2

I am analyzing Landsat data around Lagos, Nigeria going back to 1990 which includes Landsat 8, 7, and 5. I am using Google Earth Engine's code editor to access and clean the images. I am trying to get yearly composites and have been successful for Landsat 7 and 8. But I am finding incredibly spotty coverage for Landsat 5. To illustrate, the code below (Google Earth Engine code) compares the size of the resulting image collections for Landsat 7 and Landsat 5. For both Tier 1 and Tier 2 Landsat 5 has drastically fewer images. I believe both LS-5 and LS-7 circle roughly every 2 weeks so I don't understand the difference in data availability.

I am new to Earth Engine and satellite data.

// Get Landsat 5 and 7 over region of interest and print Collection size
var pt = ee.Geometry.Point(3.657968670903074, 6.691009458317496);
var ls7Collection_T1 = ee.ImageCollection("LANDSAT/LE07/C01/T1").filterBounds(pt);
var ls5Collection_T1 = ee.ImageCollection("LANDSAT/LT05/C01/T1").filterBounds(pt);
var ls7Collection_T2 = ee.ImageCollection("LANDSAT/LE07/C01/T2").filterBounds(pt);
var ls5Collection_T2 = ee.ImageCollection("LANDSAT/LT05/C01/T2").filterBounds(pt);
print("LS 5 (tier 1) collection size is: ", ls5Collection_T1.size());
print("LS 7 (tier 1) collection size is: ", ls7Collection_T1.size());
print("LS 5 (tier 2) collection size is: ", ls5Collection_T2.size());
print("LS 7 (tier 2) collection size is: ", ls7Collection_T2.size());
  • Are you looking for an specific cloud coverage percentage (e.g. <10%)? If you are, it is important for you to keep in mind that the number of images that will satisfy the cloud coverage criterion can vary depending on the year. If you would like to get a better idea on the issues with Landsat Images, I'll suggest you check out this paper by Wulder et al., 2008 Landsat continuity:Issues and opportunities for land cover monitoring – Perro Sep 18 '18 at 22:32
  • The article you linked does note that there are regions in equatorial Africa (my area of interest) where cloud-free images are in short supply, but I believe my problem occurs before that. My understanding is the all Landsat (5 or 7) raw images are classified as tier 1 or tier 2. My code above shows there are 8 tier 1 and 14 tier 2 images for LS-5 but shows there are 124 and 126 images for LS-7 over the same region. These counts are from raw images, regardless of cloud cover. Perhaps the Earth Engine datasets are already partially filtered, even 'raw' images? – cd_nederhood Sep 18 '18 at 22:59
  • 1
    You can double-check against EarthExplorer with a path/row, date range, and sensor. I don't believe GEE filters any of the images--I think they were just ingested in bulk from USGS. If you can access it, this figure nature.com/articles/nature20584/figures/5 helps understand the spatiotemporal coverage. Landsat has a storied and varied history, and although the satellites were "taking pictures" 24/7, much of the data was never saved. There is currently an effort to recover historical scenes from local datacenters. – Jon Sep 18 '18 at 23:24
2

This is because your point of interest is, as you point out, over Equatorial Africa. In the early days of the Landsat Program, much of the imagery and research was focused on the Continental United States. @Jon points out in his answer that 'although the satellites were "taking pictures" 24/7, much of the data was never saved'. This is an artifact of that.

If you modify your code to a point over the United States, you will find much more data. For example, the point

var pt = ee.Geometry.Point(-111.54765624999999, 38.450640311247184)

results in 970 T1 and 98 T2 Landsat 5 images, and 798 T1 and 67 T2 Landsat images

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.