Dear Ruben,

I do not know the cause for certain but I suspect it is this:

The tool takes a very simple approach to building climatologies: for each pixel, it calculates the specified statistic over all of the available (i.e. non-cloudy) pixels in the bin, giving each pixel equal weight. So if you’re computing a cumulative climatology, all of the pixels will be used. For a year-long period, that will typically be 365 daily images or 12 monthly images. For the daily images, there is likely to be a seasonal bias in cloud cover. In many areas it is more cloudy in winter. Therefore there will be fewer winter pixels compared to summer pixels: the statistic will be biased toward summer, because summer is less cloudy. On the other hand, for the 12 monthly images there will usually be a value a given pixel for all 12 months, except in extremely cloudy months. Thus the statistic from monthly images is less likely to be biased towards summer.

I have other ideas but I recommend you investigate that one first. One way would be to use the tool to compute a Count statistic rather than a Mean statistic. Compute the count for the daily images. Ideally you’d get 365 but it will probably be substantially less. Then do it for a monthly statistic. If my theory is correct, it would be 12 for most pixels. If you see this pattern, then start investigating which time periods have more missing daily values. Instead of a cumulative climatology, produce a monthly one from the daily images. See whether each month has around 30 images or whether colder months have substantially less. That will be the bias I was talking about.

Hope this helps,

Jason

Dear Jason,

I am interested in creating yearly mean SST (from April of one year to March of the next year) by using the  Create Climatological Rasters for Pathfinder 5.0.

I have done it 2 different ways:

1)      Using a “daily”  temporal resolution, and using the “cumulative bin type” – Start and end dates are 01/04/1998 and 31/03/1999 for example.

2)      Using a “monthly”  temporal resolution, and using the “cumulative bin type” – Start and end dates are 01/04/1998 and 31/03/1999 for example.

The results I obtain by doing this are different. For example, the average SST  that I obtain for a certain region using the daily resolution is in average 1.15 degree higher than with the monthly resolution.

And there is a r2 of 0.6 between the results which I attach in this google spreadsheet just in case you can have a look.

Do you know what might be the cause for these differences?

Thanks,

Ruben