Advances in Atmospheric Sciences最新文献

Recent studies have shown that deep learning (DL) models can skillfully forecast El Niño–Southern Oscillation (ENSO) events more than 1.5 years in advance. However, concerns regarding the reliability of predictions made by DL methods persist, including potential overfitting issues and lack of interpretability. Here, we propose ResoNet, a DL model that combines CNN (convolutional neural network) and transformer architectures. This hybrid architecture enables our model to adequately capture local sea surface temperature anomalies as well as long-range inter-basin interactions across oceans. We show that ResoNet can robustly predict ENSO at lead times of 19 months, thus outperforming existing approaches in terms of the forecast horizon. According to an explainability method applied to ResoNet predictions of El Niño and La Niña from 1- to 18-month leads, we find that it predicts the Niño-3.4 index based on multiple physically reasonable mechanisms, such as the recharge oscillator concept, seasonal footprint mechanism, and Indian Ocean capacitor effect. Moreover, we demonstrate for the first time that the asymmetry between El Niño and La Niña development can be captured by ResoNet. Our results could help to alleviate skepticism about applying DL models for ENSO prediction and encourage more attempts to discover and predict climate phenomena using AI methods.

The scarcity of in-situ ocean observations poses a challenge for real-time information acquisition in the ocean. Among the crucial hydroacoustic environmental parameters, ocean sound velocity exhibits significant spatial and temporal variability and it is highly relevant to oceanic research. In this study, we propose a new data-driven approach, leveraging deep learning techniques, for the prediction of sound velocity fields (SVFs). Our novel spatiotemporal prediction model, ST-LSTM-SA, combines Spatiotemporal Long Short-Term Memory (ST-LSTM) with a self-attention mechanism to enable accurate and real-time prediction of SVFs. To circumvent the limited amount of observational data, we employ transfer learning by first training the model using reanalysis datasets, followed by fine-tuning it using in-situ analysis data to obtain the final prediction model. By utilizing the historical 12-month SVFs as input, our model predicts the SVFs for the subsequent three months. We compare the performance of five models: Artificial Neural Networks (ANN), Long Short-Term Memory (LSTM), Convolutional LSTM (ConvLSTM), ST-LSTM, and our proposed ST-LSTM-SA model in a test experiment spanning 2019 to 2022. Our results demonstrate that the ST-LSTM-SA model significantly improves the prediction accuracy and stability of sound velocity in both temporal and spatial dimensions. The ST-LSTM-SA model not only accurately predicts the ocean sound velocity field (SVF), but also provides valuable insights for spatiotemporal prediction of other oceanic environmental variables.

This study evaluated the simulation performance of mesoscale convective system (MCS)-induced precipitation, focusing on three selected cases that originated from the Yellow Sea and propagated toward the Korean Peninsula. The evaluation was conducted for the European Centre for Medium-Range Weather Forecasts (ECMWF) and National Centers for Environmental Prediction (NCEP) analysis data, as well as the simulation result using them as initial and lateral boundary conditions for the Weather Research and Forecasting model. Particularly, temperature and humidity profiles from 3D dropsonde observations from the National Center for Meteorological Science of the Korea Meteorological Administration served as validation data. Results showed that the ECMWF analysis consistently had smaller errors compared to the NCEP analysis, which exhibited a cold and dry bias in the lower levels below 850 hPa. The model, in terms of the precipitation simulations, particularly for high-intensity precipitation over the Yellow Sea, demonstrated higher accuracy when applying ECMWF analysis data as the initial condition. This advantage also positively influenced the simulation of rainfall events on the Korean Peninsula by reasonably inducing convective-favorable thermodynamic features (i.e., warm and humid lower-level atmosphere) over the Yellow Sea. In conclusion, this study provides specific information about two global analysis datasets and their impacts on MCS-induced heavy rainfall simulation by employing dropsonde observation data. Furthermore, it suggests the need to enhance the initial field for MCS-induced heavy rainfall simulation and the applicability of assimilating dropsonde data for this purpose in the future.

The 2015/16 El Niño event ranks among the top three of the last 100 years in terms of intensity, but most dynamical models had a relatively low prediction skill for this event before the summer months. Therefore, the attribution of this particular event can help us to understand the cause of super El Niño–Southern Oscillation events and how to forecast them skillfully. The present study applies attribute methods based on a deep learning model to study the key factors related to the formation of this event. A deep learning model is trained using historical simulations from 21 CMIP6 models to predict the Niño-3.4 index. The integrated gradient method is then used to identify the key signals in the North Pacific that determine the evolution of the Niño-3.4 index. These crucial signals are then masked in the initial conditions to verify their roles in the prediction. In addition to confirming the key signals inducing the super El Niño event revealed in previous attribution studies, we identify the combined contribution of the tropical North Atlantic and the South Pacific oceans to the evolution and intensity of this event, emphasizing the crucial role of the interactions among them and the North Pacific. This approach is also applied to other El Niño events, revealing several new precursor signals. This study suggests that the deep learning method is useful in attributing the key factors inducing extreme tropical climate events.

Heavy precipitation and extreme drought have caused severe economic losses over South China and Indochina (INCSC) in recent decades. Given the areas with large gross domestic product (GDP) in the INCSC region are distributed along the coastline and greatly affected by global warming, understanding the possible economic impacts induced by future changes in the maximum consecutive 5-day precipitation (RX5day) and the maximum consecutive dry days (CDD) is critical for adaptation planning in this region. Based on the latest data released by phase 6 of the Coupled Model Intercomparison Project (CMIP6), future projections of precipitation extremes with bias correction and their impacts on GDP over the INCSC region under the fossil-fueled development Shared Socioeconomic Pathway (SSP5-8.5) are investigated. Results indicate that RX5day will intensify robustly throughout the INCSC region, while CDD will lengthen in most regions under global warming. The changes in climate consistently dominate the effect on GDP over the INCSC region, rather than the change of GDP. If only considering the effect of climate change on GDP, the changes in precipitation extremes bring a larger impact on the economy in the future to the provinces of Hunan, Jiangxi, Fujian, Guangdong, and Hainan in South China, as well as the Malay Peninsula and southern Cambodia in Indochina. Thus, timely regional adaptation strategies are urgent for these regions. Moreover, from the sub-regional average viewpoint, over two thirds of CMIP6 models agree that maintaining a lower global warming level will reduce the economic impacts from heavy precipitation over the INCSC region.

Temperature extremes over rapidly urbanizing regions with high population densities have been scrutinized due to their severe impacts on human safety and economics. First of all, the performance of the regional climate model RegCM4 with a hydrostatic or non-hydrostatic dynamic core in simulating seasonal temperature and temperature extremes was evaluated over the historical period of 1991–99 at a 12-km spatial resolution over China and a 3-km resolution over the Beijing–Tianjin–Hebei (JJJ) region, a typical urban agglomeration of China. Simulations of spatial distributions of temperature extremes over the JJJ region using RegCM4 with hydrostatic and non-hydrostatic cores showed high spatial correlations of more than 0.8 with the observations. Under a warming climate, temperature extremes of annual maximum daily temperature (TXx) and summer days (SU) in China and the JJJ region showed obvious increases by the end of the 21st century while there was a general reduction in frost days (FD). The ensemble of RegCM4 with different land surface components was used to examine population exposure to temperature extremes over the JJJ region. Population exposure to temperature extremes was found to decrease in 2091–99 relative to 1991–99 over the majority of the JJJ region due to the joint impacts of increases in temperature extremes over the JJJ and population decreases over the JJJ region, except for downtown areas. Furthermore, changes in population exposure to temperature extremes were mainly dominated by future population changes. Finally, we quantified changes in exposure to temperature extremes with temperature increase over the JJJ region. This study helps to provide relevant policies to respond future climate risks over the JJJ region.

Numerical weather prediction (NWP) models have always presented large forecasting errors of surface wind speeds over regions with complex terrain. In this study, surface wind forecasts from an operational NWP model, the SMS-WARR (Shanghai Meteorological Service-WRF ADAS Rapid Refresh System), are analyzed to quantitatively reveal the relationships between the forecasted surface wind speed errors and terrain features, with the intent of providing clues to better apply the NWP model to complex terrain regions. The terrain features are described by three parameters: the standard deviation of the model grid-scale orography, terrain height error of the model, and slope angle. The results show that the forecast bias has a unimodal distribution with a change in the standard deviation of orography. The minimum ME (the mean value of bias) is 1.2 m s⁻¹ when the standard deviation is between 60 and 70 m. A positive correlation exists between bias and terrain height error, with the ME increasing by 10%–30% for every 200 m increase in terrain height error. The ME decreases by 65.6% when slope angle increases from (0.5°–1.5°) to larger than 3.5° for uphill winds but increases by 35.4% when the absolute value of slope angle increases from (0.5°–1.5°) to (2.5°–3.5°) for downhill winds. Several sensitivity experiments are carried out with a model output statistical (MOS) calibration model for surface wind speeds and ME (RMSE) has been reduced by 90% (30%) by introducing terrain parameters, demonstrating the value of this study.

Based on the Regional Specialized Meteorological Center (RSMC) Tokyo-Typhoon Center best-track data and the NCEP-NCAR reanalysis dataset, extratropical transitioning (ET) tropical cyclones (ETCs) over the western North Pacific (WNP) during 1951–2021 are classified into six clusters using the fuzzy c-means clustering method (FCM) according to their track patterns. The characteristics of the six hard-clustered ETCs with the highest membership coefficient are shown. Most tropical cyclones (TCs) that were assigned to clusters C2, C5, and C6 made landfall over eastern Asian countries, which severely threatened these regions. Among landfalling TCs, 93.2% completed their ET after landfall, whereas 39.8% of ETCs completed their transition within one day. The frequency of ETCs over the WNP has decreased in the past four decades, wherein cluster C5 demonstrated a significant decrease on both interannual and interdecadal timescales with the expansion and intensification of the western Pacific subtropical high (WPSH). This large-scale circulation pattern is favorable for C2 and causes it to become the dominant track pattern, owning to it containing the largest number of intensifying ETCs among the six clusters, a number that has increased insignificantly over the past four decades. The surface roughness variation and three-dimensional background circulation led to C5 containing the maximum number of landfalling TCs and a minimum number of intensifying ETCs. Our results will facilitate a better understanding of the spatiotemporal distributions of ET events and associated environment background fields, which will benefit the effective monitoring of these events over the WNP.

Sea salt aerosols play a critical role in regulating the global climate through their interactions with solar radiation. The size distribution of these particles is crucial in determining their bulk optical properties. In this study, we analyzed in situ measured size distributions of sea salt aerosols from four field campaigns and used multi-mode lognormal size distributions to fit the data. We employed super-spheroids and coated super-spheroids to account for the particles’ non-spherictty, inhomogeneity, and hysteresis effect during the deliquescence and crystallization processes. To compute the single-scattering properties of sea salt aerosols, we used the state-of-the-art invariant imbedding T-matrix method, which allows us to obtain accurate optical properties for sea salt aerosols with a maximum volume-equivalent diameter of 12 µm at a wavelength of 532 nm. Our results demonstrated that the particle models developed in this study were successful in replicating both the measured depolarization and lidar ratios at various relative humidity (RH) levels. Importantly, we observed that large-size particles with diameters larger than 4 µm had a substantial impact on the optical properties of sea salt aerosols, which has not been accounted for in previous studies. Specifically, excluding particles with diameters larger than 4 µm led to underestimating the scattering and backscattering coefficients by 27%–38% and 43%–60%, respectively, for the ACE-Asia field campaign. Additionally, the depolarization ratios were underestimated by 0.15 within the 50%–70% RH range. These findings emphasize the necessity of considering large particle sizes for optical modeling of sea salt aerosols.

Ground-based microwave radiometers (MWRs) operating in the K- and V-bands (20–60 GHz) can help us obtain temperature and humidity profiles in the troposphere. Aside from some soundings from local meteorological observatories, the tropospheric atmosphere over the Tibetan Plateau (TP) has never been continuously observed. As part of the Chinese Second Tibetan Plateau Scientific Expedition and Research Program (STEP), the Tibetan Plateau Atmospheric Profile (TP-PROFILE) project aims to construct a comprehensive MWR troposphere observation network to study the synoptic processes and environmental changes on the TP. This initiative has collected three years of data from the MWR network. This paper introduces the data information, the data quality, and data downloading. Some applications of the data obtained from these MWRs were also demonstrated. Our comparisons of MWR against the nearest radiosonde observation demonstrate that the TP-PROFILE MWR system is adequate for monitoring the thermal and moisture variability of the troposphere over the TP. The continuous temperature and moisture profiles derived from the MWR data provide a unique perspective on the evolution of the thermodynamic structure associated with the heating of the TP. The TP-PROFILE project reveals that the low-temporal resolution instruments are prone to large uncertainties in their vapor estimation in the mountain valleys on the TP.