Impact of DEM Cell Size on the Accuracy of Debris Flow Modeling With RAMMS

Debris flow, which is known as the most destructive natural hazard, is a complex environmental phenomenon in which a large volume of moving mass including mud, sand, rock, soil, water, and air travels down a slope under the impact of gravitational force. Due to this phenomenon's complex physical nature, the governing equations' analytical solution is complicated and almost impossible in real situations.
In the last two decades, with the advancement in computer technologies and numerical methods, some robust software has been developed to simulate debris flow and particle movement. The Rapid Mass Movement Simulation (RAMMS) is a two-dimensional model to calculate the motion of geophysical mass movements (snow avalanches, rockslides, debris flows, and shallow landslides) from onset to runout in three-dimensional terrain. Since the numerical analysis in this software is based on the terrain parameters which are extracted from a Digital Elevation Model (DEM), the accuracy of the model might be very dependent on the resolution of the DEM. For this reason, this study's main aim is quantitively evaluating errors and uncertainties in RAMMS-Debris flow model outputs induced by the impact of DEM cell size.

RAMMS extracts all topographical parameters (i.e., slope angle, flow directions, streams, altitudes, inundation areas) from a DEM (ASCII or GeoTIFF format). Since the resolution of the DEM and the resampling process can significantly affect the physical terrain parameters of a given watershed such as the mean slope, altitudes, the cross-section of rivers, etc., therefore the RAMMS outputs may be altered with the changing of the DEM resolution which causes the propagation of errors in modeling and calculations.
To investigate the impact of DEM cell size on the accuracy of RAMMS-Debris flow modeling, a watershed with high-resolution DEM (1 m), located in northeastern Iran, is considered. Then, this basic DEM is resampled using the bilinear method in ArcGIS to build DEM maps in various cell sizes such as 2, 3, 4, 5, 10, 15, and 20 m. Since the modeling process is based on the Voellmy-fluid friction model, the viscous-turbulent and the Dry-Coulomb type friction coefficients are considered constant at 200 and 0.2 respectively. Also, all other input data such as simulation time, inflow hydrograph, fluid density, etc. are considered constant for all scenarios.

The analysis showed that by increasing the DEM cell size from 1 meter to 20 meters, the mean slope of the computational domain decreased up to 4% such that, changing in DEM cell size from 1 meter to 10 meters decreases the mean slope by only 1% but with increasing the cell size to more than 10 m, the mean slope reduces significantly.
In addition, the final results showed that the RAMMS model had an extreme sensitivity to the Dem cell size such that by keeping other parameters constant, only changing the Dem cell size from 1 meter to 20 meters caused an error of +548% in the overall inundation area, -67% in the overall maximum flow depth, and +112% in the run-out distance. It was also found that the magnitude of errors in the modeled debris flow parameters using DEMs with a cell size of less than 5 meters was almost small whereas for cell sizes larger than 5 meters, it was very large. Therefore, it seems that the RAMMS software generally works well only for DEMs with tiny cell sizes (less than 5 meters).

In this research, the effect of the cell size of the Digital Elevation Model (DEM) on the accuracy of the results of the RAMMS model has been investigated. For this purpose, a DEM with a cell size of 1 meter was selected in an area that is prone to debris flow, then using ArcGIS software, some coarser maps with cell sizes of 2, 3, 4, 5, 10, 15, and 20 meters were made using the bilinear method. based on the results obtained from the modeling of various scenarios in this study, it is found that the RAMMS model has an extreme sensitivity to the cell size of the DEM map. Although the smaller the cell size of the DEM map, the higher the accuracy of the model, the increase in the DEM cell size (considering other parameters being fixed) causes a drastic increase in modeling error. This amount of error in DEM-20 m is almost 5.5 times that of DEM-1 m. Therefore, the important point in debris flow modeling with RAMMS is to use a suitable DEM cell size to increase the accuracy of the model. According to the results, if the DEM cell size is less than 5 meters, the magnitude of error is not significant and can be ignored.