 摘要： Using multi-source meteorological observation data, the causes and forecast points of the first round of continuous heavy precipitation process during the starting period of rainy season from May 24 to June 3, 2025 in Dehong Prefecture were analyzed. The results showed that: ① this heavy precipitation was formed by the influence of multiple systems such as convergence shear line, Bay of Bengal storm, monsoon trough, and low-level jet stream. The action time of  each system did not overlap, and the precipitation was mainly stable, with significant differences in daily falling area and magnitude. ② According to the development stage of the Bay of Bengal storm, the process can be divided into three stages: convergence and shear of the South China Sea monsoon, southwest airflow outside the low-pressure system, and dominant southwest airflow in front of the monsoon trough. The strongest rainfall occurred on June 2. Combined with terrain uplift, the establishment of 850 hPa of southwest low-level jet stream provides key driving force and water vapor conditions for heavy precipitation. Its establishment and strengthening were important indicators for heavy precipitation forecasting in Dehong Prefecture. ③ The first round of heavy precipitation in Dehong Prefecture in early summer was closely related to the starting period of rainy season, and the forecast deviation was mainly due to insufficient prediction about the weakening speed of low-pressure system and the amplification effect of low-level jet stream. This paper can provide technical support for forecasting similar heavy precipitation in low-latitude plateaus.

 摘要： The climate suitability for tourism is an important natural condition that affects the development of tourism activities, the attractiveness of tourism destinations, and the level of refinement of tourism meteorological services. To explore the applicability of different tourism climate evaluation models in Hebei Province, three models were selected in this paper: tourism climate index (TCI), holiday climate index (HCI), and modified holiday climate index (MHCI) based on the tourism climate evaluation results of various regions in Hebei Province from 1981 to 2020. Combined with county-level administrative division data, spatial mapping, descriptive statistics, one-way ANOVA, and paired t-test were used to comprehensively analyze the temporal variation, spatial differentiation, and model differences of climate suitability for tourism in Hebei Province. The results indicate that all three models can reflect the seasonal changes in the suitability of tourism climate in Hebei Province, but there are significant differences in the evaluation range and regional identification ability. The annual average of Hebei Province is as follows: MHCI (73.62)>HCI (71.26)>TCI (65.04). Seasonally, the summer mean of HCI is the highest (81.39), indicating a certain overestimation of high temperature and high humidity weather. MHCI remains at a high level in spring, summer, and autumn, and can significantly correct for the effects of hot and humid summers and cold winters. TCI is generally low, especially with a winter average of only 51.64. The analysis of variance shows that there are significant differences in the evaluation results of the three models throughout the year and four seasons (P<0.001). In terms of space, the overall suitability of tourism climate in Hebei Province presents a pattern of significant differences among mountainous areas, coastal areas, and Bashang regions. Chengde has the highest MHCI in summer, Qinhuangdao performs outstandingly in TCI evaluation, and the suitability for spring and autumn in plain of central and southern Hebei is relatively high. After comprehensive comparison, it is believed that the MHCI model can better reflect the characteristics of monsoon climate, high temperature and stuffiness, cold winter, and terrain differentiation in Hebei Province, and is suitable as the preferred model for evaluating tourism climate resources and applying tourism meteorological services in Hebei Province.

 摘要： Analog Ensemble (AnEn) is a statistical interpretation method based on historical similarity, big data sample retrieval, and ensemble forecasting ideas. It can form deterministic correction results based on numerical model outputs and further provide probability ranges and uncertainty information. In this paper, the surface meteorological elements in Hebei Province were taken as the research objects. Focusing on the forecasting of 2 m temperature, 10 m wind speed, total cloud cover, and low cloud cover, with comprehensive reference to relevant AnEn research results and phased experimental data of AnEn in Hebei Province, a similar sample screening and ensemble weighting scheme suitable for the complex terrain background of Hebei Province was constructed. The results showed that AnEn had a relatively stable correction effect on 2 m temperature forecasts and can significantly reduce the systematic bias of the model. For a wind speed of 10 m, the average effect of the entire region was not yet better than EC direct forecasting, but it had strong potential for improvement in complex terrain, non urban stations, and areas with large local errors. In cloud cover forecasting, the effect of low cloud cover was better than that of total cloud cover, and the accuracy of forecasting slightly decreased with the extension of time efficiency. The sample size experiment showed that when the number of members in the ensemble increased from 5 to 10, the MAE and RMSE overall decreased, and the ranking weight method was superior to the median method. Research has shown that AnEn can serve as an important objective interpretation tool for short- and medium-term surface element forecasting in Hebei region.

 摘要： In response to practical issues such as equipment dependence, personnel scheduling risks, and disaster warning timeliness in grassroots meteorological program production, a meteorological voice broadcast auxiliary process of online and offline dual model switching is proposed in this paper. The process integrates three core functions: message parsing, model switching, and speech generation. At the same time, it is connected to the cloud volcano engine speech replication large model and the local Index-TS 2.0 open source model. Operators can manually switch as needed. The system is equipped with an updatable mandatory dictionary of high-frequency meteorological words to improve the accuracy of pronunciation of terms and place names. Subsequent tests have shown that the system can serve as a backup solution to avoid downtime in the event of equipment failure or broadcaster absence, compress the production cycle of sudden warnings to the minute level, and support mass production of multiple versions of programs. This paper provides a solution for emergency voice program synthesis for meteorological new media by complementing cloud high precision and local high reliability.

 摘要： Lightning activity is characterized by abrupt occurrence, strong spatial heterogeneity, and highly variable intensity, and traditional empirical methods are difficult to meet the needs of refined lightning warning. In this paper, the Ulanqab region was selected as the study area.VLF/LF lightning location data and ERA5 reanalysis data collected during the summers of 2020-2022 were used, and a lightning intensity level prediction model was constructed. Based on the environmental conditions of lightning occurrence, nine key predictors were selected from 25 environmental factors, including chimney index, total totals index, K index, 0-6 km of vertical wind shear, deep convection index, uplift condensation height temperature, dew point temperature, 700 hPa of pseudo equivalent temperature, and precipitable water vapor. Four machine learning methods, namely gradient boosting, random forest, neural network, and logistic regression, were used to establish a five-level lightning intensity classification model for weak, moderate, strong, very strong, and extreme lightning. The results showed that the gradient boosting model had the highest comprehensive score of 0.173 7, and performed the best in terms of average CSI, overall ACC, and comprehensive prediction ability. The DeLong test results indicated that the overall performance of gradient boosting was significantly better than other models (P<0.05). Using the typical lightning process during August 18-20, 2024, a 72-hour historical simulation and EC business prediction verification were conducted. The results showed that the model can better reflect the changes in lightning activity intensity and spatial evolution patterns, and has the business forecasting application capability.