{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,4]],"date-time":"2026-07-04T05:18:40Z","timestamp":1783142320744,"version":"3.54.6"},"posted":{"date-parts":[[2026]]},"group-title":"SSRN","reference-count":35,"publisher":"Elsevier BV","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":[],"abstract":"<jats:p>Short-duration heavy rainfall, characterized by sudden onset and strong nonlinearity, remains a major challenge for refined categorical nowcasting. Existing deep-learning models still face bottlenecks in maintaining spatial continuity of the main rainband, quantifying rainfall intensity, and fusing multi-source observations. Using radar composite reflectivity and hourly rainfall\u00a0from automatic weather stations during 2017\u20132025, this study develops FusionTempoNet (FTNet), a multi-source fusion model tailored to categorical rainfall forecasting. FTNet adopts a dual-stream residual Gated Recurrent Unit (GRU) architecture to encode radar echo sequences and surface rainfall\u00a0sequences separately, enabling synergistic representation of convective morphology and intensity. A spatial neighborhood maximum data augmentation strategy is designed to mitigate sample imbalance while preserving core extreme-rainfall intensities. An ordinal Dice loss is integrated to enforce ordinal consistency across all rainfall levels, alleviating systematic under-forecasting caused by sparse extreme samples during training. Evaluation shows that, with the integration of the aforementioned technical components, FTNet delivers the best and most stable skill in 1\u20133 h categorical nowcasting. In addition, for high rainfall intensity levels (e.g., \u226520 mm h-1), FTNet achieves substantial TS gains over classic models such as PredRNN++, PhyDNet and Fenglei, exhibiting advantages in the stability of heavy-rainfall center positioning and skill at high intensity levels. This study demonstrates the importance of multi-source spatiotemporal fusion, together with imbalance-aware sample augmentation and loss design, for improving objective categorical forecasting of heavy rainfall.<\/jats:p>","DOI":"10.2139\/ssrn.7051911","type":"posted-content","created":{"date-parts":[[2026,7,4]],"date-time":"2026-07-04T01:52:07Z","timestamp":1783129927000},"source":"Crossref","is-referenced-by-count":0,"title":["FusionTempoNet: Dual-Stream Residual Categorical Nowcasting of Short-Duration Heavy Rainfall with Data Augmentation and Ordinal Loss"],"prefix":"10.2139","author":[{"given":"Xiaoxiong","family":"You","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"He","family":"Chen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-1289-4710","authenticated-orcid":true,"given":"Zhaoming","family":"Liang","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Yang","family":"Li","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jingyu","family":"Xu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Shu","family":"Lu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Hui","family":"Zhang","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"78","reference":[{"issue":"2","key":"ref1","doi-asserted-by":"crossref","DOI":"10.3390\/rs13020246","article-title":"Fusion of rain radar images and wind forecasts in a deep learning model applied to rain nowcasting","volume":"13","author":"V Bouget","year":"2021","journal-title":"Remote Sensing"},{"issue":"3","key":"ref2","first-page":"471","article-title":"The optical flow method and its application to nowcasting","volume":"73","author":"C Y Cao","year":"2015","journal-title":"Acta Meteorologica Sinica"},{"key":"ref3","article-title":"Enhancing nowcasting with multi-resolution inputs using deep learning: Exploring model decision mechanisms","volume":"52","author":"Y Cao","year":"2025","journal-title":"2024GL113699"},{"key":"ref4","first-page":"348","article-title":"PCT-CycleGAN: Paired complementary temporal cycle-consistent adversarial networks for radar-based precipitation nowcasting","author":"J Choi","year":"2023","journal-title":"Proceedings of the 32nd ACM International Conference on Information and Knowledge Management"},{"key":"ref5","author":"A Cornelissen","year":"2025","journal-title":"Integrating Weather Station Data and Radar for Precipitation Nowcasting: SmaAt-fUsion and SmaAt-Krige-GNet"},{"key":"ref6","first-page":"1","author":"C A Doswell","year":"2001","journal-title":"Severe convective storms"},{"key":"ref7","article-title":"A Simple Approach to Ordinal Classification","volume":"2167","author":"E Frank","year":"2001","journal-title":"Machine Learning: ECML"},{"key":"ref8","doi-asserted-by":"crossref","DOI":"10.52202\/075280-3439","article-title":"PreDiff: Precipitation Nowcasting with Latent Diffusion Models","author":"Z H Gao","year":"2023","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref9","doi-asserted-by":"crossref","first-page":"74","DOI":"10.1175\/1520-0450(2004)043<0074:SDOTPO>2.0.CO;2","article-title":"Scale-dependence of the predictability of precipitation from continental radar images. Part II: Probability forecasts","volume":"43","author":"U Germann","year":"2004","journal-title":"Journal of Applied Meteorology"},{"key":"ref10","article-title":"ThoR: A Motion-Dependent Physics-Informed Deep Learning Framework with Constraint-Centric Theory of Functional Connections for Rainfall Nowcasting","volume":"15","author":"K T Gia","year":"2025","journal-title":"Scientific Reports"},{"key":"ref11","author":"J Gong","year":"2024","journal-title":"CasCast: Skillful high-resolution precipitation nowcasting via cascaded modelling"},{"issue":"4","key":"ref12","first-page":"202","article-title":"Projection of precipitation extremes for the 21st century over China","volume":"3","author":"Z H Jiang","year":"2007","journal-title":"Advances in Climate Change Research"},{"key":"ref13","first-page":"1388","article-title":"Disentangling physical dynamics from unknown factors for unsupervised video prediction","author":"V Le Guen","year":"2020","journal-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition"},{"issue":"4","key":"ref14","first-page":"443","article-title":"Temporal and spatial variations of extreme precipitation in the pre-flood period of South China in recent 40 years","volume":"15","author":"L P Li","year":"2010","journal-title":"Climatic and Environmental Research"},{"key":"ref15","first-page":"1","article-title":"MMF-RNN: A Multimodal Fusion Model for Precipitation Nowcasting Using Radar and Ground Station Data","volume":"63","author":"Q Liu","year":"2025","journal-title":"IEEE Transactions on Geoscience and Remote Sensing"},{"key":"ref16","author":"C Meo","year":"2024","journal-title":"Extreme precipitation nowcasting using transformer-based generative models"},{"key":"ref17","article-title":"M4Caster: Multi-source, multi-spatial, multi-temporal modeling for precipitation nowcasting","volume":"648","author":"D Niu","year":"2025","journal-title":"Neurocomputing"},{"key":"ref18","author":"C Pel\ufffdez-Rodr\ufffdguez","year":"2024","journal-title":"A general explicable forecasting framework for weather events based on ordinal classification and inductive rules combined with fuzzy logic. Knowledge-Based Systems"},{"issue":"10","key":"ref19","doi-asserted-by":"crossref","first-page":"4185","DOI":"10.5194\/gmd-12-4185-2019","article-title":"Pysteps: An open-source Python library for probabilistic precipitation nowcasting","volume":"12","author":"S Pulkkinen","year":"2019","journal-title":"Geoscientific Model Development"},{"issue":"7878","key":"ref20","doi-asserted-by":"crossref","first-page":"672","DOI":"10.1038\/s41586-021-03854-z","article-title":"Skilful precipitation nowcasting using deep generative models of radar","volume":"597","author":"S Ravuri","year":"2021","journal-title":"Nature"},{"key":"ref21","doi-asserted-by":"crossref","first-page":"78","DOI":"10.1175\/2007MWR2123.1","article-title":"Scale-Selective Verification of Rainfall Accumulations from High-Resolution Forecasts of Convective Events","volume":"136","author":"N M Roberts","year":"2008","journal-title":"Monthly Weather Review"},{"key":"ref22","doi-asserted-by":"crossref","first-page":"1","DOI":"10.1109\/TGRS.2025.3640078","article-title":"RadarGaugeNet: Radar-Rain-Gauge-Based AI Architecture for Real-Time Quantitative Precipitation Estimation","volume":"63","author":"R Sarafian","year":"2025","journal-title":"IEEE Transactions on Geoscience and Remote Sensing"},{"issue":"4","key":"ref23","first-page":"389","article-title":"Verification and case evaluation of the \"Fenglei\" V1 meteorological nowcasting model","volume":"51","author":"J Sheng","year":"2025","journal-title":"Meteorological Monthly"},{"key":"ref24","first-page":"802","article-title":"Convolutional LSTM network: A machine learning approach for precipitation nowcasting","volume":"28","author":"X Shi","year":"2015","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref25","first-page":"5617","article-title":"Deep learning for precipitation nowcasting: A benchmark and a new model","volume":"30","author":"X Shi","year":"2017","journal-title":"Advances in Neural Information Processing Systems"},{"key":"ref26","article-title":"Fusion of multi-source precipitation records via coordinate-based generative models","volume":"17","author":"S Sun","year":"2026","journal-title":"Nature Communications"},{"key":"ref27","first-page":"5123","article-title":"PredRNN++: Towards a resolution of the deep-in-time dilemma in spatiotemporal predictive learning","author":"Y Wang","year":"2018","journal-title":"Proceedings of the 35th International Conference on Machine Learning"},{"issue":"2","key":"ref28","first-page":"116","article-title":"Statistical analysis of temporal and spatial distribution of short-duration heavy rainfall and related environmental parameters in Hunan Province","volume":"35","author":"R Yao","year":"2020","journal-title":"Journal of Catastrophology"},{"key":"ref29","article-title":"A study on loss function against data imbalance in deep learning correction of precipitation forecasts","volume":"281","author":"X You","year":"2023","journal-title":"Atmospheric Research"},{"key":"ref30","first-page":"27758","article-title":"DiffCast: A unified framework via residual diffusion for precipitation nowcasting","author":"D Yu","year":"2024","journal-title":"Proceedings of the IEEE\/CVF Conference on Computer Vision and Pattern Recognition"},{"issue":"1","key":"ref31","first-page":"89","article-title":"Analysis of basic characteristics of severe convective weather under North China cold vortex","volume":"27","author":"Z Y Yu","year":"2011","journal-title":"Journal of Tropical Meteorology"},{"issue":"7970","key":"ref32","doi-asserted-by":"crossref","first-page":"526","DOI":"10.1038\/s41586-023-06184-4","article-title":"Skilful nowcasting of extreme precipitation with NowcastNet","volume":"619","author":"Y Zhang","year":"2023","journal-title":"Nature"},{"issue":"5","key":"ref33","first-page":"506","article-title":"Application and improvement of deep learning methods in heavy precipitation nowcasting","volume":"41","author":"Y P Zhang","year":"2022","journal-title":"Torrential Rain and Disasters"},{"issue":"6","key":"ref34","first-page":"641","article-title":"Some basic issues of severe convective weather forecasting","volume":"43","author":"Y G Zheng","year":"2017","journal-title":"Meteorological Monthly"},{"issue":"1","key":"ref35","first-page":"1","article-title":"A short-term lightning forecast method by fusing NWP and multi-source observation data using deep learning","volume":"79","author":"K H Zhou","year":"2021","journal-title":"Acta Meteorologica Sinica"}],"container-title":[],"original-title":[],"deposited":{"date-parts":[[2026,7,4]],"date-time":"2026-07-04T04:57:58Z","timestamp":1783141078000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.ssrn.com\/abstract=7051911"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026]]},"references-count":35,"URL":"https:\/\/doi.org\/10.2139\/ssrn.7051911","relation":{},"subject":[],"published":{"date-parts":[[2026]]},"subtype":"preprint"}}