Answers are grounded in this paper and cite their sources.
Predicting land surface temperature in an area is crucial as it equips stakeholders and governmental organizations with enhanced insights into the temperature dynamics, facilitating informed decision- making and planning. The primary aim of this study is to predict the daily land surface temperature of the Greater Accra region. This research first employed the Discrete Wavelet Transform (DWT) to decompose temperature data into signals representing lower and higher frequencies. Six distinct Backpropagation Artificial Neural Network machine-learning techniques were employed: Bayesian Regularization, Levenberg-Marquardt, Scaled Conjugate Gradient, Resilient algorithm, Fletcher-Reeves Conjugate Gradient, and Polak-Ribiere Conjugate Gradient. Additionally, two unsupervised machine learning methodologies, specifically the Radial Basis Function and Generalized Regression, with three conventional techniques, Autocorrelation Integrated Moving Average, Multiple Linear Regression, and Total Least Squares, were employed. The decomposed data were utilized in the training of these algorithms. The collection comprises 58 years of temporal temperature data, spanning from 1960 to 2018, obtained from the Ghana Meteorological Agency. The research examines four discrete statistical metrics to evaluate the efficacy of these algorithms. The findings indicated that the performance measures of all algorithms are promising. Upon analysis of all the methods, it was observed that the ARIMA had subpar performance for the MSE and RMSE measures. The GRANN algorithm was identified as the most effective. All approaches described in this study may be applied globally to precisely predict daily temperature.
No takes yet. Share an insight, caveat, or question.
Peprah et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: