Sensitivity of Load Removal From Trellis Wire for Static and Dynamic Prediction of Grapevine Yield.

We used an automated system of detecting tension in the load-bearing wire of a trellis to monitor the change in tension (ΔT) over the course of the growing season, and infer crop growth and development for purposes of estimating yield. Yield was predicted statically at the 'lag phase' (L) of berry growth (ΔT_L) and dynamically from continuous output of ΔT and known yields, using two commercial vineyards over three years. Yield varied nearly four-fold. Removal of uniformly distributed fruit load was detected to about 12 m to either side of the sensor, or 24 m total wire length. The response to sequential load removal was most sensitive to that nearest the sensor. Relationships between DT_L and yield were linear, with greater sensitivity for higher values of DT_L, meaning years with larger crops. The sensitivity of estimated yield to the date of DT_L was low, suggesting that there is some room for error in estimating L for applying the traditional means of hand sampling fruit at L. For dynamic estimates, the difference between estimated and observed yield depended on the choice of predictor year(s). The most precise predictions were between the two similarly large crop years (2007, 2009). Mean yields of 2008 and 2009 bracketed that of 2007, also leading to predictions with relatively low error. The lowest errors ranged from 5.8 to 9.5 %. However, an extremely small crop and high variability in 2008 made it difficult to predict that year accurately from the large-crop years.