{"paper":{"title":"MambaRain: Multi-Scale Mamba-Attention Framework for 0-3 Hour Precipitation Nowcasting","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"MambaRain combines Mamba blocks with self-attention to extend accurate precipitation nowcasting to three hours.","cross_cats":[],"primary_cat":"cs.CV","authors_text":"Boyu Liu, Chunlei Shi, Cui Wu, Dan Niu, Hao Li, Hongbin Wang, Ni Fan, Xiang Xu, Xue Han, Yanlan Yang, Yongchao Feng, Yufeng Zhu, Zengliang Zang","submitted_at":"2026-05-14T09:23:12Z","abstract_excerpt":"Accurate precipitation nowcasting over extended horizons (0-3 hours) is essential for disaster mitigation and operational decision-making, yet remains a critical challenge in the field. Existing deterministic approaches are predominantly constrained to shorter prediction windows (0-2 hours), exhibiting severe performance degradation beyond 90 minutes owing to their inherent difficulty in capturing long-range spatiotemporal dependencies from radar-derived observations. To address these fundamental limitations, we propose MambaRain, a novel multi-scale encoder-decoder architecture that synergist"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"MambaRain substantially outperforms existing deterministic methodologies in 0-3 hour nowcasting tasks, with particularly pronounced performance gains in the challenging 2-3 hour prediction range.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That the complementary combination of Mamba blocks for temporal dynamics and self-attention for spatial correlations will capture the chaotic, multi-scale nature of precipitation fields without introducing new artifacts or requiring extensive post-hoc tuning that undermines the claimed gains.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"MambaRain integrates Mamba's efficient long-sequence modeling with attention mechanisms and a spectral loss to extend accurate deterministic precipitation nowcasting from 0-2 hours to 0-3 hours.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"MambaRain combines Mamba blocks with self-attention to extend accurate precipitation nowcasting to three hours.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"bff6e3f4c2b8d7def7c5d190adc026c94ae226b58c88d57ca34bcb317bba31e0"},"source":{"id":"2605.14606","kind":"arxiv","version":1},"verdict":{"id":"24e80f9e-29c6-471f-9acd-250259a6ae59","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T04:52:47.230251Z","strongest_claim":"MambaRain substantially outperforms existing deterministic methodologies in 0-3 hour nowcasting tasks, with particularly pronounced performance gains in the challenging 2-3 hour prediction range.","one_line_summary":"MambaRain integrates Mamba's efficient long-sequence modeling with attention mechanisms and a spectral loss to extend accurate deterministic precipitation nowcasting from 0-2 hours to 0-3 hours.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That the complementary combination of Mamba blocks for temporal dynamics and self-attention for spatial correlations will capture the chaotic, multi-scale nature of precipitation fields without introducing new artifacts or requiring extensive post-hoc tuning that undermines the claimed gains.","pith_extraction_headline":"MambaRain combines Mamba blocks with self-attention to extend accurate precipitation nowcasting to three hours."},"references":{"count":38,"sample":[{"doi":"","year":2026,"title":"Wavec2r: Wavelet-driven coarse-to-refined hierarchical learning for radar retrieval,","work_id":"9f45c0f1-a620-4cdf-8fe5-1f1c8462369d","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Alphapre: Amplitude-phase disentanglement model for precipitation nowcasting,","work_id":"cf355fac-99c2-499e-b0d3-5f6020a0b365","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Lmcast: A pretrained language model guided long-term memory transformer for precipitation nowcasting,","work_id":"db0ca0d9-53f6-400a-9437-6793f7e34ff5","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"Pimmnet: In- troducing multi-modal precipitation nowcasting via a physics-informed perspective,","work_id":"baa39b60-cc17-4780-b208-41c658d830de","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2025,"title":"End-to-end data-driven weather prediction,","work_id":"ef1a78cc-8892-4c8b-9d33-14d331a1cb6d","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":38,"snapshot_sha256":"a2281e414c03d237b631e56b9e415f87c7ba93548fb8dedd19b5eb70bb86c6ef","internal_anchors":1},"formal_canon":{"evidence_count":2,"snapshot_sha256":"f64b03854227f83ca91be873dc47e189ceec353187214ab0b2cc83a0b601c9be"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}