Developing a Local Predictive Model for Geoid Undulation Values in Northwestern Sebha (Libya) Using a Multiple Linear Regression Approach

This paper develops a local predictive model for geoid undulation values in the Northwestern region of Sebha, Libya, addressing the accuracy limitations of global geopotential models in regional engineering applications. Utilizing a dataset of concurrent Global Navigation Satellite Systems (GNSS) observations and precise geometric levelling benchmarks, a Multiple Linear Regression (MLR) approach was implemented to establish an empirical mathematical relationship between ellipsoidal coordinates and geoid heights. The results of the analysis show that the created model provides a strong fit, yielding a coefficient of determination (R2= 0.932), and the model significance was confirmed by the ANOVA test (F= 31.733, p-Value<0.001), while collinearity diagnostics showed stable and acceptable Variance Inflation Factor (VIF) values for all variables.