Multiscale Very High Resolution Topographic Models in Alpine Ecology: Pros and Cons of Airborne LiDAR and Drone-Based Stereo-Photogrammetry Technologies

The vulnerability of alpine environments to climate change presses an urgent need to accurately model and understand these ecosystems. Popularity in the use of digital elevation models (DEMs) to derive proxy environmental variables has increased over the past decade, particularly as DEMs are relatively cheaply acquired at very high resolutions (VHR; <1 m spatial resolution). Here, we implement a multiscale framework and compare DEM-derived variables produced by Light Detection and Ranging (LiDAR) and stereo-photogrammetry (PHOTO) methods, with the aim of assessing their relevance and utility in species distribution modelling (SDM). Using a case study on the arctic-alpine plant, Arabis alpina, in two valleys in the western Swiss Alps, we show that both LiDAR and PHOTO technologies can be relevant for producing DEM-derived variables for use in SDMs. We demonstrate that PHOTO DEMs, up to a spatial resolution of at least 1 m, rivalled the accuracy of LiDAR DEMs, largely owing to the customizability of PHOTO DEMs to the study sites compared to commercially available LiDAR DEMs. We obtained DEMs at spatial resolutions of 6.25 cm–8 m for PHOTO and 50 cm–32 m for LiDAR, where we determined that the optimal spatial resolutions of DEM-derived variables in SDM were between 1 and 32 m, depending on the variable and site characteristics. We found that the reduced extent of PHOTO DEMs altered the calculations of all derived variables, which had particular consequences on their relevance at the site with heterogenous terrain. However, for the homogenous site, SDMs based on PHOTO-derived variables generally had higher predictive powers than those derived from LiDAR at matching resolutions. From our results, we recommend carefully considering the required DEM extent to produce relevant derived variables. We also advocate implementing a multiscale framework to appropriately assess the ecological relevance of derived variables, where we caution against the use of VHR-DEMs finer than 50 cm in such studies.

Table S2: Student t-tests for MaxEnt analyses.These were used to determine the technology-resolution combination that is able to best discriminate plant occurrence points at a) Para (n=146) and b) Martinets (n=100) from 10 000 random background points for each variable at each site.Values indicated are T-values, with significance represented by p≤0.05 = *; p≤0.01 = **; p≤0.001 = ***.The most significant technology-resolution combination for each variable is highlighted in yellow.Table S3: Summary results ranking MaxEnt species distribution models (SDM) to determine optimal parameters.Feature class (FC: L = linear only; LP = linear and product; LQ = linear and quadratic; LPQ =linear, product and quadratic) and regularization multiplier (RM: 1, 2, 5 and 10) were assessed.Arabis alpina distribution across the Para and Martinets sites were predicted based on plant presence-only points (Para: n=146; Martinets: n=100) and 10 000 random background points at each site.The optimal spatial resolution for each variable as determined in

Figure S1 :
Figure S1: Quantile-quantile (Q-Q) plots for digital elevation model (DEM) vertical error.Error (Δh; meters) was calculated as the difference between the elevation measured at assessment points at a) Para (n=157) and b) Martinets (n=110) with the elevation estimated from the DEMs acquired from LiDAR or photogrammetry (PHOTO) technologies generalised to multiple resolutions.The technology and resolution for the DEM is noted at the top of each Q-Q plot.a) Para Figure S2: Normalized median absolute deviation (NMAD; meters) of digital elevation model (DEM) vertical error.Error (Δh; meters) was calculated as the difference between the elevation measured at assessment points at a) Para (n=157) and b) Martinets (n=110) with the elevation estimated from the DEMs acquired from LiDAR or photogrammetry technologies generalized to multiple resolutions.

Figure S3 :
Figure S3: Scatterplots of Elevation and derived variables from LiDAR and photogrammetry.The values for the eight independent digital elevation model (DEM) derived variables (plot rows) for a) Para and b) Martinets, produced by LiDAR (x-axes) and photogrammetry (y-axes) at the common resolutions of 50cm, 1m, 2m, 4m, and 8m (plot columns), as assessed from 15 000 random points.Regression lines are shown in blue, and Spearman rs (rho) correlation coefficients are marked at the top left-hand corner of plots.All correlations have p-values <0.001.a) Para

Table S1 :
Description and parameters for Elevation and 23 digital elevation model (DEM)-derived variables

Table S2 :
Student t-tests for MaxEnt analyses

b) Martinets Photogrammetry LiDAR Variable 6.25cm 12.5cm 25cm 50cm 1m 2m 4m 8m 50cm 1m 2m 4m 8m 16m 32m
Table S2 was used as input environmental variables.Each MaxEnt model was run 20 times (75% training points, 25% testing points) and mean diagnostic values are shown here.Models were assessed using the mean Area Under the Receiver Operating Curve (Fielding and Bell, 1997) based on the test data (AUCtest), as well mean sample size corrected Akaike Information Criterion (AICc; Akaike, 1974).For each site, we ranked the FC-RM combination by AUCtest and AICc separately, then determined the optimal FC-RM combination as the model resulting in the lowest sum of these ranks (Overall Rank).The top three ranked models at each site are highlighted in yellow.

Table S4 : Summary statistics of digital elevation model (DEM) vertical error.
DEMs were produced using either LiDAR or photogrammetry (PHOTO) technologies and generalized to different spatial resolutions, where DEM vertical error was calculated as the difference between the elevation measured at assessment points at a) Para (n=157) and b) Martinets (n=110) with the elevation estimated from the DEMs.All values are in meters.Measures of statistics assuming a normal distribution were recalculated with outliers removed, using an outlier threshold of 3*RMSE.St dev = standard deviation.RMSE = root mean square error.NMAD = normalized median absolute deviation.