Variance Component Estimation Dilemma in Cadastral Network Adjustment

Abstract

Land record that preserve the positional accuracy is an absolute objective in cadastral implementation. To achieve that, Department of Surveying and Mapping Malaysia (DSMM) has taken drastic decision by transforming the adjustment procedure from traditional approach (i.e. Bowditch) to the most rigorous procedure (i.e. least square adjustment). However, homogenous precisions as stated in DSMM circular are not align with fundamental principle of least square adjustment (LSA). In LSA implementation, there is a knowledge known as stochastic modeling used to derive the realistic variance. With determination to verify the significance of the realistic variance, this study has designed two assessments which focus on positional errors and sensitivity towards outliers. The artificial errors have been introduced in cadastral dataset, both homogenous and realistic variance were exploited in LSA computation. Through positional accuracy examination, minor advantage has been demonstrated by the outcomes of realistic variance. The conclusion can be made when homogenous variances manage to passed the global test but failed to identify the outliers (artificial errors) in the dataset. To preserve the positional accuracy of land record, realistic variance is a crucial requirement in cadastral network adjustment.