Investigation of the Impact of Environmental Parameters on RF Signal Propagation in Eagle Island, Port Harcourt Using Artificial Neural Networks

Abstract

This study investigates the impact of environmental parameters on LTE radio frequency (RF) signal propagation in Eagle Island, Port Harcourt, using an Artificial Neural Network (ANN)-based modelling approach. Conventional empirical propagation models often struggle to represent the complex interaction between atmospheric conditions, vegetation density, urban clutter, and signal attenuation in tropical coastal environments. To address this limitation, field measurements were collected through a structured drive-test campaign conducted on the LTE 1800 MHz band across Eagle Island. A total of 600 geo-referenced measurement samples were obtained, including RF parameters such as Reference Signal Received Power (RSRP), Signal-to-Interference-plus-Noise Ratio (SINR), and path loss, alongside environmental variables including temperature, relative humidity, rainfall intensity, vegetation density, and building clutter density. The collected dataset was preprocessed and used to train a Feedforward Neural Network (FFNN) implemented in MATLAB using the Levenberg–Marquardt backpropagation algorithm. The developed ANN model achieved a Root Mean Square Error (RMSE) of 4.28 dB and a coefficient of determination (R²) of 0.9209, outperforming the COST-231 Hata model, which recorded an RMSE of 5.12 dB and R² of 0.7215. Sensitivity analysis further revealed that vegetation density and relative humidity significantly influence propagation behavior beyond distance alone. The results demonstrate that ANN-based models provide a more reliable framework for RF propagation prediction in environmentally complex tropical regions and can support more accurate network planning and optimization.