Forecasting Rainfall in Selected Cities of Southwest Nigeria: A Comparative Study of Random Forest and Long Short-Term Memory Models
Journal Title: Acadlore Transactions on Geosciences - Year 2024, Vol 3, Issue 2
Abstract
Rainfall is crucial for agricultural practices, and climate change has significantly altered rainfall patterns. Understanding the dynamic nature of rainfall in the context of climate change through Machine Learning (ML) and Deep Learning (DL) algorithms is essential for ensuring food security. ML techniques provide tools for processing large-scale data to extract meaningful insights. This study aims to compare the performance of a ML algorithm, Random Forest (RF), with a DL algorithm, Long Short-Term Memory (LSTM), in predicting rainfall in six state capitals in Southwest Nigeria: Osogbo, Ikeja, Ibadan, Akure, Ado-Ekiti, and Abeokuta. The dataset for this study was sourced from the HelioClim website archive, which offers high-quality solar radiation and meteorological data derived from satellite measurements. This archive is known for its accuracy and reliability, providing extensive and consistent historical datasets for various applications. The monthly rainfall data spanning from January 1, 1980, to December 31, 2022, were used, with 80% allocated for training and 20% for validation. As the data are time series, each model was constructed using a look-back period of five months, meaning the past five months' rainfall data served as input features. The performance of these algorithms was evaluated using Mean Square Error (MSE), Root Mean Square Error (RMSE), and Mean Absolute Error (MAE). The results indicated that the RF algorithm yielded the lowest MSE, RMSE, and MAE across all selected cities in Southwest Nigeria. This study demonstrated the superiority of RF regression over LSTM in predicting rainfall in these regions, providing a valuable tool for agricultural planning and climate adaptation strategies.
Authors and Affiliations
Timothy Kayode Samson, Francis Olatunbosun Aweda
Keywords
Related Articles
- EP ID EP752407
- DOI 10.56578/atg030202
- Views 9
- Downloads 0
Identification of Areas with Significant Flood Risks in Counties along the Danube River in Serbia and Their Risk Assessment
For countries along the Danube River, their sustainable economic and social development needs the optimum water utilization of both the Danube and its tributaries. In the context of climate change, the risks of floods an...
Inverse Analysis of Rock Mass Dynamic Parameters from Blasting Vibration Signals
The precision of determining rock mass mechanical parameters is notably impacted by mining blast activities. An advanced method for inverse analysis of these parameters, predicated upon measured blasting vibrations, has...
Morphological Distribution and Phase Composition of Rare Earth Elements in Waste Incineration Fly Ash
Utilising scanning electron microscopy (SEM) and X-ray powder diffraction (XRD), the morphological and phase composition characteristics of waste incineration fly ash were meticulously analysed. Morphological evaluations...
Stratigraphic and Structural Analysis of the Jaintia-Jaflong Monocline: Insights from Field Investigations
The Jaintiapur-Jaflong region, strategically positioned between the subsiding Surma Basin to the south and the uplifting Shillong Massif to the north, presents a unique geological setting. This study employed geological...
Comparative Analysis of Heavy Metal Concentrations and Potential Health Risks Across Varied Land-Use Zones in Ado-Ekiti, Southwest Nigeria
An investigation was conducted to determine potential threats of heavy metal contaminants in soil samples from Ado-Ekiti, Southwest Nigeria, across distinct land-use zones. Five soil specimens were systematically gathere...