Seismic Event Classification Eval

---
name: seismic-event-classification-eval
description: Evaluates a model's ability to discriminate between three types of seismic events (earthquakes, quarry blasts, and background noise) using waveform and spectral features. It probes the model's capacity to learn physically meaningful seismological signatures like P/S-wave onsets and spectral decay patterns. Use when the user wants to benchmark on Curated Seismic Dataset, or asks about evaluating this task. Reports Accuracy.
metadata:
  skill_kind: dataset_eval
  source_arxiv: 2602.15993
  bibtex_key: srivastava2026picrnn
  confidence: high
---

# seismic-event-classification-eval

> An Interpretable Physics Informed Multi-Stream Deep Learning Architecture for the Discrimination between Earthquake, Quarry Blast and Noise — Srivastava et al. (2026) (arXiv:2602.15993, 2026)

## What this evaluates

Evaluates a model's ability to discriminate between three types of seismic events (earthquakes, quarry blasts, and background noise) using waveform and spectral features. It probes the model's capacity to learn physically meaningful seismological signatures like P/S-wave onsets and spectral decay patterns.

## Datasets

- **Curated Seismic Dataset** — total ?; splits: test (900); repo https://github.com/srivastavaresearchgroup/PICRNN

## Metrics

- `Accuracy` **(primary)** — range: [0, 1]
  - Proportion of correctly classified instances out of the total test set. Calculated as (True Positives + True Negatives) / Total Instances.

## Input / output format

**Input**: Raw seismic waveform time-series data, optionally processed into multi-resolution log-mel spectrograms and physical attributes (total instantaneous energy, Vertical-to-Horizontal Ratio).

**Output**: Discrete class label: 'EQ' (Earthquake), 'Blast' (Quarry Blast), or 'Noise'.

## Scoring recipe

```python
def compute_accuracy(predictions, gold_labels):
    correct = sum(1 for p, g in zip(predictions, gold_labels) if p == g)
    return correct / len(gold_labels)
```

## Common pitfalls

- Low signal-to-noise ratio (SNR) events in the 10–15 dB range are highly prone to ambiguity and misclassification.
- The model's reliance on specific physical inductive biases (e.g., P/S-wave onsets, spectral decay) may limit generalization to datasets with different recording geometries or noise characteristics.
- Small test set size (300 per class) may lead to high variance in per-class F1 scores.

## Evidence (verbatim from paper)

> The model achieves high classification performance, as evidenced by the F1 scores in Table [1] and confusion matrix in Figure [3]. The model successfully distinguishes blasts from local earthquakes thus suggesting that the model learned the impulsive P-wave onset and rapid coda decay (typical behavior of blasts) effectively. Furthermore, the model successfully discriminates all the noise data with near perfect precision.
| Accuracy | | | 0.9756 | 900 |

## Citation

```bibtex
@misc{srivastava2026picrnn,
  title={An Interpretable Physics Informed Multi-Stream Deep Learning Architecture for the Discrimination between Earthquake, Quarry Blast and Noise},
  author={Srivastava et al. (2026)},
  year={2026},
  note={arXiv:2602.15993}
}
```

- arXiv: 2602.15993