Monthly Multiple Linear Regression Models for OMI-Absorbing Aerosol Index Estimation in Nigeria’s Sahel Region

Abstract

This study presents the development of monthly multiple linear regression (MLR) models for estimating the Absorbing Aerosol Index (AAI) from the Ozone Monitoring Instrument (OMI) over the Sahel region of Nigeria, using ground-based meteorological variables. The key predictors—wind speed, visibility, air temperature, and relative humidity—were selected for their relevance in characterizing atmospheric dust loading, particularly during the Harmattan season. Empirical regression equations were constructed with optimized coefficients derived from observed AI and meteorological data. These models were then validated using independent datasets to assess generalizability. Model performance was evaluated through the coefficient of determination (R²) and Root Mean Square Error (RMSE), both at the 95% confidence level.

The results demonstrated that the MLR models are capable of reliably predicting monthly AI values, with prediction accuracy varying seasonally. During the dry Harmattan months, when dust activity is high and AI values are more variable, the models achieved higher R² values despite larger RMSE—indicating strong model fit with greater data spread. Conversely, in the wet summer season, when aerosol levels are generally low and less variable, the models exhibited reduced R² values but also lower RMSE, due to limited variability in AI. Interestingly, transitional months such as April and October yielded the most balanced results, with both higher R² and lower RMSE values. These findings underscore the potential of ground-based meteorological data in filling spatial or temporal gaps in satellite-derived aerosol measurements. The models provide a practical tool for improving aerosol monitoring and early warning systems in dust-prone environments like the Sahel.

Keywords: Aerosol; Linear Regression; Models; TOMS AI; Sahel