Global Mapper v24.0

Comparing watersheds generated with different DEMs

JSL
JSL Global Mapper UserTrusted User
edited June 2014 in How I use Global Mapper
Hello, I’m working on a project comparing watersheds created with two different DEMs, where the difficulty lies in figuring out how to make their parameters equivalent, since the DEMs have different resolutions. I was wondering if this method makes sense:

I have two DEMS: A has a resolution of 1-arc second, and B has a resolution of 500m. A quick search showed that one arc-second = about 1860m of latitude. I know that the longitude of a square geographical unit differs depending on where on the globe one is, but for the purposes of this study, I’m just assuming that latitude and longitude are equivalent. So one cell of DEM A = 1860 squared = 3459600 sq m, and one cell of DEM B = 500 squared = 250000 sq m.

If I want to generate a watershed from DEM B with a stream cell count of 1000 for a stream threshold, that would equal 1000*250000 sq m of area. Since 1860 squared = 3459600 sq m, an equivalent flow accumulation for DEM A should be (1000*250000)/3459600, which equals 72.

Below are the results, with DEM B (500m) in the dashed red line and DEM A (1 arc second) in solid blue. Their similarity seems to confirm my calculations, but I was wondering if anyone had any comments.

Comments

  • JSL
    JSL Global Mapper User Trusted User
    edited June 2014
    Also, I've kept all the other parameters equal in the Generate Watershed command.
  • global_mapper
    global_mapper Administrator
    edited June 2014
    I would suggest making the projection on the Projection tab of the Configuration dialog the same for both calculations, then you can provide the spacing in the same units/projection for both to get the same sampling. You can then set up either the stream cell count or simply use the stream drainage area option instead and provide the same values for both.

    You may still get slightly different watersheds due to differences in the terrain, but at least the sampling of the underlying terrain will be the same. I would suggest using the projected system (i.e. whichever one is reporting the distances in meters) rather than Geographic/lat/lon.

    Thanks,

    Mike
    Global Mapper Guru
    geohelp@bluemarblegeo.com
    Blue Marble Geographics for Coordinate Conversion, Image Reprojection and Vector Translation
  • JSL
    JSL Global Mapper User Trusted User
    edited June 2014
    Thanks, Mike, for the comments. I layered the DEMs in the same workspace, so they both share the same configuration. I'm not sure what you mean by using the projected system rather than Geographic/lat/lon. Should I be using another projection besides the default?

    When I calculated flow accumulation using drainage area, the DEMs produced very similar results, but when I calculated using the same stream cell count, the watersheds were very different, which I presume is because of the difference in unit cells of the DEMs. The first image below shows two watersheds produced with 1000 sq km for flow accumulation, and the second shows the watersheds produced with 100 for stream cell. The finer resolution DEM is the dashed red line, and you can see how it's much more detailed than the coarser scale DEM in the watershed produced by using stream cells instead of square kilometers.

    Now I know all I have to do is to set the same area in square units, instead of having to find the ratio of unit cells for each DEM--thanks, that's a great timesaver.
  • global_mapper
    global_mapper Administrator
    edited June 2014
    I just suggested setting the projection system to the same for each operation just to make selecting the drainage area more consistent. Since you've got that working using the stream size by area instead, you can just stick with what you're doing.

    Thanks,

    Mike
    Global Mapper Guru
    geohelp@bluemarblegeo.com
    Blue Marble Geographics for Coordinate Conversion, Image Reprojection and Vector Translation
  • JSL
    JSL Global Mapper User Trusted User
    edited June 2014
    Thanks for the input. Can I ask your opinion of a theory I have? What’s been interesting about this study has been looking at the anomalous areas, where different DEMS have generated significantly diverging stream flows. For example, below is an image of the Congo basin, with basin fill set to the maximum of 5000m. While the finer-scale DEM, in orange, shows the Congo basin emptying through its current outlet, the coarser DEM watershed in blue shows the Congo emptying through the watershed of what is now the Ogoouee River.

    Congo 1.png

    In the next image, I raised the sea level to 520m to delineate the channel features, and I found that the area between the Congo and Ogoouee basin suddenly stands out as a lot wider than the current channel.

    Congo 2.png

    This has significance from a geomorphological aspect, because it suggests that the Congo River once reached the sea through the Ogoouee River basin, and then later switched to its current channel. This is supported by research showing that the Ogoouee River delta is very comparable in size to the Congo River delta, even though it has a much smaller basin size. That would make sense if it was once the discharge site for the Congo basin.

    I’m guessing that because of its larger cell size, the coarser DEM is selecting the older elevation channel contours as a spill point for overflow instead of the more recent channel contours, because the basin fill algorithm works top down, and the older channel contours would lie at a higher elevation than the more recent channel. Does that make sense? Can you suggest any documentation that I might read and cite that would shed more light on the subject? (I already have the 2001 Planchon paper.)

    Thanks for your help.
  • global_mapper
    global_mapper Administrator
    edited June 2014
    There could be any number of reasons for the difference, with the most likely one being actual differences in the DEM if the outflow channels are at similar heights. You might create a difference grid using the Analysis menu command to Compare/Combine Terrain and see how different they are and if maybe there are more differences in one channel than the other.

    The depression filling algorithm in GM is based on the Planchon algorithm so it should fill from the bottom up until it spills out.

    Thanks,

    Mike
    Global Mapper Guru
    geohelp@bluemarblegeo.com
    Blue Marble Geographics for Coordinate Conversion, Image Reprojection and Vector Translation
  • JSL
    JSL Global Mapper User Trusted User
    edited June 2014
    I'm sorry, I know I'm over my head mathematically here, so please correct me where I'm wrong. The abstract of the 2001 Planchon paper says, "...instead of gradually filling the depressions, it first inundates the surface with a thick layer of water and then removes the excess water." Then the Discussion and conclusions sections has this statement: "The method presented in this article is based on a radically different approach that first adds a thick layer of water over all the DEM and then drains excess water."

    That sounds like a top-down process--am I misunderstanding something here?
  • global_mapper
    global_mapper Administrator
    edited June 2014
    Ah sorry you are correct. I reviewed the implementation and indeed it does work by making everything artificially high (except edge samples), then lowering until you get valid flow everywhere.

    Thanks,

    Mike
    Global Mapper Guru
    geohelp@bluemarblegeo.com
    Blue Marble Geographics for Coordinate Conversion, Image Reprojection and Vector Translation
  • JSL
    JSL Global Mapper User Trusted User
    edited June 2014
    I think I've figured it out--I think the different stream flows are a result of both the DEM and the coarse resampling I used (Box average 9x9).

    The first image below shows the 1 arc-second DEM, with the current watershed boundary as the black line, and a 520m contour as the dark blue area, which I generated using the same 9x9 box average that I used when generating the streamflow. It's easy to see that the contour line obscured the current channel (red circled area). The red arrow points to where the generated stream flow lines cross the current watershed boundary into what is now the basin of the Ogooue River.
    congo 520 - ETOPO 9x9.png

    The second image shows the 500m DEM with a 520m contour area, and you can see that the current channel shows up this time.

    congo 520 - 500m 9x9.png

    So it seems to me that the coarseness of the DEM and the resampling are the main factors in creating the different stream flows. However, I have to admit that I'm not too clear on exactly what resampling means. Is there some documentation that explains the different sampling methods available in GM? Something that I can cite would be very helpful, as well.
  • global_mapper
    global_mapper Administrator
    edited June 2014
    You can find documentation of the different resampling modes at Display Tab.

    Thanks,

    Mike
    Global Mapper Guru
    geohelp@bluemarblegeo.com
    Blue Marble Geographics for Coordinate Conversion, Image Reprojection and Vector Translation
  • JSL
    JSL Global Mapper User Trusted User
    edited June 2014
    Thanks very much!
Sign In or Register to comment.