برآورد پارامتر شدت زلزله در منطقه‌ گسل با استفاده از داده ‏های حرارتی سنجش از دور

زلزله‏ یکی از پیش ‏بینی ‏ناپذیر‏ترین و خطرناک ‏ترین پدیده ‏های طبیعی است که هرساله خسارات مالی و جانی فراوانی را باعث می ‏شود. هنگام وقوع زلزله تنش ‏ها و فعالیت‏ های محدوده‌ گسل افزایش می ‏یابد و باعث تغییرات دمایی محسوسی نسبت به دمای نرمال می ‏شود. این تغییرات دمایی خود را به‏ صورت بی‏ هنجاری‏ هایی در مکان یا زمان نشان می‏ دهند. در این تحقیق با استفاده از محصولات حرارتی سنجنده‌ مادیس و شیپ‏ فایل گسل ‏های ایران، هفت زلزله با شدت بیشتر از شش ریشتر، که در ایران رخ داده، بررسی شده است. در این پژوهش با استفاده از تشکیل تصویر زمان دما فاصله در گسل مربوط به زلزله به ‏عنوان ورودی دو روش تشخیص بی‏ هنجاری حرارتی روی داده‏ ها بررسی شده است. در نهایت، با استفاده از نتایج حاصل از بهترین روش تشخیص بی‏هنجاری پارامتر شدت با استفاده از شبکه‌ عصبی مصنوعی برآورد شده است. نتایج الگوریتم ‏های تشخیص ناهنجاری نشان می ‏دهد هرچند هر دو روش تشخیص بی ‏هنجاری حرارتی بی‏ هنجاری حرارتی مربوط به هر زلزله را در روز زلزله در شعاع نزدیک به گسل شناسایی کرده‏اند روش چارکی (interquartile) نسبت به روش میانگین انحراف‏ معیار نتایج مناسب ‏تری را برای ورودی الگوریتم شبکه‌ عصبی فراهم می ‏کند. نتایج در مدل ‏سازی نیز نشان می ‏دهد پارامتر شدت زلزله، که با استفاده از شبکه‌ عصبی مصنوعی بررسی شد، دقت کلی 0.73 را داشته است. ذکر این نکته لازم است که پیش ‏نشانگر تغییرات دمای سطح و بی ‏هنجاری‏ های حرارتی به ‏تنهایی نمی ‏تواند برای بررسی کامل پارامترهای زلزله کافی و دقت لازم را برای تحلیل زلزله داشته باشد. ولی با توجه به حجم پایین داده‏ های حرارتی و سادگی کار با آن‏ها، توصیه می ‏شود از آن‏ها برای بررسی ‏های ابتدایی و آغازین زمین ‏لرزه استفاده شود و در صورت تایید نسبی آن برای تحلیل‏های بیشتر، از روش‏ها و پیش‏ نشانگرهای دیگر، که در آن‏ها اعمال الگوریتم ‏ها و پردازش‏ های سنگین و پیچیده نیاز است، استفاده شود.

Earthquake severity parameter estimation in fault regions using remote sensing thermal data

An earthquake is the movement of the surface of the Earth resulting from a sudden release of energy in the Earth’s lithosphere that creates seismic waves. Earthquakes are one of the most unpredictable and dangerous natural phenomena that cause many financial and human losses every year. Due to the great importance of this natural crisis, several studies have been conducted to investigate this phenomenon. Many of these studies show that the earthquakes phenomenon is highly related to the deformation of the earth, rising ground temperatures, gases and aerosols, and electromagnetic disturbances in the atmosphere. The land surface temperature is highly dependent on the interactions of the earth’s surface layers. When an earthquake occurs, stresses and activities in the fault range increase, causing significant temperature changes compared to normal temperatures. These temperature changes manifest themselves as anomalies in place or time.Regarding the materials and methods, in this research, using MODIS thermal products and shapefile of Iran’s faults, seven earthquakes with the intensity of more than 6 Ms have been investigated. First the preprocessing was performed on LST data so that thermal noise signals caused by seasonal changes be removed from the original data. This was done by using a linear model made from the previous year data which no seismic activities were reported during its 40 days of investigation. Then, using the formation of a threedimensional picture of timetemperaturedistance in the earthquakerelated fault as input, two methods for detecting thermal anomalies have been investigated on the data. The mean standard deviation method, which is a threshold method using two parameters, and the interquartile method, which is similar to the previous method but uses different statistical parameters as input, are the two algorithms used in this research. Finally, using the results of the best method for detecting anomalies, severity parameter of each earthquake is estimated using artificial neural networks.Regarding the results and discussion, it should be noted that the results of anomaly detection algorithms show that both methods of thermal anomaly detection have detected thermal anomalies related to each earthquake on the day of the earthquake in a radius closest to the fault. In some cases like fahraj earthquake some anomalies were detected aside the anomaly detected on the day of the earthquake. However, results of the meanstandard deviation method gives more false alarms as an earthquake thermal anomaly than the interquartile method. Although these anomalies could be related to the earthquake it cannot be a certain fact. So in order to have a better outcome we use the results of interquartile anomaly detection method as input for training of artificial neural network. The results in mathematical modeling have a relatively high accuracy in the case of seismic intensity parameter using artificial neural network with the total accuracy of 0.73. These results indicate that the best accuracy belongs to Azgalah and the one with least accuracy belongs to fahraj study case. Although the number of earthquakes studied for neural network training has been relatively small, but the availability of large amounts of data on each earthquake has provided appropriate accuracy. In conclusion, this study shows that thermal anomalies is one of the most significant precursors for earthquake’s investigations. Using the relevant fault and anomalies with respect to the buffer zones in different distances can help us increase the accuracy dramatically. Since many previous studies that investigated thermal anomalies connected to the earthquakes, explored areas around the epicenter, in this study we show that the corresponding fault is just as important as epicenter.Finally, it should be noted that the indicator of surface temperature changes and thermal anomalies alone cannot be sufficient to fully investigate the parameters of the earthquake or have the necessary accuracy to analyze the earthquake. However, due to the low volume of thermal data and the simplicity of working with them, it is recommended that they be used for initial earthquake surveys, and if it is partially confirmed for further analysis, use other methods and indicators that require the application of heavy and complex algorithms and processes. It is also possible to combine the results of this precursor with the results of other precursors to achieve sufficient accuracy. Regarding the results and discussion, it should be noted that the results of anomaly detection algorithms show that both methods of thermal anomaly detection have detected thermal anomalies related to each earthquake on the day of the earthquake in a radius closest to the fault. In some cases like fahraj earthquake some anomalies were detected aside the anomaly detected on the day of the earthquake. However, results of the meanstandard deviation method gives more false alarms as an earthquake thermal anomaly than the interquartile method. Although these anomalies could be related to the earthquake it cannot be a certain fact. So in order to have a better outcome we use the results of interquartile anomaly detection method as input for training of artificial neural network. The results in mathematical modeling have a relatively high accuracy in the case of seismic intensity parameter using artificial neural network with the total accuracy of 0.73. These results indicate that the best accuracy belongs to Azgalah and the one with least accuracy belongs to fahraj study case. Although the number of earthquakes studied for neural network training has been relatively small, but the availability of large amounts of data on each earthquake has provided appropriate accuracy. In conclusion, this study shows that thermal anomalies is one of the most significant precursors for earthquake’s investigations. Using the relevant fault and anomalies with respect to the buffer zones in different distances can help us increase the accuracy dramatically. Since many previous studies that investigated thermal anomalies connected to the earthquakes, explored areas around the epicenter, in this study we show that the corresponding fault is just as important as epicenter.Finally, it should be noted that the indicator of surface temperature changes and thermal anomalies alone cannot be sufficient to fully investigate the parameters of the earthquake or have the necessary accuracy to analyze the earthquake. However, due to the low volume of thermal data and the simplicity of working with them, it is recommended that they be used for initial earthquake surveys, and if it is partially confirmed for further analysis, use other methods and indicators that require the application of heavy and complex algorithms and processes. It is also possible to combine the results of this precursor with the results of other precursors to achieve sufficient accuracy. Keywords: Earthquake, Earthquake Precursor, Thermal Anomaly, Active Fault, Artificial Neural Network