Accurately forecasting wind speed on a short-term scale has become essential in the field of wind power energy. In this paper, a multi-variable long short-term memory network model (MV-LSTM) based on Pearson correlation coefficient feature selection is proposed to predict the short-term wind speed. The proposed method utilizes multiple historical meteorological variables, such as wind speed, temperature, humidity, and air pressure, to predict the wind speed in the next hour. Hourly data collected from two ground observation stations in Yanqing and Zhaitang in Beijing were divided into training and test sets. The training sets were used to train the model, and the test sets were used to evaluate the model with the root-mean-square error (RMSE), mean absolute error (MAE), mean bias error (MBE), and mean absolute percentage error (MAPE) metrics. The proposed method is compared with two other forecasting methods (the autoregressive moving average model (ARMA) method and the single-variable long short-term memory network (LSTM) method, which inputs only historical wind speed data) based on the same dataset. The experimental results prove the feasibility of the MV-LSTM method for short-term wind speed forecasting and its superiority to the ARMA method and the single-variable LSTM method.

Beijing experiences frequent PM_2.5 pollution, which is influenced by the planetary boundary layer (PBL) structure/process. Partly due to a lack of appropriate observations, the impacts of PBL on PM_2.5 pollution are not yet fully understood. Combining wind-profiler data, radiosonde measurements, near-surface meteorological observations, aerosol measurements, and three-dimensional simulations, this study investigated the influence of PBL structure and the low-level jet (LLJ) on the pollution in Beijing from 19 to 20 September 2015. The evolution of the LLJ was generally well simulated by the model, although the wind speed within the PBL was overestimated. Being influenced by the large-scale southerly prevailing winds, the aerosols emitted from the southern polluted regions could be easily transported to Beijing, contributing to ~68% of the PM_2.5 measured in Beijing on 20 September. The relative contribution of external transport of PM_2.5 to Beijing was high in the afternoon (≥80%), which was related to the strong southerly PBL winds and the presence of thermally-induced upslope winds. On 20 September, the LLJ in Beijing demonstrated a prominent diurnal variation, which was predominant in the morning and after sunset. The occurrence of the LLJ could enhance the dilution capacity in Beijing to some extent, which favors the dilution of pollutants at a local scale. This study has important implications for better understanding the complexity of PBL structure/process associated with PM_2.5 pollution in Beijing.

In this paper, L-band sounding and surface observation data are used to calculate the boundary layer height (BLH) and evaluated CMA (China Metrological Administration Numerical Forecast System) and ERA5 in Xi’an for 2017–2021 using the Richardson (Ri) and Nozaki methods. For different PM_2.5 pollution levels, the correlation between the vertical profile of meteorological factors and BLH is explored. There is a certain negative correlation between BLH and PM_2.5 concentration. The BLH mean values of Nozaki, Ri, ERA5, and CMA from high to low are ~980 m, ~640 m, ~410 m, and ~240 m, respectively. The highest correlation is between ERA5 and CMA BLH with r² > 0.85 for all pollution processes, while it between other methods is significantly lower (r² < 0.58). The observational BLH is generally higher than the model results. Nozaki has a good adaptability on the light pollution, while Ri is more applicable to the stable boundary layer. In moderate and higher pollution, the ERA5 has a slightly better performance than CMA in BLH, while in light pollution there is a significant underestimation for both. Overall, the correlation between any two BLH methods gradually increases with increasing pollution level. In this study, there is about ~30% probability of polluted weather when BLH < 200 m and only 2000 m. It is difficult to simulate the neutral boundary layer and inversion processes for CMA and ERA5, but ERA5 has higher forecasting skills than CMA. This study can provide the data and theoretical support for the development of haze numerical forecast.

The coronavirus disease 2019 (COVID-19) is spreading rapidly all over the world. The transmission dynamics of the COVID-19 pandemic is still unclear, but developing strategies for mitigating the severity of the pandemic is yet a top priority for global public health. In this study, we developed a novel compartmental model, SEIR-CV(susceptible-exposed-infectious-removed with control variables), which not only considers the key characteristics of asymptomatic infection and the effects of seasonal variations, but also incorporates different control measures for multiple transmission routes, so as to accurately predict and effectively control the spread of COVID-19. Based on SEIR-CV, we predicted the COVID-19 epidemic situation in China out of Hubei province and proposed corresponding control strategies. The results showed that the prediction results are highly consistent with the outbreak surveillance data, which proved that the proposed control strategies have achieved sound consequent in the actual epidemic control. Subsequently, we have conducted a rolling prediction for the United States, Brazil, India, five European countries (the United Kingdom, Italy, Spain, Germany, and France), southern hemisphere, northern hemisphere, and the world out of China. The results indicate that control measures and seasonal variations have a great impact on the progress of the COVID-19 pandemic. Our prediction results show that the COVID-19 pandemic is developing more rapidly due to the impact of the cold season in the southern hemisphere countries such as Brazil. While the development of the pandemic should have gradually weakened in the northern hemisphere countries with the arrival of the warm season, instead of still developing rapidly due to the relative loose control measures such as the United States and India. Furthermore, the prediction results illustrate that if keeping the current control measures in the main COVID-19 epidemic countries, the pandemic will not be contained and the situation may eventually turn to group immunization, which would lead to the extremely severe disaster of about 5 billion infections and 300 million deaths globally. However, if China’s super stringent control measures were implemented from 15 July, 15 August or 15 September 2020, the total infections would be contained about 15 million, 32 million or 370 million respectively, which indicates that the stringent and timely control measures is critical, and the best window period is before September for eventually overcoming COVID-19. Significance: COVID-19 is now posing a huge threat to global public health. The key features such as asymptomatic infection and droplet or airborne transmission make COVID-19 more easily spread and more widely distributed around the world. It is an urgent need to explore the optimal intervention strategies and measures to contain the pandemic. Our novel SEIR-CV compartmental model considers the new features of COVID-19, exhibits the impact of the intervention strategies and seasonal variations, and thus can accurately predicts its trajectory in China and the rest of the world. Our research results suggest that control measures and seasonal variations have a great impact on the development of the COVID-19 pandemic, which can only be contained by stringent strategies during the best window period before September 2020 for eventually overcoming COVID-19, otherwise it would cause a severer global catastrophe.