Tackling uncertainty in yield through weekly yield predictions

1-minute read

The need to accurately estimate yield throughout the growing season has challenged growers for centuries. It continues to be a crucial task, especially as climate change results in high year-on-year variability. 

Yield forecasting affects not only vineyard management decisions but also wine supply planning, so growers and viticulturists spend considerable time and resources on it. Despite this, conventional yield prediction methods are highly inaccurate as yield is affected by unpredictable weather conditions and other environmental factors. 

Recent developments in technology have promised to change this status quo and make accurate yield predictions using AI. 

Is technology delivering on its promise?

Conventional yield estimation: Traditionally, in-season yield estimation involves cluster sampling in the field and bunch counts and/or weighing. This method has a number of key issues:

• Inaccurate: Conventional methods have an error of at least 30%, depending on the experience of the viticulturist/grower and environmental factors, and in some cases can even go up to 500%.

• Destructive: This method requires large amounts of samples, resulting in 2% yield loss.

• Inefficient: The process itself is very time consuming and labour intensive. For a small vineyard, the manual process involves counting clusters on 4% of the vines and weighing at least 100 clusters. These numbers significantly increase as the size of the vineyard and variability between vines increase. Moreover, 95% of vineyards still use traditional spreadsheets for yield estimation which is inefficient and inaccurate.

Our approach

The latest developments in machine learning and the increased number and quality of satellites launched in the recent years have dramatically improved the ability to tackle vineyard management challenges using AI. An emerging approach is to use this increasing power of AI and satellite imagery to make data-based predictions of yield. 

Deep Planet’s VineSignal platform tackles the most pressing issues that growers and winemakers are experiencing with managing their yields. One way we are achieving this is our weekly yield prediction tool which offers a number of unique advantages:

• Accurate: We have achieved 90-95% accuracy in our yield predictions for common grape varieties, a big jump from the industry standard of 70% at best (read more about yield prediction here). 

• Dynamic: In addition to controllable factors like planting density and irrigation scheduling, yield is continuously affected by unexpected environmental factors like weather events which also lead to varying harvest dates. Our algorithms generate weekly predictions of grape maturity and yield by block and variety so you can feel in control and take action at the right time to make your vintage more efficient. (read more about maturity predictions here)

• Efficient: Our hands-off weekly yield predictions avoid yield loss by eliminating the need for destructive sampling. It also reduces the amount of time and resources necessary for yield predictions. It is also very quick and easy to set up.

To answer the question posed at the beginning, our weekly yield prediction tools are delivering on the promise to use technology to dynamically and accurately predict yield through the grape growing season. But we’d like you to answer that question on your own, so we invite you to sign up below.