I have a bunch of points in a text file that are represented using lat long (116.3158516 39.9748016). I import them into my GIS enabled database and create trajectories using st_makeline() as geometries projected in EPSG:4326. I'm using geometry here because all the GPS data falls into an area that is 50km^2: correct me if this is my first mistake.

Now I'm doing things like calculating trajectory length (st_length) and simplifying the geometry using st_simplifypreservetopology() but according to the documentation the results for the length and the tolerance for the simplification are in the coordinate system of the data; which in my case is degrees. This obviously doesn't make sense. I need it to be in meters.

Do I need to change the way I'm storing my data or do I need to convert my trajectories and points into something else before feeding them into the st_length() and st_simplifypreservetopology() functions?

Screenshot for Mike T.

enter image description here

  • Use ST_DistanceSphere if you want lat/lon and distance in meters. Where did you get the idea that ST_Intersect doesn't work in 4326? Oct 24, 2014 at 15:00
  • Hi John. I got the idea that intersect doesn't work from the possible duplicate link that links to a workshop tutorial here: 3rd paragraph. Oct 24, 2014 at 16:42

2 Answers 2


You are correct that if your data are in a small region that you should use a geometry type. However, you would also normally transform the Lat/Long coordinates to a spatial reference system for the region. Normally most folks choose a UTM Zone, which describes coordinates for Eastings and Northing with length units in metres.

Check out ST_Transform to convert your Lat/Long coordinates to Easting/Northings. You can either modify the geometry column in-place, or create another table or geometry column.

  • Thanks for the suggestion. I'm also having projection issues which may have something to do with the way I'm storing these values. When I stitch these points together to form trajectories (st_makeline) that I then add as a layer over Google Maps the trajectories are a bit off relative to where they should be on the map. My other concern is that that data source for these lat/long describes the data to be in WGS84 which if I understand correctly could mean anything. I assumed it to be 4326: the data was collected using smartphones. Oct 30, 2014 at 4:15
  • Yes, WGS84 is SRID=4326. Location estimates from smartphones are not great, so it is expected that your lines would look like zig-zags jittered around the expected locations.
    – Mike T
    Oct 30, 2014 at 4:53
  • Ya I expect them to bounce around but they seem to be off almost too perfectly. I've attached a screenshot to the original post. Have a look and let me know if it looks like a projection issue to you. The two trajectories are green and red. Thanks again. Oct 30, 2014 at 5:18
  • Looks like there is some bias, but I'm not convinced it is a projection issue since some parts match up. Also, I'm no expert with how smartphones obtain their position, but check out this thread.
    – Mike T
    Oct 30, 2014 at 5:41
  • when I created my points using ST_MakePoint() I left the elevation in it's original units: feet. I'm not certain how the elevation plays into drawing these lines. Maybe I should be storing elevation in meters? Oct 30, 2014 at 19:15

When using st_length() I can convert to geography using st_length(line.line_geometry :: geography) as was suggested or using ST_Length2D_Spheroid(line.geometry, 'SPHEROID["GRS 1980",6378137,298.257222101]'). Both answers seem to be the same as far as precision is concerned: 580139.3016159851 vs 580139.3016215176. I'm guessing the slight difference is due to a rounding error. Using either of these two methods produces my results in meters which is what I was after.

I verified that the length was correct by using the Measure Line tool in QGIS and measuring a given line.

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