Earthdata Forum

Hi everyone,

I got a little confused while compositing monthly chlorophyll-a images of MODIS-Aqua for the period 2003-2018.

While compositing the data I also was created the counts of valid pixels, that is, the number of pixels that contributed to the composite. So, in a year, I would know which points had valid pixels for the 12 months.

The issue is that when I look at the [img TOP]count image[/img] I see horizontal lines that I do not understand their origin. These lines become even clear with [img MIDDLE]more years[/img] combined. Looking at the monthly chlorophyll images it seems to me the missing values are random and I do not see any tendency for the straight lines that appear on counts.

I computed the counts using different software and this issue is present independent of the software used. However, using [img BOTTOM]Google Earth Engine[/img], the problem is not there and the distribution of the counts looks more natural to me.

So, what might be the issue?
I appreciate your kind help in understanding the reason for this issue.
Thanks in advance for your consideration.
Sincerely,
Eligio

This is what our mission composite looks like for the number of chl observations for the aqua mission.  Our composite is created like this:

l2bin - scenes -> daily
l3bin - dailies -> 8day
l3bin - 8days+dailies -> monthly
l3bin - monthlies -> yearly
l3bin - yearlies -> mission

Your count image looks the way I expected.
I forgot to mention that I am constructing the composite from monthly SMI images obtained from this website.
I have tried again but I still get the [img]strange count image[/img]

The horizontal lines artifacts give a hint that your workflow is not appropriate for these data.  MODIS (topic 10510) provides some background that may help you understand why the workflow in Don Shea's post is appropriate.  Your post, does however, provide a valuable demonstration of the importance of using appropriate (e.g. OCSSW Processing System) tools for these data.  

It is worth mentioning that the OCSSW programs Don used were originally developed for SeaWiFS data in the 1990's, which means they run well on systems that were much less powerful and orders of magnitude more expensive than today's smartphones.  Many organizations have a collection of older unloved and unwanted PC's and laptops that can make very useful headless (text only) OCSSW processing platforms for the price of a big disk.

Thanks for the background information.
But I still have cannot fully understand why I am getting the lines.
From the explanation I get the sense that the horizontal lines are caused by the difference between l3bin and l3m?
Then, not appropriate means l3m should not be used for any statistics as I got the counts from daily standard mapped images
Thanks
Eligio

Artifacts such as those you observed are not unusual for when mapped data are reprocessed.   Understanding the exact details of an artifact would require details of the original mapping as well as the final processing steps.  If you want to fully understand the artifacts, I suggest you  pick a few pixels near an artifact.  Extract the values for these locations from each input image and do your calculation by hand.

In general, mapped data are visualizations. Different workflows should be used to generate statistical products, which can then be used in visualizations.

IOCCG Report 4 may be helpful.

Hi Eligio,

As far as I am aware, we do not distribute SMI count files.
Which SMI files are you using?   If you want a proper image
showing the valid number of observations at any ___location,
then you should be using the L3 bin files as Don and George
suggested.

Regards,
Norman

Thank you folks,
I start to get the point. I may look at some of your suggestions.
The SMI I used is monthly "A*L3m_MO_CHL_chlor_a_4km.nc" provided in this website.
And yes, it does not come with counts, I calculated those myself.
Again thanks for your clarification
Eligio