First samples of high resolution vineyard yield data

In December 2017 Ackerview developed and built a yield monitor for use on a mechanical grape harvester. Once initial difficulties were overcome the yield monitor collected 23 hectares of vineyard yield data in the 2018 harvest. After fine-tuning our data processing algorithm and scaling each vineyard’s measured total yield to its total delivered harvest mass the yield distribution maps showed the coarse trends of yield distribution already known by the farmer through in-field observations. Although the yield distribution data contains some noise and some subsets of measurements are unusable due to the physical limitations of the first yield monitor, the generated yield distribution maps show clear trends over and above the broad trends observed manually by the farmer and viticulturist.

Below are some notable examples of useful insights to be gained from high resolution yield distribution data.

In the bottom centre of the vineyard a rather large area can be identified in which the yield is substantially higher than in the rest of the vineyard. Knowing the extent of this area can be used to adjust fertilizer dosage locally to save costs and avoid negative effects of over fertilization and too vigorous vine growth. The isolated green spots are resultant of the afore mentioned noise. These will be removed in further post processing. Click on the image to open a new tab showing the image in a higher resolution.

Narrow vertical strips of lower yield relative to the immediate surroundings can be seen. These are caused by sand deposits of a redirected stream impacting the nutrient and water retention of the soil locally. The lower left corner of the vineyard shows higher yield which cannot immediately be explained. Knowledge of the location and extent of areas lying outside of the modal class of performance can be used to investigate which conditional factors influence the vineyard positively or negatively. These findings can be used to improve the overall performance of the vineyard and improve profitability.

Just under halfway from the top of the vineyard five rows can be identified with much lower yields than the surrounding rows. Rows run from left to right parallel to the upper and lower edges of the vineyard. This low performance can be traced back to faulty execution of pruning instructions by one worker in the pruning team. Records are held of who pruned which set of five rows and thus future staff training efforts can be focused. This also serves as a wakeup call to the team supervisor who identified the problem too late. A different quality assurance method needs to be implemented by the supervisors in the following season to avoid losing half the potential harvest on some rows.

Numerous areas of weak performance can be identified above. The deficiency in the top left corner is caused by a blue gum tree throwing shade and leeching the soil. Low soil quality is the probable cause for the other weak areas, however, further investigation is necessary to correctly react to the deficiencies.

Depending on the cause of the yield deficiency different forms of localised process optimisation can be used to address identified problems. Lengths of dripper pipes with different delivery rates can be inserted where necessary, dosages of non-irrigation based fertilizer applications can be altered by location using the Ackerview precision farming system as seen here and very vigorous and weak vines can be pruned differently to the norm by also using the Ackerview precision farming system used on a smart phone paired to a smart watch.

Many of the physical limitations of measurement will be resolved in the fleet of yield monitors currently in development for the 2019 harvest. These devices will be always online allowing the engineers to react very quickly to possible faults and allow the device to be calibrated every morning by a minimally trained technician. These improvements will allow Ackerview to collect well above 500 hectares of yield distribution data over different farms offering diversity in conditions. Future improvements in the post processing of the raw data will reduce the noise of the resultant data and offer a cleaner and more machine usable representation of results.

Some devices of the new fleet still need to be placed on a mechanical grape harvester for the 2019 season. Interested farmers can email to get a yield monitor placed on their machine and to have the insights seen above for the 2019 harvest. The installation is non invasive to the structure of the machine and can be removed tracelessly.