Quantifying uncertainty in elevation models

Blog Post created by DWright-esristaff Employee on Aug 8, 2014

I would like to share an important article written by two of my colleagues, @Konstantin Krivoruchko and @Kevin Butler. This is a must read for those working with DEMs, especially those seeking an additional/alternative DEM-creation approach with considerable explanatory power.


To whet your appetite from the article:

"Raster based digital elevation models (DEM) are the basis of some of the most important GIS workflows: hydrologic modeling, site suitability, and cost path analysis. While there are several techniques for generating digital elevation models (DEMs), none of them can produce a true elevation surface. Locally varying measurement error and the inexactness of the interpolation methods contribute to the uncertainty of the model’s estimate of the true elevation value. Kriging models and geostatistical simulations available in the Geostatistical Analyst extension for ArcGIS 10.1 for Desktop to quantify the spatially varying uncertainty of a DEM derived from lidar data. ...


An alternative to deterministic algorithms, probablistic statistical interpolation methods such as kriging, have several advantages over deterministic methods. “Empirical Bayesian Kriging: Implemented in ArcGIS Geostatistical Analyst,” an article in the Fall 2012 issue of ArcUser magazine discusses these advantages in detail."


To read further and to contact the authors directly, please find the attached.


Happy kriging!