ارزيابي روابط كاهندگي در تطبيق با مجموعه‌ي داده‌هاي فلات ايران

Compatibility Assessment of Ground Motion Attenuation Models with the Iran Plateau Database

امروزه يكي از مهم‌ترين اهداف پژوهشگران در تحليل خطر لرزه‌يي يك ساخت‌گاه، انتخاب مناسب‌ترين رابطه‌ي كاهندگي، به‌عنوان يكي از اصلي‌ترين اجزا تحليل خطر لرزه‌يي است. به علت قرارگرفتن ايران بر روي كمربند لرزه‌خيز آلپ ـ هيماليا و احتمال وقوع زلزله‌هاي مخرب در اين منطقه، انتخاب رابطه‌ي كاهندگي مناسب اهميت بالايي دارد. هدف اصلي در اين مطالعه، بررسي ميزان قابل اعتمادبودن روابط كاهندگي با استفاده از تحليل باقيمانده‌ها توسط روش‌هاي رياضياتي و آماري كلاسيك و مدرن و همچنين بررسي ميزان پايداري روابط توسط رويكرد نوين تحليل حساسيت با استفاده از كاتالوگ لرزه‌يي مربوط به كشور ايران است. نتايج آزمون‌هاي صورت‌گرفته نشان‌گر آن است كه مدل زعفراني و همكاران ???? به‌عنوان يك مدل‌محلي، نسبت به ساير مدل‌هاي منتخب سازگاري بهتري با داده‌هاي ثبت‌شده نشان داده است. انتظار مي‌رود از نتايج حاصل از اين پژوهش بتوان در انتخاب رابطه‌ي كاهندگي مناسب در طرح‌هاي تحليل خطر لرزه‌يي داخلي استفاده كرد.

Selection of an appropriate ground motion attenuation model is a key element within any seismic hazard analysis. This issue is highlighted for regions with high levels of seismicity, e.g. Iran’s seismic plateau. The main aim of the current research is to assess different attenuation models for Iran’s ground motion database. Mathematical and statistical tools are employed in order to compare the selected models. A new re-sampling approach is also introduced to assess the stability of the models regarding the given database. Candidate ground motion prediction equations are comprised of three classes: local models, which have been developed based on Iranian data, regional models corresponding to Europe and Middle East data sets, and, finally, NGA models. After selecting the ground motion models, they are assessed based on the Iranian strong motion database. The database used in this research includes 179 ground motion records obtained from 36 earthquakes in Iran, with magnitudes ranging from 5.0-7.4 and distances ranging less than 200 km. Based on the traditional hypothesis test, such as the Z-test and the Lilliefors-test, the results show that all the residuals have normal distribution, but the majority of the models do not have zero mean residuals. Other results from different statistical and mathematical methods, such as Error terms (RMSE & MAE), coefficient of determination (R-squared), and a new approach of the information-theoretical method (LLH), indicate that Zetal12, as an Iranian model, has the best response in comparison to the others. Another important test applied in this paper is checking for no bias on residuals. In this case, Zetal12 is not heavily biased between the other models. Finally, the results based on the new approach in this study for evaluating the stability of models by re-sampling, indicate that Zetal12 is more stable in comparison to the other selected models.