154
(iv) Airborne INTA-AHS flight campaigns were conducted on 28th September
2005 (dry season) and 28th April 2008 (wet season). For the 2005 flight campaign, we acquired two flight lines in a north-east to south-west direction and
in the direction of the solar principal plane. The flight altitude was 2743 m
above sea level, which gave a pixel size of 6.5 m. For the 2008 flight campaign, we needed three flight lines with the same properties to cover the stabilised sand ecosystem as the solar azimuth was larger. All of the flight lines
were acquired using an integrated Applanix POS/AV navigation system,
which relies upon a GPS/INS for accurate determination of the instrument
position and orientation.
(v) The flight lines were processed with INTA-PAF to generate a mosaic of georefenced ground reflectance over the stabilised sand ecosystem for each date.
The sensor radiometric and spectral calibration obtained at INTA facilities
encompasses the conversion to digital numbers to at-sensor radiance (in μW/
m
2
sr nm units). The imagery was directly georeferenced by PARGE software
(Schläpfer and Richter 2002) using the GPS/INS values during the flight, the
bundle adjustment parameters calculated in the corresponding year, and the
DEM. The atmospheric correction was implemented using the ATCOR-4 software (Richter and Schläpfer 2002), which performs a Look-Up Table (LUT)
with the code MODTRAN-5 (radiative transfer model). This compensates for
the atmospheric effect in relation to flight altitude, aerosol type, visibility, and
water vapor content on a per-pixel basis. Furthermore, ATCOR-4 performs the
correction of adjacency effect on a per-pixel basis. Mosaics for the study area
can be generated using remote sensing commercial software, including the
Exelis Visual Information Solutions (ENVI).
(vi) Ground reflectance mosaics generated for both flight campaigns were evaluated for data quality. Radiometric accuracy is a function of the sensor calibration and atmospheric correction applied. The reflectance values obtained in
the INTA-AHS imagery processing were evaluated using ground reflectance
data acquired with field spectroscopy in the sand dunes in Doñana. The geometric accuracy was estimated using ground control points extracted from
digital cartography.
(vii) A Linear Spectral Unmixing Model was applied to the image mosaics using
the endmembers derived in the spectral libraries. MESMA unmixes each pixel
using different combinations of potential endmembers and was implemented
in the commercial Visualization & Image Processing for Environmental
Research (VIPER) software (Roberts et al. 2007). For vegetation mapping
studies, the recommended approach is that every pixel in the images can be
modeled by a linear combination of three land-cover types (Roberts et al.
1998): photosynthetic vegetation (Veg), non-photosynthetic vegetation
(Litter), substrate (soil), and a shade component (Shade) that is typically also
present in all pixels. The mixture model that describes Doñana’s shrubland is:
M. Jiménez and R. Díaz-Delgado
(iv) Airborne INTA-AHS flight campaigns were conducted on 28th September
2005 (dry season) and 28th April 2008 (wet season). For the 2005 flight campaign, we acquired two flight lines in a north-east to south-west direction and
in the direction of the solar principal plane. The flight altitude was 2743 m
above sea level, which gave a pixel size of 6.5 m. For the 2008 flight campaign, we needed three flight lines with the same properties to cover the stabilised sand ecosystem as the solar azimuth was larger. All of the flight lines
were acquired using an integrated Applanix POS/AV navigation system,
which relies upon a GPS/INS for accurate determination of the instrument
position and orientation.
(v) The flight lines were processed with INTA-PAF to generate a mosaic of georefenced ground reflectance over the stabilised sand ecosystem for each date.
The sensor radiometric and spectral calibration obtained at INTA facilities
encompasses the conversion to digital numbers to at-sensor radiance (in μW/
m
2
sr nm units). The imagery was directly georeferenced by PARGE software
(Schläpfer and Richter 2002) using the GPS/INS values during the flight, the
bundle adjustment parameters calculated in the corresponding year, and the
DEM. The atmospheric correction was implemented using the ATCOR-4 software (Richter and Schläpfer 2002), which performs a Look-Up Table (LUT)
with the code MODTRAN-5 (radiative transfer model). This compensates for
the atmospheric effect in relation to flight altitude, aerosol type, visibility, and
water vapor content on a per-pixel basis. Furthermore, ATCOR-4 performs the
correction of adjacency effect on a per-pixel basis. Mosaics for the study area
can be generated using remote sensing commercial software, including the
Exelis Visual Information Solutions (ENVI).
(vi) Ground reflectance mosaics generated for both flight campaigns were evaluated for data quality. Radiometric accuracy is a function of the sensor calibration and atmospheric correction applied. The reflectance values obtained in
the INTA-AHS imagery processing were evaluated using ground reflectance
data acquired with field spectroscopy in the sand dunes in Doñana. The geometric accuracy was estimated using ground control points extracted from
digital cartography.
(vii) A Linear Spectral Unmixing Model was applied to the image mosaics using
the endmembers derived in the spectral libraries. MESMA unmixes each pixel
using different combinations of potential endmembers and was implemented
in the commercial Visualization & Image Processing for Environmental
Research (VIPER) software (Roberts et al. 2007). For vegetation mapping
studies, the recommended approach is that every pixel in the images can be
modeled by a linear combination of three land-cover types (Roberts et al.
1998): photosynthetic vegetation (Veg), non-photosynthetic vegetation
(Litter), substrate (soil), and a shade component (Shade) that is typically also
present in all pixels. The mixture model that describes Doñana’s shrubland is:
M. Jiménez and R. Díaz-Delgado