GreenPhase: A Green Learning Approach for Earthquake Phase Picking

Yixing Wu; Shiou-Ya Wang; Dingyi Nie; Sanket Kumbhar; Yun-Tung Hsieh; Yun-Cheng Wang; Po-Chyi Su; C. -C. Jay Kuo

physics.geo-phFeb 23, 2026

GreenPhase: A Green Learning Approach for Earthquake Phase Picking

Authors: Yixing Wu, Shiou-Ya Wang, Dingyi Nie, Sanket Kumbhar, Yun-Tung Hsieh, Yun-Cheng Wang, Po-Chyi Su, C. -C. Jay Kuo

Organizations: Ming Hsieh Department of Electrical and Computer Engineering, University of Southern California · Center for Earth and Environmental Studies, National Central University · Department of Computer Science & Information Engineering, National Central University

Abstract

Earthquake detection and seismic phase picking are fundamental yet challenging tasks in seismology due to low signal-to-noise ratios, waveform variability, and overlapping events. Recent deep-learning models achieve strong results but rely on large datasets and heavy backpropagation training, raising concerns over efficiency, interpretability, and sustainability. We propose GreenPhase, a multi-resolution, feed-forward, and mathematically interpretable model based on the Green Learning framework. GreenPhase comprises three resolution levels, each integrating unsupervised representation learning, supervised feature learning, and decision learning. Its feed-forward design eliminates backpropagation, enabling independent module optimization with stable training and clear interpretability. Predictions are refined from coarse to fine resolutions while computation is restricted to candidate regions. On the Stanford Earthquake Dataset (STEAD), GreenPhase achieves excellent performance with F1 scores of 1.0 for detection, 0.98 for P-wave picking, and 0.96 for S-wave picking. This is accomplished while reducing the computational cost (FLOPs) for inference by approximately 83% compared to state-of-the-art models. These results demonstrate that the proposed model provides an efficient, interpretable, and sustainable alternative for large-scale seismic monitoring.

Figures & tables

Figure 1: Data preprocessing workflow prior to GreenPhase training.

Figure 2: Comparison of a seismic waveform before and after the preprocessing pipeline.

Figure 3: The architecture of P/S-wave picking

Figure 4: An example of the pseudo-label generation for a Level 1 waveform (1500-sample resolution). The top panel shows the transformed waveform with the P-arrival marked. The bottom panel displays the corresponding continuous pseudo-label for each time location.

Figure 5: The distribution of pseudo-label values for the Level 1 training set after stratified sampling. The histogram confirms that the three categories (noise, intermediate, and high-confidence) are equally represented, resulting in a balanced dataset for model training.

Model	Pr	Re	F1	Training size	Ref.
GreenPhase (Ours)	1.0	1.0	1.0	1.2M	This study
EQTransformer	1.0	1.0	1.0	1.2M	Mousavi et al. (2020)
CRED	1.0	0.96	0.98	1.2M	Mousavi et al. (2019b)
DetNet	1.0	0.89	0.94	30K	Zhou et al. (2019)
Yews	0.84	0.85	0.85	1.4M	Zhu et al. (2019)
STA/LTA	0.91	1.0	0.95	—	Allen (1978)

Table 1: Detection performance.

Model	Pr	Re	F1	Training size	Ref.
GreenPhase (Ours)	0.96	0.99	0.98	1.2M	This study
GreenPhase (Ours)	0.96	0.99	0.98	240K	This study
GreenPhase (Ours)	0.95	0.99	0.97	60K	This study
EQTransformer	0.99	0.99	0.99	1.2M	Mousavi et al. (2020)
PhaseNet	0.96	0.96	0.96	780K	Zhu and Beroza (2019)
GPD	0.81	0.80	0.81	4.5M	Ross et al. (2018)

Table 2: P-phase picking performance.

Model	Pr	Re	F1	Training size	Ref.
GreenPhase (Ours)	0.93	0.99	0.96	1.2M	This study
GreenPhase (Ours)	0.93	0.99	0.96	240K	This study
GreenPhase (Ours)	0.92	0.99	0.95	60K	This study
EQTransformer	0.99	0.96	0.98	1.2M	Mousavi et al. (2020)
PhaseNet	0.96	0.93	0.94	780K	Zhu and Beroza (2019)
GPD	0.81	0.83	0.82	4.5M	Ross et al. (2018)

Table 3: S-phase picking performance.

Model	Training Size	Model Size	FLOPs
GreenPhase (Ours)	60K	93K ( $\times$ 1.0)	4M ( $\times$ 1.0)
GreenPhase (Ours)	240K	234K (2.5 $\times$ )	12M (3.0 $\times$ )
GreenPhase (Ours)	1.2M	381K (4.1 $\times$ )	22M (5.5 $\times$ )
EQTransformer	1.2M	377K (4.1 $\times$ )	129M (32.3 $\times$ )

Table 4: Comparison of model size and testing FLOPs.

Model	Training Size	CO 2 e (g)	Energy (Wh)
GreenPhase (Ours)	60K	10.30 (1.0 $\times$ )	24.28 (1.0 $\times$ )
GreenPhase (Ours)	240K	11.44 ( $\sim$ 1.1 $\times$ )	26.98 ( $\sim$ 1.1 $\times$ )
GreenPhase (Ours)	1.2M	20.59 ( $\sim$ 2.0 $\times$ )	48.57 ( $\sim$ 2.0 $\times$ )
EQTransformer	1.2M	31.11 ( $\sim$ 3.0 $\times$ )	61.13 ( $\sim$ 2.5 $\times$ )

Table 5: Comparison of carbon footprint and energy consumption during inference on over 120,000 test waveforms.

Model	Training Size	CO 2 e (kg)	Energy (kWh)
GreenPhase (Ours)	60K	0.189 (1.0 $\times$ )	0.445 (1.0 $\times$ )
GreenPhase (Ours)	240K	0.747 (4.0 $\times$ )	1.76 (4.0 $\times$ )
GreenPhase (Ours)	1.2M	1.80 (9.5 $\times$ )	4.25 (9.6 $\times$ )
EQTransformer	1.2M	72.04 (381 $\times$ )	169.92 (382 $\times$ )

Table 6: Comparison of carbon footprint and energy consumption in model training.

Figure 6: Analysis of the RFT and SFG feature engineering steps for the Level 3 P-wave model. (a, b) The sorted RFT loss on the training and validation sets, respectively, shows that only a small fraction of features before the ”elbow” are highly discriminative. (c) The high correlation of feature ranks between the training and validation sets confirms the stability of the RFT selection process. (d) The RFT loss of the new SFG-generated features is substantially lower than that of the original RFT-selected features, demonstrating the effectiveness of SFG.

Figure 7: Prediction results at the coarsest resolution (Level 3, 375 samples).

Figure 8: Refined prediction results at the intermediate resolution (Level 2, 750 samples).

Figure 9: Final prediction results at the finest resolution (Level 1, 1500 samples).

Seismology

Seismology studies earthquakes and seismic waves to understand Earth’s structure, monitor faults, and assess geological and infrastructure hazards.

See all 43 papers

Evaluating PhaseNet on Teleseismic Data with MsPASS
May 10, 2026Jinxin Ma, Yinzhi Wang, Gary L. Pavlis +1Efficient Neural Network Inference Seismology
Texas Advanced Computing Center, The University of Texas at Austin · Department of Earth and Atmospheric Sciences, Indiana University, Bloomington, IN 47405
Real-Time Earthquake Magnitude Classification from Initial P-Waves: Models, Dataset, and Comparative Analysis for South Asia
May 10, 2026Md Nasiat Hasan Fahim, Md. Abid Ullah Muhib, Rayhanul Amin Tanvir +1Time Series Classification Seismology
Department of Computer Science and Engineering Shahjalal University of Science and Technology Sylhet, Bangladesh
Learning Earthquake Wave Arrival Time Picking from Labels with Inaccuracies
Jun 13, 2026Sen Li, Xu Yang, S. Mostafa Mousavi +5Contrastive Learning Seismology
Department of Earth and Planetary Sciences, Harvard University, Cambridge, MA, USA. · School of Computer Science and Technology, China University of Mining and Technology, Xuzhou, Jiangsu, China · School of Mines, China University of Mining and Technology, Xuzhou, Jiangsu, China +1

May 10, 2026J/K move · Enter open · S save

Evaluating PhaseNet on Teleseismic Data with MsPASS

Jinxin Ma, Yinzhi Wang, Gary L. Pavlis, Chenbo Yin

Texas Advanced Computing Center, The University of Texas at Austin · Department of Earth and Atmospheric Sciences, Indiana University, Bloomington, IN 47405

Efficient Neural Network Inference Seismology

Open paper AskPDF arXiv

Numerous studies have shown that the machine-learning picker PhaseNet produces accurate P and S picks on local earthquake signals, but its performance can degrade sharply on teleseismic signals. To address this limitation, we present a reproducible MsPASS workflow that (i) enables scalable data preparation and management for large seismic archives and (ii) supports standardized PhaseNet training and inference. We assembled a control dataset of 1.6 million waveforms linked to teleseismic P-wave picks made by analysts at the USArray Array Network Facility (ANF). The control dataset confirms that the PhaseNet model trained on regional signals performs poorly on these data. We then trained PhaseNet from scratch on the training split of the ANF control dataset and evaluated it on a non-overlapping held-out test split, increasing P-pick recall by 741.5% and yielding 683.9% more picks within a 0.1s residual window. We also evaluated PhaseNet across different model sizes on both CPUs and GPUs. Increasing the model size by about 120 times improved precision and recall by 15.6% and 23.2%, respectively. However, the scaled model reduced inference throughput by 87.2% on an NVIDIA A100 GPU and by 97.3% on a 128-core high-performance CPU node. These results indicate that scaling PhaseNet is more practical on GPUs than on CPUs, and that simply enlarging the model is not an efficient way to achieve large accuracy gains.

Abstract

Seismology

Explore similar work

Evaluating PhaseNet on Teleseismic Data with MsPASS

Real-Time Earthquake Magnitude Classification from Initial P-Waves: Models, Dataset, and Comparative Analysis for South Asia

Learning Earthquake Wave Arrival Time Picking from Labels with Inaccuracies