Season‐Net: A Deep Learning Framework for Bias Correction of Seasonal Forecasting Models
Bibliographic record
Abstract
Abstract Seasonal climate forecasts play a crucial role in decision‐making across sectors like agriculture, energy, and disaster management. However, these forecasts often exhibit spatially structured biases that undermine their reliability, but this structure also enables more effective bias correction, particularly improving performance in predicting temperature extremes. Traditional bias correction methods such as quantile mapping (QM) and linear scaling (LS) are limited by assumptions of stationarity and their inability to capture complex spatiotemporal patterns. To address these challenges, we introduce Season‐Net, a hybrid deep learning framework combining U‐Net and ConvLSTM architectures. Season‐Net is used to perform bias correction on seasonal daily temperature forecasts from the Met Office (GloSea6) and Météo‐France (System 8) by learning season‐specific spatial and temporal dependencies in a unified architecture. The model is trained with a novel sliding‐window quantile mapping loss function that introduces temporal awareness into the quantile mapping process, enhancing its ability to capture temperature distribution and evolution. Evaluations across North America and Africa show that Season‐Net consistently outperforms QM and LS in both deterministic (e.g., RMSE and Kendall's Tau) and probabilistic (e.g., Brier skill score and CRPSS) metrics. Furthermore, Season‐Net excels in impact‐based evaluations, significantly improving the prediction of extreme temperature events. These results highlight the superior capability of deep learning methods in correcting spatially structured seasonal forecast biases and enhancing the utility of climate predictions for climate‐sensitive applications. Season‐Net offers a promising pathway for advancing seasonal forecast postprocessing with high accuracy and impact relevance.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.001 | 0.001 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.001 | 0.001 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.001 |
| Research integrity | 0.001 | 0.002 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".