Earth Observing System (EOS). A description of
the planned instrument may be found in the MODIS
Instrument Panel Report (1986) and Salomonson et
al. (1989). For the nadir version of the sensor,
MODIS-N, there will be 36 spectral bands at a
variety of spectral and spatial resolutions and
ranging from the blue to the thermal infrared.
Because they are relatively narrow, the 19 solar
reflective bands (Table 2) were studied to
illustrate the potential effects of spectral band
shifts should interference filters be used to
construct these MODIS-N bandpasses.
Forward runs of the 5S atmospheric code were made
for no spectral shifts, for partial spectral
shifts of 5, 10, and (in some cases) 15 ran, and
for full spectral shifts of 5, 10, and (in some
cases) 15 nm toward shorter wavelengths. The
input conditions were the same as those used for
the TM bands in the previous section, except that
a water surface reflectance case was also used in
addition to the vegetation one. The apparent
reflectances at satellite altitude for shifted
cases were compared to the apparent reflectance at
satellite altitude obtained from the no-shift
cases.
In the absence of any specific design information,
spectral response profiles were synthesized for
the MODIS-N bands under consideration. The same
shape was used for all bands, but scaled in the
wavelength dimension according to the bandwidth.
For example, the relative response profile for the
30 nm band at 0.905 micrometers is listed in Table
3, and Figures 6 and 7 illustrate the profiles for
the unshifted and shifted bandpasses. Because the
5S code is based on a 5-nm grid, the response
values are specified every 5 nm.
The complete results are presented in Tables 4-7
and the 10-nm full-shift cases in particular are
shown in Figures 8 and 9. Substantial differences
arise between the shifted and unshifted results
for many of the bandpasses. The key results are
as follows.
CONCLUDING REMARKS
The effect of partial (long-wavelength side only)
and full spectral band shifts on radiances output
from a radiative transfer code were examined for
the reflective TM bands over vegetation and for
the reflective MODIS-N bands over vegetation and
over water. In the case of TM, the greatest
errors due to spectral band shifts occurred in the
shorter wavelength bands. For example, in TM band
1, a 25% error occurs for a full shift of 15 nm
and a 12% error results from a partial shift of 15
nm. Most MODIS-N bands in the solar reflective
spectrum can be affected by spectral filter
shifts, with errors of several hundred percent
possible in some cases. Full spectral band shifts
almost always give rise to larger errors than do
partial shifts. The main factors responsible for
the effects of spectral shifts on sensor output
vary from band to band and can include surface
reflectance, gas absorption, solar irradiance, and
molecular scattering.
ACKNOWLEDGEMENTS
The author wishes to thank P.N. Slater for
valuable discussions, as well as G. Fedosejevs and
A. Kalil for assistance in the preparation of the
manuscript.
REFERENCES
Dinguirard, M., Begni, G., and Leroy, M. (1988),
SPOT-1 results after 2 years of flight,
Proceedings, SPIE, 924:89-95.
Markham, B.L., and Barker, J.L. (1985), Spectral
characteristics of the Landsat-4 MSS sensors,
Landsat-4 Science Characterization Early Results,
1:1-23-1-56, NASA Conf. Pub. 2355.
MODIS Instrument Panel Report (1986), Earth
Observing System, Volume lib, National Aeronautics
and Space Administration, Code NIT-4, Washington,
D.C., 20546-0001.
(i) The largest errors due to spectral shifts
arise in a water vapour absorption region
in the MODIS-N bands at 0.905, 0.936, and
0.940 micrometers, regardless of whether
the background surface is water or vegeta
tion. Even a 5-nm shift of the 10-nm band
at 0.936 micrometers can alter the apparent
reflectance at the sensor by 100% (cf.
Tables 4 and 6).
(ii) In all but a few instances, partial
spectral band shifts give rise to smaller
errors than full shifts. One exception is
the 50-nm band at 0.659 micrometers when
observing vegetation, which has a
relatively narrow chlorophyll absorption
dip in that spectral region (cf. Tables 4
and 5).
Salomonson, V.V., Barnes, W.L., Mayroon, P.W.,
Montgomery, H.E. and Ostrow, H. (1989), MODIS:
Advanced Facility Instrument for Studies of the
Earth as a System, IEEE Transactions on Geoscience
and Remote Sensing, 27:145-153.
Suits, G.H., Malila, W.A., and Weller, T.M.
(1988), The prospects for detecting spectral
shifts due to satellite sensor aging, Remote
Sensing of Environment, 26:17-29.
Tanré, D., Deroo, C., Duhaut, P., Herman, M.,
Morcrette, J.J., Perbos, J., and Deschamps, P.Y.
(1986), Simulation of the satellite signal in the
solar spectrum, Laboratoire d'Optique
Atmosphérique, Université des Sciences et
Techniques de Lille, 59655 Villeneuve d'Ascq
Cédex, France, 343 pages.
(iii) Outputs from the 5S code runs were examined
to determine the principal spectral
variable causing the differences resulting
from spectral shifts (cf. Tables 4-7). In
about half the MODIS-N bands, surface
reflectance is the main factor whereas, in
other bands, the principal causative
variable can be gas absorption or solar
irradiance or, in the case of the band at
0.470 micrometers imaging water, molecular
scattering.
Teillet, P.M. (1989), Surface Reflectance
Retrieval Using Atmospheric Correction Algorithms,
Proceedings of the 1989 International Geoscience
and Remote Sensing Symposium (IGARSS'89) and the
Twelfth Canadian Symposium on Remote Sensing,
Vancouver, B.C., pp. 864-867.
60