Remote Sensing of Grape Maturity
(Sugar Content, pH, Titratable Acidity)

Precision agriculture is the use of spatial and temporal information and technology to educate and increase crop yield decisions. These practices often employ drone or satellite remote sensing equipment and provide a contactless means of assessing agricultural information like soil content/saturation, nutrient content, and maturity of the crop.

A study conducted at the Aristotle University of Thessaloniki remote sensing, spectroscopy, and GIS lab successfully estimates the sugar content, pH, and titratable acidity using Spectral Evolutions PSR+™. This study highlights the ability of in situ hyperspectral data techniques that can be used to educate crop harvest and determine crop maturity. This data can be collected and processed in a rapid and non-destructive way allowing for the rapid assessment of three different varieties of grape maturity parameters.

Study Area Description:
The study takes place in Ktima Gerovassiliou in Greece. This is one of the most prominent wine producing vineyards in the country, growing a variety of white and red grapes. The vineyard has approximately 400,000 grapevines and spans 72 hectares. The vineyard is composed of sandy textures with occasional clay substrates and calcareous rocks. The weather is described as “Mediterranean” with mild winters and hot and dry summers. (Kalopesa et. Al 2023)

Methods:
This study is conducted on four varieties of grape: Chardonnay, Malagouzia, Sauvignon Blanc, and Syrah. The primary process of this study is to collect field hyperspectral data using SEI’s PSR+ and compare it to traditional lab calculated measurements to determine if a model can be created to predict grape maturity parameters. Field hyperspectral data was collected directly on individual berries. The berries were then transported on ice to a lab where all ripening parameters (pH, ◦Brix, sugar content, and titratable acidity (TA)) could be measured using traditional methods. Hundreds of bunches were used to measure the ripening parameters for the study. This sampling took place in 7 field visits over 2020 to 2023. This data was then integrated into a machine learning algorithm best prediction model. The algorithm was a convolutional neural network with different pre-treatments. This algorithm was then used to create estimation models for ◦Brix, sugar content, pH, and TA.

Results/Conclusion:
By comparing the model to the actual measured parameters, they show great agreement and estimate the maturity indexes accurately. This methodology can predict the grape maturity parameters to a statistically significant degree. (R2 values of 0.84, 0.76, and 0.79, across all properties)This highlights the use of in situ hyperspectral imaging as a promising field within automated agricultural robots and agricultural modeling. This furthers the field of in situ remote sensing and allows for better understanding and profitability of agriculture in the future.

The ability to train models to predict different agricultural parameters is critical to improving profitability and reduce loss of harvest. Better modeling of maturity indicators like pH, TA, sugar content, and Brix is a great leap forward in agricultural studies and can lead to many more advancements using hyperspectral field sensors. Agricultural remote sensing using Spectral Evolution’s field hyperspectral sensors show great promise in a variety of crop studies. Their unmatched high-resolution sensors and signal to noise ratio allow them to target pigments, proteins, pH, and other parameters that can aid in the profitability and management of a wide range of crops and vegetation.

Kalopesa, E.; Gkrimpizis, T.; Samarinas, N.; Tsakiridis, N.L.; Zalidis, G.C. Rapid Determination of Wine Grape Maturity Level from pH, Titratable Acidity, and Sugar Content Using Non-Destructive In Situ Infrared Spectroscopy and Multi-Head Attention Convolutional Neural Networks. Sensors 2023, 23, 9536. https://doi.org/10.3390/ s23239536