Hi Mikhail,

Interesting question. I am trying to understand the details here, like which other parameters are you taking into account apart from the spectral resolutions of the two sensors? With Landsat there are a number of sensors like the MSS, TM, ETM+, OLI/TIRS. Which one are you specifically focused on? Are you also taking into consideration the spatial and radiometric resolutions of the two sensors? Sensor noise effects etc.? Is the scene area specific? Are you also looking at day/night/seasonal acquisitions? It might be also beneficial if you share the purpose if you want. Apologies if I have more questions than answers. I thought it might help build the context so that the RG community can give constructive answers.

Hi James,

Thanks for the answer and the suggestion. That was my first question on RG, I will elaborate the details in the first post after this one.

I want to emulate the reflectance values as they would be seen by the AVHRR instrument using the Landsat OLI scene as the input. The specific AVHRR channels are channels 1 and 2 (0.6 and 0.9 micrometers); OLI channels would be either 3 ,4 or 5 (0.55, 0.65, 0.86 micrometers).

The purpose of this is to analyse the signature of mixed sea ice and water in the lower resolution satellite observations, like AVHRR Global Area Coverage product (4 kilometer spatial resolution). My idea is to use Landsat OLI scene which allows to derive the fraction of sea ice with confidence and simulate how the same scene would be observed by the low resolution sensors. This only applies to the summer season, when the sun elevation is high.

The spectral windows of the two sensors (for the selected channels) are quite similar, only the OLI channels are much narrower.

What you should investigate is the aggregation of the Landsat spectral bands into AVHRR. There are papers about similar approaches for different sensors, e.g. Hyperion into Landsat. I guess you could follow the same approach.

The idea to use the GAC data, however, might cause additional problems related to geolocation accuracy and data integrity. I would recommend relying on LAC or HRPT data if possible. The way GAC data is produced is that it contains only one out of three original AVHRR scan lines. If you want to observe fractional sea ice within one GAC pixel for example, I think relying on GAC data will cause you a lot of trouble and uncertainties.

Thanks for the suggestion, Dietz! Will look into the topics you mentioned.

True, I will rather work with 1km LAC/HRPT data at the moment, from there on I am free to simulate GAC data if needed, using LAC as input. In the end I have to stick with the GAC data though, as there seems to be no consistent global historical dataset based on LAC data. At least not that I found it.

Hi Mikhail,

Thanks for adding details to your question. May be there are different ways to approach the problem. If I may, I would like to suggest the following ways:-

Another alternative might be to do the whole process using artificially created synthetic data. But before that it may be useful to look at the 3 steps and analyze whether the results are as expected.

http://earthexplorer.usgs.gov/

Thanks James!

Well, that looks like the solution I am looking for. Unfortunately I don't have access to the ENVI software, so I have to find an alternative way. Spectral resampling is the term I was looking for though, I should dig deeper in that direction. Any directions how to implement simple resampling are greatly appreciated, I could convert it into python code. Unfortunately my experience with optical remote sensing is rather limited, so suggestions would be very welcome.

Regards,

Mikhail