Towards better irrigation management for vineyards based on machine learning algorithms and SHAP analysis: a case study in New Zealand

Simple item page
dc.contributor.authorWei H-E
dc.contributor.authorGrafton MC
dc.contributor.authorBretherton M
dc.contributor.authorIrwin M
dc.contributor.authorSandoval E
dc.coverage.spatialAdelaide
dc.date.available19/09/2022
dc.date.finish-date16/09/2022
dc.date.issued19/09/2022
dc.date.start-date12/09/2022
dc.description.abstractMonitoring and management of grapevine water status (GWS) over the critical period between flowering and veraison plays a significant role in producing grapes of premium quality. GWS monitoring is critical to vineyard management, and so determines which variables are the main drivers of GWS variation. The goal of this study is to provide viticulturists with an approach to simulate the complex relationship between canopy GWS with vegetation, weather, day of the year (DOY), and soil/terrain variables, along with the interpretation of these relationships. A case study done in Martinborough, New Zealand is used for illustration. A UAV was flown over two Pinot Noir vineyards to generate aerial images with 4.3 cm resolution, and the vegetation index, Transformed Chlorophyll Absorption Reflectance Index (TCARI), was computed for every sampled grapevine. Slope and elevation were extracted from a digital elevation model, and apparent electrical conductivity (ECa) was obtained from an EM38 survey. A local weather station provided continuous air temperature, humidity, rainfall, wind speed, and irradiance data, which were computed as variables at weekly and daily intervals. DOY was used to represent temporal trends along the growing season. Hierarchical clustering and three machine learning algorithms (elastic net, random forest regression, support vector regression) were used to regress predictors against stem water potential (Ψstem), measured by a pressure bomb and used as a proxy for GWS. Shapley Additive exPlanations (SHAP) analysis (a statistical tool that weighs the importance of each variable in a model) was used to interpret the relationship between selected predictors and Ψstem. Our results showed that the coefficient of determination (R2) of the best-performed model reached 0.7 when simulated by support vector regression using TCARI, DOY, and elevation as inputs. This study has provided proof of concept of developing regression models that would be beneficial for grapevine irrigation systems via continuous GWS monitoring, while the identification and clarification of the relationship between statistically dominant variables would assist in decision-making to attain optimal grape quality.
dc.description.confidentialFALSE
dc.format.extent? - ? (10)
dc.identifier.citation2022, pp. ? - ? (10)
dc.identifier.elements-id458115
dc.identifier.harvestedMassey_Dark
dc.identifier.urihttps://hdl.handle.net/10179/17845
dc.rights(c) The author/s (CC BY 4.0)
dc.sourceIrrigation Australia Conference
dc.titleTowards better irrigation management for vineyards based on machine learning algorithms and SHAP analysis: a case study in New Zealand
dc.typeconference
pubs.notesNot known
pubs.organisational-group/Massey University
pubs.organisational-group/Massey University/College of Sciences
pubs.organisational-group/Massey University/College of Sciences/School of Agriculture & Environment
Files
Original bundle
Now showing 1 - 1 of 1
Thumbnail Image
Name:
Irrigation Australia Paper.docx
Size:
914.55 KB
Format:
Microsoft Word XML
Description:
Download
Collections
Conference Papers