Hello all,

We are looking at some multispectral images that were captured over a period of 90 minutes such that exposure settings were kept fixed. The sensor (Ultracam-D, Vexcel) has flat-fielding done i.e. the pixel values measure at-sensor radiance similarly across the focal plane. But, absolute calibration is missing and we cannot convert the pixel values into wide-band at-sensor radiance data (offset and gain are unknown). Its a CCD sensor so simple radiance = offset+gain*pixelvalue relationship should work.

Now, by using repeated observations (same camera position) with a temporal lag (of natural targets with reflectance reference values (HCRFs observations) we could derive estimates for the the offset and gain terms, and average Path radiance.

The bands are centered at wavelengths of 450 (BLU), 520 (GRN), 620 (RED), and 800 (NIR) nm. The spectral response functions are wide and overlap slightly. Its an aerial camera (eight cameras).

It was a sunny morning and aerosol optical thickness values increased mildly during the 90-minute campaign. Radiance measurements of downwelling irradiance (400-700 nm) show an increase of 25% during the campaign owing to the Sun climbing higher.

With the help of the reference reflectance targets, dark pixels (shaded waterways) and repeated obsevations we could deduce that

if Lpath (BLU) is 100 units, GRN was 90, RED was 60 and NIR was 40 (on average). The camera was 1.5 km above ground and view zenith angles were 0-25 degrees.

Are we on the right track? We don't have access to MoDTRAN or other software to verify that the proportions are of the right magnitude.

We study differences between tree species in how they response to increasing solar illumination (and decreasing solar zenith angle) and would like to turn our pixel values towards being ratio-scale observations of target radiance so that the differences we observe could be turned into proportions (response of species X is 10% from that of species Y). We should be ok if the Lpath ratios above are of the right magnitude.

br ilkka korpela

Now, our estimates