NYUS.2: an automated machine learning prediction model for the large-scale real-time simulation of grapevine freezing tolerance in North America

Abstract

Accurate and real-time monitoring of grapevine freezing tolerance is crucial for the sustainability of the grape industry in cool climate viticultural regions. However, on-site data are limited due to the complexity of measurement. Current prediction models underperform under diverse climate conditions, which limits the large-scale deployment of these methods. We combined grapevine freezing tolerance data from multiple regions in North America and generated a predictive model based on hourly temperature-derived features and cultivar features using AutoGluon, an automated machine learning engine. Feature importance was quantified by AutoGluon and SHAP (SHapley Additive exPlanations) value. The final model was evaluated and compared with previous models for its performance under different climate conditions. The final model achieved an overall 1.36 ◦C root-mean-square error during model testing and outperformed two previous models using three test cultivars at all testing regions. Two feature importance quantification methods identified five shared essential features. Detailed analysis of the features indicates that the model has adequately extracted some biological mechanisms during training. The final model, named NYUS.2, was deployed along with two previous models as an R shiny-based application in the 2022-23 dormancy season, enabling large-scale and real-time simulation of grapevine freezing tolerance in North America for the first time.

Acknowledgments

The authors would like to thank Lynn Mills (WA), Beth Ann Workmaster (WI), Katherine Benedict (NS), Alexander Campbell and Jessee Tinslay (QC), Don Smith and Meredith Persico (PA), Kim Knappenberger (Portland, NY), and Hanna Martens, Felex Pike, and Bill Wilsey (Geneva, NY) for their help in LT₅₀ data collection. This work was partially supported by the Office of the Vice Chancellor for Research and Graduate Education at the University of Wisconsin-Madison with funding from the Wisconsin Alumni Research Foundation, and through the USDA ARS appropriated project 1910-21220-006-00D and New York Wine and Grape Foundation. The NY LT₅₀ data collection was also supported by Cornell University Federal Capacity Funds Grant Program. The WA LT₅₀ data collection was supported by the Washington Wine Industry Foundation. The PA LT₅₀ data collection was supported by the USDA National Institute of Food and Agriculture (NIFA) Federal Appropriation under Projects PEN0 4794 (7003432). The NS LT₅₀ data collection was supported by a Canadian Agricultural Partnership (CAP) project (ASC-12 Wine Grape Cluster Activity 7), the Canadian Grapevine Certification Network (CGCN), and the Grape Growers’ Association of Nova Scotia (GGANS). The QC LT₅₀ data collection was supported by AgriScience program-cluster on behalf of Agriculture and Agri-Food Canada.

Author contributions

H.W. and J.P.L. assembled the dataset, conducted the modeling, did the model analysis, and designed the website. G.D.M. provided technical and biological guidance of machine learning and feature importance quantification. A.P.K. provided the code for WAUS.2 and NYUS.1. A.P.K., M.K., T.E.M., A.H.W., J.L.F., A.H.H., C.P., M.R., A.A., M.G.N., J.P.R., P.W., M.C., and J.P.L. collected LT₅₀ data and contributed the sub-datasets from different regions. H.W., J.P.L., and G.D.M. wrote most of the manuscript with contributions from all co-authors.

Data availability

The data underlying this article are available in GitHub at https://github.com/imbaterry11/NYUS.2.

Conflict of interest statement

None declared.

Supplementary data

Supplementary data are available at Horticulture Research online.

References

[1] Zabadal TJ, Dami IE, Goffinet MC. et al. Winter injury to grapevines and methods of protection. Michigan State University Extension. 2007;

[2] Poling EB. Spring cold injury to winegrapes and protection strategies and methods. HortScience. 2008; 43:1652-62

[3] Dami IE, Li S, Zhang Y. Evaluation of primary bud freezing tolerance of twenty-three winegrape cultivars new to the eastern United States. Am J Enol Vitic. 2016; 67:139-45

[4] Londo JP, Kovaleski AP. Characterization of wild North American grapevine cold hardiness using differential thermal analysis. Am J Enol Vitic. 2017; 68:203-12

[5] Pierquet P, Stushnoff C. Relationship of low temperature exotherms to cold injury in Vitis riparia Michx. Am J Enol Vitic. 1980; 31:1-6

[6] Mills LJ, Ferguson JC, Keller M. Cold-hardiness evaluation of grapevine buds and cane tissues. Am J Enol Vitic. 2006; 57:194-200

[7] Londo JP, Moyer MM, Mireles M. et al. Evaluation of sample preparation practices common with differential thermal analysis of grapevine bud cold hardiness. Am J Enol Vitic. 2023; 74:0740002

[8] Wample RL, Reisenauer G, Bary A. et al. Microcomputer-controlled freezing, data acquisition and analysis system for cold hardiness evaluation. HortScience. 1990; 25:973-6

[9] Cohen J, Agel L, Barlow M. et al. Linking Arctic variability and change with extreme winter weather in the United States. Science. 2021; 373:1116-21

[10] Ferguson JC, Tarara JM, Mills LJ. et al. Dynamic thermal time model of cold hardiness for dormant grapevine buds. Ann Bot. 2011; 107:389-96

[11] Ferguson JC, Moyer MM, Mills LJ. et al. Modeling dormant bud cold hardiness and budbreak in twenty-three Vitis genotypes reveals variation by region of origin. Am J Enol Vitic. 2014; 65:59-71

[12] North M, Workmaster BA, Atucha A. Cold hardiness of cold climate interspecific hybrid grapevines grown in a cold climate region. Am J Enol Vitic. 2021; 72:318-27

[13] Kovaleski AP, Reisch BI, Londo JP. Deacclimation kinetics as a quantitative phenotype for delineating the dormancy transition and thermal efficiency for budbreak in Vitis species. AoB PLANTS. 2018; 10:ply066

[14] Kovaleski AP, North MG, Martinson TE. et al. Development of a new cold hardiness prediction model for grapevine using phased integration of acclimation and deacclimation responses. Agric For Meteorol. 2023; 331:109324

[15] North M, Workmaster BA, Atucha A. Effects of chill unit accumulation and temperature on woody plant deacclimation kinetics. Physiol Plant. 2022; 174:e13717

[16] Rubio S, Pérez FJ. Testing the Ferguson model for the cold-hardiness of dormant grapevine buds in a temperate and subtropical valley of Chile. Int J Biometeorol. 2020; 64:1401-8

[17] Zhou Z-H. Machine Learning. 1st ed. Singapore: Springer Singapore; 2021:

[18] Ashenden SK, Bartosik A, Agapow P-M et al. Chapter 2 - Introduction to artificial intelligence and machine learning.In: Ashenden SK (eds.), The Era of Artificial Intelligence, Machine Learning, and Data Science in the Pharmaceutical Industry. Academic Press; 2021; 15-26

[19] Janiesch C, Zschech P, Heinrich K. Machine learning and deep learning. Electron Mark. 2021; 31:685-95

[20] Khaki S, Wang L. Crop yield prediction using deep neural networks. Front Plant Sci. 2019; 10:621

[21] Khaki S, Wang L, Archontoulis SV. A CNN-RNN framework for crop yield prediction. Front Plant Sci. 2020; 10:1750

[22] Shahhosseini M, Hu G, Archontoulis SV. Forecasting corn yield with machine learning ensembles. Front Plant Sci. 2020; 11:1120

[23] Gall GEC, Pereira TD, Jordan A. et al. Fast estimation of plant growth dynamics using deep neural networks. Plant Methods. 2022; 18:21

[24] Ma C, Liu M, Ding F. et al. Wheat growth monitoring and yield estimation based on remote sensing data assimilation into the SAFY crop growth model. Sci Rep. 2022; 12:5473

[25] Franczyk B, Hernes M, Kozierkiewicz A. et al. Deep learning for grape variety recognition. Procedia Comput Sci. 2020; 176:1211-20

[26] Arab ST, Noguchi R, Matsushita S. et al. Prediction of grape yields from time-series vegetation indices using satellite remote sensing and a machine-learning approach. Remote Sens Appl Soc Environ. 2021; 22:100485

[27] Liu E, Gold KM, Combs D. et al. Deep semantic segmentation for the quantification of grape foliar diseases in the vineyard. Front Plant Sci. 2022; 13:978761

[28] Qiu T, Underhill A, Sapkota S. et al. High throughput saliency-based quantification of grape powdery mildew at the microscopic level for disease resistance breeding. Hortic Res. 2022; 9:uhac187

[29] Palacios F, Melo-Pinto P, Diago MP. et al. Deep learning and computer vision for assessing the number of actual berries in commercial vineyards. Biosyst Eng. 2022; 218:175-88

[30] Mohimont L, Alin F, Rondeau M. et al. Computer vision and deep learning for precision viticulture. Agronomy. 2022; 12:2463

[31] Romero Galvan F, Pavlick R, Trolley GR. et al. Scalable early detection of grapevine virus infection with airborne imaging spectroscopy. Phytopathology. 2023; 113:1439-46

[32] Gambhir N, Paul A, Qiu T. et al. Non-destructive monitoring of foliar fungicide efficacy with hyperspectral sensing in grapevine. Phytopathology. 2023

[33] Jenkins M, Mannsfeld A, Nikzad S. et al. Novel algorithms for high-resolution prediction of canopy evapotranspiration in grapevine. OENO One. 2023; 57:3

[34] Saxena A, Pesantez-Cabrera P, Ballapragada R et al. Grape cold hardiness prediction via multi-task learning. AAAI. 2023; 37:15717-23

[35] Feng L, Zhang Z, Ma Y. et al. Alfalfa yield prediction using UAV-based hyperspectral imagery and ensemble learning. Remote Sens. 2020; 12:2028

[36] Yoosefzadeh-Najafabadi M, Earl HJ, Tulpan D. et al. Application of machine learning algorithms in plant breeding: predicting yield from hyperspectral reflectance in soybean. Front Plant Sci. 2021; 11:624273

[37] Ali I, Cawkwell F, Green S. et al. Application of statistical and machine learning models for grassland yield estimation based on a hypertemporal satellite remote sensing time series. 2014 IEEE Geoscience and Remote Sensing Symposium. Quebec City, QC, Canada, 2014,5060-3

[38] He X, Zhao K, Chu X. AutoML: a survey of the state-of-the-art. Knowl-Based Syst. 2021; 212:106622

[39] Erickson N, Mueller J, Shirkov A. et al.AutoGluon-tabular: robust and accurate AutoML for structured data. 2020

[40] Fakoor R, Mueller JW, Erickson N et al. Fast, accurate, and simple models for tabular data via augmented distillation. Advances in Neural Information Processing Systems. Vancouver: Curran Associates, Inc.; 2020; 8671-81

[41] Raza A, Razzaq A, Mehmood SS. et al. Impact of climate change on crops adaptation and strategies to tackle its outcome: a review. Plan Theory. 2019; 8:34

[42] Körner C. The cold range limit of trees. Trends Ecol Evol. 2021; 36:979-89

[43] Krantz M, Zimmer D, Adler SO. et al. Data management and modeling in plant biology. Front Plant Sci. 2021; 12:717958

[44] Ellis JL, Jacobs M, Dijkstra J. et al. Review: synergy between mechanistic modelling and data-driven models for modern animal production systems in the era of big data. Animal. 2020; 14:s223-37

[45] Azodi CB, Tang J, Shiu S-H. Opening the black box: interpretable machine learning for geneticists. Trends Genet. 2020; 36:442-55

[46] Dokoozlian NK. Chilling temperature and duration interact on the Budbreak of ‘Perlette’ grapevine cuttings. HortScience. 1999; 34:1-3

[47] Londo JP, Kovaleski AP. Deconstructing cold hardiness: variation in supercooling ability and chilling requirements in the wild grapevine Vitis riparia: cold hardiness in Vitis riparia. Aust J Grape Wine Res. 2019; 25:276-85

[48] Londo JP, Johnson LM. Variation in the chilling requirement and budburst rate of wild Vitis species. Environ Exp Bot. 2014; 106:138-47

[49] Peña Quiñones AJ, Keller M, Salazar Gutierrez MR. et al. Comparison between grapevine tissue temperature and air temperature. Sci Hortic. 2019; 247:407-20

[50] Kovaleski AP. Woody species do not differ in dormancy progression: differences in time to budbreak due to forcing and cold hardiness. Proc Natl Acad Sci. 2022; 119:e2112250119

[51] Salazar-Gutiérrez MR, Chaves-Cordoba B. Modeling approach for cold hardiness estimation on cherries. Agric For Meteorol. 2020; 287:107946

[52] Liu J, Lindstrom OM, Chavez DJ. Differential thermal analysis of ‘Elberta’ and ‘Flavorich’ peach flower buds to predict cold hardiness in Georgia. HortScience. 2019; 54:676-83

[53] Luedeling E,Fernandez E. chillR: Statistical Methods for Phenology Analysis in Temperate Fruit Trees. 2022

[54] Holt CC. Forecasting seasonals and trends by exponentially weighted moving averages. Int J Forecast. 2004; 20:5-10

[55] De Rosa V, Vizzotto G, Falchi R. Cold hardiness dynamics and spring phenology: climate-driven changes and new molecular insights into grapevine adaptive potential. Front Plant Sci. 2021; 12:591

[56] Ribeiro MT, Singh S, Guestrin C. “Why should I trust you?”: explaining the predictions of any classifier. Proceedings of the 22nd ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. San Francisco California USA: ACM; 2016;1135-44

[57] Lundberg SM, Lee S-I. A unified approach to interpreting model predictions. Proceedings of the 31st International Conference on Neural Information Processing Systems. Red Hook, NY, USA: Curran Associates Inc.; 2017;4768-77

[58] Štrumbelj E, Kononenko I. Explaining prediction models and individual predictions with feature contributions. Knowl Inf Syst. 2014; 41:647-65

[59] Chang W, Cheng J, Allaire JJ. et al.shiny: Web Application Framework for R. 2022

[60] Badr G, Hoogenboom G, Abouali M. et al. Analysis of several bioclimatic indices for viticultural zoning in the Pacific northwest. Clim Res. 2018; 76:203-23