Hi,

For a project I am currently involved with, I have an in situ dataset that I would like to match up with satellite data. The in situ data was recorded continuously during a transect with a towed body. The continuous nature of the data has presented a few problems for performing a match up with the satellite data. First of all, there are multiple in situ records for a unique lat/lon pair. In addition, there are multiple unique in situ lat/lon pairs per satellite pixel. Both of these issues make it so that for each satellite pixel in the match up, there are multiple in situ records. This is made evident in a scatter plot produced with the SeaDAS "ship track" function (see attached - note the vertical lines of data points). Is there a way to get around this with the ship track function? I have also been told that for each satellite pixel in the match up, I should check the variability in the surrounding 3x3 box. If the variability is greater than 10% then I should omit that pixel from the match up. I am not sure how I could do this with the ship track function.

Therefore I believe that I should write my own IDL script to perform the match up. I currently think I will structure the script so that for each pixel in the satellite image it first checks the variability of the surrounding 3x3 box (of course invalid with pixels on edge of the region). If the variability is greater than 10% then the next pixel is chosen. If the variability is less than 10% then the median value of all the in situ records that fall within that pixel is determined. I would be very interested in hearing other people's thoughts on how to perform this type of match up!

Of course to do this I will need to determine the lat/lon coordinates of the 4 corners of each pixel. Is there a slick way to do this? I guess the MAIN question is: does the lat/lon pair associated with each pixel correspond to the middle of the pixel, or one of its corners (if so which corner)? Thanks for any help.

Mike

Where is the attached picture? The picture should be a help.

Can you do data average for those multiple in situ records
for a unique lat/lon pair before using SeaDAS's ship-track ?

3x3 box should be better to do the match up.

SeaDAS pixel reading is associated with middle of the pixel.
You may use a zoom window to check at pixel level with
SeaDAS image display.

Long

Hi Long,

- I just attached the picture. Sorry about that.

- Yes I have attempted the match up after first averaging together the multiple in situ records for each unique lat/lon pair. However that doesn't solve the problem that there are multiple in situ lat/lon pairs with a satellite pixel.

Mike

Mike,

As Long mentioned, the lat/lon for a satellite pixel is the center of the pixel. 
You could do a crude estimation of the pixel boundaries using half the
distance between the neighboring pixels.  Or you could use the information
in this post on computing pixel area to estimate pixel boundaries
...I'll leave the exercize to the reader

I would not recommend doing validation at the individual satellite pixel level,
although the results should be OK if you do the variability test for the
surrounding pixels.  You might get more mileage using a 3x3 pixel box - or,
since you're going to calculate the pixel boundaries, you can do the validation
on a 3x3 km average (or 5x5, 7x7, etc.)  Using more satellite data values for each
matchup will allow you to estimate (if crudely) the satellite uncertainty - which
would go nicely with the in situ uncertainty you'll get with the multiple in situ
record averages.

Regards,
Sean