| name | glm-metrics |
| description | How to calculate and save GLM performance metrics (RMSE) to metrics.json. Use this whenever the user asks for RMSE checks or final model evaluation. |
GLM Performance Metrics
Metrics to Compute
Calculate the following metrics to evaluate the model performance:
overall_rmse: Total RMSE across all matched depth/time pairs.
annual_deep_rmse: RMSE for depths >= 13 m.
summer_deep_rmse: RMSE for months June-September, depths >= 13 m.
overall_n_pairs: Number of total matches.
annual_deep_n_pairs: Number of deep matches.
summer_deep_n_pairs: Number of summer deep matches.
Requirements
- Data: Match
/root/field_temp_oxy.csv with simulation output /root/output/output.nc.
- Matching method: Exact
datetime + rounded-depth merge.
- Do NOT use nearest-time matching, interpolation, or alternative binning.
Output
Save the metrics as a JSON file at /root/metrics.json with keys named exactly as above.