|
| I 100% agreee with the idea of measureing inputs and outputs, tying back to yield and developing models of what is really going on. We actually did an analysis project back in 2001 where using a techniquie known as multivariate regression (more popular in economic circles than amoung agronomists) we analyzed pooled soil test, planting, and yield data across three producers. Got specific yield responses to things like soil test P and PH, even soil "quality" properties like OM and CEC. A spreadsheet model was then developed where one could enter input prices, crop prices, and get back specific recomendations for our operation based on our own data. At the time, I thought we had unlocked the keys to the precision ag data puzzle and just needed to walk through the door. And how incredibly frustrating the last five years has been in that regard !!!!
I still believe that we caught a glimpse of what the future of data analysis will look like, but progress will be slow and one step at a time. Since Todd Pederson was mentioned, a good comparision is the NEATA talk I heard from him a few years ago on using PA data for field/equipment efficency work. A fascinating talk, but one that looked fairly far into the future, and if you tried to jump right in and apply his concepts very widely today, I bet one would quickly become very frustrated at the amount of basic data organization/massaging/formating that you had to do just to get to the point of starting to look at useful trends and results. Our sampling effort seemed the same way, the amount of legwork requred to make it all happen, and the realization of how much more we really needed to get the results we wanted made the whole project seem over-whelming and basically shut it down after one year. Esp. when amount of work is directly related to cost. In the retail sale example, only very recently has the technology reached the point where one could study the comparision of "donut sales to regular coffee customers" without breaking the bank. In many ways, such tools and systems will be much slower to arrive in commodity production agriculture.
But I digress from what prompted this discussion. Knowing all of the above, I still have this basic interest. Agronomy people (which I am not) have been making product application recommendations for a long time off a one-time soil test. Many have now moved to the "information age" and are making more standardized, equation-based recs off of a one-time geo-referenced set of samples (grids, zones, pick your flavor of the day). My interest is all of this soil test data is now being stored digitally, and is there any value in previous years of soil test results when doing the rec? If there is, do we have the tools/processes available to make it work on a commercial level? Forget yield monitors even exist at the moment.. we were able to make application recs for quite some time without them.
| |
|
Thinking about data..
