Loading...
Machine learning-based projections of earth skin temperature anomalies in Nigeria using ERA5-land data
Lateef, Mayowa Benjamen Ayanlade, Oluwatoyin Seun Adepiti, Awodayo Oluwatoyin Ayanlade, Ayansina
Lateef, Mayowa Benjamen
Ayanlade, Oluwatoyin Seun
Adepiti, Awodayo Oluwatoyin
Ayanlade, Ayansina
Title / Series / Name
Publication Volume
2
Publication Issue
1
Pages
Editors
Keywords
SDG 13 - Climate Action
Files
URI
https://hdl.handle.net/20.500.14018/28853
Abstract
This study examines spatiotemporal dynamics of Earth Skin Temperature anomalies across Nigeria using combination of machine learning and remote sensing technologies. Based on XGBoost models trained on ERA5-Land historical-reanalysis temperature dataset of 1993–2023, the study estimates temperature projection for 2028–2058. The study provides detailed insights into future temperature patterns through systematic sampling across 70 geographical sites and inverse distance weighting interpolation. The results demonstrate significant geographical variation in temperature, with southwestern parts displaying continuous positive anomalies and coastal areas demonstrating milder changes. Warm temperature are predicted to be more intensified from 2043 onward while extreme heat is widespread across almost the entire country, much more between 2053 and 2058. The model’s reliability was tested by RMSE analysis, providing values between 0.30 - 0.36 °C, with high predictive power. These findings give valuable information for Nigeria’s climate adaptation strategies and environmental management, notably emphasizing regions requiring targeted attention due to expected temperature extremes. The results from the projections in this study indicate that extreme heat, with anomalies exceeding +1.0 °C, will pose significant climate challenge in the nearest future and this necessitates critical climate adaptation strategies such as heat mitigation, improved water management, and climate-resilient agriculture in may part of the country.
Topic
Publisher
Place of Publication
Type
Journal article
Date
2026-03
Language
ISBN
Identifiers
10.1080/29966876.2026.2624855