Data Mining and NIR Spectroscopy in Viticulture: Applications for Plant Phenotyping under Field Conditions

Abstract

Plant phenotyping is a very important topic in agriculture. In this context, data mining strategies may be applied to agricultural data retrieved with new non-invasive devices, with the aim of yielding useful, reliable and objective information. This work presents some applications of machine learning algorithms along with in-field acquired NIR spectral data for plant phenotyping in viticulture, specifically for grapevine variety discrimination and assessment of plant water status. Support vector machine (SVM), rotation forests and M5 trees models were built using NIR spectra acquired in the field directly on the adaxial side of grapevine leaves, with a non-invasive portable spectrophotometer working in the spectral range between 1600 and 2400 nm. The ν-SVM algorithm was used for the training of a model for varietal classification. The classifiers’ performance for the 10 varieties reached, for cross- and external validations, the 88.7% and 92.5% marks, respectively. For water stress assessment, the models developed using the absorbance spectra of six varieties yielded the same determination coefficient for both cross- and external validations (R2 = 0.84; RMSEs of 0.164 and 0.165 MPa, respectively). Furthermore, a variety-specific model trained only with samples of Tempranillo from two different vintages yielded R2 = 0.76 and RMSE of 0.16 MPa for cross-validation and R2 = 0.79, RMSE of 0.17 MPa for external validation. These results show the power of the combined use of data mining and non-invasive NIR sensing for in-field grapevine phenotyping and their usefulness for the wine industry and precision viticulture implementations.

Discrimination of Camellia japonica cultivars and chemometric models: An interlaboratory study

“…Moreover, the results obtained in the best PLS-DA models for both laboratories (approximately 80% of correct predictions) demonstrate the capacity of NIR spectroscopy to discriminate C. japonica cultivars. These results were slightly lower than the results obtained in similar works referred in the introduction section (Gutierrez et al, 2016;Lang et al, 2017;Machado et al, 2018;Páscoa et al, 2018) but in this work several cultivars from different locations were included. In this sense, it is known that different types of soil have a significant impact over the leaves of the plant (Pascoa et al, 2016) and this could somewhat explain the lower percentage of correct predictions.…”

“…In this sense, it is essential to find an easy, low-cost and effective method for the discrimination of C. japonica cultivars. Near infrared (NIR) spectroscopy presents all these features and has already been successfully applied in combination with chemometric models for the discrimination of other plants, namely citrus species (Páscoa et al, 2018), hops varieties (Machado et al, 2018), Amazonian plant species (Lang et al, 2017), and grapevine varieties (Gutierrez et al, 2016). In all these works, around 90% of correct predictions were obtained considering air-dried powdered leaves (with exception of the latter work where the leaves were scanned directly on field) which demonstrate the suitability and efficiency of the technique.…”

Plant-Based Methods for Irrigation Scheduling of Woody Crops

“…For whole canopies of plants under field conditions, near-infrared (NIR, 700-1100 nm) spectral indices are useful water stress indicators [80,81]. Thus, field-measured hyperspectral remote sensing data is being successfully used to estimate leaf water content and leaf water potential in vineyards [82][83][84][85], cotton fields [86] and maize [87], among other crops.…”

“…Recent advances both in hardware (NIR spectrometers and related systems for data storage) and software (mathematical models for identifying different components in the spectrum of the sampled material, chemometrics) have allowed the development of on-the-go or in-field NIR spectroscopy methods for assessing plant water status, among other crop characteristics [85]. Although robust correlations between observed and predicted water potential have been reported [84,85,90], the price of the equipment, the cost of data acquisition (portable NIR spectrometers and related systems are usually mounted on vehicles than run through the orchard) and the need for substantial data processing are limiting factors for the use of this approach in a context of precision irrigation.…”

Modelling Water Stress in a Shiraz Vineyard Using Hyperspectral Imaging and Machine Learning

“…The Savitzky-Golay filter has become a popular algorithm for smoothing spectroscopic data (for example, see [43,47,60]). In this study, the filter proved adept at smoothing the hyperspectral signature without significantly altering the originality of the input data.…”