What works better with LSTM, decomposition or deseasonalisation for rainfall forecasting?

What works better with LSTM, decomposition or deseasonalisation for rainfall forecasting? | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Research Article What works better with LSTM, decomposition or deseasonalisation for rainfall forecasting? Achal Lama, Debopam Rakshit, K N Singh, Pankaj Das, Ritwika Das, and 2 more This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-5155959/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Forecasting rainfall is crucial for countries like India where farming is the livelihood for around half of the population and rainfall is their most important water source. The intensity of rainfall varies for different seasons and is not spread evenly across the country. Over the years, different researchers used various statistical models for rainfall forecasting. This article uses the monthly rainfall series for all India and five sub-divisions, namely Central North East, North East, North West, Peninsular and West Central for modeling and forecasting. It has been observed that, as a time series, these series follow nonlinearity and non-normality but are stationary. Again, all the series are seasonal. Decomposition or deseasonalised (DS) based hybrid models may be useful for this type of complex scenario. The variational mode decomposition (VMD) is applied to the both actual rainfall series and deseasonalised series (DS) and the Intrinsic Mode Functions (IMFs) are obtained. The Long Short-Term Memory (LSTM) model is fitted to these IMFs. LSTM is also applied to the deseasonalised series leading to DS-LSTM model. The traditional seasonal autoregressive integrated moving average (SARIMA) and LSTM model also fitted directly to the actual rainfall series. The DS- -LSTM hybrid model established its superiority in forecasting compared to the standalone SARIMA, LSTM, VMD-LSTM and DS-VMD-LSTM models based on Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) in the model testing set for all the series. Hence, establishing the importance of deseasonalisation of the series before applying appropriate model to it. Deep learning Nonlinearity Rainfall Deseasonalisation Time series Full Text Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-5155959","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":364981041,"identity":"2035de20-38c3-4442-8b2d-aad30f9db675","order_by":0,"name":"Achal Lama","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA90lEQVRIiWNgGAWjYFACHgaGBwwWQAbzgQNwwQRCWhIYJIAMtgSStfAYEOcs/vazBz8Atcjrtp/5eOBHzT0GefceA4aHO3BrkTiTlywB1GK47UzuhoM9x4oZDM+cMWBIPIPHmhs8BiAtjNsO5G44DPQPg+GMtASGxDbcOuRv8Bj/AGqx33b+zYPDDP+I0GJwg8cMZEviths5DIcZ2xIY5CWSD+DVYngmx8wiwUAieduNZwYHe/sSeAx4Dh84gE+L3PEzxjc+VNjYbjuf/PjDj28JcvLtjY0Pf+LRAnUegsljcICB4QAhDahAvoE09aNgFIyCUTD8AQAGglPLELRaqgAAAABJRU5ErkJggg==","orcid":"https://orcid.org/0000-0002-5376-3760","institution":"ICAR Indian Agricultural Statistics Research Institute","correspondingAuthor":true,"prefix":"","firstName":"Achal","middleName":"","lastName":"Lama","suffix":""},{"id":364981042,"identity":"b5149c15-a0bb-4bdb-9835-e0ad6b342496","order_by":1,"name":"Debopam Rakshit","email":"","orcid":"","institution":"Indian Veterinary Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Debopam","middleName":"","lastName":"Rakshit","suffix":""},{"id":364981043,"identity":"69184aa7-4f63-44d7-a5e2-e020487637f1","order_by":2,"name":"K N Singh","email":"","orcid":"","institution":"Indian Agricultural Statistics Research Institute","correspondingAuthor":false,"prefix":"","firstName":"K","middleName":"N","lastName":"Singh","suffix":""},{"id":364981044,"identity":"fbee59ac-220e-42b8-9561-f5ea956eb0f9","order_by":3,"name":"Pankaj Das","email":"","orcid":"","institution":"Indian Agricultural Statistics Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Pankaj","middleName":"","lastName":"Das","suffix":""},{"id":364981045,"identity":"78cf33b2-2b92-45c0-85fc-5dbe548cee1a","order_by":4,"name":"Ritwika Das","email":"","orcid":"","institution":"Indian Agricultural Statistics Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Ritwika","middleName":"","lastName":"Das","suffix":""},{"id":364981046,"identity":"547b1a8b-0547-4f66-95ec-5a1c802704ed","order_by":5,"name":"Satyam Verma","email":"","orcid":"","institution":"Indian Agricultural Statistics Research Institute","correspondingAuthor":false,"prefix":"","firstName":"Satyam","middleName":"","lastName":"Verma","suffix":""},{"id":364981047,"identity":"ea1465f1-939b-484e-8afa-1297ab491d8b","order_by":6,"name":"Bishal Gurung","email":"","orcid":"","institution":"North Eastern Hill University","correspondingAuthor":false,"prefix":"","firstName":"Bishal","middleName":"","lastName":"Gurung","suffix":""}],"badges":[],"createdAt":"2024-09-26 05:58:58","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-5155959/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-5155959/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104399453,"identity":"a858b8c9-67d2-4fa9-8e60-ea6744f91c55","added_by":"auto","created_at":"2026-03-11 12:06:12","extension":"pdf","order_by":1,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":859887,"visible":true,"origin":"","legend":"","description":"","filename":"DSLSTMWaterResourceManagement.pdf","url":"https://assets-eu.researchsquare.com/files/rs-5155959/v1_covered_a170849a-8188-4169-b227-4debc85522bf.pdf"}],"financialInterests":"","formattedTitle":"What works better with LSTM, decomposition or deseasonalisation for rainfall forecasting?","fulltext":[],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":false,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":true,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":true,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Deep learning, Nonlinearity, Rainfall, Deseasonalisation, Time series","lastPublishedDoi":"10.21203/rs.3.rs-5155959/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-5155959/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eForecasting rainfall is crucial for countries like India where farming is the livelihood for around half of the population and rainfall is their most important water source. The intensity of rainfall varies for different seasons and is not spread evenly across the country. Over the years, different researchers used various statistical models for rainfall forecasting. This article uses the monthly rainfall series for all India and five sub-divisions, namely Central North East, North East, North West, Peninsular and West Central for modeling and forecasting. It has been observed that, as a time series, these series follow nonlinearity and non-normality but are stationary. Again, all the series are seasonal. Decomposition or deseasonalised (DS) based hybrid models may be useful for this type of complex scenario. The variational mode decomposition (VMD) is applied to the both actual rainfall series and deseasonalised series (DS) and the Intrinsic Mode Functions (IMFs) are obtained. The Long Short-Term Memory (LSTM) model is fitted to these IMFs. LSTM is also applied to the deseasonalised series leading to DS-LSTM model. The traditional seasonal autoregressive integrated moving average (SARIMA) and LSTM model also fitted directly to the actual rainfall series. The DS- -LSTM hybrid model established its superiority in forecasting compared to the standalone SARIMA, LSTM, VMD-LSTM and DS-VMD-LSTM models based on Root Mean Squared Error (RMSE) and Mean Absolute Error (MAE) in the model testing set for all the series. Hence, establishing the importance of deseasonalisation of the series before applying appropriate model to it.\u003c/p\u003e","manuscriptTitle":"What works better with LSTM, decomposition or deseasonalisation for rainfall forecasting?","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2024-10-15 09:45:33","doi":"10.21203/rs.3.rs-5155959/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"8e4adc95-f5bc-48a7-9116-70ef3f5726e0","owner":[],"postedDate":"October 15th, 2024","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-02-27T16:03:11+00:00","versionOfRecord":[],"versionCreatedAt":"2024-10-15 09:45:33","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-5155959","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-5155959","identity":"rs-5155959","version":["v1"]},"buildId":"qtupq5eGEP_6zYnWcrvyt","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}