Research On DEM Error Model

Digital Elevation Model (DEM) is known as the most important GIS fundamental data used in geo-simulation and spatial analysis, and its quality has gained widespread attention and extensive research ever since DEM concept was proposed in 1958. With the development of geoinformatics, a great progress has been made in DEM sampling method and accuracy assessment. However, due to the multiple data resources of DEM, uneven terrain surface, and complexity of DEM modeling, the research on DEM uncertainty relatively fell behind its application requirements. The various types of DEM errors, to some extent, reduce the precision of the GIS analysis and its application, especially when DEM and its derivatives serve as the input data of various types of topographical models.Nowadays, spatial information services (SIS) which is based on 4D products and especially on the combination of GIS and geo-models have become the main development trend of GIS. The increasingly complex geo-models ask for requirements of multi-feature data geocomputation as well as the quality of the GIS fundamental data. DEM as one of the important 4D products, its application has expanded into many fields, which simultaneously put forward higher requirements to DEM quality, In order to offer a sound quality standard for SIS, spatial analysis and geo-simulation, further theoretical study on the DEM accuracy is extremely essential.This thesis firstly gave a systematic review and assessment on literature, and pointed out the following limitations: global DEM accuracy model like root mean square error (RMSE) can not describe spatial distribution of DEM error, which can not provide overall accuracy information for users. Traditional error models can not give precised error prediction or lack of universality in use. Based on these limitations, this thesis started from the conventional DEM interpolation theory, analyzed the DEM interpolation mechanism and deduced DEM unified interpolation model (UIM). Under the framework of UIM, also by using geostatistic and error propagation theory, this thesis modeled DEM local accuracy (Pixel accuracy), proposed DEM accuracy field concept and model, which aimed at revealing the spatial characteristics of the DEM error distribution (spatial structure of DEM error), and investigating DEM error propagation in terrain analysis (take slope and aspect extraction as example) considered the impact of spatial structure of DEM error.The major contributions are as the following:1) DEM unified interpolation model (UIM)After introducing conventional DEM interpolation models, this thesis analyzed the mechanisms of the DEM interpolation models, investigated the nature and general character of the DEM interpolation algorithms, and lastly presented UIM. UIM demonstrates that conventional DEM interpolation model actually is a weight allocation process of each reference point, that is, each reference point has its own base function, and its base function relies on interpolation kernel. UIM is an abstract model that links each conventional DEM interpolation models.2) DEM error model based on UIMThe existing DEM error models are all based on specific interpolation model, thus, they lack the universality in use. Under the framework of the UIM, and by using the spatial statistics and the error propagation theory, this thesis constructed the DEM accuracy field model (including noise accuracy field: A field; approximated accuracy field: e field; synthesis accuracy field:αfield), DEM error model based on UIM is independent from specific interpolation model, thus its universality in use has been enhanced. Lastly, this thesis carried on experimental research analyzing the practicality of the DEM error model, results showed DEM error model presented in this thesis was sound in accuracy prediction. 3) The spatial structure of the DEM errorBased on DEM error model, this thesis theoretically testified that DEM error existed specific spatial autocorrelation, by using the theory of geostatistic, this thesis presented the methodology on how to describe spatial structure of DEM error, and constructed the spatial autocorrelation model of the DEM accuracy field, lastly formed the DEM accuracy evaluation criterion from the global to local description as well as spatial structure description.4) Terrain parameters accuracy analysis considered spatial structure of DEM accuracy fieldConsidered the spatial structure of the DEM accuracy field, this thesis took slope and aspect as study objects, inferred the theoretical precision model for the slope and aspect, and calculated slope and aspect accuracy field (βfield ) based on various DEM difference algorithms, testified the correctness of the slope and aspect accuracy model through experimentation, results further revealed the impact that spatial autocorrelation of DEM error on accuracy of digital terrain analysis was remarkable.This study enhanced the evaluation method of DEM accuracy, provided more comprehensive description information of the DEM accuracy for its users, gave a theoretical support to the security and reliability of DEM when used in terrain analysis and engineering application, and simultaneously this study provided a useful reference for the study of the uncertainty of spatial data and reliability analysis of the geo-models.