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.

This study demonstrates how model bias can adversely affect the quality assessment of an ensemble prediction system (EPS) by verification metrics. A regional EPS [Global and Regional Assimilation and Prediction Enhanced System-Regional Ensemble Prediction System (GRAPES-REPS)] was verified over a period of one month over China. Three variables (500-hPa and 2-m temperatures, and 250-hPa wind) are selected to represent “strong” and “weak” bias situations. Ensemble spread and probabilistic forecasts are compared before and after a bias correction. The results show that the conclusions drawn from ensemble verification about the EPS are dramatically different with or without model bias. This is true for both ensemble spread and probabilistic forecasts. The GRAPES-REPS is severely underdispersive before the bias correction but becomes calibrated afterward, although the improvement in the spread’s spatial structure is much less; the spread–skill relation is also improved. The probabilities become much sharper and almost perfectly reliable after the bias is removed. Therefore, it is necessary to remove forecast biases before an EPS can be accurately evaluated since an EPS deals only with random error but not systematic error. Only when an EPS has no or little forecast bias, can ensemble verification metrics reliably reveal the true quality of an EPS without removing forecast bias first. An implication is that EPS developers should not be expected to introduce methods to dramatically increase ensemble spread (either by perturbation method or statistical calibration) to achieve reliability. Instead, the preferred solution is to reduce model bias through prediction system developments and to focus on the quality of spread (not the quantity of spread). Forecast products should also be produced from the debiased but not the raw ensemble.

This study experimented with a unified scheme of stochastic physics and bias correction within a regional ensemble model [Global and Regional Assimilation and Prediction System–Regional Ensemble Prediction System (GRAPES-REPS)]. It is intended to improve ensemble prediction skill by reducing both random and systematic errors at the same time. Three experiments were performed on top of GRAPES-REPS. The first experiment adds only the stochastic physics. The second experiment adds only the bias correction scheme. The third experiment adds both the stochastic physics and bias correction. The experimental period is one month from 1 to 31 July 2015 over the China domain. Using 850-hPa temperature as an example, the study reveals the following: 1) the stochastic physics can effectively increase the ensemble spread, while the bias correction cannot. Therefore, ensemble averaging of the stochastic physics runs can reduce more random error than the bias correction runs. 2) Bias correction can significantly reduce systematic error, while the stochastic physics cannot. As a result, the bias correction greatly improved the quality of ensemble mean forecasts but the stochastic physics did not. 3) The unified scheme can greatly reduce both random and systematic errors at the same time and performed the best of the three experiments. These results were further confirmed by verification of the ensemble mean, spread, and probabilistic forecasts of many other atmospheric fields for both upper air and the surface, including precipitation. Based on this study, we recommend that operational numerical weather prediction centers adopt this unified scheme approach in ensemble models to achieve the best forecasts.

Accurate wind speed prediction plays a crucial role in wind power generation and disaster avoidance. However, stochasticity and instability increase the difficulty of wind speed prediction. In this study, we proposed a dual-attention mechanism multi-channel convolutional LSTM (DACLSTM), collected European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis v5 (ERA5) near-ground element-grid data from some parts of North China, and selected elements with high correlations with wind speed to form multiple channels. We used a convolutional network for the feature extraction of spatial information, a Long Short-Term Memory (LSTM) network for the feature extraction of time-series information, and used channel attention with spatial attention for feature extraction. The experimental results show that the DACLSTM model can improve the accuracy of six-hour lead time wind speed prediction relative to the traditional ConvLSTM model and fully connected network long short-term memory (FC_LSTM).

With mesoscale data from a high-resolution numerical model, this paper analyzes the impact of the complex mesoscale terrain at the eastern edge of the Tibetan Plateau on the formation and development of a local severe storm. The result shows that the dynamic blocking effect of the Tibetan Plateau terrain affects the moisture transfer channel in this heavy rain scenario and brings about the formation of a strong water-vapor convergence center in northwest Sichuan and the upper branch of Huang He (Yellow River). The terrain structure of the east side of the Tibetan Plateau places a high energy tongue and energy front in the lower layer of the troposphere of Sichuan Basin and helps establish unstable potential energy stratification in the lower troposphere of the Sichuan Basin. The accumulation of water vapor in the lower troposphere in the Sichuan Basin is the major reason for the establishment of unstable potential energy stratification. The steep terrain of the east side of Tibetan Plateau streng...