Global sensitivity and Uncertainty analyses of a dynamic model for the biosystem Lettuce (Lactuca sativa L.) Crop-Greenhouse

Mathematical models help understand a system's behaviour, evaluate hypotheses, control it and use it as a virtual environment for training on the possible outcomes that arise in real systems. Nevertheless, it is not enough to have a model; it should be analysed by considering its components. The research objectives were to perform global sensitivity and uncertainty analyses on a dynamic mathematical model for the output variables: air temperature, thermal mass temperature, relative humidity, and accumulated lettuce biomass inside a greenhouse. The model was based on mass and energy balances at a non-steady state. The global sensitivity analysis of the model's parameters was realised using Standard Regression Coefficient (SRC), EFAST, and Sobol. The first-order indices were found with the three methods, and the total effects indices were found with the EFAST and Sobol methods. The model parameters were analysed to determine their influence over the output variables. The uncertainty analysis considered the variability in the parameters, assuming a uniform distribution for each with a 20% variation from its nominal value. The Monte Carlo and Latin hypercube samplings were used with 5000 samples. The more influential parameters for air temperature were related to the physical characteristics of the greenhouse; for the thermal mass, it was found to be the soil temperature; for relative humidity and biomass, the more significant parameters were those related to the leaf area index. None of the output variables showed a normal distribution. The highest uncertainty was linked with the biomass, followed by the thermal mass temperature, air temperature, and relative humidity. •Global sensitivity analysis helps to determine the influence of the model parameters.•The measurements and focus should be placed on the more influential parameters.•The uncertainty analysis helps address the output uncertainty of the model's inputs.•There is no difference between analyzing the output variables' mean or the integral.•Most parameters of a greenhouse climate model are not normally distributed.