Testing Linear Regressions by StatsModel Library of Python for Oceanological Data Interpretation

Abstract

The study area is focused on the Mariana Trench, west Pacific Ocean. The research aim is to investigate correlation between various factors, such as bathymetric depths, geomorphic shape, geographic location on four tectonic plates of the sampling points along the trench, and their influence on the geologic sediment thickness. Technically, the advantages of applying Pythonprogramming language for oceanographic data sets were tested. The methodological approaches include GIS data collecting, data analysis, statistical modelling, plotting and visualizing. Statistical methods include several algorithms that were tested: 1) weighted least square linear regression between geological variables, 2) autocorrelation; 3) design matrix, 4) ordinary least squareregression, 5) quantile regression. The spatial and statistical analysis of the correlation of thesefactors aimed at the understanding, which geological and geodetic factors affect the distributionof the steepness and shape of the trench. Following factors were analysed: geology (sedimentthickness), geographic location of the trench on four tectonics plates: Philippines, Pacific, Marianaand Caroline and bathymetry along the profiles: maximal and mean, minimal values, as well as thestatistical calculations of the 1st and 3rd quantiles. The study revealed correlations between thesediment thickness and distinct variations of the trench geomorphology and sampling locationsacross various segments along the crescent of the trench.