For farmers and vineyard owners, the wealth of data provided by these machines sometimes can be a bit overwhelming. Some wineries even think they are a fad. But could they genuinely provide added value? Once the initial novelty has worn off, it is time to take a close look at what they can do, together with their limitations.
Currently, the most common models are fixed-wing and multirotor. The former can fly at higher altitudes with less precision, while the second can cover less ground, but usually with higher accuracy, below 8 cm GSD per pixel. Both types of drones can carry multispectral and thermal cameras that provide images at different wavelengths, unnoticed to the naked eye. While drones can be used for bird deterrence or even delivering pesticides and fertilizers, their imaging capabilities enable them to:
- Detect diseases like leaf roll or red blotch
- Measure weed infestation
- Predict yields
- Assess moisture
Vigor maps
Most of those diagnostics can be accessed through the Normalized Difference Vegetation Index (NDVI), which is a measurement of the live vegetation in a vineyard. The images, through specific algorithms, are transformed into colored maps that show variations across the plots. As drones take hundreds of photos, these also need to be stitched to create a composite image, also known as orthomosaic. Now, with the full picture, it is time for the interpretation. For instance, the amount of reflection of the plants is a telltale sign of the amount of moisture. Then, the irrigation sectors can be fine-tuned to balance water intake.
The NDVI can also enable selective harvesting. Once ripeness and sugar levels of the grapes are related to color classes, it is possible to define uniform blocks in the vineyard. Those areas can be harvested first or later to achieve more consistent wines. Some winemakers even prefer to mix the grapes from different plots to obtain a favored blend. Zeroing on diseased plants can also prevent the harvesting of bad quality grapes, which can affect the final flavor of the wine.
Multispectral and hyperspectral imaging alone will not cut it
However useful, the NDVI is a bit like a map without names or cardinal directions. Why is there a patch of underdeveloped or dry vines? Is it a lack of irrigation? Could it be nutrient deficiencies? Maybe pests? Up till now, specialists were called in to make sense of the images. Still, human diagnostics take time and are frequently inaccurate. This is where new artificial intelligence and machine learning tools come into play. By training the software over several seasons, it can start assessing the cause of variations. Mildew spread or vine moths can have different effects on the vines. The system will then be able to flag those outbreaks with increasing precision.
One of the holy grails of modern viticulture is yield predictions. These are basic to plan harvests or to apply green harvesting -the removal of grape clusters before they are fully developed- to achieve higher quality wines. In vineyards like Concha y Toro in Chile, they are already using drones and artificial intelligence to improve their yield forecasts. Their goal is to reach 90 percent precision and above in the coming years.
Drones, while powerful devices are only one element within the precision viticulture ecosystem. Other data sources are necessary to get the big picture. And sophisticated software tools to boot.
Integration is the name of the game
Bodegas Ayuso is a household name in Spanish winemaking. Yearly producing more than 4 million liters of wine, their Estola flagship wine can be found in every Spanish supermarket and plenty of restaurants. They also export to 40 countries around the world. Over the last decade, they have made considerable progress in the modernization of their facilities. “Our winery is almost fully automated,” says María José Jerez, sales manager at this company based in La Mancha, the largest single wine-producing region in the world. However, their high-tech approach had not reached their vineyards yet, says Jerez. Thus, they recently decided to move forward with a test vineyard covered by drone flights.
At their vineyard, they are combining drone flights with a local weather station and ground sensors, as well as satellite imagery. All this data is then put through an artificial intelligence platform that integrates all the sources and crunches through large amounts of data. A single multispectral drone flight can collect several hundred gigabytes worth of data, so there is plenty of information to process. Once the system analyses all that information, it can provide real-time diagnostics to act early, before diseases or irrigation issues damage the harvest.
