terclim by ICS banner
IVES 9 IVES Conference Series 9 Comparison of imputation methods in long and varied phenological series. Application to the Conegliano dataset, including observations from 1964 over 400 grape varieties

Comparison of imputation methods in long and varied phenological series. Application to the Conegliano dataset, including observations from 1964 over 400 grape varieties

Abstract

A large varietal collection including over 1700 varieties was maintained in Conegliano, ITA, since the 1950s. Phenological data on a subset of 400 grape varieties including wine grapes, table grapes, and raisins were acquired at bud break, flowering, veraison, and ripening since 1964. Despite the efforts in maintaining and acquiring data over such an extensive collection, the data set has varying degrees of missing cases depending on the variety and the year. This is ubiquitous in phenology datasets with significant size and length. In this work, we evaluated four state-of-the-art methods to estimate missing values in this phenological series: k-Nearest Neighbour (kNN), Multivariate Imputation by Chained Equations (mice), MissForest, and Bidirectional Recurrent Imputation for Time Series (BRITS). For each phenological stage, we evaluated the performance of the methods in two ways. 1) On the full dataset, we randomly hold-out 10% of the true values for use as a test set and repeated the process 1000 times (Monte Carlo cross-validation). 2) On a reduced and almost complete subset of varieties, we varied the percentage of missing values from 10% to 70% by random deletion. In all cases, we evaluated the performance on the original values using normalized root mean squared error. For the full dataset we also obtained performance statistics by variety and by year. MissForest provided average errors of 17% (3 days) at budbreak, 14% (4 days) at flowering, 14.5% (7 days) at veraison, and 17% (3 days) at maturity. We completed the imputations of the Conegliano dataset, one of the world’s most extensive and varied phenological time series and a steppingstone for future climate change studies in grapes. The dataset is now ready for further analysis, and a rigorous evaluation of imputation errors is included.

DOI:

Publication date: May 31, 2022

Issue: Terclim 2022

Type: Article

Authors

Luca Brillante1, Greg Jones2 and Diego Tomasi3

1Department of Viticulture & Enology, California State University, Fresno, USA
2Abacela Vineyard and Winery, Roseburg, OR, USA
3CREA-VE Research Centre for Viticulture and Enology, Conegliano, Italy

Contact the author

Keywords

phenology, climate change, time series, imputation methods, recurrent neural networks

Tags

IVES Conference Series | Terclim 2022

Citation

Related articles…

Enhancing hydric stress tolerance by editing the VviMYB60 promoter with CRISPR/Cas9 

Climate change presents increasing challenges to viticulture, particularly with rising water stress contributing significantly to yield losses and damages. The identification of the MYB60 transcription factor, which regulates stomatal opening and closing in Arabidopsis thaliana and Vitis vinifera, offers potential solutions. Notably, knockout studies in Arabidopsis have shown reduced stomatal opening and increased drought tolerance in myb60 mutants. Additionally, the grapevine ortholog, VviMYB60, can restore the wild-type phenotype of Arabidopsis myb60 mutants. Further investigation of the Arabidopsis promoter region has revealed that mutations in DOF motifs lead to reduced expression of AtMYB60.

Physiological behavior of the Chasselas grape variety under water deficit: 30 years of experiments in Switzerland

In the context of increasingly hot and dry summers, the adoption of innovative irrigation technologies has become essential for maintaining grape production while minimizing water use.

Caracterización sensorial y preferencias de los consumidores en vinos de crianza biológica elaborados a diferente graduación alcohólica

Las tendencias actuales del mercado apuntan hacia el consumo de vinos con menor contenido en alcohol, y, por otra parte, de vinos con características especiales y diferenciadoras, siendo los vinos con indicación geográfica o denominación de origen los más demandados.

FOLIAR APPLICATION OF METHYL JASMONATE AND METHYL JASMONATE PLUSUREA: INFLUENCE ON PHENOLIC, AROMATIC AND NITROGEN COMPOSITION OFTEMPRANILLO WINES

Phenolic, volatile and nitrogen compounds are key to wine quality. On one hand, phenolic compounds are related to wine color, mouthfeel properties, ageing potential. and are associated with beneficial health properties. On the other hand, wine aroma is influenced by hundreds of volatile compounds. Fermentative aromas represent, quantitatively, the wine aroma, and among these volatile compounds, esters, higher alcohols and acids are mainly responsible for the fermentation bouquet.

Variations of soil attributes in vineyards influence their reflectance spectra

Knowledge on the reflectance spectrum of soil is potentially useful since it carries information on soil chemical composition that can be used to the planning of agricultural practices. If compared with analytical methods such as conventional chemical analysis, reflectance measurement provides non-destructive, economic, near real-time data. This paper reports results from reflectance measurements performed by spectroradiometry on soils from two vineyards in south Brazil. The vineyards are close to each other, are on different geological formations, but were subjected to the same management. The objective was to detect spectral differences between the two areas, correlating these differences to variations in their chemical composition, to assess the technique’s potential to predict soil attributes from reflectance data.To that end, soil samples were collected from ten selected vine parcels. Chemical analysis yield data on concentration of twenty-one soil attributes, and spectroradiometry was performed on samples. Chemical differences significant to a 95% confidence level between the two studied areas were found for six soil attributes, and the average reflectance spectra were separated by this same level along most of the observed spectral domain. Correlations between soil reflectance and concentrations of soil attributes were looked for, and for ten soil traits it was possible to define wavelength domains were reflectance and concentrations are correlated to confidence levels from 95% to 99%. Partial Least Squares Regression (PLSR) analyses were performed comparing measured and predicted concentrations, and for fifteen out of 21 soil traits we found Pearson correlation coefficients r > 0.8. These preliminary results, which have to be validated, suggest that variations of concentration in the investigated soil attributes induce differences in reflectance that can be detected by spectroradiometry. Applications of these observations include the assessment of the chemical content of soils by spectroradiometry as a fast, low-cost alternative to chemical analytical methods.