Bias Correction and Multiensemble in the NAEFS Context or How to Get a “Free Calibration” through a Multiensemble Approach
Bibliographic record
Abstract
Abstract Previous studies have shown that the raw combination (i.e., the combination of the direct output model without any postprocessing procedure) of the National Centers for Environmental Prediction (NCEP) and Meteorological Service of Canada (MSC) ensemble prediction systems (EPS) improves the probabilistic forecast both in terms of reliability and resolution. This combination palliates the lack of reliability of the NCEP EPS because of the too small dispersion of the predicted ensemble and the lack of probabilistic resolution of the MSC EPS. Such a multiensemble, called the North American Ensemble Forecast System (NAEFS), especially shows bias reductions and dispersion improvements that could only come from the combination of different forecast errors. It is then legitimate to wonder whether these improvements in terms of biases and dispersions, and by extension the skill improvements, are only due to the balancing between opposite model errors. In the NAEFS framework, bias corrections “on the fly,” where the bias is updated over time, are applied to the operational EPSs. Each model of the EPS components (NCEP/MSC) is individually bias corrected against its own analysis with the same process. The bias correction improves the reliability of each EPS component. It also slightly improves the accuracy of the predicted ensembles and thus the probabilistic resolution of the forecasts. Once the EPSs are combined, the improvements due to the bias correction are not so obvious, tending to show that the success of the multiensemble method does not only come from the cancellation of different biases. This study also shows that the combination of the raw EPS components (NAEFS) is generally better than either the bias corrected NCEP or MSC ensembles.
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 distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.001 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.001 | 0.000 |
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 teacher head, 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".