Reconstructing global PM_2.5 monitoring dataset from OpenAQ using a two-step spatio-temporal model based on SES-IDW and LSTM

Abstract Fine particulate matter (PM 2.5 ) is widely concerned for its harmful impacts on global environment and human health, making air pollution monitoring so crucial indispensable. As the world’s first open, real-time, historical quality platform, OpenAQ collects provides government measurement research-level data from various channels. However, despite OpenAQ’s innovation in providing us with ground-measured PM worldwide, we find significant gaps time series most of sites. The incompleteness directly affects public perception concentration levels hinders progress research related to pollution. To address these issues, a two-step hybrid model named ST-SILM, i.e. spatio-temporal single exponential smoothing-inverse distance weighted (SES-IDW) long short-term memory (LSTM), proposed repair missing sites worldwide collected 2017 2019. Both correlation neighborhood fields are considered established model. be specific, SES-IDW were firstly used values, secondly, LSTM network was employed reconstruct continuous data. After reconstructed, light gradient boosting machine applied remote sensing estimation original reconstructed further verify performance ST-SILM. Experiment results show that accuracy dataset better ( R 2 2019 increased by 0.02, 0.01 compared dataset). Therefore, it concluded can effectively worldwide.