Multiscale analysis of surface roughness for the improvement of natural hazard modelling

Abstract

Abstract. Surface roughness influences the release of avalanches and the dynamics of rockfall, avalanches and debris flow, but it is often not objectively implemented in natural hazard modelling. For two study areas, a treeline ecotone and a windthrow-disturbed forest landscape of the European Alps, we tested seven roughness algorithms using a photogrammetric digital surface model (DSM) with different resolutions (0.1, 0.5 and 1 m) and different moving-window areas (9, 25 and 49 m2). The vector ruggedness measure roughness algorithm performed best overall in distinguishing between roughness categories relevant for natural hazard modelling (including shrub forest, high forest, windthrow, snow and rocky land cover). The results with 1 m resolution were found to be suitable to distinguish between the roughness categories of interest, and the performance did not increase with higher resolution. In order to improve the roughness calculation along the hazard flow direction, we tested a directional roughness approach that improved the reliability of the surface roughness computation in channelised paths. We simulated avalanches on different elevation models (lidar-based) to observe a potential influence of a DSM and a digital terrain model (DTM) using the simulation tool Rapid Mass Movement Simulation (RAMMS). In this way, we accounted for the surface roughness based on a DSM instead of a DTM, which resulted in shorter simulated avalanche runouts by 16 %–27 % in the two study areas. Surface roughness above a treeline, which in comparison to the forest is not represented within the RAMMS, is therefore underestimated. We conclude that using DSM-based surface roughness in combination with DTM-based surface roughness and considering the directional roughness is promising for achieving better assessment of terrain in an alpine landscape, which might improve the natural hazard modelling.

Modeling deadwood for rockfall mitigation assessments in windthrow areas

“…In contrast, in W2 the two V UAS distributions differed distinctly: the initial bimodal distribution pattern of V Jan UAS changed into the left-skewed pattern of V Nov UAS . This effect can be explained by the removal of deadwood in W2 between the two DSM acquisition dates and is in line with the findings of Brožová et al (2021) that deadwood increases the VRM. The deviations to the slightly higher deadwood VRM values (VRM = 0.17-0.30) can be explained by a different DSM resolution (1.0 m) and a larger moving window size (49 m 2 ), and it matches the presented sensitivity analysis.…”

“…We documented the state of the entire area twice with fixed-wing drone (unmanned aircraft system, UAS) surveys: after the windthrow event with the eBee + RTK (24 January 2020, average flight altitude 200 m above ground) and after the forest service operations with the WingtraOne PPK (25 November 2020, average flight altitude 290 m above ground). Orthophotos with a spatial resolution of 4 cm and DSM UAS with a spatial resolution of 10 cm were produced based on the photogrammetric workflow within the Agisoft software suite (Bühler et al, 2016).…”

“…Longterm observations and modeling of snow avalanches after natural disturbances suggest that the protective effect reaches a minimum at 10-15 years, followed by an increase due to the combined effect of post-disturbance regeneration and the remaining terrain roughness created by deadwood (Caduff et al, 2022). However, this effect may vary strongly according to site conditions, the initial stand structure, and species composition of a forest before the disturbance, as well as external management factors, such as partial removal of deadwood, planting of additional trees, or browsing (Bebi et al, 2015;Brožová et al, 2021;Caduff et al, 2022). An ADG that considers the initial stand structure of a forest, as well as other relevant drivers of post-disturbance development, may serve as a valuable tool to develop spatially explicit and locally adapted long-term scenarios of protection against rockfall after different disturbances and under different management scenarios.…”

“…After forest fires of high severity, the amount of woody debris is strongly reduced, and rockfall simulations after such fires may even assume the cleared condition (Maringer et al, 2016). In addition, the DSM gen used here as an intermediate product only could serve as input for avalanche simulations (Brožová et al, 2021) and for the identification of their release areas in forested terrain (Bühler et al, 2018(Bühler et al, , 2022. Extending the ADG to other disturbance regimes and including additional time steps of deadwood protective effects would thus provide a more comprehensive basis for optimized deadwood management in forests with a protection function against rockfall.…”

Automated avalanche hazard indication mapping on a statewide scale

“…The forest cover threshold takes into account the coverage at different spatial scales and optimizes for the detection of trees, especially in a critical range between 3 and 5 m (see details in Table 2 in Bebi et al, 2021). In order to account for increased protection forest capacities of high terrain roughness compared to smooth surfaces, we delineate areas with a high surface roughness with the "vector ruggedness measure" (VRM) according to Sappington et al (2007) based on the 2 m DTM (swissALTI3D) and a moving window of 5 × 5 m. Areas with a value > 0.02 and no lateral convex curvature are considered to be rough (Brožová et al, 2021). Likewise, we account for reduced protection forest capacities of shrubs (e.g., green alder, Alnus viridis, or the shrub form of mountain pine, Pinus mugo Turra subsp.…”