Geographic Information Systems Stack Exchange is a question and answer site for cartographers, geographers and GIS professionals. Join them; it only takes a minute:

Sign up
Here's how it works:
  1. Anybody can ask a question
  2. Anybody can answer
  3. The best answers are voted up and rise to the top

For one of my projects using PostGIS 2 I have a table with GPS tracks (igc_files). It holds records including a MULTIPOINT() column that has about 1,000 points for each record and a column that has an array of the corresponding timestamps. I also have a raster table (elevations) that should at some point hold all of the SRTM elevation data. For now I have only imported a few relevant areas though.

What I want to do is get the elevation for each point of one certain record and the corresponding timestamps. I have implemented an naive approach and it does work, but I'm wondering about its performance implications. This is how it currently looks:

SELECT timestamps[location_id] AS timestamp,
       ST_VALUE(elevations.rast, subq.location) AS elevation 
FROM elevations, (
    SELECT (ST_DUMP(igc_files.locations)).path[1] AS location_id, 
           (ST_DUMP(igc_files.locations)).geom AS location 
    FROM igc_files 
    WHERE = 1
) AS subq, igc_files 
      ST_INTERSECTS(subq.location, elevations.rast) AND
      ST_VALUE(elevations.rast, subq.location) IS NOT NULL

Since this is a GPS track and the points are not that far apart, PostGIS could assume that the next point is in the same raster tile and check that tile first before checking the index again from the top. Is that actually done by PostGIS or is it just starting to search the index from the top for each of the points?

I have tried to use ST_INTERSECTION() instead of ST_VALUE() and feeding in the MULTIPOINT directly, but that was actually even slower by a few orders of magnitude. Is there a feature/function of PostGIS that I am missing to do this more efficiently?

share|improve this question
i have seen people to recommend using dwithin(0.1, geom ) to speed up intersection searches but i have no idea does it speed up process. – simplexio Nov 16 '12 at 11:53

Did you tile your raster when you loaded it in the db (using the -t raster2pgsql option)? Smaller tiles size = better performance. Check this Blog entry:

Make sure to also create the index with the -I option...

Also in your query:

1) Don't call ST_DUMP() two times. Assign the result to a name and use (name).path[1] and (name).geom in the first SELECT

2) "ST_VALUE(elevations.rast, subq.location) IS NOT NULL" in the WHERE clause is useless as ST_INTERSECTS will return FALSE if the point intersects with a nodata value. This test is quite slow if there are a lot of nodata values in your raster. Is this the case?

share|improve this answer
Yes, I used 100x100 tiles and created an ST_ConvexHull() index by using the -I option. This code will ultimately run in a web application so the code should be as fast as possible, thats why I asked about the possible optimizations. In a sister project we are using GeoJP2 for the terrain lookups and there we are first checking the current tile for the next coordinate before using the index again. I hoped that PostGIS would/could do the same. – TBieniek Nov 15 '12 at 23:23

Your Answer


By posting your answer, you agree to the privacy policy and terms of service.

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