How to Correctly Select an L2 Scenes to Generate an L3 Image？

[mercury_1106]

I encountered a problem while generating L3 images from MODIS-Aqua L2 images.
Suppose I want to generate an L3 image for July 1, 2018. My approach was to aggregate all L2 images from the time period between July 1, 2018, 00:00:00 Z and July 1, 2018, 23:59:59 Z into a single L3 image.
However, after comparing this with the official L3 image provided on Ocean Data, I found that, in fact, the official product actually selected all L2 images from the time period between July 1, 2018, 00:15:01 Z and July 2, 2018, 02:55:00 Z. After comparing the official L3 products for July 2, 2018, I noticed that they all span 17 orbits. Is this a coincidence?
I know that OCSSW uses l2bin to generate L3 images in the bin format, but if I want to use a different method—processing the original L2 images and then aggregating them to create my own L3 products—I need to know how to obtain the same file selection range as the official products. This will make it easier to compare my products with the official ones.

[OB.DAAC - SeanBailey]

What constitutes a "day" for the L3 binned products is not a 24hr period starting at midnight. We define the "data day" based on the equatorial node crossing time for our sun-synchronous polar orbiting satellites. For MODIS (at least before the orbit maintenance was stopped), that time was 1330 local. So, for your example day of July 1st, the day begins at 0130Z on July 1st and ends at 0130Z on July 2nd (roughly - remember that time is at the equator crossing). To ensure we select all the files that would contribute to that data day, we buffer the search window by +/-90 minutes.

Now, not ALL the data from those file contribute to the data day. For scenes that cross the dateline, the binner will exclude data east of the dateline for the granules "entering" the day and west of the dateline for granules "leaving" the day. And since the search window is wider than it needs to be, some of the files won't contribute at all.

I hope this helps,

Sean