Determining vine leaf water stress by functional data analysis

Abstract

The control of water stress in vines is key to ensuring satisfactory harvest yields and quality wines. Most indirect methods used to evaluate water stress from leaf reflectance are based on calculating vegetation indexes. However, since these methods only consider reflectance values for specific wavelengths, results are disappointing as no existing index is capable of accurately calculating water stress from reflectance. We describe an alternative approach based on the construction of a linear regression model for functional data. The method is based on considering the spectral signature of leaves as a function rather than as a set of discrete reflectance values. A differential operator can also be included in the mathematical model to penalize excessively high local fluctuations. The proposed method was applied to a set of real data for Californian vine leaves, producing results that were better than those resulting from vegetation indexes, and particularly when the differential operator was included.

Keywords:

• functional approach
• water stress
• linear regression
• differential operator

2000 AMS Subject Classifications :

• 62J05
• 62J99
• 68T35
• 92F99