In: Wagner W., Székely, B. (eds.): ISPRS TC VII Symposium - 100 Years ISPRS, Vienna, Austria, July 5-7, 2010, IAPRS, Vol. XXXVIII, Part7B 
448 
The multispectral sensor TM captures high spatial resolution 
scenes onboard the Landsat-5 platform. The instrument has the 
capability to observe earth’s surface via a set of six reflective 
bands with wavelengths ranging from 0.45 to 2.35pm (Figure 
5) and a geometric resolution of 30m at nadir. With a revisit 
cycle of 16 days, one of the sensor’s major applications is the 
monitoring of land cover and land cover changes, although the 
operational use of the data is often limited by a restricted 
availability of cloud-free observations. 
For our analysis, the following acquisition dates have been 
selected (by reason of clear-sky conditions and availability of 
both Terra MODIS and TM data): 19 July 2003, 4 August 2003 
and 28 May 2005. Changes of atmospheric and surface 
conditions due to different overpass times of Terra MODIS and 
Landsat-5 TM were assumed to be negligible. At all dates, data 
were taken from the same satellite paths. 
The study site was located in western Rhineland-Palatinate 
(Germany) close-by Luxembourg and had a size of approxi 
mately 2000 km 2 (Figure 1). The heterogeneous land cover 
included different types of forests, grassland, cropland and 
some urban settlements. 
Figure 1. Subset of the MODIS data set (19 July 2003) with the 
TM scene near the edge of the scan swath. The red line 
illustrates the MODIS nadir line 
Apparent albedo is defined as the ratio of upwelling irradiance 
to downward irradiance in dependency of solar zenith angle and 
wavelength. By the influence of the total downward radiance 
(direct and diffuse), apparent albedo is obviously a function of 
atmospheric conditions (Liang et al., 1999). This implies that 
apparent albedo received via remote sensing systems under 
specific conditions of the atmosphere may not be suitable for an 
application to other atmospheric conditions. The distribution of 
the downward flux after passing through the atmosphere is the 
weighting function for converting spectral albedo to broadband 
albedo. Apparent albedo is equivalent to the measurands 
registered by albedometers or pyranometers in the field (Liang, 
2000). In absence of an atmosphere inherent albedo would be 
equivalent to apparent albedo. 
To retrieve the broadband albedo from narrowband sensors a 
method based on radiative transfer simulations has been 
developed by Liang (2000). The formulae of Liang (2000), 
presented by Eq. (1) and (2), are used to compute MODIS and 
TM total shortwave broadband albedo (0.25-2.5pm) from its 
spectral albedos. Assuming Lambertian surfaces, surface 
inherent narrowband albedos are equal to bidirectional spectral 
surface reflectances received from atmospherically corrected 
satellite observations. 
MODIS: 
otshort = 0.16a i + 0.291 a 2 + 0.243a 3 + 0.166a 4 + 
0.122a 5 + 0.081a 7 -0.0015 
(1) 
TM: 
a S hort~ 0.356ai + 0.13a 3 + 0.373a 4 + 0.085a 5 + 
0.072a 7 -0.0018 
To test the performance of broadband albedo conversion 
formulae, ideal MODIS and TM data have been simulated 
using a HyMap dataset (cf. chapter 3.1). The airborne hyper- 
spectral scanner HyMap rM operates over a wavelength range 
from 0.44 to 2.5pm. With a narrow bandwidth of 10-20nm the 
sensor is used as a nearly continuous spectrometer for tasks like 
environmental pollution monitoring, agriculture and forestry as 
well as soil and natural vegetation mapping (Cocks et al., 
1998). The HyMap data (126 spectral bands, acquisition date 
28 May 2005) covered an area that is 2.8km wide, 13km long 
and extends from the city of Trier (Rhineland-Palatinate, Ger 
many) northwestbound. After an across-track illumination 
correction, the atmospheric correction was performed with the 
FLAASH (Fast Line of-sight Atmospheric Analysis of Spectral 
Hypercubes) module of ENVI™. The geometric correction was 
accomplished with the Parge™ software (see Vohland et al., 
2010). 
The implemented radiometric correction of the Landsat-5 TM 
data comprised sensor calibration and full radiative transfer 
modelling; earth’s surface was assumed to be a Lambertian. 
Pre-processing of the TM data from 28 May 2005 was validated 
with the HyMap dataset; the radiometric consistency of all 
three TM scenes was validated using a set of pseudo-invariant 
objects. 
3. METHODOLOGY 
Different terminologies of albedo exist why a systematic 
definition is necessary. “Inherent albedo” can be distinguished 
from “apparent albedo”. Inherent albedo at a given solar zenith 
angle and wavelength solely depends on surface properties and 
thus is independent from the current atmospheric conditions. 
where a S h or t = total shortwave albedo (0.25-2.25pm) 
oij = spectral albedo from spectral band i 
Simulated Data 
To avoid effects caused by sensor-specific scan geometries, 
“ideal” data have been simulated from the HyMap dataset. Due 
to the nearly continuously spectral coverage of the complete 
VIS/NIR/SWIR-region with narrow bands, the hyperspectral 
dataset is appropriate to generate both synthetic MODIS and 
TM data as a weighted combination of HyMap bands. This was 
achieved by integrating the relevant HyMap bands according to 
the spectral response function of each MODIS and TM band 
(Figure 5). Finally, each channel of MODIS and TM was 
represented as a linear combination of specifically weighted 
HyMap bands (Figure 2). 
Besides the simulation of the spectral characteristics, an 
additional processing step was necessary to account for the 
spatial characteristics of both instruments. For this purpose, the 
Point Spread Function (PSF) of the scanning system was 
applied. Every image collected by an instrument onboard a 
satellite or aircraft platform is affected by current atmospheric 
conditions, blurring caused by the instrument's optics, detectors 
and electronics and platform motion. These influences lead to 
an image with low contrasts and loss of details (Kavzoglu, 
2004). This degradation can be described by the PSF, which has 
the approximate shape of a Gaussian function in both scan- and 
track-direction (Huang et al., 2002; Townshend et al., 2000). To 
fulfill a realistic spatial simulation of the MODIS and TM data 
based on HyMap, each band was convolved with the according 
sensor PSF and afterwards aggregated to the pixel size of 
MODIS (463m) and TM (30m), respectively. This step was