Certain HLS (S30) granule(s) lack imagery

[mitchbon]

Hello,

I am working with the HLS product and have noticed cases where certain granules, more so S30 than L30, lack imagery compared to surrounding granules.

Take this example from Hudson Bay Lowlands in Canada (around 54.50479,-83.08301):

S30_granule_issue.jpg (147.62 KiB) Not viewed yet

This shows a 180 x 180 km squared area, with the raster image representing HLS (S30 & L30) clear observation frequency over a period of time (white = more clear observations, black = fewer clear observations). Green in the background is Sentinel-2s orbit paths and red in the foreground are MGRS tiles, with tile T17ULA noted.

The vertical slices with less clear observations are expected. These line up with Sentinel-2 (and to the left Landsat 8) orbits, where the vertical slices are areas outside overlap areas as the orbits converge at the poles.

However, the specific granule (T17ULA) with less clear observations is unexpected. To confirm this was not due to my workflow, I checked NASA EarthData Search for this granule. What I found was that this granule in particular had very few available scenes during the winter period for S30. For example, T17ULA has only 2 S30 (S2A and B) scenes between Nov 12, 2019 and Apr 10, 2020. L30 (Landsat 8 only) has 7 scenes over that period. This situation is true for all the years I checked (i.e., not just one year) and only effects this granule in this area (i.e., if you move to a nearby granule you get a bunch more late Jan - March scenes that are missing from this one), so this is not a winter night situation.

The Earthdata Search for this granule on S30: https://search.earthdata.nasa.gov/search/granules?p=C2021957295-LPCLOUD&pg[0][v]=f&pg[0][id]=HLS.S30.T17ULA.*&pg[0][gsk]=-start_date&q=hls&tl=1727357823!3!!&lat=54.79982287302141&long=-84.69580078125&zoom=7

I am curious as to why this specific granule lacks images? I am also not sure if this is an isolated case, or if I will come across more granules like this as I expand the spatial scope of my work. I assume there are Sentnel-2 scenes here that for some reason did not get processed by HLS.

Note that these data were accessed well before the recent HLS STAC issues that have been documented recently, so it is not related to that.

Thank you!

[LP DAACx - dgolon]

Hi @mitchbon Our science team is taking a look at this. We'll report back when we have additional information. Thanks -- Danielle

[waltersdan]

Following!

[kristianbodolai]

Following!

[miketaku111]

Following!

[ioyler]

Also following

[LP DAACx - dgolon]

Hello all! ( @ioyler @miketaku111 @kristianbodolai @waltersdan )@mitchbon

We have heard back from the science team.

The HLS team applies cloud cover filter to filter out any data with more than 95% cloud cover and a solar zenith angle filter to filter out any data with a solar zenith angle of more than 76 to the Landsat and Sentinel-2 data prior to selecting which data will be selected to be processed into HLS. So, for example, while there were originally 61 Sentinel acquisitions over 17ULA, after filtering out for clouds and solar zenith angle, the amount of Sentinel scenes that made it into HLS processing was only 21 scenes. For Landsat, there were originally 11 scenes, and only 4 made it into HLS processing. Noting the filter information is also mentioned in the FAQ section of our website, just in case any users were not aware we have an FAQ section :https://lpdaac.usgs.gov/resources/faqs/#what-are-the-criteria-for-selecting-acceptable-sentinel-2-tiles-for-hlss30-v20-processing)

Additionally, any other gaps in data that should have met these requirements may be due to processing failures or any failures that occurred while transferring the ESA pre-Collection 1 data. The HLS team does plan on going back to revisit previous scenes that had any processing or transferring failures at some point, but currently we do not have an ETA on that effort. When it does occur, we will post a news announcement on our website.

Thank you all for your patience while we investigated into the answer.

Danielle

[mitchbon]

Thanks Danielle for the answer!