Activity regimes inferred from automatic classification of volcanic tremor at Mt. Etna, Italy
20
0
0
Testo completo
(2) Mt. Etna Mount Etna is the largest active volcano in Europe: Type: Basaltic stratovolcano Location: Sicily, Italy (3350 m a.s.l.) Latest eruptions: 01, 02-03, 04-05, 06 Mount Etna’s volcanic monitoring is a key issue.
(3) Volcanic Tremor For basaltic volcanoes (e.g., Mount Etna)… . Volcanic tremor is a persistent seismic signal marking different states of the volcano’s activity:. Pre-eruptive Lava fountain Eruptive . Post-eruptive. Volcanic tremor provides reliable information for alerting governmental authorities during a crisis and permits surveillance even when direct access to the eruptive theatre is not possible.
(4) Automatic Classification In [Masotti, Geo. Res. Lett., 33 (20) (2006)], a system able to automatically classify different states of the volcano’s activity from the analysis of its volcanic tremor was proposed: Data: Volcanic tremor. Features: Spectrogram-based. Classification: Support Vector Machine. Here, what if: Support Vector Machine is optimized (e.g., using Genetic Algorithm)? A different (optimized) classifier is used (e.g., Artificial Neural Networks or Cluster Analysis)? A different eruption is considered (e.g., 2006)?.
(5) Data Seismograms are labeled according to their recording date… Lava fountain (FON) seismograms Jul 01, 01. Jul 16, 01. Aug 08, 01. Aug 15, 01. Jul 01, 06. Jul 15, 06. Jul 28, 06. Aug 12, 06. Pre-eruptive (PRE) seismograms. Eruptive (ERU) seismograms. Post-eruptive (POS) seismograms. Number of patterns for each class PRE. FON. ERU. POS. 2001. 153. 55. 180. 37. 2006. 84. 0. 168. 84.
(6) Feature Extraction Features are computed by… 1. Calculating the spectrogram of each seismogram (10 min., 0-15 Hz) 2. Averaging the rows of each spectrogram (62-dimensional feature vector). Pre-eruptive Lava fountain. Eruptive. Post-eruptive.
(7) Classification A comparison is performed among:. Supervised classification based on Support Vector Machine (SVM) + Genetic Algorithm (GA) VS Supervised classification based on Artificial Neural Network (ANN) + GA Unsupervised classification based on Cluster Analysis (CA).
(8) Classification :: SVM Supervised classification.
(9) Classification :: ANN Supervised classification.
(10) Classification :: CA Unsupervised classification.
(11) Classification :: GA Some of the SVM and ANN parameters are tuned using GA:.
(12) Results :: SVM + GA. 2001. Overall classification error: 22/425 = 5% patterns. 2006. Overall classification error: 22/336 = 7% patterns.
(13) Results :: SVM + GA. 2001 + 2006. Overall classification error: 68/761 = 9% patterns.
(14) Results :: ANN + GA. 2001. Overall classification error: 124/425 = 29% patterns. 2006. Overall classification error: 64/336 = 19% patterns.
(15) Results :: ANN + GA. 2001 + 2006. Overall classification error: 196/761 = 26% patterns.
(16) Results :: CA. 2001. Number of clusters: 5. 2006. Number of clusters: 2.
(17) Results :: CA. 2001 + 2006. Number of clusters: 4.
(18) Conclusions The improvement achieved using SVM+GA rather than SVM is not significant, i.e., < 1% SVM+GA performs significantly better than ANN+GA, i.e., overall classification error is equal to 5% on 2001 and 7% on 2006, versus 29% on 2001 and 19% on 2006, respectively Individually, SVM+GA and ANN+GA achieve quite similar classification results regardless of the data considered, i.e., 2001, 2006, or 2001+2006 CA: separation of data is quite close to what expected.
(19) Advertisement… The translation of the SVMbased system (TREMOrEC) from Matlab to Visual C++, to make it available to our collaboration and scientific community for validation, is completed. For more information, and to see a demo of TREMOrEC, join us today at poster A-05120!.
(20) Thank you.
(21)
Documenti correlati
Le considerazioni ed i numeri analizzati finora rendono ancora fortemente obiettabili le affermazioni circa l'esistenza di una vera e propria selezione dei casi effettuata in
The scenario is the opposite if we consider the chimneystack on the downstream position on the pitched roof: for the smallest height h/H = 0.25 (Figure 15) the plume
We first discuss a prior summability condition for the posterior to accumulate around the densities in the model closest in the Kullback– Leibler sense to the data generating
We also performed a series of controlled potential electrolysis with different Brønsted acids, showing that their strengths can drive the selectivity of the catalyst towards
I am from Jamaica where the official language is Jamaican English, a respected language, and the popular language is Jamaican Creole, commonly called Patwa, the
Conceptual Integration or Blending (see Fauconnier and Turner, “Conceptual Inte- gration Networks” and The Way We Think) makes use of Fauconnier’s theory of Mental Spaces,
In una simile fluvialità (e varietà di interessi: Vollmann ha scritto anche di Af- ghanistan, teatro Nō, identità femminile – come già Duchamp con Rrose Selavy, anch’egli ha un
Tornando ora alla questione della narratività dei primi libri manganelliani, sono convinto che uno dei fattori che a lungo ha contribuito ad un’iscrizione quasi automatica dei
