Submitted to: Groupe d'Etude des Systèmes de Conduite de la Vigne
Publication Type: Proceedings
Publication Acceptance Date: June 1, 2005
Publication Date: August 23, 2005
Citation: Tarara, J.M., Blom, P.E. 2005. Continuous crop monitoring and yield estimation via measurements of trellis tension. In: Proceeding of Groupe d'Etude des Systèmes de Conduite de la Vigne, August 23-27, 2005, Geisenheim, Germany. 2:655-660. Technical Abstract: Accurate yield predictions are desired for scheduling harvest, fermentation, and storage. The current "industry standard" method of estimating yield by hand sampling is labor intensive and produces only an instantaneous prediction, but it is ubiquitous among large operations because there has yet to be developed a less labor-intensive alternative. With its limited sampling frequency, the standard method cannot provide information about the dynamics of berry growth or other vineyard attributes, which could allow growers to make equally timely adjustments to cultural practices. We devised a novel approach to monitor growth and estimate fruit mass (i.e., yield) in grapevines by exploiting the support structure of a single-wire trellis. Our approach involves direct, continuous measurement of tension in the trellis wire, correction for environmental influences on the wire (e.g., temperature), and conversion of the output to an estimate of crop mass. On a single-wire trellis, load cells were installed in-line with the cordon wire near the end post. Each installation was calibrated with known mass. Load cell output and wire temperature were monitored continuously from budburst through leaf fall. Yield per vine was recorded. There was a linear relationship between known mass hung on the wire and wire tension. Results suggest a linear relationship between wire tension and yield that varies by vineyard row, but not within a row during a single season. The sensitivity of the system was a function of initial wire tension. Post-processing accounted for the effects of wire temperature on wire tension. The signal averaging period was sufficient to render the effects of wind on wire tension below the measurement threshold. In year 1, tension-based yield estimates were within 15% of actual harvested mass. Subsequently, we mechanically isolated a defined length of trellis wire to address the uncertainty in the length of trellis wire detected by the load cell (open-ended sample space), a refinement that also addressed concerns about the practicality of implementing the method in commercial vineyards, where calibration along great lengths of row is intractable. Mechanical isolation of a finite trellis length defined the sample space, reduced transient noise in the tension signal, and facilitated the setting of uniformly high initial tension before bud burst, thus increasing the sensitivity of the system. The mechanical isolation of a trellis segment also may obviate the need to measure simultaneously the temperature of the trellis wire, which also would simplify implementation in commercial vineyards.