0

How not to calculate land mask when 'extract using vector' in Temporal Profiles (ETM)?

idham khalil

February 08, 2013 09:08

Answered

Follow

Hi Guys,

In the Earth Trend Modeller (ETM), we have an option to understand a time series's trend using Explore Temporal Profiles. We can do that by either draw a circular sample region or select our own vector sample. Both work perfectly if our image does not have any mask pixels (e.g cloud or land area). If a vector was used as a boundary to calculate the trends, all those land/cloud mask(normally zero values) will be counted and will result in very low slope values. It might hard to understand, I've tried to attach picture as an example but failed.

Any idea to get rid of all those land/cloud pixels or not to count them during temporal profile's calculation?

Thanks in advance.

• Facebook
• Twitter
• LinkedIn

1 comment

• 0
  
  Clark Labs February 11, 2013 14:20

  Hi Idham,

  You have addressed a fairly common issue. There a couple of ways that this can be addressed. The easiest way is to fill in the missing data using one or more of the Missing Data Interpolation routines in ETM. If you missing data in the majority of your imagery, then this may not be a viable option.

  A more precise method would be to mask out the missing pixels by creating a 'vector' mask. This is not so straight-forward, but is easily doable using Macro Modeler.

  What you would need to do is to take your initial region vector mask file, and rasterize it. Then multiply that mask with all your imagery, add them all then create a Boolean mask of non-zero values. A tedious task that you can automate in Macro Modeler.

  The resulting file need to be vectorized and used as the vector sample feature file. 

  I hope this helps. 

  James Toeldano

  Clark Labs