Estimating the Information VALUE of EACH Additional Soil Tensiometer in BOTH Uniform and Variable RATE Management of Potatoes.

Wireless tensiometers are now available that provide a grower with real-time measurements of crop-water conditions, but are expensive. A grower hoping to use this technology for irrigation decisions in a spatially variable field must decide how many sensors will be sufficient. At some point, the marginal sensor will cost more than the benefit of the information it provides the grower. A grower that has the infrastructure for variable rate irrigation might wonder whether there are consistently wet or consistently dry regions of the field that could be monitored and managed. The variable-rate grower might also wonder what the tension at a measured site can tell him about tension at an unmeasured location in the field. In this study we assess the value of information gained by each marginal sensor, and we assess spatial stability by comparing predicted tension values to actual tension values.

We set up three experiments in three different locations using these wireless tensiometers to observe soil-water tension over time. We also took several samples throughout the field to assess spatial variability including bulk density, texture, and complete soil water retention curves from soil cores.

In order to assess value of the information that the grower gainds from the marginal sensor, we compared hypothetical management scenarios, using fewer than all of the sensors, with hypothetical management based on all of the sensors. Specifically, we counted “false positives” and “false negatives” of our estimate being above or below a given tension threshold.

If a grower intends to infer something about a non-measured point or manage a field in zones, the field must be spatially stable: point A must have a tension that is dependably higher than point B. We use linear regression to predict tension values over randomly selected time periods at randomly selected sites, (it is adequate to do so if soil tension is a tru re andom variable? Do you have data to support this assumptions in both time and space? The specify it. Otherwise a space-time model maybe more appropriate. d compared the predicted tension with the actual tension measured at that site.

We are in preliminary stages in our data analysis. We have found that the additional benefit of tensiometers seems to drop sharply after one or two, implying that there is little economic justification for more than two sensors in a field. (as long as the net profit increases, there is still an incentive to put more sensors. Is the addtional benefitsross or a net margin? We have found some cases in which a regression-based prediction of tension out-performs an average-based prediction, implying that some spatially variable management might be appropriate. We have found some correlations between tension and yield.