How to master berry forecasting

If you’ve got a single-picking crop like apples, total tonnage is more straightforward to measure based on fruit weights and number of fruits. However, when it comes to berries and other crops that require multiple harvests over an extended period, the situation becomes far more complicated.

Take the example of strawberries. Depending on the specific variety, strawberries can produce fruit over a prolonged period—often around 15 weeks per season. To effectively forecast the yield from the plant, it's essential to understand both the total potential harvest, and also how much we can expect to pick in any given week.

Growing degree days vs calendar days?

One of the most significant challenges in accurately forecasting berry yields is understanding the ripening process. For many growers, the instinct is to rely on calendar days, reasoning that strawberry ripening typically takes about six weeks. However, this method often leads to inaccuracies, particularly with the increasing unpredictability brought on by climate change.

One key to better forecasting lies in understanding growing degree days (GDD) or growing degree hours (GDH). This approach involves monitoring the accumulation of heat units over time, which plays a crucial role in a plant's development and fruit ripening. By approaching forecasting based on growing degree days rather than calendar days, we can account for the changing climatic conditions and the varying impacts this has on plant growth.

Understanding the ripening process

To adopt the growing degree days method effectively, we need to examine how strawberries ripen in relation to different variables:

• Flower development: Different flowers will ripen at varying rates based on environmental conditions. Understanding this variability helps in accurately predicting harvesting times.
• Weight changes of berries: As strawberries develop, their weights will fluctuate across the season. Forecasting models must incorporate this data to provide realistic yield estimates.
• Ripening curves: There are typically three distinct ripening curves to consider throughout the season. These curves help in understanding the dynamics of berry production and when fruit will be at its peak for harvesting.
• Solar intensity: The amount of sunlight a plant receives can significantly affect ripening speed and fruit weight. By factoring in solar intensity along with growing degree days, growers can create a more accurate forecast.

With over 50 different ripening curves tailored to various berry varieties, our AI continuously evolves to account for changing conditions.?

The importance of berry weight in forecasting

As we know, berry weights are not static; they fluctuate throughout the growing season. For instance, if we consider a flower on a strawberry plant that begins its journey in the first 100 growing degree hours, the initial weight of the berry may be around 20 grams. Fast forward to when that flower has developed into a ripe fruit after approximately 1,000 growing degree hours, and the weight could increase to 30 grams. This variance in berry weight necessitates a detailed modelling approach throughout the season.

Berry weight brings a considerable amount of variability to forecasts, second only to weather conditions. One of the key challenges in measuring berry weight is the inherent difficulty of taking accurate samples in the field. Human error often biases the sampling process. For example, when conducting random sampling, pickers might unintentionally favour either the largest or smallest berries, leading to skewed data. Or simply picking berries from a single row might not reliably represent the overall performance of the entire crop.

Another common technique involves taking a sample during a picking event—calculating the average weight based on the number of berries collected during that time. While this method can provide some insights, it can also be misleading if pickers are focused on specific size grades. For example, in blueberry harvesting, if pickers target only the "jumbo" size category, smaller fruits may remain on the plant, thereby overestimating the average berry weight in the dataset. This distortion can have cascading effects when forecasting the remaining yield on the plants.

It’s imperative to smooth out these averages and perform a thorough analysis of berry weights, continuously working closely with our customers to refine our forecasts.

Weather conditions: working with a variable beyond our control??

Weather variables add another layer of complexity in forecasting berry yields. Although we are not in the business of weather forecasting, we utilise valuable data from in-field sensors that monitor temperature and solar radiation in both tunnels and open fields. This real-time information allows us to gain a better understanding of current weather conditions compared to local weather services. By establishing a correlation between our data and historical weather patterns, we can improve the accuracy of our forecasts.

We’ve observed significant temperature discrepancies within the same farm, sometimes equivalent to weeks of growing degree days difference within a single batch of fruit. Such variances can drastically impact how we approach yield forecasting. We leverage data partnerships with companies like WayBeyond that enable us to analyse and assimilate valuable field-level insights that greatly inform our yield predictions.

We also look at long-term forecasts from weather services to anticipate the number of growing degree days expected over the next two weeks. These forecasts allow us to project future harvest availability, helping growers to plan effectively.

Recognising behavioural factors in harvesting?

Variations in picker availability can influence harvesting outcomes. For instance if pickers don’t show up on a designated picking day, growers could miss the opportunity to harvest fruit that may otherwise go to waste. Additionally, fluctuations in berry pricing may lead growers to leave ripe fruit on the plant longer than ideal, further complicating forecast models.

We focus on providing customers with dynamic forecasts of available fruit, which helps them make informed decisions regarding their harvesting strategies. This approach allows them to evaluate their specific circumstances and make real-time choices based on economic realities.

Comprehensive data: The foundation of accurate forecasting

Forecasting berry yields is a complex task that extends far beyond simple calculations of fruit quantity and weight. At Bitwise Ag, we don’t just look at the immediate variables affecting berry yields; we incorporate historical data alongside new insights to build a robust forecast model. This includes analysing plant dates, pruning dates, and the specific management practices employed across different blocks on a farm.

High-fidelity data—derived from advanced GoPro technology and AI systems—enables us to track berry growth accurately. Our object detection algorithms can assess fruit condition every two weeks, enabling us to see the movement of fruits through phenological stages with precision. This data-driven approach ensures we can create reliable forecasts.

Find out more about Bitwise Agronomy solutions at: https://bitwiseag.com/?