| Re: Convolve a spectrum (from MODTRAN) with a spectral response function [message #69346] |
Mon, 04 January 2010 14:39  |
![Paul Van Delst[1] is currently offline](theme/default/images/xoffline.png.pagespeed.ic.XRkd1fkXye.png "Paul Van Delst[1] is currently offline")
Paul Van Delst[1]
Messages: 1157
Registered: April 2002
|
Senior Member |
|
|
Your SRF (spectral response function) seems to only increase (looks like only half an
SRF?), and your computed radiances do not span the bit of SRF that you do have.
Also, in that part of the IR spectrum you have some water vapour line absorption
contamination so I would be suspect of a convolution at such a coarse resolution - I would
think you would have a bias that varies with the column water amount in your MODTRAN calc.
Easy enough to determine - just plot the convolved result as a function of TPW for a
couple hundred(or thousand) profiles.
I would start by doing the radiative transfer calc at a finer resolution and linearly
interpolating the results to the same frequency spacing as the SRF. Then it's just a case of
Convolved_Radiance = TOTAL(Interpolated_Radiance*SRF,/DOUBLE)
You could do a "proper" convolution using INT_TABULATED or CONVOL, but given the paucity
of your dataset, a simple summation is the better option IMO.
At least in the case of INT_TABULATED, I know from experience it does not (cannot) perform
well for too few SRF data points as it uses spline interpolation.
I reckon the FFT approach is overkill here - and you would have to interpolated the SRF to
do it... again, not a good idea due to the lack of data.
cheers,
paulv
Mat wrote:
> On Dec 30, 4:41 pm, Mat <m...@waikato.ac.nz> wrote:
>
> Sorry I will try again!
>
> Hi all
>
> I would like to convolve a spectrum (output from MODTRAN) with a
> spectral response curve (Landsat7 B62) and then integrate the
> convoluted result. As you can see from the example below the spectral
> resolution does not match.
>
> Eg. Spectrum:
>
> Wavenumber Radiance
> 800.0 7.52E-004
> 801.0 7.66E-004
> 802.0 7.58E-004
> 803.0 7.25E-004
> 804.0 7.23E-004
> 805.0 7.62E-004
>
> Eg. Response
>
> Wavenuber Relative Spectral response
> 800.00000 0.2
> 800.64051 0.29
> 801.28205 0.27
> 801.92462 0.28
> 802.56822 0.36
> 803.21285 0.36
> 803.85852 0.39
> 804.50523 0.41
> 805.15298 0.42
> 805.80177 0.43
> 806.45161 0.45
>
> Any help would be much appreciated.
|
|
|
|
|
|
|
|
|
|
| Re: Convolve a spectrum (from MODTRAN) with a spectral response function [message #69501 is a reply to message #69346] |
Mon, 18 January 2010 13:51  |
Mat
Messages: 14
Registered: December 2009
|
Junior Member |
|
|
On Jan 5, 11:39 am, Paul van Delst <paul.vande...@noaa.gov> wrote:
> Your SRF (spectral response function) seems to only increase (looks like only half an
> SRF?), and your computed radiances do not span the bit of SRF that you do have.
>
> Also, in that part of the IR spectrum you have some water vapour line absorption
> contamination so I would be suspect of a convolution at such a coarse resolution - I would
> think you would have a bias that varies with the column water amount in your MODTRAN calc.
> Easy enough to determine - just plot the convolved result as a function of TPW for a
> couple hundred(or thousand) profiles.
>
> I would start by doing the radiative transfer calc at a finer resolution and linearly
> interpolating the results to the same frequency spacing as the SRF. Then it's just a case of
>
> Convolved_Radiance = TOTAL(Interpolated_Radiance*SRF,/DOUBLE)
>
> You could do a "proper" convolution using INT_TABULATED or CONVOL, but given the paucity
> of your dataset, a simple summation is the better option IMO.
>
> At least in the case of INT_TABULATED, I know from experience it does not (cannot) perform
> well for too few SRF data points as it uses spline interpolation.
>
> I reckon the FFT approach is overkill here - and you would have to interpolated the SRF to
> do it... again, not a good idea due to the lack of data.
>
> cheers,
>
> paulv
>
>
>
> Mat wrote:
>> On Dec 30, 4:41 pm, Mat <m...@waikato.ac.nz> wrote:
>
>> Sorry I will try again!
>
>> Hi all
>
>> I would like to convolve a spectrum (output from MODTRAN) with a
>> spectral response curve (Landsat7 B62) and then integrate the
>> convoluted result. As you can see from the example below the spectral
>> resolution does not match.
>
>> Eg. Spectrum:
>
>> Wavenumber Radiance
>> 800.0 7.52E-004
>> 801.0 7.66E-004
>> 802.0 7.58E-004
>> 803.0 7.25E-004
>> 804.0 7.23E-004
>> 805.0 7.62E-004
>
>> Eg. Response
>
>> Wavenuber Relative Spectral response
>> 800.00000 0.2
>> 800.64051 0.29
>> 801.28205 0.27
>> 801.92462 0.28
>> 802.56822 0.36
>> 803.21285 0.36
>> 803.85852 0.39
>> 804.50523 0.41
>> 805.15298 0.42
>> 805.80177 0.43
>> 806.45161 0.45
>
>> Any help would be much appreciated.- Hide quoted text -
>
> - Show quoted text -
Thanks to all who replied. Yes I did not need convolve, just interpol
and simple multiplication as paulv suggested.
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
