Fusion Based on Geostatistics to Improve the Altimetry Accuracies of Digital Elevation Models

Fusions based on geostatistical methods are used in this article to improve the accuracy of the altimetry attribute of Digital Elevation Models (DEMs). Ordinary kriging, kriging with external drift, regression kriging and cokriging procedures, are applied to assess uncertainty representations from which it is possible to get altimetry predictions and other information. The fusion data modeling is performed from existing DEMs, mainly available for free in the internet, and additional high accurate set of 3D sample points. Although the freeware DEMs are dense and generally have good spatial distributions, the accuracy of their altimetry information
might not be suitable for many applications. A way of mitigating this problem is to combine, in the data modeling processes, the available DEM data along with additional information coming from various other sources and having better quality. Usually, high accurate altimetry data are collected in field works, with higher cost,
at specific point locations inside the spatial region of interest. In short, this work aims to integrate, through geostatistical methods, spatial elevation information of different sources, data structures and elevation accuracies to obtain better accurate DEMs. The methodology addressed in this research was applied to a case
study in a Brazilian Southeast geographical region. Quantitative and qualitative validations were performed using an independent high accurate data set and comparisons based on DEM differences and drainage network automatic extraction. For the considered study area, the kriging with external drift and the regression kriging have led to similar quantitative and qualitative improvements, better than the co-kriging approach.