ILLUSTRATIVE · NASA / SVS · NOT ADITYA-L1 DATA
Model cardsclimatology
Model card · reference point
Climatology
Predicts the historical base rate, ignoring the observation.
What is this?
A model card for Climatology, one of eight detectors evaluated on the frozen Aditya-L1 flare benchmark.
Why should I trust it?
Every figure below is read at build time from the committed benchmark artifact; none is transcribed. The judgements are traceable to the written evaluation protocol and the dataset limitations record.
Where can I verify it?
- The validation evidenceComputed from 192,541 held-out predictions
- Dataset limitationsL-1 … L-10
- Evidence traceabilityclaim → artifact → commit
- The raw payloadfindings/models.json
Section 1
Purpose and intended use
Why it is in the benchmark. Separates 'knows about flares in general' from 'knows about this minute'.
Intended use. Reference point only.
Out of scope. Everything else.
Section 2
Inputs, outputs, training
Inputs. The training-period base rate.
Outputs. A constant probability.
Training protocol. Estimated on the training period only.
Operating point. Decision threshold 0.5000, selected on the training period and then frozen.
Section 3
Evaluation
Both tasks, both reported.
M/X NOWCAST · 192,541 held-out minutes · base rate 1.24%
| Metric | Value | What it measures |
|---|---|---|
| ROC-AUC | 0.500 | Ranking quality across all thresholds |
| PR-AUC | 0.012 | The informative curve at a low base rate |
| Precision | 0.000 | Of the minutes flagged, how many were flares |
| Recall | 0.000 | Of the flare minutes, how many were flagged |
| F1 | 0.000 | Harmonic mean of the two above |
| MCC | 0.000 | Correlation, robust to imbalance |
| Brier | 0.013 | Probability accuracy; undefined for hard classifiers |
| Balanced accuracy | 0.500 | Mean of the two class recalls |
- Events detected
- 0 / 82
- Event recall 0.000 · 95% CI 0.000–0.000
- False alarm runs
- 0
- Of 0 alarm runs raised. A run, not a minute — the unit an operator actually responds to.
- ROC-AUC interval
- 0.500–0.500
- 95%, day-block bootstrap.
M/X 30-MIN PREDICTION · 192,541 held-out minutes · base rate 1.29%
| Metric | Value | What it measures |
|---|---|---|
| ROC-AUC | 0.500 | Ranking quality across all thresholds |
| PR-AUC | 0.013 | The informative curve at a low base rate |
| Precision | 0.000 | Of the minutes flagged, how many were flares |
| Recall | 0.000 | Of the flare minutes, how many were flagged |
| F1 | 0.000 | Harmonic mean of the two above |
| MCC | 0.000 | Correlation, robust to imbalance |
| Brier | 0.013 | Probability accuracy; undefined for hard classifiers |
| Balanced accuracy | 0.500 | Mean of the two class recalls |
- Events detected
- 0 / 82
- Event recall 0.000 · 95% CI 0.000–0.000
- False alarm runs
- 0
- Of 0 alarm runs raised. A run, not a minute — the unit an operator actually responds to.
- ROC-AUC interval
- 0.500–0.500
- 95%, day-block bootstrap.
Confusion matrix · M/X NOWCAST
True negative
190,148
False positive
0
False negative
2,393
True positive
0
Section 4
Strengths and failure modes
Strengths
- A well-calibrated model that has no discriminative skill whatsoever.
Failure modes
- Cannot distinguish any minute from any other.
Section 5
Limitations and ethical considerations
Ethical considerations. None — it is a ruler, not a detector.
Governing limitation clauses
Clause text is not restated here. It is published in full on the dataset card, from the frozen DATASET_LIMITATIONS_FOR_ML.md report, so there is exactly one copy of every caveat.