Spectral library of glacier surface reflectances

Background and study area

The current research activities find their roots in the EU-funded OMEGA project (http://omega.utu.fi) that dealt in particular with glaciology and climate change in a framework of remote sensing, GIS and visualisation methods. The study site at hand is the glacier Hintereisferner located near the border of Austria and Italy (46° 48´ N, 10° 46´ E) in the Tyrolean Alps (Figure 1).

Figure. 1: 3-Dimensional view (towards NE) of the location of Hintereisferner (HEF) in the Austrian Alps. The image consists of a Landsat TM image (1999) in band combination 5/4/3 that was resolution-merged with its panchromatic band and draped over a 30-m DEM.

Studies of the mass balance of small glaciers in the Alps have drawn attention during the last decades. The worldwide retreat of glaciers provides one of the clearest signals of a change in global climate (Knap 1997). A glacier surface may comprise snow, firn, ice and various intermediate mixtures of them. It has been shown by Zeng et al. (1984) that a glacier surface reflects differently in certain wavelength regions of the visible and near-infrared part of the solar spectrum. From an Enhanced Thematic Mapper (ETM+) image, surface reflectances can be computed after performing radiometric, atmospheric and topographic corrections. In addition, ground-based spectral measurements provide supportive data at point specific locations.

 

Aims of the research

Within this project remote sensing techniques are applied and developed to conduct research on the temporal behaviour of the glacier Hintereisferner in terms of glacier dynamics, snow/ice distributions and reflective properties using optical remote sensing datasets. The objective of this project to make a quantitative comparisons between satellite-derived reflectances and in situ nadir reflectances obtained during fieldwork. Parameterised functions that use the ETM+ bands 2 and 4 are used to estimate the reflectances from snow, firn and ice.


Recent results

On 02.09, 03.09 and 04.09 of the year 2002, nearly 600 spectra were acquired from horizontal parts of Hintereisferner using the FieldSpec® Pro FR spectrometer (Figure 2). The spectrometer has a spectral sampling interval of 1.0 nm in the spectral interval 350 nm - 2500 nm. On the glacier distinct surface covers wer chosen that are likely to have different and easy to discern reflection patterns (Zeng et al. 1984). For snow, firn and ice the following classification scheme was made: New Wet Snow (NWS), Old Dirty Snow (ODS), Old Clean Snow (OCS), Firn (FI), Clean Ice (CI) and Dirty Ice (DI).

Figure 2: Spectrometer measurement on Hintereisferner

 

Per location, approximately 6-7 measurements together with 2 reference measurements were made. Reflectances from the resulting spectra were computed by dividing the spectral response by the average of the reference measurements before and after each measurement. Using this method, parameters like changing illumination conditions and optical throughput of the spectrometer were mathematically eliminated (Goetz 1992).


The broadband reflectance (Equation 1) was calculated according to the equation proposed by Knap (1997). The indices in the equation refer to the spectral range of the ETM+ bands.

Substitution of the values 0.43 and 0.40 (a and b) showed the best result (Figure 3). In total 65 points were used to perform the analysis with no regard to different glacier surface.

Figure 3: Single relationship between measured and modelled average spectrometer measurements for different glacier surfaces (R2 = 0.977).

 

The separate weighing functions for different surfaces are shown in Table 1. Predicting the average reflectance between 350 nm - 1200 nm using the spectral intervals of ETM+ bands 2 and 4 works very well. R 2 values of the relationship between measured and modelled spectrometer measurements are extremely high and almost perfect (Hendriks & Pellikka 2004). For relations between measured and modelled broadband albedo, Knap (1997) found very high correlation coefficients as well.

 

Table 1: Weighing functions based on spectrometer measurements to estimate average reflectance using the spectral intervals 530-610 nm and 780-910 nm.

Weighting function

Surface

No. measurements

R 2 measured-modelled

α = 0.328α2 + 0.635α4

New wet snow (NWS)

14

0.999

α = 0.221α2 + 0.645α4

Firn (FI1-2)

16

0.998

α = 0.430 α2 + 0.407α4

Clean ice (CI1)

9

0.998

α = 0.247α2 + 0.582α4

Dirty ice (DI2-3)

13

0.999

α = 0.743α2 - 0.148α4

Old clean snow (OCS1)

4

0.999

α = -0.228α2 - 1.229α4

Old dirty snow (ODS2)

9

0.997

α = 0.400α2 + 0.474α4

Entire glacier surface

65

0.999

 

 

Contact

 

References

Goetz, A.F.H. (1992). Principles of Narrow Band Spectrometry. In F. Toselli & J. Bodechtel (Eds.), The Visible and IR: Instruments and Data Analysis in Imaging Spectroscopy: Fundamentals and Prospective Applications (pp. 21-32). Brussels and Luxembourg.

Hendriks, J.P.M. & P. Pellikka, 2004. Estimation of reflectance from a glacier surface by comparing spectrometer measurements with satellite-derived reflectances. Zeitschrift für Gletscherkunde und Glazialgeologie 38(2) : 139-154.

Knap, W.H. (1997). Satellite-derived and ground-based measurements of the surface albedo of glaciers. Ph.D. thesis, Institute of Marine and Atmospheric Sciences. Rijksuniversiteit Utrecht, ISBN90-393-1983-9, 175 pp.

Zeng, Q., Cao, C.M., Feng, X., Liang, F., Chen, X., & Sheng, W. (1984). Study on spectral reflection characteristics of snow, ice and water of northwest China. In B. Goodison (Ed.), Hydrological applications of remote sensing and remote data transmission (pp. 451-462). IAHS Publication No. 145.

Cooperation partners

  • Department of Geography, University of Innsbruck
  • Department of Meteorology and Geophysics, University of Innsbruck