Improving IDW and Sample Model using ModelBuilder?

I want to create a dataset of estimated values using inverse distance weighting. I am starting with a dataset of daily values of measurements for specific coordinates. I created a model (below) which iterates through daily values and creates raster layers of interpolated values (IDW tool). Then, values from a set of specific coordinates are extracted from each raster later and are output as a table (sample tool). Finally, tables are converted to .xls and I will use a different data management software to complete the final merge.

As far as I can tell, this model works. But, it takes some time and I have three decades worth of daily measurements to run this for.

How can I make this model more efficient? -delete the intermediate raster and info table files after they have been converted to .xls --> i tried selecting "delete intermediate data" but files weren't deleted.

-use collect values, can't figure how this works

-use the batch option to run this model in batches by 1..5..10 years.

-merge in arcmap

• Idw is very simple, so compute n nearest neighbors for each point of interest and use them for interpolation for that point. Can be done in Excel in no time, because you skip raster interpolation. – FelixIP Jan 29 at 20:06
• this sounds good. can you please elaborate on this? let's say i want to compute n neighbors within 20km of the points of interest. and only use these for interpolation. how would i incorporate this into the model? thank you! – suchgreatheights22 Jan 29 at 20:28

Use Generate Near Table with N closest matches. Add 3 highlighted fields to output: Compute inverse distance (in unlikely case convert 0 near distance to 0.00001), note you have to do it just once.

1. Use join by attributes (NEAR_FID) and bring observed values into 2nd highlighted column.
2. Compute 3rd column as INV_DIST*VALUES.
3. Use summary statistics on IN_FEAT to compute interpolated values at every point of interest: I rarely use models but with Python this is a very simple task to code.