Xavier Barthelemy

3. Results and Discussion This section presents the results from our evaluation of the BaySurcls model compared to the USEPA correction method (as baseline) and other standalone DL/ML models (as benchmark) for correcting hourly PAS2.5 data for both the collocation and noncollocation scenarios with the relevant AQMS (reference) data. The evaluation was based on several metrics, including root mean square error (RMSE), Nash–Sutcliffe Efficiency (NSE) and Pearson correlation (R).

Air pollution is becoming a pressing issue affecting human health, with 7.3 billion people worldwide facing air pollution levels considered unsafe, according to WHO statistics.1 In Australia, the government has strongly committed to a Nature Positive Plan designed to improve the national environment with a major target “to provide an authoritative source of high-quality environmental information.”2 Globally, ensuring the availability of reliable information for monitoring and forecasting air...