[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Bounding

Brian wrote:
I am looking for some help with interpolating a surface from ungridded
data points.  I have several thousand measurements in a river with
corresponding latitudes and longitudes for each of the points.  I would
like to make a surface of these data, but have run into a small
problem.  I have been using a combination of TRIANGULATE and TRIGRID to
grid the data into a surface, however I end up with data being
interpolated outside the bounds of the river.  Is there any way to bound
the resulting grid to only include data within the river banks?

I have bumped into the same problem with marine surveys around irregular coastlines.   I usually have difficulty with concavity in the horizontal scatter of the data.    If the data is scattered in a convex hull pattern, use the boundary nodes (from boundary keyword to triangulate ....

    "B An optional, named variable that, upon return, contains a list of the indices of the boundary points in counterclockwise order.")

 to limit the extrapolation.    Unfortunately, river bends will introduce a concavity into that outer hull.  The best solution I have come up with is to manually digitize a polygon shape around the ROI that I want to keep and mask all points outside the polygon.   If the riverbanks are well mapped (vectorized), you maybe able to lift the riverbank coordinates from your dataset and use those values.   It's a brute force solution but has worked well so far.  I really would like to solve this problem differently.

Also,is there any way to determine a variance associated with the
I'm not sure what you mean.   You can interpolate the gridded value for the original X, Y locations and then perform statistics on the interpolated values vs. the original data values.

Hope it helps,

Ben Tupper

Bigelow Laboratory for Ocean Science

Pemaquid River Company