By Laurens van der Burgt
ABSTRACT
The regular square grid is a standardized data format in current GIS. From a hydrological point of view it has been expected that the regular hexagonal grid can offer advantages. The main reason for this is the equality in distances to all neighbouring cells.
The goal of this explorative research was to investigate the effects for a hydrologist of using square and hexagonal regular grids for representing DEMs and secondary data derived from DEMs.
Elevation points from two locations in the Netherlands with contrasting topography were used as source data: Schokland and Zuid-Limburg.
A vector based method is presented for the implementation of the regular hexagon grid with accompanying algorithms for this grid type. Algorithms to calculate slope, smoothing, flow direction, flow accumulation and upstream catchment area are included.
DEMs were generated for different interpolation methods, grid rotations and cell areas and
the results were statistically analyzed. Secondary data was compared between the grid types and with reference data. Dual triangular surfaces were used as additional data format to derive secondary data from.
Resulting hexagonal DEMs show that they can be advantageous for the representation of circular features. It was found that grid rotation on the generation of a DEM does not have a considerable effect on the aggregated error. Also the hexagonal and square grid yield a similar error. For varying cell area, the resulting error is similar as well for both grid types. However, the choice of interpolation method does affect the error. Rotation does affect the representation of hydrological features.
It is concluded that the use of hexagonal grids for hydrological applications offers more symmetry than square grids. This symmetry applies both to the visualisation of grids and to the computations leading to this result.
For secondary data, the effects are more clear than for primary elevation data. The most remarkable differences between grid type occur for multi-directional flow accumulation and upstream catchment area. For other applications of models using cell-to-cell interaction, it is to be expected that additional advantages and disadvantages of the choice for grid type can be obtained.
This research ends with the presentation of recommendations for further research on this topic. It is recommended to investigate the performance of regular hexagonal grids on additional GIS models. Secondly it is worthwhile to do additional research on conversions from a regular hexagonal grid to a regular square grid and vice versa and on the accompanying error.