Earthquake Forecast Verification for Northwestern Turkey

Yıl 2020, , 118 - 127, 30.06.2020

Hakan Karaca

https://doi.org/10.7240/jeps.526906

Öz

Seismic events have a pattern of recurrence in magnitude,
time and space. Considerable effort is being spent to identify seismic patterns
and successfully predict future earthquakes by using the recognized patterns.
As a result of these intensive efforts, a variety of methods has been proposed.
As the knowledge and experience in the field accumulated in parallel to the
variety of the methods proposed, it was deemed necessary to test the
performance of some of the highlighted methods, especially considering the wide
reception of methods utilizing spatially smoothed seismicity (SSS), pattern
informatics (PI) and Relative Intensity (RI). The performance of these methods
in the prediction of future earthquakes has been selected for investigation.

The investigated area is the region bounded by 27⁰-33⁰ in longitudes
and 39.8⁰-41⁰
in latitudes, well known for the North Anatolian Fault.
The period of coverage has been selected such as to maximize the length with
the minimum magnitude of completeness. As a result of such optimization, the
period from 1973 to 2019 has been selected with minimum magnitude of
completeness being determined as 3.8. In order to measure the relative
performance of the methods, relative operating characteristic (ROC) analysis has
been utilized. The method based on SSS has been adapted to the related ROC
procedures, while the results of PI and RI methods are already suitable for the
evaluation by ROC procedures.   

After the analysis was completed, according to the ROC
procedures, none of the methods were singled out in forecast performance. However,
when the ratio of hits versus total alarms and the area covered by the alarms,
PI method outperforms two other methods by its efficiency. 

Anahtar Kelimeler

Pattern informatics, relative intensity, earthquake forecast

Kuzeybatı Türkiye için Deprem Tahminleri Doğrulama Çalışması

Öz

Sismik hareketlerin büyüklük, zaman ve oluşum yerleri
bağlamında, tekrarlama desenlerine sahip olduğu varsayılır. Sismik desenlerin
belirlenerek geleceğe yönelik tahminlerin başarılı biçimde yapılabilmesi
uğrunda dikkate değer ölçüde çaba harcanmıştır. Bu çabaların sonucu olarak,
birçok yöntem geliştirilmiştir. Geliştirilen yöntemlerin çeşitliliği bağlamında
bilgi ve deneyim artışı süregelmiş ve özellikle çoğunlukla kabul gören
yöntemlerden, düzleştirmeye dayalı yöntemler (spatially-smoothed seismicity), desen bilgi (pattern informatics) ve
göreli yoğunluk (relative
intensity) yöntemlerinin yeni verilerle denenmesi gerekli
olmuştur. Bahsi geçen yöntemler, deprem tahmin performansları açısından
incelenmek üzere seçilmiştir.

İnceleme alanı 27⁰-33⁰ boylam ve
39.8⁰-41⁰ enlemleri arasında kalan, Kuzey Anadolu Fay
Hattı ile ünlü bölge olarak belirlenmiştir. Geçmiş verilerin seçilmesi
sırasında tamlık ölçütlerini sağlayan ve aynı zamanda en uzun dönemi kapsayacak
şekilde belirlenmiştir. Bu kapsamda yapılan analiz sonucuna göre, tamlık
ölçütünü sağlayan en küçük deprem büyüklüğü 3.8 olarak bulunmuş, tamlık
ölçütlerine uyan dönem ise 1973 ile 2019 yılları arasında kalan dönem olarak belirlenmiştir.
Tahmin yöntemlerinin başarısı ise göreli işletim ölçütü (ROC) analizi
kullanılarak değerlendirilmiştir. Düzleştirme yöntemi uygulamaları ROC veri
girdisi formatına göre uyarlanmış olup, PI ev RI yöntemleri is halihazırda ROC
girdilerine uygun olarak veri üretmekte olduğundan herhangi bir uyarlamaya
gerek kalmamıştır.

Çözümlemeler sonucunda, ROC değerlendirmeleri sonucunda
yöntemlerden hiçbirisi öne çıkmamış ancak, başarılı tahminlerin toplam
tahminlere ve tahminlerin kapladıkları alanlara göre değerlendirmesi sonucunda
PI yönteminin diğer iki yönteme göre daha verimli bir yöntem olduğu
belirlenmiştir.

Anahtar Kelimeler

Örüntü bilişim, göreli yoğunluk, deprem tahminleri, kuzey anadolu fay hattı

Kaynakça

• Akkar S, Cağnan, Z, Yenier E, Erdogan E, Sandıkkaya MA, Gülkan P (2010). The Recently Compiled Turkish Strong-Motion Database: Preliminary Investigation for Seismological Parameters, J of Seismol 14, 457-479.
• Choi E, Hall P (1999). Nonparametric Approach to The Analysis of Spacetime Data On Earthquake Occurrences, J Comput Graph Stat 8,733–748.
• Field EH (2007). Special Issue: Regional Earthquake Likelihood Models. Seismological Research Letters Vol. 78, No.1
• Helmstetter A, Werner JM (2014). Adaptive Smoothing of Seismicity in Time, Space and Magnitude For Time-Dependent Earthquake Forecasts for California, Bull Seism Soc Am 104:2, 809-822.
• Helmstetter A, Kagan YY, Jackson DD (2006). Comparison of Short-Term and Time-Independent Earthquake Forecast Models for Southern California, Bull Seism Soc Am 96(1), 90-106.
• Helmstetter A, Kagan YY, Jackson DD (2007). High-Resolution Time-Independent Grid-Based Forecast for M ≥ 5 Earthquakes In California, Seismol Res Lett 78:1,78–86.
• Hiemer S, Jackson DD, Wand Q (2013). A Stochastic Forecast of California Earthquakes Based On Fault Slip and Smoothed Seismicity, Bull Seism Soc Am, 103:2A, 799-810
• Holliday J, Chen C, Tiampo KF, Rundle JB, Turcotte DL. (2007). A RELM Earthquake Forecast Based On Pattern Informatics, Seismol Res Lett. 78:87–93.
• Holliday JR, Nanjo KZ, Tiampo KF, Rundle JB, Turcotte DL. (2005). Earthquake Forecasting and Its Verification. Nonlinear Processes Geophys. 12:965–977.
• Holliday JR, Rundle JB, Tiampo KF, Klein W, Donnellan A. (2006b) Modification of The Pattern Informatics Method for Forecasting Large Earthquake Events Using Complex Eigenvectors. Tectonophysics. 413(1–2):87–91.
• Kagan YY, Jackson D (1994). Long-Term Probabilistic Forecasting of Earthquakes. J Geophys Res. 99(B7):13685–13700.
• Kagan YY, Knopoff L (1977). Earthquake Risk Prediction as a Stochastic Process, Phys Earth Planet in 14:2, 97–108.
• Mohanty WK, Mohapatra AK, Verma AK, Tiampo KF, Kislay K (2016) Earthquake Forecasting and Its Verification in Northeast India, Geomat Nat Haz Risk, 7:1, 194-214, DOI: 10.1080/19475705.2014.883441
• Schorlemmer D, Zechar JD,Werner and The RELM Working Group. (2010) First Results of the Regional Earthquake Likelihood Models Experiment, Pure Appl Geophys, 167, 859-876
• Stock C, Smith EGC (2002). Adaptive Kernel Estimation and Continuous Probability Representation Of Historical Earthquake Catalogs, Bull Seism Soc Am, 92(3), 904–912.
• Werner MJ, Helmstetter A, Jackson DD, Kagan YY (2011). High-Resolution Long-Term and Short-Term Earthquake Forecasts for California, Bull Seism Soc Am, 101:4, 1630–1648
• Woo G (1996). Kernel Estimation Methods for Seismic Hazard Area Source Modeling, Bull Seism Soc Am, 86:2, 353–362.
• Zechar JD, Schorlemmer D,Werner MJ, Gerstenberger MC, Rhoades DA, Jordan TH (2013). Regional Earthquake Likelihood Models I: Firstorder Results, Bull Seism Soc Am, 103(2A), 787–798.
• Deniz A, Yücemen MS (2008). Processing Earthquake Catalog Data for Seismic Hazard Analysis, 8th International Congress on Advances in Civil Engineering, North Cyprus
• Moore EF. (1962). Machine models of self reproduction. In: Belman R, editor. Proceedings of the Fourteenth symposium on Applied Mathematics; Providence (RI): American Mathematical Society.
• Kandilli Rasathanesi ve Deprem Araştırma Enstitüsü (2017) http://www.koeri.boun.edu.tr/sismo/zeqdb/, Last Date of Access: January 2019