ESTIMATION OF NO2 CONCENTRATION VALUES IN A MONITORING SENSOR NETWORK USING A FUSION APPROACH

This study is focused on calculation of a reliable estimation of the hourly concentration value of NO2 at a monitoring station based on a data fusion approach. Different feature selection procedures have been tested and their results were used as inputs to an artificial neural network (ANN) two-stage approach. The final aim is to develop a data fusion estimation tool in order to aid the decision-making in the monitoring process. ANN models were trained using backpropagation and early stopping in order to avoid overfitting. Furthermore, the study compares the different combinations of methods to estimate hourly NO2 concentration values. The comparison was made using different performance indexes (R, d, MAE, MSE). The data from the Algeciras Bay was used as a case study. This approach may become a supporting tool for different purposes such as missing data imputation, automatic detection of decalibration or imprecise data in monitoring networks.