Digital landscape modeling allows us to better understanding of landscape what it is a very important for prediction of landscape processes in future. Acquisition and processing of such spatially localized data on landscape became widely accessible in the last decade. This fact relates to the development of modern technologies allowing for a remarkable effective recording of geodata with a high level of detail and accuracy as, for example, laser scanning, hyperspectral scanning, digital photogrammetry or automatized recording of climate or hydrological variables by data loggers. It especially concerns geographic information systems (GIS) which were developed for data processing and analyses to make the research and landscape modelling more effective. The SPATIAL3D project is focused on the development of new methods for effective handling of massive data sets in a GIS environment and 3D modeling based on such data sets.
The research is performed on a case study a karst landscape. The karst landscape presents a complex system with a marked horizontal and vertical interaction of its components. For that reason, karst is an appropriate type of landscape to demonstrate applicability and benefits of mapping the surface and subsurface landforms by modern methods and also 3D modeling with acquired data using GIS platform.
Fig. 1 Millions of points in a 3D point cloud acquired by terrestrial laser scanning. The measurements depict surface and subsurface landforms of the Marble Arch Cave area, Northern Ireland. The figure originated within the cooperation with Queen’s University of Belfast and it is courtesy of John Meneely.
Physical-geographical specifics of the karst landscape in the region of Central Europe determine relatively difficult accessibility of such areas. Traditional mapping methods such as tachymetry or aerial photogrammetry are not applicable in an efficient way. Dense scrubs or forests, steep slopes, and specific landforms (dolines, rock cliffs, overhangs) are the main factors hindering the ability to capture terrain in with a high level of detail. It is for the difficulties of photogrammetry to map terrain under the vegetation canopy. Forest pose difficulties also for mapping with global navigation satellite systems (GNSS), because of the multi-path signal or sky visibility. In addition, mapping with GNSS or tachymetry in areas larger than few hectares becomes inefficient.
Fig. 2 Demonstration of capabilities for terrain mapping by aerial photogrammetry and airborne laser scanning. The photogrametrically derived contour lines (A) do not adequately depict landforms under the forest canopy (green colour), while the terrain details are better portrayed by contours derived from airborne laser scanning data (B).
The outlined problems of mapping and studying the landscape can be overtaken by laser scanning. It is an active remote sensing method based on measuring the travelling time between the emitted laser pulse when it leaves the transmitter and is scattered back from the object and is detected. This makes it capable of collecting altitude of several surface levels. The number of the recorded laser echoes (returns) depends on the penetration of laser beam down through the ground. However, they can also represent non-terrain objects. The ultimate result of laser scanning is a so called point cloud containing millions of 3D point measurements depending primarily on the measurement density and the mapped area size. Laser scanning is also applicable for mapping other features then terrain surface, for instance, the vegetation structure, buildings, and technical infrastructure. One of the main challenges of using laser scanning in the research of landscape is the difficult processing of the immense amount of point measurements (point cloud) representing the scanned surface so that the form of representation allows for further processing of the data in a geographic information system (GIS). Massive data sets which are the result of scanning have to be filtered to extract points representing a particular surface (e.g. terrain, vegetation canopy, buildings).
Fig. 3. 3D point cloud representing altitude measurements acquired by laser scanning of a part of the Slovak Karst (100m x 250m). A – original unfiltered point cloud, B – filtered point cloud containing terrain points only.
In case of mapping terrain with altering open grassland and areas covered by dense vegetation, filtering the original point cloud to remove the non-terrain objects can yield a spatially heterogeneous terrain point field with a highly variable data density. Areas of high point density altering with areas of low point density cause difficulties for fitting mathematical function when generating a digital terrain model using methods of spatial prediction.
Fig. 4. Digital terrain model of a portion of the Slovak Karst based on airborne laser scanning data where local depresions as karst surface microforms are depicted. The smaller section of the model portrays spatial heterogeneity of the terrain altitude measurements after filtering the original point cloud (A). The interpolated surface before optimizing the interpolation contains artefacts (B) which are eliminated (C) after optimizing the selection of input points (C).