Spatial variability of snow depth using UAS technology: Study from Ny-Ålesund, Svalbard, Norway B.C. Bhattarai 1, J.F. Burkhart 1 R. Storvold 2 1 Department of Geoscience, University of Oslo, Norway 2 Northern Research Institute (norut), Tromsø, Norway
Presentation outline Introduction Study area Methods Results Conclusion
Introduction Snow cover is an extremely dynamic surface that is continuously varying over space and time (Scipiòn, et al., 2013), and this variability has a critical role in climate, ecological and hydrological systems both on a local and on a global scale. It is expected the need to efficiently and accurately estimate snow distribution over time for improving several environmental research sectors, but also for increasing the efficiency of water resources management systems. However, despite its relevance, traditional and state-of-the-art methods for estimating snow depth present some serious drawbacks and limitations.
Digital photogrammetry has emerged as a alternative tool to perform such measurements in diverse topography. This research investigates a new alternative method for estimating snow depth spatial distribution by combining two emerging technologies in the geosciences research sectors which are Structure from Motion (SfM) digital photogrammetry and the use of Unmanned Aerial Vehicles (UAVs) in Ny-Ålesund, Svalbard Norway.
Study Area Study is carried out in Ny-Ålesund, Svalbard, North to the Ny-Ålesund airport. Tundra climate Average temp: -5.2 0C Average Preip : 375 mm Overlapping area in both images (area of difference DEM) is = 3.16 Km2 Bare ground (with out snow) elevation ranges from 31 to 68 m asl
Methods We have designed two field survey Aug-2014 Beginning of snow accumulation season Ground images without snow April-2015 At the end to the snow accumulation season Images with snow cover GCPs
Software used Agisoft s Photoscan Pro structure from motion (SfM) algorithm UAV platforms norut cryowing micro cameras utilized were a Canon EOS M 22 mm fixed focal length.
Accuracy Assessment GCPs elevation vs DEM elevation. Statistics (RMSE, Mean difference and SD) were used to evaluate
Results -DEM with 5cm resoultion
Table 1: Accuracy parameter calculated from GCPs altitude and Dem altitude Accuracy Parameters 2015 2014 RMSE (m) 0.16 0.10 SD (m) 0.11 0.07 Mean Difference (m) 0.11 0.08 Figure 2: DEM elevation and GCP elevation comparision for the year 2014 Figure 3: DEM elevation and GCP elevation comparision for the year 2015
Table 2: Measured and calculated snow depth GCPs Lat Lon Altitude Measured Snow depth snow depth from difference DEM Bb7 78.940209 11.83753 62.34902 0.06 0.20 Bb4 78.943138 11.841812 61.87485 0.05 0.08 Ostwest 78.942966 11.832353 64.51966 0.19 0.23 Bb3 78.943092 11.830581 65.28956 0.11 0.14 Varde 78.944866 11.827695 63.55404 0 0.22 Bb2 78.943175 11.815726 58.77866 0.11 0.09 B3 78.943714 11.804485 58.17315 0.19 0.21 Mean difference of 0.08 m RMSE of 0.10
Figure 5 : Difference DEM Figure 6: Histogram for Difference DEM Highest and lowest values ranges from -4.0 to 7.5 The ve and extreme value (~7.5m) is might be due to the construction activity in the study area and is also source of error in this study. Distribution of the pixel value is shown in figure 6, indicating that most of the area covered by snow with depth ranges from 0-2 m.
Some issues South of Ny-Ålesund Airport site Contrast issue???
(CLAH method) Histogram equalization by Contrast Limited Adaptive Histogram Equalization
With out upper
With out lower
With out right
With out left
With out Gcps
Question Where is the proble? In contrast?? Number of GCPs?? Accuracy within GCPs??...???
Conclusions DEM with 5cm resolution for the study area is successfully generated. Result show that the U.A.S. technique provides an accurate estimation of point snow depth value (the mean difference with reference to manual measurements of 0.08 m and RMSE of 0.10 ), and distributed evaluation of the snow accumulation patterns. Spatial distribution of Snow has some issues that need to be justified. This UAV born snow depth data can be used as input data for snow distribution modelling
Thank You Question/ Suggestion