Search

Air quality plays a vital role in people's health, and air quality forecasting can assist in decision making for government planning and sustainable development. In contrast, it is challenging to multi-step forecast accurately due to its complex and nonlinear caused by both temporal and spatial dimensions. Deep models, with their ability to model strong nonlinearities, have become the primary methods for air quality forecasting. However, because of the lack of mechanism-based analysis, uninterpretability forecasting makes decisions risky, especially when the government makes decisions. This paper proposes an interpretable variational Bayesian deep learning model with information self-screening for PM2.5 forecasting. Firstly, based on factors related to PM2.5 concentration, e.g., temperature, humidity, wind speed, spatial distribution, etc., an interpretable multivariate data screening structure for PM2.5 forecasting was established to catch as much helpful information as possible. Secondly, the self-screening layer was implanted in the deep learning network to optimize the selection of input variables. Further, following implantation of the screening layer, a variational Bayesian gated recurrent unit (GRU) network was constructed to overcome the complex distribution of PM2.5 and achieve accurate multi-step forecasting. The high accuracy of the proposed method is verified by PM2.5 data in Beijing, China, which provides an effective way, with multiple factors for PM2.5 forecasting determined using deep learning technology. [ABSTRACT FROM AUTHOR]

Lidar is a powerful active remote sensing technique to monitor the planetary boundary layer (PBL), and there have already been many algorithms to retrieve the planetary boundary layer height (PBLH) using aerosol lidars. However, the algorithm suitable for all atmospheric conditions doesn't exist. This study evaluates the capability of nine lidar algorithms under different pollution scenarios in terms of the instantaneous performance and the diurnal evolution, which are derived from classical theories and expressed as GM, LGM, CRGM, IPM, VAR, VGM, WCT, WCTD, and POLARIS in this paper. Based on the nine algorithms, the continuous PBLHs are obtained from the observation dataset of a Beijing urban site in 2017. In order to verify the lidar results, the statical comparison analysis with radiosonde is also presented. On the premise of the inversion hypothesis, such as sufficient mixing in the PBL, the consistency of lidar results was the best, with the difference being not more than 200 m. And what's different from our cognition is that the concentration of pollutants near the ground under the light pollution condition makes the lidar results tend to be consistent. In fact, LGM prefers to get higher PBLHs because of filtering the information near the ground, whereas POLARIS is apt to underestimate unless the actual morphology can be captured by the depolarization ratio. VAR and IPM are more vulnerable to the conspicuous stratification. GM, CRGM, and WCTD have better consistency with radiosonde and capture the evolution characteristic no matter in clean or polluted days. Compared with the clean days, PBLH from lidar algorithms under the polluted condition reduced about 500 m overall, as well as the standard deviation. Generally speaking, the comprehensive comparison provides a reference to choose the proper lidar algorithm when retrieving the PBLH. GM is appropriate for the detection of clean days, while CRGM is robust for polluted days. WCTD is more suitable for the judgment of PBLH in the morning and evening and will underestimate at noon. As to the situation with cloud, WCTD is also the best choice. • The nine algorithms are compared both internally and with radiosonde. • The difference of lidar results with sufficient mixing is not more than 200 m. • GM, WCT and WCTD are robust to meet the retrieval requirements. • WCTD is much better for the inversion of residual layer height at night. [ABSTRACT FROM AUTHOR]