I have calculated flow accumulation (img file) using this DEM (img file), which is 0.5 degree resolution, using arcmap.

I am wondering if the cell accumulation value can directly be translated into drainage areas in square kilometers or not, and if these values reflect true catchment areas?

For example, if the flow accumulation value is 140 and my resolution is 0.5 * 0.5 degree = 56km^2, the upstream area is 140*56 = 7840km^2.

  • 3
    There seem to be two mistakes in the calculation "0.5 * 0.5 degree = 56 km^2". One is that the area of a square degree varies with latitude, which is not accounted for here. The other--which is a huge mistake--is that a cell of 1/2 degree will have dimensions (approximately) 111/2 km = 56 km in the north-south direction and (approximately) 111/2 * cos(latitude) in the east-west direction, for a total area of about 3080 * cos(latitude) km^2, which is orders of magnitude greater than 56 km^2!
    – whuber
    Nov 5, 2012 at 15:49
  • 1
    A more fundamental mistake reflected here is to compute flow accumulation using such a poor projection (the Plate Carree or "geographic"). This biases the results (see Calculating Flow Direction and Delineating Basins from Projected vs. Unprojected DEM Data). Rather than translating the results into areas, it is far better to redo the flow accumulation using a better projection, which will provide flow accumulation areas directly and with less bias.
    – whuber
    Nov 5, 2012 at 15:56
  • Thanks ,how did you know the projection is geographic?what is better projection?
    – Jonsson
    Nov 5, 2012 at 16:30
  • The link in my second comment goes to a detailed, extensive thread in which all these things are discussed and your question about a better projection is thoroughly answered.
    – whuber
    Nov 5, 2012 at 22:02
  • 1
    I'm sorry, I am not going to take a look at it. In my first comment I have already explained why you are experiencing these differences in areas. The links in my second comment fully answer your other questions. More generally, this is not a place to ask others to do your work for you (although many kind community members often go a long way in that direction). Questions here are expected to be self-contained insofar as they provide sufficient context and showing what preliminary efforts and research have been done towards resolving the problem.
    – whuber
    Nov 6, 2012 at 17:21

1 Answer 1


It will work as a rule of thumb, what you have to consider is the method used by arcmap to calculate flow accumulation. It is based off the D8 method, which assumes that 100% of a cell's input flows out to one of the 8 surrounding cells at the compass points: N, NE, E, SE etc.

This works as an approximation but you can't say that such a model would reflect the true catchment areas, as it is quite common for water flow to not follow 45 degree angles and to branch and recombine as it flows across the surface. This is one of the limitations of trying to model a landscape in a raster grid, but one which is being overcome to some degree with the use of LiDAR data and more complex analysis techniques which trace vectors of probabilistic water flow between cells based on work by Lea (1992).

General reading about these hydrology techniques should always include work by David Tarboton, especially his 1997 paper, which has been the groundwork for most DEM based hydrology for the last 15 years. If you are more inclined towards programming (particularly c++) you could also check out Jon Pelletier's book which has great algorithms of all the classic techniques in hydrology and DEM processing in easy to follow c++ samples.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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