Glacier inventory using airborne laser scanner data

Document Sample
Glacier inventory using airborne laser scanner data Powered By Docstoc
Glacier inventory using
airborne laser scanner data
by Christoph Knoll and Hanns Kerschner, University of Innsbruck, Austria

Airborne laser scanner (ALS) data is a useful tool for facilitating the compilation of glacier topographies.

       he observation of glaciers and        aspect and the size of the individual           is a derivative of an already existing
       their changes in the past, present    glaciers. With these parameters which           one [7] it was necessary to adapt this
       and future has provided valuable      are calculated from the ALS data by             glacier delineation algorithm so that it
information on the climate and on            using the adopted delineation algorithm         works not just for a single glacier but
                                             of Kodde et al [7] the outlines are             for all possible glaciers in a raster data
how these changes affect the local
                                             derived. The necessary data for this            file.
economy, natural hazards and water
supply. In this study glacier outlines are   study were acquired for the whole               The raw data of the ALS DEM was
calculated half-automatically with the       area of the Autonomous Province of              recorded by the company Compagnia
                                             Bolzano-South Tyrol (Italy) in the year         Generale Ripreseaeree S.p.A. (CGR) as
help of airborne laser scanning raster
                                             2005.                                           ordered by the Autonomous Province
data and digital orthophotos.
                                                                                             of Bolzano–South Tyrol for the whole
Besides photogrammetry ALS has               Data acquisition and data                       South Tyrolean territory (see Fig. 1).
become one of the standard methods           processing                                      The ALS DEM was calculated from ALS
for the acquisition of topographic data                                                      point data with a nearest neighbour
                                             All glaciers and remote firn areas in
for many applications during the last                                                        interpolation method on the last pulse
                                             the research area of South Tyrol (see
decade [1]. This specific method using                                                       returns [12].
                                             Fig. 1.) are located in high mountain
ALS technology is characterised by a
                                             regions which entails certain problems          The initial ALS point data was surveyed
high degree of automation in terms
                                             for the application of ALS. The rugged          using "Optech ALTM 3033" and
of data recording and computer-aided                                                         "TopoSys Falcon II" scanner systems
                                             topography of these areas with large
data analysis methods, both in                                                               at the end of the ablation period
                                             differences in the relief can cause
open source software (i.e. with                                                              in 2005. Basic information for the
                                             serious troubles on ALS systems with
GRASS GIS) or Windows-based                                                                  georeferencing of the ALS data was
                                             a limited laser range, point density
software (i.e. with ArcGIS) [2]. The                                                         recorded during the scanning flight with
                                             and spectral resolution for the data
results of these analyses provide a                                                          an integrated Global Positioning System
                                             acquisition (e.g. [10, 2, 11]).
broader spatial coverage than in situ                                                        (GPS) and an inertial measurement
measurements.                                As the algorithm used in this study             unit (IMU). An average point density

Various attempts have been made
worldwide in the last few years to
utilise ALS for glaciological purposes
[3, 4, 5, 6,7,8,9].

Airborne laser scanner data is a
useful and powerful tool to facilitate
the compilation of recent, past or
historic glacier topographies, because
the spatial extent of glaciers can be
monitored semi-automatically with
the help of raster data calculated
from ALS data. The topographies of
glaciers can be derived from ALS raster
data by a classification of the terrain
model into the classes “glacier” and
“non-glacier” area [7]. The single
parameters that are used for the
derivation of the glacier topographies
are the topographical smoothness,
the connectivity, the hydrological
constraints as well as the slope, the        Fig. 1: Glacier covered mountain ranges in South Tyrol in 2006.

PositionIT - June 2009                                                                                                                    27
     SURVEYING              technical

                                                                                        of 8 points per 25 m² for areas
                                                                                        below 2000 m a.s.l. with an elevation
                                                                                        accuracy of ±0,4 m and 3 points per
                                                                                        25 m² for areas above 2000 m a.s.l.
                                                                                        with an elevation accuracy of ±0,55 m
                                                                                        was achieved. The vertical accuracy
                                                                                        over a control area is stated with
                                                                                        0,095 m [13]. The full details of the
                                                                                        collected data have not been released

                                                                                        The derived ALS DEM is a 2,5 m raster
                                                                                        dataset which is calculated from ALS
                                                                                        point data using a nearest neighbour
                                                                                        interpolation method on the last
                                                                                        pulse returns. The resulting raster
                                                                                        data forms the input for the updated
                                                                                        algorithm presented in the following
                                                                                        section [7].

                                                                                        Orthophotos of the year 2006 were
                                                                                        used for the verification of the glacier
                                                                                        outlines as for the recording year of
                                                                                        the DEM no orthophotos are available.
                                                                                        This provides a sufficient accuracy for
                                                                                        further calculations.

                                                                                        The glacier algorithm

                                                                                        UNESCO [14] defines a glacier as “a
                                                                                        mass of ice with a minimum size of
                                                                                        1 hectare”. For the identification of the
                                                                                        debris covered parts of the glaciers,
                                                                                        the surface shape and roughness of the
                                                                                        DEM, as well as a hillshade of the DEM
                                                                                        and existing orthophotos of the years
                                                                                        2000, 2003 and 2006 were used. The
                                                                                        final identification and correction of the
                                                                                        glacier boundaries was done manually
                                                                                        using QGIS and ArcGIS software.
     Fig. 2: Workflow for the glacier delineation algorithm and outline verification.   In Fig. 2 the workflow of the used
                                                                                        algorithm is pictured.

                                                                                        In a first step, the southern Stubai
                                                                                        Mountains in the north of South Tyrol
                                                                                        were chosen for testing the original
                                                                                        delineation algorithm by Kodde et al
                                                                                        [7]. There several different glacier
                                                                                        types such as valley and mountain
                                                                                        glaciers or glacierets can be found.
                                                                                        The classification of glaciers is done
                                                                                        with three criterions: smoothness,
                                                                                        connectivity and hydrological
                                                                                        constraints. In the original delineation
                                                                                        rule described in Kodde [15] the laser
                                                                                        intensity map was used as well.

                                                                                        After checking the results of the
                                                                                        delineation procedure (Fig. 2) some
                                                                                        improvements in the operation method
                                                                                        of this calculation rule were made, as
                                                                                        the original method was developed for
                                                                                        a different dataset based on different
                                                                                        raster resolutions.

                                                                                        The improvement of the used algorithm
     Fig. 3: Calculated (blue) and corrected (red) glacier boundaries of                of Kodde [15] and Kodde et al [7]
     Übeltalferner (Stubai Alps).                                                       is based on the thorough testing of

28                                                                                                       PositionIT - June 2009
                                                                                                 SURVEYING               technical

different raster datasets as input data
(resolutions of 1 m, 2,5 m and 5 m)
and the effects of the different raster
resolutions, the adaption of the default
values (threshold and window size) for
the work with the 2,5 m ALS raster
DEM or even bigger, the enlargement
of the script for the calculation of all
possible glaciers and the setting of the
minimum glacier area to 1 ha. In a
second step the revised algorithm was
used to calculate all glaciers in South

The results of this calculation (blue
in Fig. 3) are good but not accurate
enough especially in the ablation areas
of the glaciers and so the results of the
derivation needed to be verified with
orthophotos of 2006 (red in Fig. 3).

For the recording year of the DEM no
orthophotos were available. Therefore
the glacier outlines have been corrected
with the hillshade which was calculated
from the DEM and the corresponding
orthophotos of the survey flight of

Additional information such as drainage
area, identification numbers, glacier
name, glacier area, aspect of the
ablation area etc. were added to the        Fig. 4: Results of the delineation algorithm (blue) and the correction via orthophotos and
                                            hillshades (red) of Übeltalferner (Stubai Alps).
corrected glacier boundaries. Once
the glacier boundaries had been
                                            parts or have a different hydrological           Glacier changes between 1983, 1997
determined, information such as
                                            constraint are not taken into account            and 2006
area-elevation distributions and other
                                            by the algorithm. The corrected glacier
statistical values of the glaciers were                                                      In line with a global trend, the effects
                                            extent (red in Fig. 4) shows the result
derived.                                                                                     of glacier recession in the investigation
                                            for the glacier that is just limited in its
                                            accuracy by the resolution of the raster
Problems with the used algorithm
                                            dataset and the orthophotos.
As this study is an advancement of
the work of Kodde et al [7] certain         Results of the glacier inventory
problems still remain unsolved. The
                                            Comparison of different datasets
glacier delineation is based on criteria
like the smoothness of the surrounding      The glacier inventory data for the year
area, the connectivity and the              1983 is not available digitally so a
hydrological constraints [7], it is still   direct comparison and quantification
not possible to detect glaciers without     of the glacier changes for that year is
human supervision and error handling.       not possible. There is no information
This algorithm was tested for the first     available regarding which glacier
time for more than a single glacier and     definition was used and whether the
for more glacier types than just a valley   firn boundaries are included.
glacier. Good results were achieved for     For the inventory of the year 1997
all three glacier types (valley glacier,    shapefiles and tables are available.
mountain glacier and glacierete) that       These shapefiles have been calculated
occur in the South Tyrolean Alps.           by the Technical University of Munich
However, Fig. 4 (Übeltalferner in           with a Plancomp P1 system using aerial
                                            photographs of the year 1996.
the Ridnaun valley) shows the
main problems which need further            For 2006 the whole dataset of the
improvement. The results of the             research area was calculated by the
algorithm (blue in Fig. 4) are satisfying   Institute of Geography at the University
but two extensive parts (see 1 and 2        of Innsbruck and is described above.
in Fig. 4) which are connected to the       The results of the inventory are
main glacier through debris covered         shapefiles and tables.

PositionIT - June 2009                                                                                                                   29
     SURVEYING             technical

     area shows problems in compiling                 third glacier inventory of 2006 for the     raster data is a highly accurate tool
     glacier inventories, i.e. the multiple           research area (-14,7%, from 109,7           for monitoring glaciers but still needs
     parting of a former single glacier into          km² to 94,1 km²). Our results show          human supervision. The glacier
     two or more parts. A comparison of               large variations for the extent of small    boundaries are delineated from ALS
     the three inventories (1983, 1997                glaciers (< 0,1 km²). These glaciers        data of a raster resolution of 2,5 m
     and 2006) is difficult because of the            lost more than 72% of their area on         but they need to be verified with
     disintegration of previously continuous          average. For the larger glaciers (> 0,1     corresponding orthophotos as especially
     glaciers. The biggest glacier in South           km²) the mean area loss was 31%.            on the small South Tyrolean glaciers
     Tyrol, the Übeltalferner, split up into          Similar observations for small glaciers     the amount of the relative error (for
     four individual parts of different               have been made by Lambrecht and             glaciers > 0,5 km² between 5 and
     sizes. In order to compare the three             Kuhn [16] for the new Austrian glacier      39%) is not acceptable.
     inventories it is necessary to consider          inventory and Paul et al [17] for the
     all parts of the former glacier as               new Swiss glacier inventory.                One important conclusion can be
     one single entity. The World Glacier                                                         drawn from the use of ALS data in
                                                      A comparison of the area-altitude           the compilation of glacier inventories.
     Monitoring Service (WGMS) assigns
                                                      distribution is only possible for the       ALS technology is a powerful tool
     each glacier a worldwide unique
     inventory number which makes it                  glacier inventories of 1997 and 2006        which opens new possibilities in areas
     possible to summarise these glacier              (Fig. 5) as no elevation data for the       with predominantly small glaciers
     parts. A total of 205 glaciers were              year 1983 is available. The results show    (< 0,1 km²) but the ALS survey
     identified in the research area for the          the maximum glacier covered area at         missions are very cost-intensive.
     year 1983, in 2006 only 186 glaciers             elevations of 3000 to 3100 m a.s.l.
                                                      in 1997 and 3100 to 3200 m a.s.l.           As a direct consequence of climate
     are left so in the last 23 years 19
                                                      in 2006. The largest area reduction         change, the South Tyrolean glaciers
     glaciers collapsed and melted away
                                                      between the inventories of 1997 and         show a significant negative trend in
                                                      2006 is shown in altitudes ranging from     terms of area and volume changes.
     The analysis for the three different             2700 m to 3000 m a.s.l. where glacier       During the last 23 years glaciers have
     datasets shows an overall glacier                area decreased by 9,9 km². The mean         shown a strong reduction in both area
     area reduction of -31,6%, from 136,8             glacier elevation in 1997 has been at       and number. Nineteen of them have
     km² to 93,4 km², for the period from             an altitude of 2980 m a.s.l., in the year   melted away completely, some of them
     1983 to 2006. In the 14 years from               2006 it was at 3020 m a.s.l..               have shrunken to an extent that is per
     the first glacier inventory in 1983 to                                                       definition not a glacier anymore and
     the second inventory in 1997, the                Conclusions and perspectives                some are in danger of degeneration.
     loss was (-19,7%, from 136,6 km² to                                                          Almost all altitudes are affected by the
                                                      The calculation method presented
     109,7 km²) and in the last nine years
                                                      in this study shows that ALS based          melting process but elevations between
     from the inventory of 1997 to the new
                                                                                                  2600 m and 2900 m are especially


                                                                                                  The project is funded by a PhD
                                                                                                  scholarship from the University of
                                                                                                  Innsbruck. The authors want to thank
                                                                                                  Martin Kodde for his guidance and
                                                                                                  help in improving the glacier outline
                                                                                                  algorithm. Thanks also to Georg Kaser,
                                                                                                  Michael Kuhn, Roberto Dinale, Astrid
                                                                                                  Lambrecht, Jakob Abermann, Markus
                                                                                                  Tusch and the Glaciological Seminar
                                                                                                  Innsbruck for their support.

                                                                                                  This paper was presented at the
                                                                                                  Free and Open Source Software for
                                                                                                  Geospatial Conference 2008 (FOSS4G
                                                                                                  2008) and is published here with the
                                                                                                  permission of the authors.

                                                                                                  [1] N Arnold, W Rees, B Devereux and G
                                                                                                      Amable: Evaluating the potential of
                                                                                                      high resolution airborne LiDAR data
                                                                                                      in glaciology. Int. J. Rem. Sens., Vol.
                                                                                                      27(6), pp. 1233-1251, 2006.
                                                                                                  [2] T Geist and J Stötter: First results on
                                                                                                      airborne laser scanning technology as
                                                                                                      a tool for the quantification of glacier
     Fig. 5: Area-altitude distribution for 1997 (orange) and 2006 (grey) and the absolute area       mass balance. EARSeL eProc., Vol.
     change between the two inventories in 100 m elevation bands.                                     2(1), pp. 8-14, 2003.

30                                                                                                                   PositionIT - June 2009
                                                                                                  SURVEYING               technical

[3] T Geist and J Stötter: Documentation       [8] C Hopkinson and MN Demuth: Using                Geographisches Informationssystem.
    of glacier surface elevation change with       airborne lidar to assess the influence          Beschreibung der kartographischen
    multitemporal airborne laser scanner           of glacier downwasting on water                 Daten. Bozen, Autonome Provinz Bozen
    data - case study: Hintereisferner and         resources in the Canadian Rocky                 - Südtirol. (Abteilung Raumordnung,
    Kesselwandferner, Tyrol, Austria. Z            Mountains. Can. J. Rem. Sens., Vol.             Amt für überörtliche Raumordnung
    Gletscherkd. Glazialgeol., Vol. 41, pp.        32(2), pp. 212-222, 2006.                       27.1.), 2006.
    77-106, 2007.                              [9] E Schiefer, B Menounos and R Wheate:        [14] UNESCO/IASH: Perennial ice and
[4] J Lippert, M Wastl, J Stötter, A.              A inventory and morphometric analysis            snow masses: a guide for compilation
    Moran, Th Geist and C Geitner:                 of British Columbia glaciers, Canada.            and assemblage of data for a world
                                                   J. Glaciol., Vol. 54(186), pp 551–560,           inventory. Paris, France, UNESCO/
    Measuring and modelling ablation and
                                                   2008.                                            IASH. (Technical Papers in Hydrology 1,
    accumulation on glaciers in Northern
                                               [10] J Albertz: Einführung in die                    A2486.) 1970.
    Iceland. Z Gletscherkd. Glazialgeol.,
    39, 2003/2004, pp. 87-98, 2006.                 Fernerkundung. Grundlagen der              [15] M Kodde: Glacier surface analysis:
                                                    Interpretation von Luft- und                    airborne laser scanning for monitoring
[5] B Höfle, T Geist, M Rutzinger and N
                                                    Satellitenbildern. Third Edition.               glaciers and crevasses. (M.Sc. thesis,
    PfeiferL: Glacier surface segmentation          Darmstadt, Wissenschaftliche                    Department of Earth Observation and
    using airborne laser scanning point             Buchgesellschaft, 2007.                         Space Systems, Delft University of
    cloud and intensity data. ISPRS – Int.                                                          Technology, Delft, The Netherlands.)
                                               [11] E Lutz, T Geist and J Stötter:
    Arch. Photogramm. Rem. Sens. Spatial                                                            2006.
                                                    Investigations of airborne laser
    Inf. Sci., Vol. 36(3/W52), pp. 195-200,
                                                    scanning signal intensity on glacial       [16] A Lambrecht and M Kuhn: Glacier
                                                    surfaces – utilizing comprehensive laser        changes in the Austrian Alps during the
[6] T Geist, H Elvehøy, M Jackson                   geometry modelling and orthophoto               last three decades, derived from the
    and J Stötter: Investigations on                surface modelling (a case study                 new Austrian glacier inventory. Ann.
    intra-annual elevation changes using            Svartisheibreen, Norway). ISPRS – Int.          Glaciol., Vol. 46, pp.177-184, 2007.
    multi-temporal airborne laser scanning          Arch. Photogramm. Rem. Sens. Spatial       [17] F Paul, A Kääb, M Maisch, T
    data: case study Engabreen, Norway.             Inf. Sci., Vol. 34(3/W13), pp. 143-148,         Kellenberger and W Haeberli: Rapid
    Ann. Glaciol., Vol. 42, pp. 195-201,            2003.                                           disintegration of Alpine glaciers
    2005                                       [12] R Wack and H Stelzl: Laser DTM                  observed with satellite data. Geophys.
[7] M Kodde, N Pfeiffer, B Gorte, T Geist           generation for South-Tyrol and                  Res. Lett., Vol. 31(21), L21402.
    and B Höfle: Automatic Glacier Surface          3D-visualization. ISPRS - Int. Arch.            (10.1029/2004GL020816.), 2004.
    Analysis from Airborne Laser Scanning.          Photogramm. Rem. Sens., Vol. 36(3/
                                                    W19), pp. 48-53, 2005.                     Contact Christoph Knoll,
    ISPRS – Int. Arch. Photogramm. Rem.
    Sens. Spatial Inf. Sci., Vol. 36(3/W52),   [13] G Zanvettor, D Colmano and M               University of Innsbruck,
    pp. 221-226, 2007                               Iellici: Landeskartographie und   

PositionIT - June 2009                                                                                                                        31