查看更多>>摘要：Thunderstorm wind gusts are small in scale,typically occurring within a range of a few kilometers.It is extremely challenging to monitor and forecast thunderstorm wind gusts using only automatic weather stations.Therefore,it is necessary to establish thunderstorm wind gust identification techniques based on multisource high-resolution observations.This paper introduces a new algorithm,called thunderstorm wind gust identification network(TGNet).It leverages multimodal feature fusion to fuse the temporal and spatial features of thunderstorm wind gust events.The shapelet transform is first used to extract the temporal features of wind speeds from automatic weather stations,which is aimed at distinguishing thunderstorm wind gusts from those caused by synoptic-scale systems or typhoons.Then,the encoder,structured upon the U-shaped network(U-Net)and incorporating recurrent residual convolutional blocks(R2U-Net),is employed to extract the corresponding spatial convective characteristics of satellite,radar,and lightning observations.Finally,by using the multimodal deep fusion module based on multi-head cross-attention,the temporal features of wind speed at each automatic weather station are incorporated into the spatial features to obtain 10-minutely classification of thunderstorm wind gusts.TGNet products have high accuracy,with a critical success index reaching 0.77.Compared with those of U-Net and R2U-Net,the false alarm rate of TGNet products decreases by 31.28％and 24.15％,respectively.The new algorithm provides grid products of thunderstorm wind gusts with a spatial resolution of 0.01°,updated every 10 minutes.The results are finer and more accurate,thereby helping to improve the accuracy of operational warnings for thunderstorm wind gusts.
查看更多>>摘要：Existing traditional ocean vertical-mixing schemes are empirically developed without a thorough understanding of the physical processes involved,resulting in a discrepancy between the parameterization and forecast results.The uncertainty in ocean-mixing parameterization is primarily responsible for the bias in ocean models.Benefiting from deep-learning technology,we design the Adaptive Fully Connected Module with an Inception module as the baseline to minimize bias.It adaptively extracts the best features through fully connected layers with different widths,and better learns the nonlinear relationship between input variables and parameterization fields.Moreover,to obtain more accurate results,we impose KPP(K-Profile Parameterization)and PP(Pacanowski-Philander)schemes as physical constraints to make the network parameterization process follow the basic physical laws more closely.Since model data are calculated with human experience,lacking some unknown physical processes,which may differ from the actual data,we use a decade-long time record of hydrological and turbulence observations in the tropical Pacific Ocean as training data.Combining physical constraints and a nonlinear activation function,our method catches its nonlinear change and better adapts to the ocean-mixing parameterization process.The use of physical constraints can improve the final results.
查看更多>>摘要：It is fundamental and useful to investigate how deep learning forecasting models(DLMs)perform compared to operational oceanography forecast systems(OFSs).However,few studies have intercompared their performances using an identical reference.In this study,three physically reasonable DLMs are implemented for the forecasting of the sea surface temperature(SST),sea level anomaly(SLA),and sea surface velocity in the South China Sea.The DLMs are validated against both the testing dataset and the"OceanPredict"Class 4 dataset.Results show that the DLMs'RMSEs against the latter increase by 44％,245％,302％,and 109％for SST,SLA,current speed,and direction,respectively,compared to those against the former.Therefore,different references have significant influences on the validation,and it is necessary to use an identical and independent reference to intercompare the DLMs and OFSs.Against the Class 4 dataset,the DLMs present significantly better performance for SLA than the OFSs,and slightly better performances for other variables.The error patterns of the DLMs and OFSs show a high degree of similarity,which is reasonable from the viewpoint of predictability,facilitating further applications of the DLMs.For extreme events,the DLMs and OFSs both present large but similar forecast errors for SLA and current speed,while the DLMs are likely to give larger errors for SST and current direction.This study provides an evaluation of the forecast skills of commonly used DLMs and provides an example to objectively intercompare different DLMs.
查看更多>>摘要：Potential vorticity(PV)streamers are elongated filaments of high PV intrusions that generally exhibit three distinct shapes:ordinarily southwestward,hook,and treble-clef,each with significant influences on weather.These PV streamers are most frequent over arid and semi-arid Central Asia in the mid-high latitudes.This study applied the Mask Region-based Convolutional Neural Network algorithm(Mask R-CNN)to PV streamers on the dynamical tropopause during the warm season(May to September)over the years 2000-04 to train a weighted variational model capable of identifying these different shapes.The trained model demonstrated a strong ability to distinguish between the three shapes.A climatological analysis of PV streamers over Central Asia spanning 2000 to 2021 revealed an increasingly deep and pronounced reversal of circulation from ordinary to treble-clef shapes.The treble-clef shape featured a PV tower and distinct cut-off low in the troposphere,but the associated upward motions and precipitation were confined within approximately 1200 km to the east of the PV tower.Although the hook-shape PV streamers were linked to a weaker cut-off low,the extent of upward motion and precipitation was nearly double that of the treble-clef category.In contrast,the ordinary PV streamer was primarily associated with tropopause Rossby wave breaking and exhibited relatively shallow characteristics,which resulted in moderate upward motion and precipitation to 500 km to its east.
查看更多>>摘要：Three-dimensional ocean subsurface temperature and salinity structures(OST/OSS)in the South China Sea(SCS)play crucial roles in oceanic climate research and disaster mitigation.Traditionally,real-time OST and OSS are mainly obtained through in-situ ocean observations and simulation by ocean circulation models,which are usually challenging and costly.Recently,dynamical,statistical,or machine learning models have been proposed to invert the OST/OSS from sea surface information;however,these models mainly focused on the inversion of monthly OST and OSS.To address this issue,we apply clustering algorithms and employ a stacking strategy to ensemble three models(XGBoost,Random Forest,and LightGBM)to invert the real-time OST/OSS based on satellite-derived data and the Argo dataset.Subsequently,a fusion of temperature and salinity is employed to reconstruct OST and OSS.In the validation dataset,the depth-averaged Correlation(Corr)of the estimated OST(OSS)is 0.919(0.83),and the average Root-Mean-Square Error(RMSE)is 0.639℃(0.087 psu),with a depth-averaged coefficient of determination(R2)of 0.84(0.68).Notably,at the thermocline where the base models exhibit their maximum error,the stacking-based fusion model exhibited significant performance enhancement,with a maximum enhancement in OST and OSS inversion exceeding 10％.We further found that the estimated OST and OSS exhibit good agreement with the HYbrid Coordinate Ocean Model(HYCOM)data and BOA_Argo dataset during the passage of a mesoscale eddy.This study shows that the proposed model can effectively invert the real-time OST and OSS,potentially enhancing the understanding of multi-scale oceanic processes in the SCS.
查看更多>>摘要：Grid forecasting can be used to effectively enhance the spatial and temporal density of forecast products,thereby improving the capability of short-term marine disaster forecasting and warnings in terms of proximity.The traditional method that relies on forecasters'subjective correction of station observation data for forecasting has been unable to meet the practical needs of refined forecasting.To address this problem,this paper proposes a Transformer-enhanced UNet(TransUNet)model for wave forecast AI correction,which fuses wind and wave information.The Transformer structure is integrated into the encoder of the UNet model,and instead of using the traditional upsampling method,the dual-sampling module is employed in the decoder to enhance the feature extraction capability.This paper compares the TransUNet model with the traditional UNet model using wind speed forecast data,wave height forecast data,and significant wave height reanalysis data provided by ECMWF.The experimental results indicate that the TransUNet model yields smaller root-mean-square errors,mean errors,and standard deviations of the corrected results for the next 24-h forecasts than does the UNet model.Specifically,the root-mean-square error decreased by more than 21.55％compared to its precorrection value.According to the statistical analysis,87.81％of the corrected wave height errors for the next 24-h forecast were within±0.2 m,with only 4.56％falling beyond±0.3 m.This model effectively limits the error range and enhances the ability to forecast wave heights.
查看更多>>摘要：Traditional meteorological downscaling methods face limitations due to the complex distribution of meteorological variables,which can lead to unstable forecasting results,especially in extreme scenarios.To overcome this issue,we propose a convolutional graph neural network(CGNN)model,which we enhance with multilayer feature fusion and a squeeze-and-excitation block.Additionally,we introduce a spatially balanced mean squared error(SBMSE)loss function to address the imbalanced distribution and spatial variability of meteorological variables.The CGNN is capable of extracting essential spatial features and aggregating them from a global perspective,thereby improving the accuracy of prediction and enhancing the model's generalization ability.Based on the experimental results,CGNN has certain advantages in terms of bias distribution,exhibiting a smaller variance.When it comes to precipitation,both UNet and AE also demonstrate relatively small biases.As for temperature,AE and CNNdense perform outstandingly during the winter.The time correlation coefficients show an improvement of at least 10％at daily and monthly scales for both temperature and precipitation.Furthermore,the SBMSE loss function displays an advantage over existing loss functions in predicting the 98th percentile and identifying areas where extreme events occur.However,the SBMSE tends to overestimate the distribution of extreme precipitation,which may be due to the theoretical assumptions about the posterior distribution of data that partially limit the effectiveness of the loss function.In future work,we will further optimize the SBMSE to improve prediction accuracy.
万方数据
维普