MAY 20: SYMPOSIUM DAY 2 — THE NEXT 50 YEARS

​

Future-Proofing Vineyards​

​

‘Money-vine’: Ensuring Long-Term Economic Sustainability in Vineyards | Brent Sams, Research Scientist, GALLO Viticulture Research

​

Simply put, the best way to future-proof vineyards is to ensure that grape production is economically sustainable. Challenges from extreme weather events, water scarcity, and reduced labor availability must be addressed to ensure long-term vineyard profitability. Technology advancements are therefore critical to optimizing future vineyard performance. Today most vineyard managers employ traditional applications such as manual control of irrigation, random vine and fruit sampling methodologies, and single-rate block-level applications for inputs such as water, nutrients, and standard cultural practices. Future economic success will depend on growing more and higher quality grapes on less land with fewer inputs. Vineyards and other perennial crop systems inherently require more intensive farming methods than commodity crops due to their greater establishment costs and longer production life. Vineyard trellis and irrigation systems increase establishment costs and contribute to the difficulty of industry-wide scaling of automation and mechanization due to the multitude of the systems employed. Vineyard profitability currently suffers from high fixed and annual production costs, as well as few options for optimizing management practices based on spatial variability. The spatial quantification of yield and production input costs are the first steps for optimizing management practices and vineyard performance. Once spatial yield and net revenue variability within a vineyard block are characterized, managers can optimize productivity by identifying production zones and applying differential practices within each zone based on spatial variability. The additional revenue generated by variable rate management must exceed the expense of its implementation to achieve economic viability. While the agronomic knowledge for variable rate management and other precision farming practices currently exists, viable tools and practical applications must be developed for their implementation to ensure vineyard economic sustainability for the future. 

​

New Multi-scale Data Streams and Machine Learning Approaches for Precision Irrigation Management: Lessons from GRAPEX | Andrew McElrone, Research Plant Physiologist, USDA-ARS; Adjunct Professor, UC Davis 

​

Premium winegrape production thrives in Mediterranean climates, which are characterized by hot and dry summers, but can also experience prolonged droughts and heat waves that increase crop water demands. Under these conditions, most winegrape growers in California utilize irrigation strategies that support water demands of the vine while maximizing fruit quality.  Growers face increasing scrutiny over water use associated with the Sustainable Groundwater Management Act that mandates a long-term, sustainable approach to managing groundwater to prevent overdraft and protect water resources. Precision irrigation management of vineyards must address three core questions: how much water to apply, when to apply it, and where to apply it? Our large and interdisciplinary research group has developed technologies that integrate ground-based and remotely sensed data streams to answer these questions and guide irrigation management. Here, I will share details on how these data streams were developed in partnership with growers within the framework of the GRAPEX project, and how machine learning/AI approaches have been used to predict near term crop water demands and map root-zone soil moisture and vine stress using thermal and multispectral reflectance. I will also touch upon plans to develop computer vision tools for additional agriculture trait targets.

​

Tomorrow’s Grapevines: AI-Powered Genomics and Breeding

​

Building Out Our Information and Technology Stacks: Genomics-Enabled Breeding for Crop Quality and Stress Tolerance | Christine Diepenbrock, Associate Professor in Plant Sciences, UC Davis

​

Genetics and genomics provide an important set of tools for use in breeding. Of these, one common tool has been marker-assisted selection, which involves assaying specific positions in the genome that have a large effect on priority traits. Another common tool has been genomic selection, which involves leveraging genome-wide markers to predict breeding values for priority traits. Genomics can also be helpful for understanding how plant materials relate to each other, and (once paired with trait data) how that variation gives rise to traits of interest. AI, and AI-powered genomics, are bringing a complementary set of opportunities to next-generation plant breeding. We can increasingly uncover relationships between multi-omic data and both classical and novel trait measurements, to understand productivity and quality (and factors underlying differential performance for these traits) across genotypes, environments, and management scenarios. I will discuss recent and ongoing work in our lab, including through collaborative partnerships, to measure, dissect, and predict nutritional quality and stress tolerance traits to inform breeding. A couple of the examples I will describe have taken place in diverse crop and/or environment types, and a couple will involve phenolic compounds given their cross-relevance in the wine industry. I will also discuss where we (humbly, and thus far) have found the value added to be with regards to the use of AI in plant breeding. Finally, I will describe why we are still keen on both foreground and background selection in breeding, and a couple to few challenges that have seemed critical to address in co-design/build/test/learn cycles even while building out our information and technology stacks.

​

Leveraging AI to Unlock the Genetics of Rootstock Adaptation | Luis Diaz-Garcia, Assistant Professor, Department of Viticulture and Enology, UC Davis 

​

Grapevine rootstocks are a critical but historically underserved target for genetic improvement, yet their role in determining vine performance under drought, salinity, and biotic stress has never been more consequential as growers navigate increasingly water-limited and challenging production environments. AI is now reshaping how we approach rootstock breeding, enabling the integration of high-dimensional genomic, phenotypic, and environmental data to accelerate the development of superior and more diverse germplasm. This presentation offers a broad perspective on how AI-driven tools are transforming rootstock research and breeding programs, drawing on advances from the wider field as well as ongoing work at UC Davis. Key topics include the application of genomic prediction and machine learning to complex quantitative traits such as drought tolerance and root system architecture; the use of genome-environment association approaches to identify adaptive loci across diverse growing regions; and the emerging role of computer vision in capturing root morphology and physiological responses that were previously intractable at breeding scale. By synthesizing recent developments across research programs, this presentation aims to illustrate how AI is not simply accelerating conventional breeding pipelines but enabling fundamentally new questions about the genetic basis of rootstock adaptation, with direct implications for the resilience and sustainability of viticulture worldwide.

​

Towards the Adoption of Modern Breeding Tools Based on High-Throughput Genomic and Phenomic Data to Speed Up Grapevine Variety Release | Vincent Segura, INRAE Researcher in Quantitative Genetics, AGAP Institute

​

In the context of climate change, it is urgent to release new grapevine varieties that meet the ever-growing expectations of winegrowers. However, traditional breeding programs are long, typically spanning around 16 years between the original cross and the registration of a new variety. While modern tools, like molecular markers, are being routinely used to select disease-resistant varieties, their application to more complex traits—such as those typically targeted for grapevine adaptation to the constraints imposed by climate change—remains limited. Genomic and phenomic selection, which make use of high-throughput genomic and phenomic data, have been proposed as promising modern tools in several species to complement marker-assisted selection, but their application in perennial fruit crops like grapevine is still in its infancy. Here, I will review the few studies reporting the use of phenomic and genomic selection in grapevine and discuss how they can be integrated to boost grapevine breeding programs.

​

Data Science Tools for the Future of Wine

​

Foundation Models in Agri-Food Science: A Quick Tour and Two Case Studies | Xin Liu, Professor of Computer Science, UC Davis 

​

Foundation models are emerging as a powerful new paradigm for learning from large, diverse biological and agricultural data, with growing potential across agri-food applications. This talk will briefly review how foundation models are being used in agri-food science and highlight both their promise and current challenges. I will then present two ongoing research projects: one on multimodal foundation models for bacteria detection, and another on developing genome foundation models for lettuce. Together, these examples illustrate how foundation models may support more scalable, data-driven discovery in agri-food science.

​

The Effect of GLP-1 Drugs on the Demand for Beverages | Jill McCluskey, Regents Professor and Director of the School of Economics, Washington State University 

​

In recent years, many consumers are consuming less alcohol for several possible reasons, including health-conscious moderation and the widespread adoption of glucagon-like peptide-1 (GLP-1) drugs, such as Ozempic. To understand this better, we classify grocery store scanner data for a wide variety of drink products, including wine and other alcoholic beverages, into categories based on their levels of food processing. We present sales trends of selected products.  Then, we focus on households where someone started taking a GLP-1 medication. By comparing their shopping habits before and after starting the drug—and comparing them to people who have not started taking them—we are able to measure exactly how much these medications change the way people buy drinks.

​

From Finding a Bottle to Global Wine Market Intelligence: 27 Years of Wine-Searcher Data | Jules Perry, CEO, Wine-Searcher

​

For 27 years, Wine-Searcher has helped consumers find wines and compare prices. Along the way, it has also built one of the world’s most extensive datasets on wines, spirits, beers, prices, availability, and consumer interest across 133 countries. This talk will show how it has grown from a practical search tool into a source of global wine market intelligence. I will share a few highlights, trends, quirks, and stories from the dataset, and explore how these insights can be useful to both the wine industry and consumers. I will finish with some thoughts on the future, as AI and other new technologies change the way information is searched, analysed, and trusted.