While the influence of basin‐wide sea surface temperature (SST) anomalies in the Indian Ocean on tropical cyclones (TCs) in the Western North Pacific (WNP) has been extensively studied, the impact of the Indian Ocean Dipole (IOD) mode on WNP TC activity remains less well understood. In this study, we use best‐track and reanalysis data to investigate the effects of the IOD on WNP TC activity during the peak typhoon season (July–October) over the period 1965–2023. Given the strong interconnection between the El Niño–Southern Oscillation (ENSO) and the IOD, we remove the influence of ENSO through linear regression to isolate the independent impact of the IOD. Results show that during the positive phase of the IOD, TC genesis is associated with an increase over the South China Sea but suppressed in the ocean to the east of the Philippines. This spatial shift in TC genesis, along with an anomalous cyclonic steering flow over the western WNP, leads to increased TC activity influencing northern China. The suppressed TC genesis east of the Philippines is largely driven by enhanced vertical wind shear, resulting from strengthened low‐level westerlies and upper‐level easterlies associated with suppressed convective heating over the eastern pole of the IOD. Meanwhile, the positive IOD phase enhances the WNP monsoon and increases low‐level vorticity over the South China Sea, favoring TC development in this region. These findings suggest that the IOD can modulate spatial distribution and regional characteristics of TC activity over the WNP, offering important insights into the climatic variability of WNP TCs.
The estimation of height of the planetary boundary layer (hPBL) is essential for weather prediction, air quality assessment, and atmospheric research, particularly in complex terrain where boundary-layer dynamics strongly influence temperature structure, cloud development, and convective processes. In data-scarce regions such as the tropical Andes, routine hPBL monitoring is often constrained by the high operational costs and logistical limitations of radiosonde observations. Low-altitude drones offer a flexible and cost-effective alternative for sampling lower atmospheric conditions, although their restricted vertical range necessitates complementary inference approaches. This study presents a methodological proof-of-concept through a site-specific case study in Quito, Ecuador, exploring the feasibility of combining drone-derived meteorological profiles (0–300 m above ground level) with machine learning (ML) techniques to estimate hPBL. Four models were evaluated: neural networks (NNs), random forest (RF), support vector regression (SVR), and multiple linear regression (MLR). Among these, RF and MLR achieved the strongest performance when trained using the most relevant predictors identified through RF-based feature importance analysis. Model performance was generally higher under unstable atmospheric conditions, characterized by elevated convective available potential energy (CAPE; ≈1400 J kg−1) and deeper hPBL (≈2700 m agl), and lower under more stable conditions (CAPE, ≈300 J kg−1; hPBL, ≈2200 m agl). Surface wind speed (WS), mixing ratio (MR), and the interaction between MR and temperature emerged as the most influential predictors, offering preliminary insight into the physical drivers of boundary-layer development in this environment. While the findings are constrained by limited sample size, vertical extent, and geographic specificity, this exploratory framework demonstrates the potential of integrating unmanned aerial vehicle (UAV) observations with artificial intelligence (AI)-based methods as a low-cost strategy for hPBL estimation in observationally limited, topographically complex regions, and provides a foundation for future validation and broader application.
Since the State Council of China launched the “National Air Quality Action Plan” in 2013, the air quality in China has improved markedly in recent years. Disentangling the respective influences of meteorological conditions and emission reduction on the evolution of air pollution, along with evaluating the effectiveness of pollution reduction measures, is of great significance for formulating more scientific and targeted prevention and control policies. In this study, based on daily average concentrations of PM2.5 and PM10 and NO2 and the maximum daily 8 h average ozone (O3) concentration in Nan’an City of China from 2018 to 2022, together with meteorological factors and ERA5 reanalysis data, we quantify the respective contributions of meteorological conditions and emission reduction to the observed trends in these pollutants. The results show that during 2018–2022, the annual decline rates of PM2.5, PM10, and O3 are 2.9, 6.0, and 2.4 μg m−3 yr−1 in Nan’an City, respectively. Seasonally, the most pronounced decline of all pollutants is in autumn and spring. PM2.5 exceedances are mainly observed in winter, whereas O3 exceedances primarily occur in spring and summer. Meteorological normalization results show that the annual decline rates of PM2.5, PM10, and O3 in Nan’an City from 2018 to 2022 are 1.9, 3.7 and 1.0 μg m−3 yr−1, respectively. After the pollution-reduction turning point triggered by the COVID-19 outbreak in 2020, the concentration variations induced by emission-source changes were limited. Meteorological conditions contributed 55%, 57% and 77% to the daily-scale variations of PM2.5, PM10, and O3, all higher than the contributions from anthropogenic-emission-related temporal variables. Ozone exhibits the strongest dependence on meteorological factors including ultraviolet radiation, relative humidity and temperature, which amplifies its interannual fluctuations. Overall, favorable meteorological conditions facilitated regional air-quality improvement, yet driving mechanisms differ markedly across pollutants. The long-term declines of PM2.5 and PM10 are predominantly governed by anthropogenic emission control, with corresponding control contributions of 66.0% and 61.2%. In contrast, the control contribution for long-term O3 variation is only 42.5%. Ozone is comodulated by anthropogenic emissions and meteorology, whereby meteorology exerts a more prominent influence on its short-term fluctuations. Driven by the substantial drop in NO2 concentrations, the local O3/NO2 ratio increased and the O3-NO2 correlation coefficient reversed from negative to positive, indicating a potential shift of ozone-formation regime from VOC-limited toward NOx-limited conditions. These findings can provide references for coordinated ozone pollution control in cities of southeast coastal China.
Accurate shortwave radiation (SWR) forecasting is essential for meteorological modeling, climate analysis, renewable-energy assessment, and environmental decision support, yet it remains challenging due to strong nonlinearity, multivariate dependencies, and pronounced temporal variability. Although recent Transformer-based models have shown promising performance in long-horizon forecasting, many existing approaches remain limited in adaptively modeling variable importance, oscillatory patterns, and nonlinear temporal dynamics. To address these limitations, this paper proposes a novel CAIT model, which is a coordinate-attention–enhanced inverted Transformer framework for dealing with multivariate SWR forecasting. The proposed architecture adopts a variable-centric representation to explicitly model intervariable dependencies among meteorological observations. A coordinate-attention mechanism adaptively recalibrates informative variable-token and feature representations, while a SIREN-based sinusoidal transformation enhances the modeling of periodic and oscillatory behavior. In addition, a KAN-based nonlinear forecasting head improves the mapping of learned representations to future radiation values. Extensive experiments on satellite-derived regional datasets demonstrate that our proposed CAIT model consistently outperforms representative deep learning (DL) and Transformer-based baselines. The cross-period and horizon-specific evaluations further confirm stable generalization across extended forecasting horizons. Comprehensive ablation studies verify the complementary contributions and robustness of the proposed architectural components. Overall, the results demonstrate that our introduced CAIT model provides an effective framework for robust long-horizon SWR forecasting under heterogeneous meteorological conditions.
As a major economic and population center in northern China, the Beijing–Tianjin–Hebei region experiences winter climate influenced by both natural variability and urbanization. Using monthly meteorological observations from 2000 to 2024 at the Beijing, Tianjin, and Shijiazhuang stations, this study focuses on the conventional heating season (November to March of the following year). It systematically analyzes the long-term evolution and regional differences in mean air temperature, relative humidity, rainfall, number of snowfall days (SD), number of snow cover days (SCD), and maximum snow cover depth (MSCD). The results show that from 2000 to 2024: (1) Regional mean temperature during the heating season showed an increasing tendency in the later period, with the most pronounced warming in February (+1.2°C averaged across the three cities), while relative humidity showed a general decrease of 4%–6%; (2) The number of SD decreased (by 15%–37%), whereas MSCD showed a slight, non-significant increase in the later period at some stations, suggesting a possible tendency towards fewer but more intense snowfall events; (3) Rainfall during the heating season showed no significant trend but exhibited increased interannual variability. A comparison among the three cities reveals that Tianjin shows the most pronounced decrease in the number of SD, while Shijiazhuang exhibits the largest fluctuations in MSCD. By providing an observational description of historical changes in meteorological indicators during the heating season, this study offers preliminary data and generates testable hypotheses for understanding regional climate change, assessing urbanization impacts, and developing sustainable urban winter management strategies.
