I recently extracted coastlines (for my master thesis) for a Bangladesh region for the years 2013 and 2014. Regardless of the method I used, the distance between the 2 coastlines was always the same: 42 m.

This distance is way too big for a one year difference (same time of day, so tides were more or less the same). I realized that the 42 m are exactly the length of a diagonal with 30 m sides (which is exactly the resolution of these Landsat images).

In conclusion, no distances smaller than the pixel resolution can be measured. Does anyone know of any method to improve this, or free higher resolution alternatives to Landsat?

  • 2
    Have you tried pansharpening the L8 images to 15m? – PyMapr Mar 14 '16 at 10:19
  • @Thunderbolt I was just trying that, but I think 15m is still too low relative to my expected results. – Alin Radu Mar 14 '16 at 10:30

30m - 45m is a lot of change in a coastline over one year and only very dynamic areas see that kind of change rate. As such, you are correct in your assessment of the impact of imagery resolution on the analysis.

However, your assumption about "same time of day, so tides were more or less the same" is not a very good assumption as tides are more variable than that.
This doesn't change the overall conclusion being Landsat 8 is too coarse for accurate coastline change detection. Pansharpend data might help a bit, but 15m is still coarse resolution for coastline studies. Sentinel-2 data may also be an option, providing you with 10m resolution. Higher resolution data than that is generally commercial data.

As for getting better data for free, given that you're doing academic research, you could try to get a DigitalGlobe imagery grant. I don't know the success-rate on applications, but it might be worth a try. There are other similar setups, where you can apply for data for use in research.

  • I would note that the statement "Landsat 8 is too coarse for accurate coastline change detection" is not strictly true, or rather not qualified sufficiently. I use 30 m resolution Landsat images to detect riverbank changes frequently. The quantity of interest is the total amount of displacement/resolution of image. If you have a coastline from 1985 and one from 2015, there very well might be sufficient change to measure accurately from Landsat (see the Wax Lake Delta). However, Landsat 8 only goes back to 2014 (I believe), so it's unlikely that enough change has occurred to accurately measure. – Jon Sep 7 '18 at 23:05

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.