Enhancing PM2.5 Predictions in Dakar Through Automated Data Integration into a Data Assimilation Model

Abstract

The objective of this work is to predict daily PM2.5 air quality in Dakar, Senegal using data from an automated measurement station integrated into a server using a data assimilation model. Initially, a 3-year data set was used to identify and validate an appropriate ARIMA data assimilation model. The data was split into an 80% training set and a 20% test set. The Augmented Dickey-Fuller (ADF) test was used to check the normality of the data series. Subsequently, we used the AutoArima method to determine the optimal model to represent the time series. Preliminary results show that a model with order (2,1,1) accurately represents the series. Additional analysis using model fit tests showed that the (3, 0, 1) model was most effective in representing and predicting the data. The statistical validation performance of this model demonstrates its capability to forecast PM2.5 concentrations for up to 72 h (3 days), achieving correlation coefficients exceeding 80%. However, after three days, the predictions returned to background levels. In the final stage of the study, data from automatic stations were integrated into a server hosting the assimilation model to improve daily PM2.5 forecasts for Dakar. An interactive platform was developed to visualize measurements and forecasts over two days. The results show that by integrating the data with the assimilation model, predictions are significantly improved.