MSF-PhyDRNN: A Physics-Driven Multi-Source Fusion Recurrent Neural Network for Short-Term Thunderstorm Gale Nowcasting

Deep convective clouds, such as towering cumulus and Cumulonimbus, can endanger lives and property, also being a major hazard to aviation. This study presents the convective index (IndexCON) used operationally at the Portuguese Meteorological Watch Office. Moreover, IndexCON is evaluated against lightning and precipitation data for two years, between January 2022 and December 2023, over mainland Portugal and its surrounding areas. This index combines several European Center for Medium-Range Weather Forecasts (ECMWF) prognostic variables, such as stability indices, cloud water content, relative humidity and vertical velocity, using a fuzzy-logic approach. IndexCON performs well in the warm season (May–October), with a probability of detection (POD) of 70%, a false alarm ratio (FAR) of 30% and a probability of false detection (POFD) less than 5%, leading to a Critical Success Index (CSI) above 0.55. However, IndexCON performs worse in the cold season (November–April), when dynamical drivers are more relevant, mainly due to overestimating the convective activity, resulting in CSI and Heidke Skill Score (HSS) values below 0.3. Optimizing the membership functions partially reduces this overestimation. Finally, the added value of IndexCON was illustrated in detail for a thunderstorm episode, using satellite products, lightning and precipitation data.

Data-to-data translation-based nowcasting of specific sea fog using geostationary weather satellite observation

Many previous studies have attempted to distinguish fog from clouds using low-orbit and geostationary satellite observations from visible (VIS) to longwave infrared (LWIR) bands. However, clouds and fog have often been misidentified because of their similar spectral features. Recently, advanced meteorological geostationary satellites with improved spectral, spatial, and temporal resolutions, including Himawari-8/9, GOES-16/17, and GeoKompsat-2A, have become operational. Accordingly, this study presents an improved algorithm for detecting daytime sea fog using one VIS and one near-infrared (NIR) band of the Advanced Himawari Imager (AHI) of the Himawari-8 satellite. We propose a regression-based relationship for sea fog detection using a combination of the Normalized Difference Snow Index (NDSI) and reflectance at the green band of the AHI. Several case studies, including various foggy and cloudy weather conditions in the Yellow Sea for three years (2017–2019), have been performed. The results of our algorithm showed a successful detection of sea fog without any cloud mask information. The pixel-level comparison results with the sea fog detection based on the shortwave infrared (SWIR) band (3.9 μm) and the brightness temperature difference between SWIR and LWIR bands of the AHI showed high statistical scores for probability of detection (POD), post agreement (PAG), critical success index (CSI), and Heidke skill score (HSS). Consequently, the proposed algorithms for daytime sea fog detection can be effective in daytime, particularly twilight, conditions, for many satellites equipped with VIS and NIR bands.

Weather radar echo prediction method based on convolution neural network and Long Short-Term memory networks for sustainable e-agriculture

ABSTRACTIn this paper, we have proposed a method of sea fog prediction using model sea surface wind (SSW) data and hourly optical satellite data with 500 m spatial resolution. Three sea fog cases in the Yellow Sea were selected from Geostationary Ocean Color Imager (GOCI) images, and their shifts were determined from the fog cluster centroid displacements at successive hours. The 4 km resolution Weather Research and Forecasting (WRF) model SSW were used to predict sea fog shifts. The sea fog prediction results were verified using validation indices like probability of detection (POD), false alarm ratio (FAR), and critical success index (CSI). The predictions started with high initial skill scores (POD = 0.76, FAR = 0.18, and CSI = 0.66), and showed values 0.59, 0.30, and 0.46 for POD, FAR, and CSI, respectively for predictions after two hours. After three hours their values still remained close to the medians though decreased, but afterwards decreased considerably for POD and CSI. Thus the method is found suitable for short-term prediction of sea fog. Additionally, trajectory sensitivities were compared between 20 and 4 km resolution WRF which, in general, resulted in less errors for 4 km WRF model SSW.

India Meteorological Department (IMD) issues heavy rainfall warning for a meteorological sub-division when the expected 24 hours rainfall over any rain gauge station in that sub-division is likely to be 64.5 mm or more. Though these warnings have been provided since the inception of IMD, a few attempts have been made for quantitative evaluation of these warnings. Hence, a study is undertaken to verify the heavy rainfall warnings over 36 meteorological sub-divisions of India during monsoon months (June-September) and season as a whole. For this purpose, data of recent 5 years (2002-2006) has been taken into consideration. In this connection, the day when heavy rainfall is recorded over at least one station in a sub-division, has been considered as a heavy rainfall day for that sub-division.
 
 There is a large spatial and temporal variability in skill scores of heavy rainfall warnings over India during summer monsoon season. Considering the monsoon season as a whole, the Heidke Skill Score (HSS) is relatively less (<0.20) over the regions with less frequent heavy rainfall like Lakshadweep, southeast peninsula, Vidarbha, Marathwada, Jammu & Kashmir, Arunachal Pradesh and Nagaland, Manipur, Mizoram & Tripura (NMMT). It is higher (> 0.50) over Konkan & Goa, Madhya Maharashtra and Gujarat region. There has been improvement in the forecast skill with gradual increase in the critical success index and Heidke skill score over the years mainly due to the Numerical Weather Prediction (NWP) models' guidance available to the forecasters. However, the false alarm rate and missing rate are still very high (> 0.50), especially over many sub-divisions of northwest India, southeast peninsula and NMMT.

Air pollution, particularly fine particulate matter (PM2.5), poses significant public health and environmental risks. This study explores the effectiveness of spatiotemporal graph neural networks (ST-GNNs) in forecasting PM2.5 concentrations by integrating remote-sensing hyperspectral indices with traditional meteorological and pollutant data. The model was evaluated using data from Switzerland and the Gauteng province in South Africa, with datasets spanning from January 2016 to December 2021. Key performance metrics, including root mean squared error (RMSE), mean absolute error (MAE), probability of detection (POD), critical success index (CSI), and false alarm rate (FAR), were employed to assess model accuracy. For Switzerland, the integration of spectral indices improved RMSE from 1.4660 to 1.4591, MAE from 1.1147 to 1.1053, CSI from 0.8345 to 0.8387, POD from 0.8961 to 0.8972, and reduced FAR from 0.0760 to 0.0719. In Gauteng, RMSE decreased from 6.3486 to 6.2319, MAE from 4.4891 to 4.4066, CSI from 0.9555 to 0.9560, and POD from 0.9699 to 0.9732, while FAR slightly increased from 0.0154 to 0.0181. Error analysis revealed that while the initial one-day ahead forecast without spectral indices had a marginally lower error, the dataset with spectral indices outperformed from the two-day ahead mark onwards. The error for Swiss monitoring stations stabilized over longer prediction lengths, indicating the robustness of the spectral indices for extended forecasts. The study faced limitations, including the exclusion of the Planetary Boundary Layer (PBL) height and K-index, lack of terrain data for South Africa, and significant missing data in remote sensing indices. Despite these challenges, the results demonstrate that ST-GNNs, enhanced with hyperspectral data, provide a more accurate and reliable tool for PM2.5 forecasting. Future work will focus on expanding the dataset to include additional regions and further refining the model by incorporating additional environmental variables. This approach holds promise for improving air quality management and mitigating health risks associated with air pollution.

Evaluations on gridded precipitation products spanning more than half a century over the Tibetan Plateau and its surroundings

ABSTRACTFour years (2011–2014) of the summer period (May–September) dual‐polarization Doppler C‐band weather radar data for Sürgavere, Estonia, were examined to determine the best indicator for cloud‐to‐ground lightning activity. Furthermore, the legacy radar‐derived hail indicator probability of hail and the storm maximum reflectivity were compared with the polarimetric hydrometeor classification algorithm (HCA) to establish a link between them and ensure data continuity. The study is based on convective storm cells identified from radar reflectivity data, using a 35 dBZ reflectivity threshold. Applying this threshold to the radar data resulted in 123 360 individual storm cells. It was found that 33.9% of the identified cells produced lightning and 25.9% hail. The number of individual storm cells on average is highest in the afternoon at 1600 local time and the mean storm cell area is largest in the evening at 2100 local time. Echo top 15 dBZ (ET15), ET20 and ET25 achieved the highest scores in terms of critical success index (0.39, 0.39 and 0.38 respectively), Heidke skill score (0.24, 0.26, 0.27) and equitable threat score (0.14, 0.15, 0.16). The graupel class of the dual‐polarization radar hydrometeor classification showed a similar performance to the 35 dBZ echo top. It was also shown that the reflectivity value just below 50 dBZ indicates the best skill for detection of hail, if HCA hail detection is used as the ground truth. The probability of hail value that shows the best agreement with the HCA hail class is 0.2.

Prior-Guided gated convolutional networks for rainstorm forecasting

Abstract. Although many multi-source precipitation products (MSPs) with high spatiotemporal resolution have been extensively used in water cycle research, they are still subject to various biases, including false alarm and missed bias. Precipitation merging technology is an effective means to alleviate this uncertainty. However, how to efficiently improve precipitation detection efficiency and precipitation intensity simultaneously is a problem worth exploring. This study presents a two-step merging strategy based on machine learning (ML) algorithms, including gradient boosting decision tree (GBDT), extreme gradient boosting (XGBoost), and random forest (RF). It incorporates six state-of-the-art MSPs (GSMaP, IMERG, PERSIANN-CDR, CMORPH, CHIRPS, and ERA5-Land) and rain gauges to improve the accuracy of precipitation identification and estimation from 2000 to 2017 over China. Multiple environment variables and spatial autocorrelation are combined in the merging process. The strategy first employs classification models to identify wet and dry days and then combines regression models to predict precipitation amounts based on classified wet days. The merged results are compared with traditional methods, including multiple linear regression (MLR), ML regression models, and gauge-based Kriging interpolation. A total of 1680 (70 %) rain gauges are randomly chosen for model training and 692 (30 %) for performance evaluation. The results show that (1) the multi-source merged precipitation products (MSMPs) outperformed all original MSPs in terms of statistical and categorical metrics, which substantially alleviates the temporal and spatial biases. The modified Kling–Gupta efficiency (KGE), critical success index (CSI), and Heidke Skill Score (HSS) of original MSPs are improved by 15 %–85 %, 17 %–155 %, and 21 %–166 %, respectively. (2) The spatial autocorrelation plays a significant role in precipitation merging, which considerably improves the model accuracy. (3) The performance of MSMPs obtained by the proposed method is superior to MLR, Kriging interpolation, and ML regression models. The XGBoost algorithm is recommended more for large-scale data merging owing to its high computational efficiency. (4) The two-step merging strategy performs better when higher-density gauges are used to model training. However, it has strong robustness and can also obtain better performance than original MSPs even when the gauge number is reduced to 10 % (237). This study provides an accurate and reliable method to improve precipitation detection accuracy under complex climatic and topographic conditions. It could be applied to other areas well if rain gauges are available.

Droughts severely impact agriculture, food security, and water resources, particularly in data scarce regions like the Genale Dawa River Basin (GDRB). This study evaluates the performance of five satellite rainfall products including CHIRPSv2.0, RFE2.0, TAMSATv3.1, PERSIANN-CDR, and ARC2 over 2001-2020 using metrics such as Mean Bias, Absolute Error (MAE), Root Mean Square Error (RMSE), Correlation Coefficient (CC), Probability of Detection (POD), False Alarm Ratio (FAR), and Critical Success Index (CSI). Furthermore, the studies examined the relationship between climate indices including the Pacific Decadal Oscillation (PDO), North Atlantic Oscillation (NAO), El Niño Southern Oscillation (ENSO), and Atlantic Multidecadal Oscillation (AMO) and drought variability. CHIRPSv2.0 demonstrated the highest accuracy, especially at Ginir (CC = 0.98) and Chewbet (CC = 0.75), and the lowest errors at Teferekella (MAE = 45, RMSE = 65). Based on a multi-criterion ranking approach, CHIRPSv2.0 was identified as the most reliable satellite-based product for daily rainfall estimation across the entire basin, exhibiting strong performance metrics (Lp = 0.433, CC = 0.84, POD = 0.85, CSI = 0.46). CHIRPSv2.0 was used to analyze drought spatiotemporal patterns using SPI-3 and SPI-6. Drought severity, frequency, and duration were highest in the west-central and northern GDRB, with major events occurring in 1988, 1991-1993, 2000, 2004, 2011, and 2012. Climate index analysis showed that AMO and NAO positive phases were associated with wetter conditions, while negative PDO and ENSO phases corresponded with drier periods, especially in central and eastern areas. These findings highlight CHIRPSv2.0's reliability for drought monitoring and its value for early warning and mitigation planning in the GDRB.

A statistical-empirical decision aid for forecasting morning fogs at Mount Gambier Airport based on pre-dawn data is presented. Of main concern are those fog events significantly impacting on aviation operations, although the model is applicable to other fog events of lesser impact. The aid is restricted to precipitation-absent fogs. The probability of fog occurrence was modelled on a ten-year dependent dataset using the logistic equation. By converting probabilistic forecasts to categorical forecasts of fog events, skill scores of Probability of Detection, False Alarm Rate and Critical Success Index were determined based upon three years of independent data. For operationally significant fogs, the Probability of Detection, False Alarm Rate and Critical Success Index were 73%, 23% and 60% respectively on the independent data. These scores were very similar to those obtained from the dependent dataset of ten years. When fog events were defined less stringently as either fogs per se or shallow fogs, the respective skill scores on the independent data were 84%, 20% and 70%; again with similar scores on the dependent data.

Deep learning techniques for radar echo extrapolation and prediction have become crucial for short-term precipitation forecasts in recent years. As the extrapolation leading time extends, radar echo intensity attenuates increasingly, and the forecast performance on strong echoes declines rapidly. These are two typical characteristics contributing to the current inaccurate results of radar extrapolation. To this end, we propose a novel diffusion radar echo extrapolation (DiffREE) algorithm driven by echo frames in this study. This algorithm deeply integrates the spatio-temporal information of radar echo frames through a conditional encoding module, and then it utilizes a Transformer encoder to automatically extract the spatio-temporal features of echoes. These features serve as inputs to the conditional diffusion model, driving the model to reconstruct the current radar echo frame. Moreover, a validation experiment demonstrates that the proposed method can generate high-precision and high-quality forecast images of radar echoes. To further substantiate the model performance, the DiffREE algorithm is compared with the other four models by using public datasets. In the radar echo extrapolation task, the DiffREE demonstrates a remarkable improvement in the evaluation metrics of critical success index, equitable threat score, Heidke skill score and probability of detection by 21.5%, 27.6%, 25.8%, and 21.8%, respectively, displaying notable superiority.

Tornadoes are highly destructive small-scale extreme weather processes in the troposphere. The weather radar is one of the most effective remote sensing devices for the monitoring and early warning of tornadoes. The existing tornado detection algorithms based on radar data are unsupervised and have strict multi-altitude constraints, such as the tornado detection algorithm based on tornado vortex signatures (TDA-TVS), which may lead to high false alarm rates, and the performance of the detection algorithm is greatly affected by the radar data quality control algorithm. A novel TDA-RF algorithm based on the random forest (RF) classification algorithm is proposed for real-time tornado identification of the S-band China new generation of Doppler weather radar (CINRAD-SA). The TDA-RF algorithm uses velocity features to identify tornadoes and adds features related to reflectivity and velocity spectrum width in radar level-II data. Historical CINRAD-SA tornado data from 2006–2015 are used to construct the tornado dataset and train the TDA-RF model. The performance of TDA-RF is evaluated using CINRAD-SA data from five tornadoes of 2016–2020 with enhanced Fujita(EF) scale ratings ranging from EF0 to EF4 and distances from 10 to 130 km to the radar. TDA-RF performs well overall with the probability of detection (POD), false alarm ratio (FAR), and critical success index (CSI) of 71%, 29%, and 55%, respectively. Moreover, the TDA-RF improves POD and CSI, and reduces FAR compared to the TDA-TVS. The maximum tornado early-warning time of TDA-RF is 17 min, and the average is 6 min; TDA-RF can provide classification probability according to the tornado generation and development process to facilitate tracking ability.