Interpolation with barriers providing unrealistic coastal values

[coryhodes]

I am working with trace metal data collected from the Chukchi Sea, Alaska.  I have tried numerous interpolation techniques (from both the Geostatistical and Spatial Analyst toolbox) and have so far had the most success using the Spline with Barriers interpolation technique from the Spatial Analyst toolbox.  The problem I have with my results is that I continually receive unrealistic interpolated coastal values (where no data points exist) and my contour lines run into the coastline instead of following the Alaskan coastline.  I am currently using my 'Alaska coast' shapefile as my barrier.  I have experimented with creating a Buffer along the coast and using that as my barrier, again with unsatisfactory results.  I have a hunch that these unrealistic interpolated coastal values are a result of a lack of data points near the coastline, or possibly that the near-coast data points that I have are not close enough in value. 

I have attached a sample image of my results.  The black dots represents the point features (of known value) from which the interpolation was created.  The min and max values on the color ramp represent the highest and lowest known values (entered manually).  Any assistance with this issue would be greatly appreciated.

[JeffreySwain]

In general with contour generation at the shore, I recommend break lines with a constant elevation for an area that small.  That way the contours will match the data.  I assume you are modeling the subsurface in the bay rather than ocean elevations, but I always prefer breaklines to points when generating surfaces such as you describe unless the point cloud is great like lidar.  When you add the breakline or 3d line, the values for the interoplation should line up correctly.  You should probably create a TIN or Terrain from the combination of lines and points.  Simple point distributions like described will probably keep producing the same results despite the interpolation method chosen.