Developing machine learning algorithms for meteorological temperature and humidity forecasting at Terengganu state in...

Area of Science:

• Meteorology
• Environmental Science
• Data Science

Background:

• Accurate prediction of meteorological parameters like air temperature and humidity is vital for effective air quality management.
• Traditional forecasting methods may have limitations in capturing complex atmospheric dynamics.

Purpose of the Study:

• To evaluate the performance of various machine learning algorithms for predicting daily and monthly air temperature (T) and relative humidity (Rh).
• To compare the efficacy of Gradient Boosting Tree (G.B.T.), Random Forest (R.F.), Linear Regression (LR), and Artificial Neural Network (ANN) architectures, specifically Multi-Layered Perceptron (MLP-NN) and Radial Basis Function (RBF-NN).

Main Methods:

• Utilized 24 years of daily meteorological data from the Kula Terengganu station, Malaysia.
• Trained and validated multiple machine learning models including G.B.T., R.F., LR, MLP-NN, and RBF-NN.
• Assessed model performance using correlation coefficients (R) and standard deviation, and validated on an unseen dataset.

Main Results:

• MLP-NN demonstrated superior performance in predicting daily T (R=0.7132) and Rh (R=0.633).
• For monthly T prediction, MLP-NN yielded a high correlation (R=0.8462) and closer standard deviation to actual values.
• RBF-NN showed higher efficiency in predicting monthly Rh (R=0.7113) compared to other models.

Conclusions:

• Artificial Neural Network architectures, including MLP-NN and RBF-NN, exhibit significant potential for accurate daily and monthly prediction of air temperature and relative humidity.
• The study validates the applicability of ANNs for meteorological forecasting with acceptable accuracy.
• Findings support the use of advanced machine learning techniques in environmental and air quality management applications.