131 
Along-track reflectance (%) Along-track reflectance (%) 
(a) Retrieved albedo (b) Retrieved visibility 
Figure 4. Sensitivity of retrieved parameters to deviatimi from Lambertian reflectance 
32. Sensitivity to variation in aerosol composition 
The atmospheric correction method uses a single parameter to describe variation in atmospheric state, namely 
he boundary layer aerosol loading expressed as ground visibility. An assumption must be made about the optical 
properties of the aerosols existing in the boundary layer. The aerosol optical properties of phase function and 
single scattering albedo however depend on the types of particle present in the boundary layer. In general the 
aerosol composition will be unknown, although some approximation may be made of the basis of location, and 
meteorological conditions. Mie theory or interpolation from databases allow approximate relation of the 
boundary layer composition to aerosol optical properties. In this section we explore how deviation of boundary 
layer aerosols from expected type effects the atmospheric correction. 
Four standard atmospheres are considered, which reflect typical aerosol compositions for continental rural, 
urban, desert and maritime locations. Radiance tables for correction assuming a rural atmosphere were first 
generated. To explore sensitivity to deviation from the implicit assumption of aerosol type, simulated top-of- 
atmosphere radiances were generated for each of the four atmospheric classes while holding surface reflectance 
and visibility constant. Figure 3b shows algorithm operation for the case of an urban atmosphere of 10km 
ground visibility. The figure shows the superimposition of contours formed on the nadir and along-track 
radiance surfaces for channel 1. Both contours are shifted, resulting in intersection at an albedo of 16.26% and a 
visibility of 32.96km. 
The results of parameter retrieval on all channels for a surface albedo of 20% and a visibility of 20km are shown 
in table 1, and for an albedo of 20% and visibility of 10km in table 2. The results show the retrieval of albedo to 
be reasonably robust to deviation from expected aerosol type. The most significant errors occur for the urban 
aerosol model. Anthropogenic aerosols have relatively strong absorption and hence give rise to lower top-of- 
atmosphere radiances. The same radiances are thus explained using a rural atmosphere but with lower surface 
albedo. The retrieved values for ground visibility are much poorer. Although discussion is beyond the scope of 
this paper, the variation in retrieved visibility across channels indicates the possibility of discrimination of 
aerosol types through multi-spectral analysis. 
33. Sensitivity to sensor noise and quantisation 
For the sensitivity analysis of noise and quantisation effects, it is assumed that the dynamic range of the new 
channels will be from zero to the expected radiance at 100% albedo. This is the current default specification, 
although the individual gain and offset for each channel may be set in flight. The signal to noise ratio is 
stipulated at 20:1 at 0.5% albedo. The signal is digitised to 12 bits, and so quantisation errors dominate over 
signal noise. However bandwidth requirements may reduce the number of bits available. To study sensitivity of 
atmospheric correction to a given quantisation, radiances derived from the simulation were corrupted over the 
range equivalent to that quantisation, and subsequently fed into the correction algorithm. Figures 5a and 5b 
show R.M.S. errors in retrieved albedo and visibility for quantisation error ranging from 0.0 to 0.5% of the 
channel dynamic range. Surface albedo was fixed at 20% and visibility at 10km. Use of 8 bit digitisation