Initially trained on: 2020-03-17

This is the latest version of the model. Older versions were used in 2018 and 2019.

Source: United States Geological Survey (USGS)

Training Set: 1973-01-01 to 2017-12-31

Test Set: 2018-01-01 to 2018-12-31

Update Set: 2019-01-01 to 2020-03-16

Seed Set: Last 1 Day (Previous 24 Hours)

Temporal Resolution: 1 Day (24 Hours)

Spatial Resolution: 2 Degrees Latitude and 2 Degrees Longitude

Details: Earthquakes were reorganized into tensors, with each slice representing a single day and consisting of a matrix with the rows and columns representing approximate latitudes and longitudes. Each position in the matrix was filled with a magnitude determined by converting all the earthquakes occurring at that approximate location and time from logarithmic scale magnitudes into linear energies, summing them together, and then converting the resulting value back into the logarithmic scale. Some preprocessing is required to convert between different types of magnitude scales as well. This model is designed for the moment magnitudes (Mw). A baseline magnitude of -1.0 is assumed for areas where no event takes place.

Long Short Term Memory (LSTM) Recurrent Neural Network (RNN)

Hidden Layers: 5

Neurons Per Hidden Layer: 512 nodes

Timesteps: 1

Epochs of Training: 1

Loss Function: Asymmetric (Exponential Or Logarithmic)

Library: Keras in Tensorflow 2.1

Notes: Model is fully Stateful and utilizes an Online Learning schedule. Utilizes a dual path architecture that combines dense connectivity with residual skip connections.

- -1.0--0.5: 3204425
- -0.5-0.0: 92150
- 0.0-0.5: 44708
- 0.5-1.0: 25858
- 1.0-1.5: 19453
- 1.5-2.0: 17694
- 2.0-2.5: 19633
- 2.5-3.0: 19833
- 3.0-3.5: 17503
- 3.5-4.0: 13259
- 4.0-4.5: 10300
- 4.5-5.0: 5650
- 5.0-5.5: 1034
- 5.5-6.0: 138
- 6.0-6.5: 4
- 6.5-7.0: 0
- 7.0-7.5: 0
- 7.5-8.0: 0
- 8.0-8.5: 0
- 8.5-9.0: 0
- 9.0-9.5: 0
- 9.5-10.0: 0

- A True Positive is scored when the magnitude prediction is within +/-1.0 of the actual earthquake event magnitude, and the actual earthquake event magnitude is greater than 0.0.
- A True Negative is scored when the magnitude prediction is within +/-1.0 of the actual earthquake event magnitude, and the actual earthquake event magnitude is less than 0.0.
- A False Positive is scored when the magnitude prediction is outside of +/-1.0 of the actual earthquake event magnitude, and the actual earthquake event magnitude is less than the predicted magnitude.
- A False Negative is scored when the magnitude prediction is outside of +/-1.0 of the actual earthquake event magnitude, and the actual earthquake event magnitude is greater than the predicted magnitude.

- True Positive (TP) = 23415.0
- True Negative (TN) = 3293568.0
- False Positive (FP) = 164801.0
- False Negative (FN) = 11082.0
- Total Population (TP + TN + FP + FN) = 3492866.0
- Positive (TP + FN) = 34497.0
- Negative (FP + TN) = 3458369.0
- Predict Positive (TP) = 188216.0
- Predict Negative (TP) = 3304650.0
- Prevalence (PRE = positive / total_pop) = 0.009876416673299233
- True Positive Rate OR Sensitivity OR Recall (TPR = TP / positive) = 0.6787546743195061
- False Negative Rate OR Miss Rate (FNR = FN / positive) = 0.32124532568049396
- False Positive Rate OR Fall-Out (FPR = FP / negative) = 0.0476528097493356
- True Negative Rate OR Specificity (TNR = TN / negative) = 0.9523471902506644
- Accuracy (ACC = (TP + TN) / total_pop) = 0.949645076564632
- Positive Predictive Value OR Precision (PPV = TP / predict_positive) = 0.12440493900624815
- False Discovery Rate (FDR = FP / predict_positive) = 0.8755950609937518
- False Omission Rate (FOR = FN / predict_negative) = 0.0033534564931233263
- Negative Predictive Value (NPV = TN / predict_negative) = 0.9966465435068766
- Positive Likelihood Ratio (PLR = TPR / FPR) = 14.243749275014569
- Negative Likelihood Ratio (NLR = FNR / TNR) = 0.337319550022445
- Diagnostic Odds Ratio (DOR = PLR / NLR) = 42.2262785363813
- Jaccard Index (IoU) (J = TP / (TP + FP + FN)) = 0.11748738070627904
- F1-Score = 0.21027061734160107
- Matthews Correlation Coefficient = 0.2763979311120408
- Informedness = 0.6311018645701705
- Markedness = 0.12105148251312481
- Balanced Accuracy = 0.8155509322850852

- True Positive (TP) = 578.0
- True Negative (TN) = 0.0
- False Positive (FP) = 829.0
- False Negative (FN) = 1678.0
- Total Population (TP + TN + FP + FN) = 3085.0
- Positive (TP + FN) = 2256.0
- Negative (FP + TN) = 829.0
- Predict Positive (TP) = 1407.0
- Predict Negative (TP) = 1678.0
- Prevalence (PRE = positive / total_pop) = 0.7312803889789303
- True Positive Rate OR Sensitivity OR Recall (TPR = TP / positive) = 0.25620567375886527
- False Negative Rate OR Miss Rate (FNR = FN / positive) = 0.7437943262411347
- False Positive Rate OR Fall-Out (FPR = FP / negative) = 1.0
- True Negative Rate OR Specificity (TNR = TN / negative) = 0.0
- Accuracy (ACC = (TP + TN) / total_pop) = 0.1873581847649919
- Positive Predictive Value OR Precision (PPV = TP / predict_positive) = 0.4108031272210377
- False Discovery Rate (FDR = FP / predict_positive) = 0.5891968727789624
- False Omission Rate (FOR = FN / predict_negative) = 1.0
- Negative Predictive Value (NPV = TN / predict_negative) = 0.0
- Positive Likelihood Ratio (PLR = TPR / FPR) = 0.25620567375886527
- Negative Likelihood Ratio (NLR = FNR / TNR) = 0
- Diagnostic Odds Ratio (DOR = PLR / NLR) = 0
- Jaccard Index (IoU) (J = TP / (TP + FP + FN)) = 0.1873581847649919
- F1-Score = 0.3155883155883156
- Matthews Correlation Coefficient = 0
- Informedness = -0.7437943262411347
- Markedness = -0.5891968727789623
- Balanced Accuracy = 0.12810283687943264