Vineyard water status assessment using on-the-go thermal imaging and machine learning

Area of Science:

• Agricultural Engineering
• Remote Sensing
• Machine Learning

Background:

• Irrigation significantly impacts grapevine yield and quality, making water status monitoring crucial for sustainable viticulture.
• Existing methods for assessing plant water status can be time-consuming and may not be suitable for large-scale vineyard management.

Purpose of the Study:

• To develop and validate an on-the-go system for estimating vineyard water status using thermal imaging and machine learning.
• To assess the performance of prediction models with and without the inclusion of reference temperatures.

Main Methods:

• An all-terrain vehicle equipped with a thermal camera collected on-the-go thermal images of grapevine canopies.
• Stem water potential was measured as the reference method for water status.
• Crop Water Stress Index (CWSI) and Jones index (Ig) were calculated using thermal indices and reference temperatures (Tdry, Twet).

Main Results:

• Machine learning models achieved high accuracy in predicting vineyard water status, with determination coefficients (R2) up to 0.65.
• Models performed comparably whether or not reference temperatures were included in the training data.
• The developed on-the-go method demonstrated reliable estimation of vineyard water status.

Conclusions:

• Thermal imaging combined with machine learning offers a fast and reliable method for estimating vineyard water status.
• The system can effectively monitor water status without the need for supervised acquisition of reference temperatures.
• This on-the-go approach provides a valuable tool for the grape and wine industry for vineyard water management and mapping.