Snow depth distribution by airborne LiDAR data

Abstract:
Airborne LiDAR (Light Detection And Ranging) is powerful tool for surveying large area within short period. This attribute is quite usable taking snapshot of variable terrain information such as snow depth of non-permanent snow area like in Japan. In this study a series of airborne LiDAR survey was performed around Tateyama mountain range (Toyama prefecture in Japan) during with/without snow coverage season. The extent of survey area is more than 60 square kilometers. The main purpose of these surveys is precise and high density digital elevation model (DEM) of this area. To eliminate the effect of miss-classification of tree height, we chosen defoliation period as survey season (October 2003) for ground elevation data. According to past meteorological record, we chosen March as maximum snow depth season around this area (in 2007). We also performed careful geometric calibration based on the adjustment with building shape which coordinate was previously survey by ordinary survey method. We present the distribution of snow depth around this area which were estimated by subtraction of those DEMs. The shape of mountain ridge covered by snow implies the regional wind condition.

Introduction
Snow depth distribution is important parameter for snow mountain administration. Appling air borne LiDAR system to snow covered area was report in several papers already [1,2,3]. Tateyama mountain area is located in Toyama prefecture adjacent the coast line of the Sea of Japan. And one of the most prominent heavy snowfall zone in Japan. Airborne LiDAR survey of this area has continuously executed by Ministry of Land, Infrastructure and Transport, Hokuriku Regional Development Bureau, Tateyama SABO Office since 2003. The main objects of these surveys are river erosion control and research of snow depth distribution of this area.

As a part of these LiDAR surveys, two independent data were taken under different season, such as maximum snow time and non snow time. Snow depth distribution of this area is retrieved using both data.

Tendency of snow depth distribution was visualize using superimpose of two type of images. Shape of bare earth is displayed with shaded relief and snow depth distribution is displayed with color gradation.

Using superimposed snow distribution map, we can efficiently understanding brief overview of this area. And also detail of terrain feature such as snow cornice, snow accumulating along to the valley, the trajectory of snow removal along Tateyama alpine rote. The cross sections of two surfaces indicate the shape of snow cornice too. An overview of this study will be presented.

Airborne LiDAR survey and data processing
Airborne LiDAR survey is one of an application of laser range finder on an aircraft platform. The basic mechanism of Airborne LiDAR system is combination of laser range finder and GPS and IMU (Inertial Measurement Unit). Laser range finder measures distance between aircraft and surface. It combined with scanner mirror then laser beam sweep ground level surface left and right perpendicular to the flight direction. GPS determines location of the aircraft (1Hz). IMU detect attitude (rolling, pitching and heading) of the aircraft and interpolate GPS data within short time scale (200Hz). Elaborate 3d ground/snow surface information is finally retrieved by post-processing which combines ground based reference GPS station data and above LiDAR data. In addition to the LiDAR, color digital camera (4,000 x 4,000 pixels) is also integrated in this system.

Airborne LiDAR survey of Tateyama mountain area were conducted October 2003 and March 2007. Each of survey specifications are summarized in Table 1.

Two different LiDAR systems were used in those surveys. The state of the art Leica ALS50-II improves laser pulse frequency up to 150 kHz. But its high laser pulse frequency makes the limitation of flight altitude tighter in mountainous zone. Elevation difference in this area more than 2500 meters. To keep clearance between aircraft and terrain constant, flight altitude was adjusted according to the terrain height. The point density of each LiDAR pulse was better than 1 point/m2 at high elevations and about 1.5 point/m2 in valley bottoms. In order to minimize interference of vegetation, just after the defoliation season was selected as bare earth elevation data acquisition period.

In this area usually snow begins to cover the ground in late around November and reaches maximum depth in middle of March. Almost all snow disappeared in May. The intention of this survey is acquiring maximum snow depth information. So the date of survey was carefully selected from snow depth record of ordinary year and vicissitude of weather condition. Three days were required due to bad weather condition in 2003. But Optech system also had an ability to finish this area in one day.

During the entire of survey GPS condition was acceptable. PDOP was less than 3 and number of satellite is more than 5.

Computed XYZ coordinates of returned laser pulse are distributed as scanning pattern so call point cloud. These XYZ data holds every return from all surface attribute (e.g. forest, artificial structures, snow, bare earth). For subsequent processing these XYZ data classified into bare earth and another by TerraScan (Terrasolid Limited) software. Then bare earth data were girded to generate DEMs. In winter season bare earth data is surface of snow instead of bare earth. Figure 1 is shaded relief using snow surface elevation data and flight courses in March 21, 2007.

LiDAR data refinement
Prior to the classification of XYZ coordinate data, we had performed calibration and refinement of LiDAR data. The flow of this refinement is as follows.

1. Check and adjust the consistency of adjacent data strip. Put elevation check circle inside the every overlapping zone of LiDAR data strip. The number of check circle depends on the length of flight course 2 points shorter than 10 kilo-meters length, 3 points equal and longer than 10 kilo-meters. In this check maximum RMS was 0.18 meters. This check made entire LiDAR data set as smooth surface. No seam, no elevation breaking between every data strip. IMU-laser alignment, scanner offset and scanner scale adjustment were the principal way to resolve this discrepancy [4].
2. Compare the elevation value with reference point. Paved roadway or parking and with sky opened position was selected as reference point. In this study, number of reference point was 6 which mainly distributed in west part of survey area due to terrain condition. These reference points constantly removed snow even at the winter season. Static GPS survey was performed before the snowfall. GPS derived elevation offset was resolved this adjustment. Maximum RMS was 0.10 meters after this adjustment.
3. Compare the horizontal location of artificial structures of both LiDAR data (2003, 2007). We selected 7 buildings in this survey area. Each of them was stable and clearly visible with LiDAR point cloud data. Building shapes traced using both point cloud were located inside of plus or minus 0.6 meters range each other. These actual geographic coordinate was not available but we could confirmed horizontal consistency of both LiDAR data.

Snow depth analysis
Snow depth was calculated by subtracting bear earth elevation from snow surface elevation. Snow depth calculation using independent elevation data is sensitive to horizontal correspondence of each data, especially with steep terrain condition. Even small horizontal error becomes cause of large error of snow depth.

After the careful calibration two type of elevation data were grided to generate 1 meter raster DEMs. Finally 1m girded snow depth distribution was calculated. The entire area of this data is about 72 km2.

These DEMs were effectively visualized using superimpose of several images. In this time we use shaded relief for represent terrain condition of this area. On the other hand, color gradation was chose for snow depth distribution. Superimposing the both images makes understanding of processing result easy (Figure 2). Snow depth varies from 0 meter to 10 meters and above. It is increasing with elevation as an entire tendency. This trend agree with another research of this area [5,6]. Local accumulation is observed inside a valley, basin and behind a ridge.

The various terrain & snow depth features which are show in Figure 2 could be measure by utilizing the CAD and GIS software. Figure 3 illustrates snow the shape of snow cornice and its profiles across the ridge at the east side of this area. Left side of Figure 3 is shaded relief of bare earth and middle is snow surface. In both shaded reliefs red curved lines are mountain ridge line under no snow condition (traced by bare earth data). The middle of Figure 3 indicates peak of mountain ridge was shifted to east about 40 meters. Right side of Figure 3 is cross-section across the mountain ridge. In these figures blue lines are bare earth elevation, red lines are snow surface and green lines show the snow depth along cross-sections. Both cross-sections express snow cornice was formed around east side of this mountain ridge. The maximum snow depth in this snow cornice was about 16 meters. In contrast average snow depth was about less than 1 meter around west side of this mountain ridge. Similar tendency about snow cornice was reported and the relation with regional wind condition way suggested [7].

Conclusion
In this paper airborne LiDAR survey and its application in large snow area (72 km2) within short period (1-3 days) has demonstrated. Airborne LiDAR survey was conducted in two seasons with and without snow condition. The survey area was mountain zone and its elevation varies from 415 to 3003 meters. The point density of each LiDAR pulse was better than 1 point/m2 at high elevations and about 1.5 point/m2 in valley bottoms. At this moment rapid data acquisition and high point density is most useful features of airborne LiDAR survey of this application.

Processed point cloud data of every flight courses were carefully checked and adjusted its elevation consistency. Verification of elevation with pre-surveyed ground based reference points and comparison using artificial structure keep the vertical and horizontal consistency of two seasonal data. These verification ensure the reliability of successive snow depth analysis and also effective to all DEM subtracting analysis.

Snow depth distribution of this area shows noticeable terrain & snow depth features like that of snow cornice in Figure 3. Airborne LiDAR survey is almost only technology to get those high dense information in mountain area which closed in the deep snow coverage.

Acknowledgement
The authors would like to thank Mr. Mitsuo Fukuda, Ministry of Land, Infrastructure and Transport, Hokuriku Regional Development Bureau, Tateyama SABO Office. He pleasantly allows us to use the LiDAR dataset.

Reference

1. W. B. Krabill, R. H. Thomas, C. F. Martine, J. G. Sonntag, R. N. Swift, E. B. Frederick, S. S. Manizade, J. K. Yungel, Airborne Laser Mapping of Greenland, Second International Airborne Remote Sensing Conference and Exhibition, Sanfransisco, California, 24-27, 1996
2. Robert Bindschadler, Mark Fahnestock, Angela Sigmund, Comparison of Greenland Ice Sheet Topography Measured by TOPSAR and Airborne Laser Altimetry, Geoscience and Remote Sensing, IEEE Transactions, 37, 1999
3. Chris Hopkinson, Mike Demuth, Mike Sitar, Laura Chasmer, Applications of Airborne LiDAR Mapping in Glacierised Mountainous Terrain, Geoscience and Remote Sensing Symposium, 2001. IGARSS '01. IEEE 2001 International, 2, 949-951, 2001
4. Dr.-Ing. Rolf Katzenbeisser, ABOUT THE CALIBRATION OF LIDAR SENSOR, ISPRS
5. Workshop "3-D Reconstruction form Airborne Laser-Scanner and InSAR data", 8 – 10 October 2003, Dresden
   • Masaaki Hanaoka, Shinichi HOMMA, Yoichi ITO, Masaaki WATARI, Hajime IIDA, Research on the elevation dependability of Snow depth around the Tateyama caldera by airborne laser scanner, Proceedings of Japanese Society of Snow and Ice, 2006.
   • Hajime IIDA, Osamu WATANABE, Masaaki WATARI, Masaaki HANAOKA, Masanori KANEKO, Shinichi HOMMA, Snow distribution on Mt. Tateyama by using Airborne Lidar, Proceedings of Japanese Society of Snow and Ice, 2006.
   • Masaaki WATARI, Naomi NAGATA, Hajime IIDA, Hiroshi MATSUDA, Masaaki HANAOKA, Creation of Snow depth distribution map around the Tateyama caldera by airborne laser scanner, Proceedings of Japanese Society of Snow and Ice, 2006.