GLOBAL SENSITIVITY ANALYSIS OF GREENHOUSE CROP MODELS
Sensitivity analysis is part of the study of mathematical model behaviour. For greenhouse crop models, mostly local approaches based on the calculation of partial derivatives are applied. However, the main drawback of local approaches is that derivatives provide only information at the base point where they are computed and do not take into account variation range of the input factors. Nowadays, several global sensitivity analysis approaches such as standardized regression coefficients, scatter plots, elementary effect test, variance-based methods and Monte Carlo Filtering, are in developing. The aim of the present work is to show a procedure to carry out a global sensitivity analysis based on variance calculations, applied to the parameters of two greenhouse crop models. Firstly, uniform probability density functions (PDFs) were assigned to each of the models parameters. Secondly, several thousands of Monte Carlo simulations were carried out in order to calculate both the first-order and the total-order sensitivity indices. A combination of the SimLab (ver 3.2) software and Matlab environments was used to carry out all the simulations. Also the scatter plots for all model parameters were generated. The first test case was a simplified lettuce growth model, which considers dry weight and leaf area index (LAI) as state variables and only synthetic climatic data as its input variables. The second test case was a lettuce crop model with structural and non-structural dry weight as state variables. However, unlike to the first model actual weather data collected from a Mexican greenhouse with natural ventilation were used as inputs. Results showed that the global approach is feasible and advantageous over the local sensitivity methods, as long as one can determine PDFs to all model parameters.
López-Cruz, I.L., Rojano-Aguilar, A., Salazar-Moreno , R. and Ruiz-García, A. (2012). GLOBAL SENSITIVITY ANALYSIS OF GREENHOUSE CROP MODELS . Acta Hortic. 952, 103-109
variance-based methods, scatters plots, crop model, probability functions, FAST method