Spatial data quality in digital visibility models

Ing. Ondřej Lagner

Doctoral thesis

Abstract:

Digital terrain models and applied visibility analyses are widely used tools of geoinformatics. They have an irreplaceable position in landscape evaluation as well as in evaluation of individual interventions. Similarly, they can facilitate assessment of projects in the fields of landscape ecology, animal ecology, urban planning, architecture and civil engineering, urbanization, planning of tourist paths, lookout towers, etc. The present day modelling of terrain and visibility however faces a problem of existence of a large amount of input geodata of varying quality.The thesis consists of a set of three commented studies focused on selection and processing of data suitable for visibility analyses. The studies aim to identify spatial data that would lead to the most accurate results of visibility analysis. The accuracy of spatial data is affected by multiple factors such as the method of data acquisition, spatial scale and data processing. All these variables affect, along with the terrain characteristics and the degree of forestation, the accuracy of the visibility results. The three presented studies imply that the greatest effect on the accuracy of the result of visibility analysis can be attributed to the accuracy and spatial scale of the input geodata. This influence is so great that it by far exceeds the effect of other variables such as the degree of forestation, terrain complexity, the number of obstacles to visibility, buildings, etc. Data acquired through airborne laser scanning are the most suitable input data for visibility analysis. Use of fine scale vector data however may serve as an alternative where LiDAR data are not available. Coarsening the scale however leads to overestimating the visible areas and to the higher false positive rate, i.e., the error where the model predicts more pixels to be visible than really are, which is most likely caused by “flattening” the surface and reducing the number of obstacles when using coarser data. Another factor having a great effect on the accuracy of the digital models is the degree of pre-processing of input geodata. If using raw data that are specifically processed with the aim of the analysis in mind, the results are significantly better than when using ready-made data that are pre-processed with a general purpose algorithm as such data may suffer from processing flaws unknown to the researcher or with loss of detail information caused by an unsuitable processing method.

Keywords

Digital terrain model, Uncertainty, Viewshed, Visibility, Accuracy

Language used: English

Date on which the thesis was submitted / produced: 30. 8. 2019

Thesis defence

• Supervisor: doc. Ing. Petra Šímová, Ph.D.
• Reader: Rudolf Urban, Ing., externi

Czech University of Life Sciences Prague

Faculty of Environmental Sciences

Doctoral programme / field:
Environmental Sciences / Applied and Landscape Ecology