Interpolation of Values in ArcGIS

I wanted to interpolate the trend values (trend of sunshine hour calculated over 20 years) throughout the study area from the available station points. My confusion is in the selection of interpolation tools I shall use in ArcGIS to do that (IDM, Kriging, Natural Neighbor, spline and all that).

• Can you describe in more detail what your input data is? What software (Spatial Analyst, Geostatistical Anlayst) do you have? – Jens Jan 16 '13 at 8:56
• the input is trend values for various stations that I calculated in excel values like (-2.7,-2.7,-2.6,3.9,-4.8,0.7,1.6,0.6,0.4,0.2,0.9,0.1,3.4,3.6,1.2,0.9,) I have 3D analyst, Geostatistical analyst, as well as spatial analyst in my ArcGIS 10 – Uttam Paudel Jan 16 '13 at 9:00
• do you have ONE value for each station? or do you have many values (e.g. one for each year) for each station? – Jens Jan 16 '13 at 9:03
• its one value for one station – Uttam Paudel Jan 16 '13 at 9:03
• do you have barriers? e.g a mountain. on each side of the mountain there may be different climate conditions so that you may want do use barriers. – Jens Jan 16 '13 at 9:09

I want to give you some hints about the differences in the methods.

More information can also be found on the esri help pages An overview of the Interpolation toolset

Because your variable (sunshine) depends on a second variable (level of pollution) Kriging may be a good method. You can use your second variable as an “external drift”. Kriging requires very deep knowledge of geostatistic.

Natural Neighbor does not support barriers. I believe that barriers are important for you.

So let's look at some differences between IDW and Spline with barriers:

Spline:

• using a two-dimensional minimum curvature spline technique
• creates a smooth surface
• The resulting surface passes exactly through the input points. But: Usually some values of the resulting surface can be higher than the maximum of the input values (and lower than the minimum of the input values). You have to decide if that is possible for you. I think that it's good for you because your stations are probably not right there where in reality the minimum or maximum values occurs. Try it and see if the calculated minima and maxima are meaningful.

IDW:

• less smooth surface than spline
• best result when our station distribution is dense (regard to the local variation of sunshine)
• The maximum value of the resulting surface is never higher than the maximum of the input value. The minimum value of the resulting surface is never lower than the minimum of the imput value.
• The resulting raster is a convex hull. Therefore, the border of your area of interest may not be entire covered by the resulting raster.

I would start with Spline with barriers. You can use Spatial Analyst for this. The profile tool of 3D Analyst can help to examine the results.

• Thanks @Jens, with your explanation i think i should go with Spline with barriers as you suggested, but again, setting the barriers is equally tough... I will give it a try and compare the outputs and their acceptability.... Thanks again. – Uttam Paudel Jan 16 '13 at 15:44