I have two different point layers, where points represent centroids of areas covering a stream segment. One layer (lets call it layer A) is coarser, with fewer point (each point represents a larger area), then layer B. There is no commonality between how the points are distributed, except for the boundaries. Each layer contains multiple attributes (ex. depth, velocity, areas). I want to create a common grid, but I'm hesitant about rasterizing this data, as I need to retain the multiple attributes.

I have tried converting each set of points to voronoi diagrams, then spatially joining layer A to layer B, and converting back to points, but there have been problems retaining area values required.

Also I have joined layer B's voronoi areas to layer A's points. This gives me layer B's areas in terms of layer A's points, but leaves holes in the coverage.

Ideally, I'd like layer A to fit within layer B, or have a new mesh fitted to the stream segment that both these layers can be applied to.

Does anyone have suggestions for optimizing this process, or ideas I've failed to see/am not familiar with?

I've included pictures of the point layers, as well as a visual goal.

Layer A picture

Layer A points

Layer B picture

Layer B points


Example of possibility

I'm using ArcGIS 10.3 with Spatial Analyst, and QGIS 2.4 Chugiak.

  • 1
    What is the purpose of the common grid you are trying to create?
    – whuber
    Oct 17 '14 at 17:43
  • 1
    This is part of a time series analysis, where layer B comes from a model that tracks flow during the summer, while layer A comes from a model that can track winter under ice flow, which layer B's model cannot. So at a certain point in my process, I'll need to switch from the summer points to the winter points. Oct 17 '14 at 17:46
  • I've still been searching for an answer, but most recently I've been thinking about vector grids. Is there a way to incorporate both data sets (A,B) into creating a grid? From there possibly creating points from the grid intersections? I'm currently running a QGIS vector grid analysis but processing time has proven an issue. Oct 24 '14 at 16:39
  • More details would be helpful, Andy, in order to develop a clearer picture of what you are attempting to do and the magnitude of the problem. A natural thought is to create a combined model grid (or mesh), but I gather that is not possible (or is prohibitively costly). An illustration showing the two layers could help, too.
    – whuber
    Oct 24 '14 at 17:34
  • 1
    Essentially, constant points need to be retained year round. The complexity comes from the switch in which model provides data, because when that switch occurs the geography of the points changes (due to how the models work differently from each other-beyond my control). Somehow the geography needs to stay the same though the models switch, whether that be to join one models point layer to the other, or to develop a new grid. I'll add visuals to the original post. Oct 24 '14 at 18:24

I solved this problem using this process.

For Layer A or Layer B

  1. Convert points to a grid with the point statistics tool. Point statistics uses a neighborhood analysis that can interpolate values from known cells to unknown cells.
  2. Convert raster to point, which recreates vector points on a common grid.
  3. Repeat steps 1-2 for each desired attribute. (This is where you have to bite the bullet a little bit)
  4. Join all re-gridded vector point layers into one layer.
  5. I also added X,Y coordinates for this compiled layer.

When your processing extent is standardized, and cell size of the created raster is the same, Layer A and Layer B can be transformed to perfectly overlapping, matching point grids.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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