Computational Models Growth & Development

Improving wine yield and quality by improving carbohydrate modelling

A new study compares the accuracy of carbohydrate allocation and transport models for grapevine.
You can listen to this podcast as an audio file.

High quality grapes make high quality wine. The allocation of carbohydrates between the vegetative and reproductive organs in perennial fruiting crops is crucial for reaching and maintaining high productivity and quality.

Photosynthesis produces carbohydrates that are transported and distributed throughout the plant, affecting its survival, vegetative growth, and reproductive development. Leaves are therefore considered the carbohydrate source, while the other organs are sinks.

In viticulture, modelling can inform management inputs that affect quality and quantity. Management inputs including leaf removal, fruit thinning, and winter pruning are commonly conducted with the goal of manipulating the carbohydrate allocation between the different plant organs, through attempting to optimize the ratio between the source and sink organs during the growing season.

Junqi Zhu, researcher at The New Zealand Institute for Plant and Food Research Limited Marlborough Research Centre, and his colleagues compared two types of carbohydrate allocation and transport models to determine which simulates organ biomass variability and carbohydrate distribution in perennial fruit crops better. This work was recently published by in silico Plants.

Common assimilate pool (CP) models assume that the ability of developing plant organs to attract carbohydrates is based on sink strength alone. Mechanistic phloem carbohydrate transport (CT) models assume that the ability of developing plant organs to attract carbohydrates is based on sink strength AND topological position of the organs AND carbohydrate concentration gradient.

Illustration of the similarities and differences between the common assimilate pool (CP) model and a coupled phloem/xylem transport (CT) model.

“The simpler CP model has been widely and successfully used in various modelling platforms when simulating the biomass production and final yield of many annual crops. However, many studies involving tree or vine plant species show that organ growth on a particular branch is dependent on the carbohydrate status of that branch, in conjunction with the carbohydrate status of the whole plant. This is not captured by the CP model, but can be captured by the CT model,” says Zhu.

The authors first improved the CP model, GrapevineXL, which they developed previously. Some of the improvements they made were:

  • Adding an existing phloem/xylem transport model that includes the hydraulic properties of the stem segments (internode, cordon and trunk) and the changes of carbohydrate concentration along the carbohydrate transport pathway,
  • Incorporating the temperature response of all carbohydrate loading and unloading, and
  • Enhancing the canopy architecture representation to optimize leaf orientation to represent leaf heliotropism.

They calibrated the improved model by ensuring that non-structural carbohydrate reserve and total dry mass in each organ were captured when carbon allocation was altered. This experiment was calibrated using data from a previous study where leaves were removed during the berry ripening period to alter source strength. The three treatments were 100 leaves retained per vine, 25 leaves retained per vine, and no leaves (i.e., all leaves removed at the start of the experiment). The dynamics of non-structural carbohydrate reserve and total dry mass in each organ simulated by the improved model closely matched the observed values.

The dynamics of daily mean carbohydrate concentration c(x) under different carbon allocation via leaf treatments. The CP simulation results are in orange.

The authors then compared the CT and CP model’s abilities to capture final berry mass and carbohydrate concentration when the canopy architecture was homogenous. For this simulation, all fruits were equally close to the carbohydrate sources or evenly distributed within the canopy. They found that the carbohydrate concentration gradient along the carbon transport pathway was relatively small. For this reason, the performances of the two models were very similar.

Next, they compared the models when canopy architecture was heterogenous to elucidate the effects of the proximity of bunches to the carbohydrate sources. The distance between fruit and carbohydrate sources increased with crop load. The uniform bunch dry matter obtained by the CP model was statistically equal to the mean bunch dry matter value obtained by the CT model for each crop load treatment. However, the simpler CP model was not able to capture the increasing coefficient of variation in bunch dry matter as the crop load increased as revealed by the CT model.

The setup of the simulation for treatments representing an increase in bunch number (A) and the effects of heterogeneous architecture and crop load on bunch weight (B).

Zhu concludes, “our whole-plant model showed that the most limiting factor for fruit growth varies under different plant source/sink status, and the model can, therefore, help with disentangling the effects of different processes on fruit growth, and offer practical suggestions for vineyard and orchard management.”


Junqi Zhu, Fang Gou, Gerhard Rossouw, Fareeda Begum, Michael Henke, Ella Johnson, Bruno Holzapfel, Stewart Field, Alla Seleznyova, Simulating organ biomass variability and carbohydrate distribution in perennial fruit crops: a comparison between the common assimilate pool and phloem carbohydrate transport models, in silico Plants, 2021;, diab024,

This manuscript is part of in silico Plant’s Functional Structural Plant Model special issue.

The model and data are publicly available in the git repository:

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.