Reflectance Jump Between Two Instruments

[FCarotenuto]

I've been working on the atmospheric correction of a flight that carried both HySpex VNIR1800 and SWIR384 sensors. In order to optimize the atmospheric correction I've used the SPECTRA module taking as reference a series of spectra that were taken on the ground simultaneously to the flight on an area covered by asphalt. This allowed me to get me quite satisfying results when considering the whole spectrum of the two instruments versus the ground reference measurements as it is possible to see from the following figures. 
   

However I found many pixels, even with similar surfaces such as asphalt, where a massive jump in reflectance between the two instruments was clearly visible. The following image shows the VNIR and SWIR spectra for another asphalt pixel in the flight. The top plots show the at-sensor radiance and the bottom plot the reflectance outputted by ATCOR. While radiances appear to be quite overlapped, the reflectance shows a massive jump. 

   

Any clue at what could be causing this?

[reseadmin]

I've been working on the atmospheric correction of a flight that carried both HySpex VNIR1800 and SWIR384 sensors. In order to optimize the atmospheric correction I've used the SPECTRA module taking as reference a series of spectra that were taken on the ground simultaneously to the flight on an area covered by asphalt. This allowed me to get me quite satisfying results when considering the whole spectrum of the two instruments versus the ground reference measurements as it is possible to see from the following figures. 


However I found many pixels, even with similar surfaces such as asphalt, where a massive jump in reflectance between the two instruments was clearly visible. The following image shows the VNIR and SWIR spectra for another asphalt pixel in the flight. The top plots show the at-sensor radiance and the bottom plot the reflectance outputted by ATCOR. While radiances appear to be quite overlapped, the reflectance shows a massive jump. 



Any clue at what could be causing this?

This is a known problem of the dual-sensor Hyspex system (as well as with other dual sensor / detector systems) This jump in the spectrum is due to the different spatial resolution of the VNIR vs the SWIR instrument of Hyspex in combination with residual co-registration errors. Even with high accuracy rectification, such residuals exist on the pixel level as accuracy/offsets can be about 0.3 pixels and also the PSF of the two instruments is different. The only way to get around this is to average a number of pixels for such comparisons  - or to adjust the spectra empirically against each other. We're working on such an empirical correction right now and may add this soon to an atcor or droacor release.

[FCarotenuto]

I've been working on the atmospheric correction of a flight that carried both HySpex VNIR1800 and SWIR384 sensors. In order to optimize the atmospheric correction I've used the SPECTRA module taking as reference a series of spectra that were taken on the ground simultaneously to the flight on an area covered by asphalt. This allowed me to get me quite satisfying results when considering the whole spectrum of the two instruments versus the ground reference measurements as it is possible to see from the following figures. 


However I found many pixels, even with similar surfaces such as asphalt, where a massive jump in reflectance between the two instruments was clearly visible. The following image shows the VNIR and SWIR spectra for another asphalt pixel in the flight. The top plots show the at-sensor radiance and the bottom plot the reflectance outputted by ATCOR. While radiances appear to be quite overlapped, the reflectance shows a massive jump. 



Any clue at what could be causing this?

This is a known problem of the dual-sensor Hyspex system (as well as with other dual sensor / detector systems) This jump in the spectrum is due to the different spatial resolution of the VNIR vs the SWIR instrument of Hyspex in combination with residual co-registration errors. Even with high accuracy rectification, such residuals exist on the pixel level as accuracy/offsets can be about 0.3 pixels and also the PSF of the two instruments is different. The only way to get around this is to average a number of pixels for such comparisons  - or to adjust the spectra empirically against each other. We're working on such an empirical correction right now and may add this soon to an atcor or droacor release.

Thank you for your prompt response! I would just like to ask a further explanation to understand better these kinds of issues. The at-sensor radiance images have the same co-registration as the reflectance ones (they have already undergone radiometric and geometric correction) so why isn't the jump present there if it's an issue of co-registration?
Thank you again for any further information!

[reseadmin]

The detector split wavelength is in the middle of the strongly absorbing 940nm water vapor band. Therefore, the jump is not as obviously visible in the radiance spectrum but appears only after reflectance correction.