Accurate long-term air temperature prediction - Machine Learning models and data reduction techniques

Aug 4, 2026

Long-term summer air temperature forecasting is addressed through three customised AI frameworks combining deep learning, classical machine learning, and data reduction techniques. Average air temperature for the first and second August fortnights is predicted from previous-month inputs at Paris (France) and Córdoba (Spain), with the target reflecting potential extreme-heatwave signals such as 2003. Models include a CNN via video-to-image translation, ML baselines (Lasso, Decision Trees, Random Forest), and a CNN using Recurrence Plots that map time series to images. Strong prediction skill supports these methods for seasonal climate prediction.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages20

Reading Time50 min

AccessFree to download

Introduction

Seasonal Climate Prediction and motivation

Related work on air temperature forecasting

Proposed frameworks and methods

CNN video-to-image translation

Classical ML baselines

Recurrence plots with CNN

| Accurate long-term air temperature prediction with Machine Learning models and data reduction techniques | | | | --- | --- | --- | | D. Fister a , J. Pérez-Aracil a ,∗, C. Peláez-Rodríguez a , J. Del Serb,c , S. Salcedo-Sanz aa Department of Signal Processing and Communications, Universidad de Alcalá, 28805, Madrid, Spain
b TECNALIA, Basque Research & Technology Alliance (BRTA), 48160 Derio, Spain
c University of the Basque Country (UPV/EHU), 48013 Bilbao, Spain | | | | a r t i c l e i n f o | a b s t r a c t
In this paper, three customised Artificial Intelligence (AI) frameworks, considering Deep Learning, Machine Learning (ML) algorithms and data reduction techniques, are proposed for a problem of longterm summer air temperature prediction. Specifically, the prediction of the average air temperature in the first and second August fortnights, using input data from previous months, at two different locations (Paris, France) and (Córdoba, Spain), is considered. The target variable, mainly in the first August fortnight, can contain signals of extreme events such as heatwaves, like the heatwave of 2003, which affected France and the Iberian Peninsula. Three different computational frameworks for air temperature prediction are proposed: a Convolutional Neural Network (CNN), with video-to-image translation, several ML approaches including Lasso regression, Decision Trees and Random Forest, and finally a CNN with pre-processing step using Recurrence Plots, which convert time series into images. Using these frameworks, a very good prediction skill has been obtained in both Paris and Córdoba regions, showing that the proposed approaches can be an excellent option for seasonal climate prediction problems.
© 2023 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license ([http://creativecommons.org/licenses/by-nc-nd/4.0/](http://creativecommons.org/licenses/by-nc-nd/4.0/)). | | | Article history:
Received 9 October 2022
Received in revised form 3 January 2023 Accepted 8 February 2023
Available online 13 February 2023 | | | | Keywords:
Deep Learning Temperature prediction Recurrence plots
Data reduction techniques | | | 1. Introduction Seasonal Climate Prediction (SCP) has gained momentum in the last decade [1], becoming an important field of study, with applications in very different areas such as agriculture, risk management, long-term energy planning or climate change and extreme events modelling [2,3], among others. SCP problems are specially interesting in the current context of climate change, since they may have important consequences in the future [4]. One of such climate change effects are the constantly rising long-termed average temperature, coined as the so-called global warming, and associated greenhouse gases [5,6]. However, the constantly changing weather conditions not only affect the longterm temperature averages, but also stipulate temporally much shorter periods with drastically large deviations from steady levels, producing extreme phenomena such as heatwaves and severe droughts. Evidence shows that these extreme weather events can cause worldwide consequences and impacts in natural resources (agriculture, construction, renewable energy) [7,8], financial sector [9] ∗ Corresponding author. E-mail addresses: [dusan.fister@uah.es](dusan.fister@uah.es) (D. Fister), [jorge.perezaracil@uah.es](jorge.perezaracil@uah.es)[ ](jorge.perezaracil@uah.es)(J. Pérez-Aracil), cesar.pelaez@uah.es (C. Peláez-Rodríguez), [javier.delser@tecnalia.com](javier.delser@tecnalia.com) (J. Del Ser), [sancho.salcedo@uah.es](sancho.salcedo@uah.es) (S. Salcedo-Sanz). and of course human’s health [10, 11]. Also, one of the effects of climate change is to produce warmer summers [12], which can be further studied by predicting average summer months temperature at a long-term basis. SCP related to air temperature are, therefore, extremely important and challengin

FAQs

What locations and time windows are used for long-term summer air temperature prediction?

The study predicts average air temperature for the first and second August fortnights using previous-month inputs at Paris (France) and Córdoba (Spain).

Which AI frameworks are proposed for forecasting?

Three frameworks are proposed: a CNN with video-to-image translation, several ML approaches including Lasso regression, Decision Trees, and Random Forest, and a CNN using Recurrence Plots to convert time series into images.

Why is predicting August fortnight temperatures challenging in the context of climate change?

The target period can include signals of extreme events like heatwaves; such extremes cause large deviations from long-term temperature averages and make long-horizon forecasting difficult.

Aug 4, 2026

Upload by Lucas Martin