{"paper":{"title":"Predicting Forecast Error for the HRRR Using LSTM Neural Networks: A Comparative Study Using New York and Oklahoma State Mesonets","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"LSTM networks trained on mesonet data predict HRRR precipitation forecast errors with 48 percent average improvement.","cross_cats":[],"primary_cat":"physics.ao-ph","authors_text":"Chris D. Thorncroft, David Aaron Evans, Jay C. Rothenberger, Kara J. Sulia, Lauriana C. Gaudet, Nick P. Bassill","submitted_at":"2025-12-16T20:22:41Z","abstract_excerpt":"Long Short-Term Memory (LSTM) models are trained to predict forecast errors for the High-Resolution Rapid Refresh (HRRR) model using the New York State Mesonet and Oklahoma State Mesonet near-surface weather observations as ground truth. When evaluated using mean-absolute-error and percent improvement relative to the HRRR, LSTMs predict precipitation error most accurately, providing, on average, a 48% improvement relative to the HRRR forecast, followed by wind error, providing, on average, a 15% improvement, and then temperature error, providing, on average, a 25% improvement. Precipitation er"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"LSTMs predict precipitation error most accurately, providing, on average, a 48% improvement relative to the HRRR forecast, followed by wind error, providing, on average, a 15% improvement, and then temperature error, providing, on average, a 25% improvement.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The LSTM trained on historical mesonet-HRRR pairs will continue to predict future forecast errors accurately without major changes in model behavior or observation quality.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"LSTM networks predict HRRR forecast errors with average improvements of 48% for precipitation, 25% for temperature, and 15% for wind using mesonet ground truth.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"LSTM networks trained on mesonet data predict HRRR precipitation forecast errors with 48 percent average improvement.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"515910c831b93133bb9da0b9ed5c009175eb14b4f632d3eb7dfca59a0cd3e06a"},"source":{"id":"2512.14898","kind":"arxiv","version":2},"verdict":{"id":"fe3e4a88-265e-476c-b016-2add161809c1","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T21:58:03.994793Z","strongest_claim":"LSTMs predict precipitation error most accurately, providing, on average, a 48% improvement relative to the HRRR forecast, followed by wind error, providing, on average, a 15% improvement, and then temperature error, providing, on average, a 25% improvement.","one_line_summary":"LSTM networks predict HRRR forecast errors with average improvements of 48% for precipitation, 25% for temperature, and 15% for wind using mesonet ground truth.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The LSTM trained on historical mesonet-HRRR pairs will continue to predict future forecast errors accurately without major changes in model behavior or observation quality.","pith_extraction_headline":"LSTM networks trained on mesonet data predict HRRR precipitation forecast errors with 48 percent average improvement."},"references":{"count":72,"sample":[{"doi":"10.1190/1.3073005","year":2009,"title":"Case history inversion and interpretation of a 3d seismic data set from the ouachita mountains, oklahoma","work_id":"779d730b-f2b3-4366-af12-c57a620d2a0d","ref_index":1,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"","year":2023,"title":"New york state climate change projections methodology report","work_id":"eb21f1c6-9e56-4681-9743-d58399be528a","ref_index":2,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1175/1520-0442(1992)005","year":1992,"title":"A. G. Barnston and C. F. Ropelewski. Prediction of enso episodes using canonical correlation analysis. J. Climate, 5: 0 1316--1345, 1992. doi:10.1175/1520-0442(1992)005<1316:POEEUC>2.0.CO;2","work_id":"5932a88e-eda0-45e9-92f9-03b672fe9ffd","ref_index":3,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1007/978-3-031-45468-4","year":2023,"title":"Bishop and Hugh Bishop","work_id":"fe1cde39-66b7-4d47-a588-f99b6e722990","ref_index":4,"cited_arxiv_id":"","is_internal_anchor":false},{"doi":"10.1016/j.cageo.2017.08.005","year":2017,"title":"B. K. Blaylock, J. D. Horel, and S. T. Liston. Cloud archiving and data mining of high-resolution rapid refresh forecast model output. Computers & Geosciences, 109: 0 43--50, 2017. doi:10.1016/j.cageo","work_id":"25570a41-bc11-4dea-9659-28c7344bc21e","ref_index":5,"cited_arxiv_id":"","is_internal_anchor":false}],"resolved_work":72,"snapshot_sha256":"dde01a150d690ba49fe3a1339318fbc9ba07c7579146342fd96a7847d87ecb0d","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}