Orchard Studies: 2008 - 09
Experiments were conducted for both almond and pistachio. Almond leaves were collected on two dates in 2008 and four dates in 2009. Samples were taken from a Kern County orchard where a four-level N fertigation trial was being conducted, thus a wide range of leaf N concentration was expected. Multiple trees were sampled at each N level. Pistachio leaves were collected on three dates in 2009 from a Kings County orchard.
In the laboratory,leaf area, fresh weight and chlorophyll index were determined for each leaf. The optical absorption spectra in the visible and near infrared (NIR) regions were measured by placing the central region (parallel to and including the main vein stem) of each whole leaf in a customized holder. A high-resolution spectrophotometer was then used to record the absorbance spectrum.
After scanning, leaf samples were dried and the leaf dry mass were measured. Subsequently, samples were sent for N analysis to the UC Davis Stable Isotope facility. Nitrogen concentration was determined for all samples on a unit leaf surface area (ug N /cm2) basis.
Data Analysis
The optical (derivative spectra) and leaf N concentration datawere merged and a partialleast regression (PLS) and a multiple linear regression (MLR) analyses were conducted. Separate NIR prediction models for leaf N content were initially developed for each of the 2009 almond and pistachio datasets. The 2009 almond model was then evaluated for its ability to predict the leaf N content of the almond leaves from 2008. In addition, model development was done to evaluate the feasibility of developing a single unified model that could be applied to either almond or pistachio leaves.
Results
The results show that it is feasible to develop NIR prediction models for individual leaf N content using either PLS or MLR modeling techniques. For both almond and pistachio a single PLS model for each species could be developed that accurately predicted the leaf N content across all sampling dates (see figure at right).
The models predicted leaf N content on a unit leaf area basis in order to develop models that were not impacted by the changing ratio of non-structural to structural composition of the leaf as the leaves mature. The models developed here were able to predict leaf N on a leaf area basis across all sampling dates and were not affected by the phenological age of the leaves.
Continued Work
Orchard trials were conducted in 2010 and are planned for 2011, to continue to refine and validate the models. The challenge is to develop a predictive tool for growers that is low-cost, fast and sensitive.
Fig.2. 'Predicted N' developed through spectral measurements appears to be a good predictor of N content of the leaves for both almond & pistachio