Research On DEM Interpolation Assessing Model And Its Application System Development

Interpolation is an approximation problem in mathematics and an estimation problem in statistics.It is also one of the core techniques in digital terrain modeling as it is involved in the various stages of the modeling process such as quality control,surface reconstruction,accuracy assessment,terrain analysis,and applications.Before interpolation being used,it has to be decided to choose an interpolation method including its models and parameters.It is believed that skills of exploratory spatial data analysis(EDA) and experiment/understanding with interpolation algorithms cause good construction of interpolated surface.While GIS being widely used,many people,especially beginners in GIS or researchers in others fields,feel perplexed about spatial data interpolation and even make terrible mistakes.This research makes no effort in developing a new interpolation algorithm,but explores a methodology about how to interpolate while people are in "bad" situation described above,that they do not know much about kinds of interpolation methods and EDA and/or maybe they have no accuracy description about the input data.A software formed as a plug-in of ESRI ArcMap is developed to realize this assessment system of interpolation methods.There are two implicit assumptions behind interpolation techniques:(a)the terrain surface is continuous and smooth and(b)there is a high correlation between the neighboring data points.Each interpolation method obeys these rules,and the difference of interpolation algorithm to each other is how the predicted point correlates with the input points from the near to the distant. Considering with this route,a uniform conceptual model of interpolation is presented,which is the base of a derivation of the spatial distribution of interpolated DEM error.This kind of error map using in this assessing system is considered to be an extension of the Kriging error map to other interpolation algorithms,and provides people the error distribution no matter which method they choose to use.Besides this,subset validation is introduced to the system to compare some common interpolation methods with Root Mean Square Error and Mean Error as its indexes,which provides people the appropriate algorithm even without "true value" dataset.