Roger,

I have been busy harvesting.  It just rained so I had a little time to scan the site.  I will try to comment briefly on some responses to your question.  First off I would not aggregate yield data at all because it is the most dense data set you have and it goes straight to revenue.  Why "dummy down" such a good data set.  It may not be good in the absolute but it is good in the relative except for the bad spots brought up about having to stop in the field etc.  High density data sets take a lot of argument about proper surfacing techniques out simply because they are so dense.  Then if they are being used in a prescription you have better transitions to rate changes that keep the controller from hunting and oscillating as I think Ted and Tim aluded to.  I tend to use the yield layer to do an initial sluthing of soil test values to see if fertility has anything to do with yield variability.  Often times on my fields it simply does not.  It often is something else like elevation related data that relates better to yield variations.  Does that make VRT unpractical?  No but it is then driven more from a profit perspective rather than a fertility perspective.  Try to stay away from the color book approach and keep the resolution higher for a more digital picture approach.  Other "dense" data sets depending on the field that may help bring the digital picture more into focus would be veris, elevation, and other contineous collected data sets.  I keep cheering on Slava at the U of N to develope compaction sensors etc. 

I won't even start on grid sampling except to agree with the statement that currently we can't afford to sample at the proper spacing to get a "representative picture" of the true fertility out in the field.  Which is why I selectively sample based on yield data and then go from there.  No matter how you slice it soil sampling is not very dense but even with that you do have to transition from one value area to another.....so what ever interpolation and surfacing method you do use try to keep it closer to a digital picture rather than a color book approach so that if nothing else you have better transitions for the controller to work with.  It used to be that a lot of systems could not handle such intense data sets and thus the reason for a "few" values.  That probably has changed by now which if it has brings us back to the question of why aggregate yield data?

I really respect my agronomist that I consult with.  I think he would agree that general agronomic approaches have not reached the digital picture resolution yet but that does not mean we should drag the data sets down to a low resolution of a few zones.  Rather we should keep the resolution high and pressure the agronomic approaches to reach a higher resolution.  Fertility is very important but because it is a corse data set does not mean it should be the main driver in the fertilizer application process and thus the reason....in my opinion.....to not aggregate yield data.

One final "common sense" look at yield data.  If yield data is so bad and always needs to be "fixed" or contoured or aggregated etc.....why do many of my yield maps show the circles of pivot tracks?  Dense data sets combined with high resolution techniques might have something to do with it.

Thanks for the chance to ramble.

Arnie