Chemical thinning is one of the most important and unpredictable practices in apple production. Thinners are variable in their efficacy, and visible signs that a fruitlet will drop may not appear for weeks after a spray and by then, it is too late to decide whether a second application is needed.

For years, the Fruit Growth Rate (FGR) model has been our recommended tool for getting ahead of this problem. The FGR model compares the growth rate of individual fruitlets to the fastest-growing fruitlets in the tree; those growing at 50% or less of the rate of the top 10% fastest growing fruitlets are predicted to abscise (Greene et al. 2013). The FGR model can deliver a prediction within about 5 to 6 days of a thinner application, which is early enough to consider a follow-up spray. The MaluSim (https://malusim.org/) has been a great tool to utilize the FGR model in user-friendly way.

However, the limitation of the FGR model is the time it takes to implement. It requires repeated on-tree caliper measurements of the same individually marked fruitlets every 2 to 4 days, which has limited its adoption in commercial orchards (Hillmann et al. 2025).

The FSD Model

The Fruitlet Size Distribution (FSD) model, developed by Laura Hillmann and Dr. Todd Einhorn at Michigan State University, was designed to produce comparable predictions with considerably less time in the orchard. Rather than tracking growth rate over multiple visits, the FSD model uses the weight distribution of a destructively harvested sample of fruitlets collected on a single date. The principle is the same as the FGR: fruitlets whose size is 50% or less of the largest 10% of the sample are predicted to abscise (Hillmann et al. 2025).

How the Model Works

The process has three phases (Hillmann et al. 2024):

Before and at bloom, you collect some baseline orchard data: the average number of flowering clusters per tree in 2 to 5 representative trees, the average number of flowers per cluster, and your target fruit number. You also flag 120 flowering spurs on representative trees at this time, these are the spurs you will harvest for sampling later. Pre-tagging at bloom is important; selecting spurs randomly at sampling time has been shown to bias results by inadvertently favoring spurs with visible fruit and underrepresenting blank spurs.

After your thinner application, you collect 40 of your pre-tagged spurs on each of three sampling dates: 3, 6, and 9 days after the application. On each date, every fruitlet from the harvested spurs is weighed individually. The first sampling date should ideally begin three days after average fruitlet diameter in the orchard reaches 6 mm. Harvested spurs can be stored at refrigerator temperature for up to two days before processing if needed.

The FS-Predict Excel spreadsheet (free download at https://pacman.extension.org) does the rest. Once fruitlet weights are entered, built-in macros sort the data, calculate the size distribution, and generate a prediction graph showing estimated percent fruit set relative to your target.

Benefits and Limitations

Benefits:

• Higher prediction success rate than the FGR model across multiple regions and cultivars (83% vs. 67%).
• Requires only one orchard visit per sampling date rather than repeated measures of the same fruit.
• Lab processing of a full sample takes roughly half the time of the FGR method.

Limitations:

The model requires some orchard preparation at bloom (cluster counts and spur flagging) to work correctly.

Predictions arrive about 2.8 days later on average than the FGR model, which matters most when thinning applications are delayed past 6 mm fruitlet diameter.

Like the FGR model, the FSD model can overestimate final fruit set under high fruit set conditions, where intense competition between fruitlets drives additional natural abscission after the prediction window has closed.

Want to Learn More?

Full details on the science behind the model, performance across cultivars and regions, and complete step-by-step instructions for implementation are available at the links below.

Research article: 

Hillmann, L., Gonzalez Nieto, L., Kon, T., Larson, J., Musacchi, S., Robinson, T., Serra, S., & Einhorn, T. C. (2025). An Apple Fruit Set Prediction Model From Distributions of Fruitlet Mass Accurately Estimates Abscission in Four Disparate Regions of the United States. HortScience, 60(11), 2007–2017. https://doi.org/10.21273/HORTSCI18854-25

Greene, D. W., Lakso, A. N., Robinson, T. L., & Schwallier, P. (2013). Development of a Fruitlet Growth Model to Predict Thinner Response on Apples. HortScience, 48(5), 584–587. https://doi.org/10.21273/HORTSCI.48.5.584

How-to guide and free FS-Predict spreadsheet: Hillmann L, Einhorn T. 2024. The fruitlet size distribution (FSD) model: A how-to guide (2024 update). PACMAN. https://pacman.extension.org/