PDF(Pubmed)
Abstract：
The ground-based gravity data reveals diverse anomaly signatures in areas of the Main Ethiopian rift where active volcanic and tectonic activities are dominant. In such a region ground-based data collection is restricted to existing roads and relies on accessible stations. These resulted in gaps in data, either missing, uneven, or insufficient spatial coverage that must be estimated with proper interpolation techniques. Comparison and evaluations of the spatial interpolation methods that are commonly used in potential field geophysical data analysis were made for the terrestrial gravity and elevation data of the central Main Ethiopian rift. In this research, two widely used interpolation techniques, minimum curvature interpolation, and Ordinary Kriging were compared and assessed. A 10 % hold-out validation was employed, where 90 % of the data points were used to generate interpolated surfaces, which were then evaluated against the remaining 10 %. Following interpolation with each technique, the generated grid was converted into discrete data points (estimated values). These are then compared with the available gravity data, which were deliberately excluded from the gridding process (10 % remaining dataset). The accuracy of each method was assessed by evaluation metrics such as mean value, variance, Mean Absolute Error (MAE), Root Mean Square Error (RMSE), correlation coefficient (r), and R-squared. The results showed that the ordinary Kriging interpolation method outperformed the minimum curvature interpolants for gravity data with all performance metrics, while both interpolants seem to perform equally well for the elevation dataset. Therefore, it is proposed to use the Kriging interpolation method for potential field gravity studies conducted in the central Main Ethiopia rift.
摘要：
地面重力数据揭示了埃塞俄比亚主要裂谷地区活跃的火山和构造活动占主导地位的各种异常特征。在这样的地区，基于地面的数据收集仅限于现有道路，并且依赖于可访问的站点。这些导致了数据的差距，要么失踪，不均匀,或空间覆盖不足，必须使用适当的插值技术进行估计。对埃塞俄比亚中部主要裂谷的地面重力和高程数据进行了比较和评估，这些方法通常用于势场地球物理数据分析。在这项研究中,两种广泛使用的插值技术，最小曲率插值，和普通克里格法进行了比较和评估。采用了10%的保持验证，其中90%的数据点用于生成插值曲面，然后对剩余的10%进行评估。在使用每种技术进行插值之后，生成的网格被转换成离散的数据点（估计值）。然后将这些数据与现有的重力数据进行比较，它们被故意从网格化过程中排除(剩余10%的数据集)。通过均值、方差，平均绝对误差(MAE)，均方根误差(RMSE),相关系数(r)，和R平方。结果表明，普通的Kriging插值法在所有性能指标下都优于重力数据的最小曲率插值法，而对于高程数据集，这两种插值似乎都表现得同样好。因此,建议使用Kriging插值法进行埃塞俄比亚中部主要裂谷的势场重力研究。
