Articles
MODELLING CHEMICAL THINNING IN PEACH
Article number
499_12
Pages
123 – 128
Language
Abstract
A statistical model was developed to make chemical thinning possible in the peach growing industry.
The new concept behind this model is that, in addition to the obvious factors – active compound, concentration, volume and timing of the thinning application – an accurate prediction of the tree response should consider the plant’s physiological status, and the environmental conditions following the application.
The method of model building is presented as a factor analysis on three different sets of parameters – chemical thinning application, environmental conditions and physiological status of the tree – followed by a multivariate linear regression.
The number of fruitlets remaining on the trees two weeks after the application was used as the predicted variable, while the predictors were the factors resulting from the factor analysis.
Unnecessary variables were taken out of the factor analysis by a repetitive method.
The influence of their removal was evaluated by checking the linear regression: prediction coefficient R2, Marlow’s total squared error Cp, and regression significance p.
The correlations between the predictions of the model and observed results were highly significant.
They varied between r=0.66 and r=0.85 when the model equation was applied on same-year observations, and between r=0.42 and r=0.81 when it was applied to those from another year or to data from the whole study.
The new concept behind this model is that, in addition to the obvious factors – active compound, concentration, volume and timing of the thinning application – an accurate prediction of the tree response should consider the plant’s physiological status, and the environmental conditions following the application.
The method of model building is presented as a factor analysis on three different sets of parameters – chemical thinning application, environmental conditions and physiological status of the tree – followed by a multivariate linear regression.
The number of fruitlets remaining on the trees two weeks after the application was used as the predicted variable, while the predictors were the factors resulting from the factor analysis.
Unnecessary variables were taken out of the factor analysis by a repetitive method.
The influence of their removal was evaluated by checking the linear regression: prediction coefficient R2, Marlow’s total squared error Cp, and regression significance p.
The correlations between the predictions of the model and observed results were highly significant.
They varied between r=0.66 and r=0.85 when the model equation was applied on same-year observations, and between r=0.42 and r=0.81 when it was applied to those from another year or to data from the whole study.
Publication
Authors
E. Szafran, Z. Kizner, I. David, S. Zilkah
Keywords
Prunus, factor analysis, multivariate linear regression
Online Articles (34)