Object-based analysis of grapevine canopy relationships with winegrape composition and yield in two contrasting vineyards using multitemporal high spatial resolution optical remote sensing

Abstract

Predictive mapping of within-vineyard winegrape yield, quality, and ripeness, using high spatial resolution optical remote sensing, relies upon relationships between image-derived canopy vigour metrics and fruit composition and yield components. Regular image acquisition of two contrasting vineyard sites enabled a temporal analysis of variation in these relationships. An image processing algorithm was developed to segment vineyard imagery into single grapevine objects. Various remote-sensing vegetation indices, calculated for each grapevine object, revealed that indices sensitive to high vegetation densities performed significantly better at predicting fruit composition and yield elements than the commonly used normalized difference vegetation index. The strength and direction of correlations between canopy vigour and season-end fruit descriptors varied by phenological stage and vineyard type. The ability of optical remote sensing to successfully map within-vineyard winegrape composition and yield may vary depending upon vineyard characteristics, management, and temporal variability in overall vineyard production.

Acknowledgements

This work was supported by the Wine Growing Futures Program, a joint initiative of the Grape and Wine Research and Development Corporation and the National Wine and Grape Industry Centre. On-ground assistance was provided by John Hornbuckle and David Smith at the Riverina vineyard and Louise Eather and Rod Muldoon at the Hunter Valley vineyard. Additional field work was carried out by Chris Haywood and Tony Sommers. We are grateful to the Hancock Farm Company Pty Ltd and Wyndham Estate Wines for field site access and generous in-kind project support. Bruno Holzapfel provided helpful comments on an earlier draft of this article. The authors appreciate ongoing technical support provided by the Charles Sturt University Spatial Data Analysis Unit (CSU-SPAN).