Scholarlink

<p><b>Abstract</b>—The accuracy of deep temperature forecasting from seismic velocity data and model temperature logs is studied as a function of distance to the point for which a prediction is made (a forecast point). For this purpose, the velocity sections from seismic tomography of the subsurface along a sublatitudinal profile in the Northern Tien Shan and the temperature model, previously constructed for this profile down to a depth of 27 km, are used. The accuracy assessment of temperature forecast using artificial neural network technology shows that at distances up to 16 km from the forecast point, the residual between the predicted and model temperature is 7.4, 5.7, and 4.6% for forecasts from longitudinal and shear wave velocities and their combination, respectively. With a fourfold increase in the distance to the forecast point, the residuals increase by a factor of 2–3. In general, it can be concluded that neural network forecasting of the temperature of the Earth’s interior based on seismic velocities can be performed with acceptable accuracy at large distances from measurements of input data and can be used as a “seismological geothermometer.”</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Modeling a Seismological Temperature Prediction at Upper Crustal Depths

  • 期刊论文

  • 发表日期：

    2025年12月1日

查看全文
  • O. K. Zakharova,
  • V. V. Spichak

摘要

Abstract—The accuracy of deep temperature forecasting from seismic velocity data and model temperature logs is studied as a function of distance to the point for which a prediction is made (a forecast point). For this purpose, the velocity sections from seismic tomography of the subsurface along a sublatitudinal profile in the Northern Tien Shan and the temperature model, previously constructed for this profile down to a depth of 27 km, are used. The accuracy assessment of temperature forecast using artificial neural network technology shows that at distances up to 16 km from the forecast point, the residual between the predicted and model temperature is 7.4, 5.7, and 4.6% for forecasts from longitudinal and shear wave velocities and their combination, respectively. With a fourfold increase in the distance to the forecast point, the residuals increase by a factor of 2–3. In general, it can be concluded that neural network forecasting of the temperature of the Earth’s interior based on seismic velocities can be performed with acceptable accuracy at large distances from measurements of input data and can be used as a “seismological geothermometer.”