Efficient Fine-Tuning of 37-Level GraphCast with the Canadian Global Deterministic Analysis
Bibliographic record
Abstract
Abstract This work describes a process for efficiently fine-tuning the GraphCast data-driven numerical weather prediction model to be consistent with another analysis system, here the Global Deterministic Prediction System (GDPS) of Environment and Climate Change Canada (ECCC). The GDPS system was significantly upgraded in 2019, so there is only a limited period of operational data to use for model training. This work considers the effect of using 2 years of training data (July 2019–December 2021) and a restricted computational budget to tune the 37-level, quarter-degree version of GraphCast with an empirically determined vertical weighting of model error. The GDPS-tuned model significantly outperforms both the operational (traditional) forecast and the unmodified GraphCast model when initialized with the GDPS analysis, showing significant forecast skill in the troposphere over 1- to 10-day lead times. This fine-tuning is accomplished through an abbreviation of the original training curriculum, relying on a shorter single-step forecast stage to complete most of the adaptation, followed by a consolidation of the forecast-lengthening stages into separate 12-h, 1-day, 2-day, and 3-day stages. In addition, a “control” run trained with ERA5 data shows that fine-tuning on recent data improves forecast skill on a going-forward basis. Significance Statement This work represents the first published guideline for the adaptation of the full, 37-level, quarter-degree GraphCast model to another analysis system. Without this fine-tuning, the accuracy of GraphCast forecasts suffers when it is fed by initial conditions from analysis systems that have systematic differences compared to GraphCast’s training data, and fine-tuning is essential to make the best operational use of GraphCast.
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 | 0.002 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.000 | 0.002 |
| Science and technology studies | 0.001 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.002 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 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".