongside NWP forecasts is also shown to contribute significant skill improvements in many cases.
Errors in numerical weather prediction (NWP) systems limit their skill. It is therefore desirable to combine multiple NWP systems to enhance skill, particularly at extended range. Here, a new set of schemes—sequential learning algorithms—is shown to offer significant benefits over existing combination schemes. The schemes are well‐suited to operational use and, in contrast to more traditional schemes, allow weights to evolve dynamically and the transparent inclusion of statistical forecasts. Skill improvements of up to 5–15% are found.</abstract><cop>Chichester, UK</cop><pub>John Wiley & Sons, Ltd</pub><doi>10.1002/qj.4177</doi><tpages>14</tpages><orcidid>https://orcid.org/0000-0003-0154-0087</orcidid><oa>free_for_read</oa></addata></record> |
fulltext | fulltext |
identifier | ISSN: 0035-9009 |
ispartof | Quarterly journal of the Royal Meteorological Society, 2021-10, Vol.147 (741), p.4269-4282 |
issn | 0035-9009 1477-870X |
language | eng |
recordid | cdi_proquest_journals_2609674642 |
source | Access via Wiley Online Library |
subjects | Algorithms Climate prediction Climatic data expert advice extended range Learning algorithms Machine learning Methods multimodel online prediction s2s forecasting sequential learning subseasonal Weather Weather forecasting Wind power |
title | A new approach to extended‐range multimodel forecasting: Sequential learning algorithms |
url | https://sfx.bib-bvb.de/sfx_tum?ctx_ver=Z39.88-2004&ctx_enc=info:ofi/enc:UTF-8&ctx_tim=2024-12-24T17%3A29%3A59IST&url_ver=Z39.88-2004&url_ctx_fmt=infofi/fmt:kev:mtx:ctx&rfr_id=info:sid/primo.exlibrisgroup.com:primo3-Article-proquest_cross&rft_val_fmt=info:ofi/fmt:kev:mtx:journal&rft.genre=article&rft.atitle=A%20new%20approach%20to%20extended%E2%80%90range%20multimodel%20forecasting:%20Sequential%20learning%20algorithms&rft.jtitle=Quarterly%20journal%20of%20the%20Royal%20Meteorological%20Society&rft.au=Gonzalez,%20Paula%20L.%20M.&rft.date=2021-10&rft.volume=147&rft.issue=741&rft.spage=4269&rft.epage=4282&rft.pages=4269-4282&rft.issn=0035-9009&rft.eissn=1477-870X&rft_id=info:doi/10.1002/qj.4177&rft_dat=%3Cproquest_cross%3E2609674642%3C/proquest_cross%3E%3Curl%3E%3C/url%3E&disable_directlink=true&sfx.directlink=off&sfx.report_link=0&rft_id=info:oai/&rft_pqid=2609674642&rft_id=info:pmid/&rfr_iscdi=true |