Artificial Neural Network to Predict Vine Water Status Spatial Variability Using Multispectral Information Obtained from an Unmanned Aerial Vehicle (UAV)

Abstract

Water stress, which affects yield and wine quality, is often evaluated using the midday stem water potential (Ψstem). However, this measurement is acquired on a per plant basis and does not account for the assessment of vine water status spatial variability. The use of multispectral cameras mounted on unmanned aerial vehicle (UAV) is capable to capture the variability of vine water stress in a whole field scenario. It has been reported that conventional multispectral indices (CMI) that use information between 500–800 nm, do not accurately predict plant water status since they are not sensitive to water content. The objective of this study was to develop artificial neural network (ANN) models derived from multispectral images to predict the Ψstem spatial variability of a drip-irrigated Carménère vineyard in Talca, Maule Region, Chile. The coefficient of determination (R2) obtained between ANN outputs and ground-truth measurements of Ψstem were between 0.56–0.87, with the best performance observed for the model that included the bands 550, 570, 670, 700 and 800 nm. Validation analysis indicated that the ANN model could estimate Ψstem with a mean absolute error (MAE) of 0.1 MPa, root mean square error (RMSE) of 0.12 MPa, and relative error (RE) of −9.1%. For the validation of the CMI, the MAE, RMSE and RE values were between 0.26–0.27 MPa, 0.32–0.34 MPa and −24.2–25.6%, respectively.

A Review of Current and Potential Applications of Remote Sensing to Study the Water Status of Horticultural Crops

“…In the context of precision agriculture, the application of UAS, their future prospects, and knowledge gaps are discussed in [53,[78][79][80][81]. While many horticultural crops have been studied using UAS technology, the most studied horticultural crops are vineyards [31,[82][83][84], citrus [85,86], peach [32,33], olive [18,87,88], pistachio [89,90], and almond [91][92][93][94], among others [95][96][97][98][99]. Some of the UAS types used for water status studies of horticultural crops are shown in Figure 2.…”

“…Thus, such stress is reflected in the vegetation profile and easily captured by indices, such as NDVI. For this reason, NDVI has shown correlations with a wide array of crops response including vigour, chlorophyll content, leaf area index (LAI), crop water stress, and occasionally yield [34,[82][83][84]127]. The rededge band covers the portion of the electromagnetic spectrum between the red and NIR bands where reflectance increases drastically.…”

“…In the pixel-based approach, techniques, such as applying global threshold and masking, have been used. Binary masks, such as NDVI, eliminates non-canopy pixels from the sampling [82,84]. Combining the NDVI mask with the canopy height mask can exclude the pixels associated with non-vegetation, as well as vegetation that does not meet the height threshold.…”

“…Thermal water stress indices have also shown to be useful to distinguish between water use strategies of different grapevine cultivars [83,281], which is useful for customising irrigation scheduling based on the specific water needs of a given cultivar. More recently, studies have used UAS-based multispectral-based VIs to train an artificial neural network (ANN) models to predict spatial patterns of Ψ stem [84,282]. Using UAS-based multispectral data, the authors showed that ANN estimated Ψ stem with higher accuracy (RMSE lower than 0.15 MPa) as compared to the conventional multispectral indices based estimation (RMSE over 0.32 MPa).…”

“…It is also expected that the data processing pipeline of remote sensing images will become automated to be 'fit for purpose' for crop water status measurements. The ideal solution may lie in the use of satellites (or sometimes manned aircraft) for regional estimation and planning [55,260], UAS for seasonal monitoring and zoning [32,100,197,286], proximal sensors for continuous measurement [287], and artificial intelligence to derive decision-ready products [84,282] that can be used for making irrigation scheduling decisions [31,[288][289][290][291][292][293][294][295]. Continued technological developments in this space will enable growers to acquire actionable data with ease, and eventually transition towards semi-automated or fully-automated irrigation applications.…”