Search

Accurate and timely precipitation forecasts are critical in modern society, influencing both economic activity and daily life. While deep learning methods leveraging remotely sensed radar data have become prevalent for precipitation nowcasting, longer-term forecasting remains challenging. This is due to accumulated errors in deep learning models and insufficient information about precipitation systems over longer time horizons. To address these challenges, we introduce the Short-Term Precipitation Forecast Network (STPF-Net), a recurrent neural network designed for longer-term precipitation prediction. STPF-Net uses a multi-tier structure with varying temporal resolutions to mitigate the accumulated errors during longer forecasts. Additionally, its transformer-based module incorporates larger spatial contexts, providing more complete information about precipitation systems. We evaluated STPF-Net on radar data from southeastern China, training separate models for 6 and 12 h forecasts. Quantitative results demonstrate STPF-Net achieved superior accuracy and lower errors compared to benchmark deep learning and numerical weather prediction models. Visualized case studies indicate reasonably coherent 6 h predictions from STPF-Net versus other methods. For 12 h forecasts, while STPF-Net outperformed other models, it still struggled with storm initiation over longer forecasting time.

Locally-originated typhoons in the South China Sea (SCS) are characterized by long duration, complex track, and high probability of landfall, which tend to cause severe wind, rainstorm, and flood disasters in coastal regions. Therefore, it is of great significance to conduct research on typhoon precipitation microphysics in the SCS. Using GPM satellite observations, the precipitation microphysics of typhoons in the SCS are analyzed by combining case and statistical studies. The precipitation of Typhoon Ewiniar (2018) in the SCS is found to be highly asymmetric. In the eyewall, the updraft is strong, the coalescence process of particles is distinct, and the precipitation is mainly concentrated in large raindrops. In the outer rainbands, the “bright-band” of melting layer is distinct, the melting of ice particles and the evaporation of raindrops are distinct, and there exist a few large raindrops in the precipitation. Overall, the heavy precipitation of typhoons in the SCS is composed of higher concentration of smaller raindrops than that in the western Pacific (WP), leading to a more “oceanic deep convective” feature of typhoons in the SCS. While the heavy precipitation of typhoons in the SCS is both larger in drop size and number concentration than that in the North Indian Ocean (NIO), leading to more abundant rainwater of typhoons in the SCS. For the relatively weak precipitation (R < 10 mm h−1), the liquid water path (LWP) of typhoons in the SCS is higher than that of the NIO, while the ice water path (IWP) of the locally-originated typhoons in the SCS is lower than that of the WP. For the heavy precipitation (R ≥ 10 mm h−1), the LWP and IWP of typhoons in the SCS are significantly higher than those in the WP and NIO.

Autonomous driving has received enormous attention from the academic and industrial communities. However, achieving full driving autonomy is not a trivial task, because of the complex and dynamic driving environment. Perception ability is a tough challenge for autonomous driving, while 3D object detection serves as a breakthrough for providing precise and dependable 3D geometric information. Inspired by practical driving experiences of human experts, a pure visual scheme takes sufficient responsibility for safe and stable autonomous driving. In this paper, we proposed an anchor-free and keypoint-based 3D object detector with monocular vision, named Keypoint3D. We creatively leveraged 2D projected points from 3D objects’ geometric centers as keypoints for object modeling. Additionally, for precise keypoints positioning, we utilized a novel self-adapting ellipse Gaussian filter (saEGF) on heatmaps, considering different objects’ shapes. We tried different variations of DLA-34 backbone and proposed a semi-aggregation DLA-34 (SADLA-34) network, which pruned the redundant aggregation branch but achieved better performance. Keypoint3D regressed the yaw angle in a Euclidean space, which resulted in a closed mathematical space avoiding singularities. Numerous experiments on the KITTI dataset for a moderate level have proven that Keypoint3D achieved the best speed-accuracy trade-off with an average precision of 39.1% at 18.9 FPS on 3D cars detection.

We used the Constellation Observing System for Meteorology, Ionosphere, and Climate-2 (COSMIC-2) dry temperature profile data from December 2019 to November 2021 to study the vertical wavenumber spectra of the potential energy of stratospheric gravity waves (GWs Ep) in the Asian monsoon region (15–45°N, 70–150°E). The GW Ep decreases with increasing vertical wavenumber, and the spectral slope varies with wavenumber. The spectral slope becomes smaller over a wavenumber range of 0.1–0.45 km−1, and larger from 0.45–1 km−1, with increasing wavenumber. The energy density distribution at middle and low latitudes shows seasonal variations. Over a wavenumber range of 0.05–0.5 km−1, the energy density in winter is higher at middle latitudes than at low latitudes, and the opposite is observed in summer over a wavenumber range from 0.1 to 1 km−1. Both the spectral amplitude and characteristic wavelength exhibit band-like patterns, and the large-value bands and their centers vary significantly with the season. In winter, the middle latitude spectral amplitude is larger than that at low latitudes, and the significant large-value band-like distribution is at ~40°N. In summer, the distribution is opposite, with large-value band regions over the Bay of Bengal and Indo-China Peninsula. The large-value region of the middle latitude spectral amplitude corresponds to a longer characteristic wavelength, while the large-value region of the low latitude spectral amplitude corresponds to a shorter characteristic wavelength. There is also significant seasonal variation in the distribution of spectral slopes. Over a wavenumber range of 0.1 to 0.5 km−1, the slope is smaller at middle latitudes and larger at low latitudes in winter; the opposite is observed in summer. There is a significant annual cycle of spectral amplitude at middle and low latitudes, and a 4.8 month cycle at middle latitudes.

COSMIC-2 (Constellation Observing System for Meteorology, Ionosphere and Climate- 2) dry temperature profile data from December 2019 to November 2021 are used to study stratospheric gravity waves (GWs) in the Asian monsoon region. The stratosphere between 20 and 50 km is divided into the lower, middle, and high layers based on the vertical distribution of the mean potential energy (Ep) and the horizontal distribution of GW Ep in these three layers, and their seasonal changes are analyzed. The source and propagating mechanism of GWs in middle latitudes in winter are revealed. The results show that GWs in the stratosphere have distinct distribution features during different seasons. The significant Ep in winter appears mainly in middle latitudes north of 30°N, whereas in summer, it appears in the low latitudes south of 30°N. There are significant areas of GW activity in both low and middle latitudes in spring and autumn, but their intensity is significantly weaker than in winter and summer. Areas with significant GWs and the seasonal variation of their intensity are accompanied by the Asian monsoon activity. In winter, there is a northward and upward propagating column for GWs above the Sichuan Basin, and in summer, there is an eastward and upward propagating column for GWs in the zonal band 15–25°N. The occurrence of GWs in northwestern China in winter is the result of the subtropical jet stream and topography. Once GWs enter the stratosphere, they are regulated by the winter stratospheric environment, and the GWs acquire a northerly component by the wind shear. The meridional wind shear in the background field is an important factor affecting the development and propagation of GWs.

Typhoon In-Fa hit continental China in July 2021 and caused an unprecedented rainfall amount, making it a typical case to examine the ability of numerical models in forecasting landfalling typhoons. The record-breaking storm was simulated using a 3-km-resolution weather research and forecast (WRF) model with spectral bin microphysics scheme (BIN) and two-moment seven-class bulk parameterization scheme (BULK). The simulations were then separated into three different typhoon landfall periods (i.e., pre-landfall, landfall, and post-landfall). It was found that typhoon intensity prediction is sensitive to microphysical schemes regardless of landfall periods, while typhoon track prediction tends to be more (less) sensitive to microphysical schemes after (before) typhoon landfall. Moreover, significant differences exist between BIN and BULK schemes in simulating the storm intensity, track, and rainfall distribution. BIN scheme simulates stronger (weaker) typhoon intensity than BULK scheme after (before) landfall, while BULK scheme simulates typhoon moving faster (slower) than BIN scheme before (after) landfall. BIN scheme produces much more extensive and homogeneous typhoon rainbands than BULK scheme, whereas BULK scheme produces stronger (weaker) rainfall in the typhoon inner (outer) rainbands. The possible reasons for such differences are discussed. At present, the ability of WRF and other mesoscale models to accurately simulate the typhoon precipitation hydrometeors is still limited. To evaluate the performances of BIN and BULK schemes of WRF model in simulating the condensed water in Typhoon In-Fa, the observed microwave brightness temperature and radar reflectivity from the core observatory of Global Precipitation Mission (GPM) satellite are directly used for validation with the help of a satellite simulator. It is suggested that BIN scheme has better performance in estimating the spatial structure, overall amplitude, and precise location of the condensed water in typhoons before landfall. During typhoon landfall, the performance of BIN scheme in simulating the structure and location of the condensate is close to that of BULK scheme, but the condensate intensity prediction by BIN scheme is still better; BULK scheme performs even better than BIN scheme in the prediction of condensate structure and location after typhoon landfall. Both schemes seem to have poorer performances in simulating the spatial structure of precipitation hydrometeors during typhoon landfall than before/after typhoon landfall. Moreover, BIN scheme simulates more (less) realistic warm (cold) rain processes than BULK scheme, especially after typhoon landfall. BULK scheme simulates more cloud water and larger convective updraft than BIN scheme, and this is also reported in many model studies comparing BIN and BULK schemes.

In July 2021, Typhoon In-Fa attacked eastern China and broke many records for extreme precipitation over the last century. Such an unrivaled impact results from In-Fa’s slow moving speed and long residence time due to atmospheric circulations. With the supports of 66 networked surface disdrometers over eastern China and collaborative observations from the advanced GPM satellite, we are able to reveal the unique precipitation microphysical properties of the record-breaking Typhoon In-Fa (2021). After separating the typhoon precipitation into convective and stratiform types and comparing the drop size distribution (DSD) properties of Typhoon In-Fa with other typhoons from different climate regimes, it is found that typhoon precipitation shows significant internal differences as well as regional differences in terms of DSD-related parameters, such as mass-weighted mean diameter (Dm), normalized intercept parameter (Nw), radar reflectivity (Z), rain rate (R), and intercept, shape, and slope parameters (N0, µ, Λ). Comparing different rain types inside Typhoon In-Fa, convective rain (Nw ranging from 3.80 to 3.96 mm−1 m−3) shows higher raindrop concentration than stratiform rain (Nw ranging from 3.40 to 3.50 mm−1 m−3) due to more graupels melting into liquid water while falling. Large raindrops occupy most of the region below the melting layer in convective rain due to a dominant coalescence process of small raindrops (featured by larger ZKu, Dm, and smaller N0, µ, Λ), while small raindrops account for a considerable proportion in stratiform rain, reflecting a significant collisional breakup process of large raindrops (featured by smaller ZKu, Dm, and larger N0, µ, Λ). Compared with other typhoons in Hainan and Taiwan, the convective precipitation of Typhoon In-Fa shows a larger (smaller) raindrop concentration than that of Taiwan (Hainan), while smaller raindrop diameter than both Hainan and Taiwan. Moreover, the typhoon convective precipitation measured in In-Fa is more maritime-like than precipitation in Taiwan. Based on a great number of surface disdrometer observational data, the GPM precipitation products were further validated for both rain types, and a series of native quantitative precipitation estimation relations, such as Z–R and R–Dm relations were derived to improve the typhoon rainfall retrieval for both ground-based radar and spaceborne radar.

The Qilian Mountains (QLMs), an important ecological protective barrier and major water resource connotation area in the Hexi Corridor region, have an important impact on ecological security in western China due to their ecological changes. However, most existing studies have investigated vegetation changes and their main driving forces in the QLMs on the basis of a single scale. Thus, the interactions among multiple environmental factors in the QLMs are still unclear. This study was based on normalised difference vegetation index (NDVI) data from 2000 to 2019. We systematically analysed the spatial and temporal characteristics of the QLMs at multiple time scales using trend analysis, ensemble empirical mode decomposition, Geodetector, and correlation analysis methods. At different time scales under single-factor and multi-factor interactions, we examined the mechanisms of the vegetation changes and their drivers. Our results showed that the vegetation in the QLMs showed a trend of overall improvement in 2000–2019, at a rate of 0.88 × 10−3, mainly in the central western regions. The NDVI in the QLMs showed a short change cycle of 3 and 5 years and a long-term trend. Sunshine time and wind speed were the main drivers of the vegetation variation in the QLMs, followed by temperature. Precipitation affected the vegetation spatial variation within a certain altitude range. However, temperature and precipitation had stronger explanatory powers for the vegetation variation in the western QLMs than in the eastern part. Their interaction was the dominant factor in the regional differences in vegetation. The responses of the NDVI to temperature and precipitation were stronger in the long time series. The main drivers of vegetation variation were land surface temperature and precipitation in the east and temperature and evapotranspiration in the west. Precipitation was the main driver of vegetation growth in the northern and southwestern QLMs on both the short- and long-term scales. Vegetation changes were more significantly influenced by short-term temperature changes in the east but by a combination of temperature and precipitation in most parts of the QLMs on a 5-year time scale.

The Advanced Himawari Imager (AHI) onboard the Himawari-8 geostationary satellite provides continuous observations every 10 min. This study investigates the assimilation of every-10-min radiance from the AHI with the POD-4DEnVar method. Cloud detection is conducted in the AHI quality control procedure to remove cloudy and precipitation-affected observations. Historical samples and physical ensembles are combined to construct four-dimensional ensembles according to the observed frequency of the Himawari-8 satellite. The purpose of this study was to test the potential impacts of assimilating high temporal resolution observations with POD-4DEnVar in a numerical weather prediction (NWP) system. Two parallel experiments were performed with and without Himawari-8 radiance assimilation during the entire month of July 2020. The results of the experiment with radiance assimilation show that it improves the analysis and forecast accuracy of geopotential, horizontal wind field and relative humidity compared to the experiment without radiance assimilation. Moreover, the equitable threat score (ETS) of 24-h accumulated precipitation shows that assimilating Himawari-8 radiance improves the rainfall forecast accuracy. Improvements were found in the structure, amplitude and location of the precipitation. In addition, the ETS of hourly accumulated precipitation indicates that assimilating high temporal resolution Himawari-8 radiance can improve the prediction of rapidly developed rainfall. Overall, assimilating every-10-min AHI radiance from Himawari-8 with POD-4DEnVar has positive impacts on NWP.

High spatial resolution estimating of exposure to particulate matter 2.5 (PM2.5) is currently very limited in China. This study uses the newly released nationwide, hourly PM2.5 concentrations to create a nationwide, geographically weighted regression (GWR) model to estimate ground-level PM2.5 concentrations in China. A3 km resolution aerosol optical depth (AOD) product from MODIS is used as the primary predictor. Fire emissions detected by MODIS fire count were considered in the model development process. Additionally, meteorological features were used as covariates in the model to improve the estimation of ground-level PM2.5 concentrations. The model performed well and explained 81% of the daily PM2.5 concentration variations in model predictions, and the cross validations R² is 0.79. The cross-validated root mean squared error (RMSE) of the model was 18.6 μg/m³.Annual PM2.5 concentrations retrieved by the MODIS 3 km AOD product indicated that most of the residential community areas exceeded the new annual Chinese PM2.5 National Standard level 2. Estimated high-resolution national-scale daily PM2.5 maps are useful to identify severe air pollution episodes and determine health risk assessments. These results suggest that this approach is useful for estimating large-scale ground-level PM2.5 distributions, especially for regions without PM monitoring sites. [ABSTRACT FROM AUTHOR]