Ecological Modelling 221 (2010) 1897–1906 Contents lists available at ScienceDirect Ecological Modelling journal homepage: www.elsevier.com/locate/ecolmodel Comparison of sensitivity analysis techniques: A case study with the rice model WARM R. Confalonieri a,∗ , G. Bellocchi b , S. Bregaglio a,b , M. Donatelli b,c , M. Acutis a a Università degli Studi di Milano, Department of Plant Production, via Celoria 2, 20133 Milan, Italy b Agriculture Research Council, Research Centre for Industrial Crops, via di Corticella 133, 40128 Bologna, Italy c European Commission Joint Research Centre, Institute for Security and Protection of the Citizen, MARS Unit, AGRI4CAST Action, via E. Fermi 2749-TP 483, I-21027 Ispra (VA), Italy article info Article history: Received 19 February 2010 Received in revised form 27 April 2010 Accepted 30 April 2010 Available online 31 May 2010 Keywords: Morris method Latin hypercube Quasi-Random LpTau Sobol’ FAST Fourier Amplitude Sensitivity Test abstract The considerable complexity often included in biophysical models leads to the need of specifying a large number of parameters and inputs, which are available with various levels of uncertainty. Also, models may behave counter-intuitively, particularly when there are nonlinearities in multiple input–output rela- tionships. Quantitative knowledge of the sensitivity of models to changes in their parameters is hence a prerequisite for operational use of models. This can be achieved using sensitivity analysis (SA) via methods which differ for specific characteristics, including computational resources required to perform the analysis. Running SA on biophysical models across several contexts requires flexible and compu- tationally efficient SA approaches, which must be able to account also for possible interactions among parameters. A number of SA experiments were performed on a crop model for the simulation of rice growth (Water Accounting Rice Model, WARM) in Northern Italy. SAs were carried out using the Mor- ris method, three regression-based methods (Latin hypercube sampling, random and Quasi-Random, LpTau), and two methods based on variance decomposition: Extended Fourier Amplitude Sensitivity Test (E-FAST) and Sobol’, with the latter adopted as benchmark. Aboveground biomass at physiological maturity was selected as reference output to facilitate the comparison of alternative SA methods. Rank- ings of crop parameters (from the most to the least relevant) were generated according to sensitivity experiments using different SA methods and alternate parameterizations for each method, and calculat- ing the top-down coefficient of concordance (TDCC) as measure of agreement between rankings. With few exceptions, significant TDCC values were obtained both for different parameterizations within each method and for the comparison of each method to the Sobol’ one. The substantial stability observed in the rankings seem to indicate that, for a crop model of average complexity such as WARM, resource intensive SA methods could not be needed to identify most relevant parameters. In fact, the simplest among the SA methods used (i.e., Morris method) produced results comparable to those obtained by methods more computationally expensive. © 2010 Elsevier B.V. All rights reserved. 1. Introduction Biophysical crop models play a key role in studies on agro- biophysical systems, in agricultural and environmental planning and management (e.g., Donatelli et al., 2002). Crop models are used to estimate the behaviour of the soil-plant system as affected by weather and agricultural management, thus standing at the core Abbreviations: FAST, Fourier Amplitude Sensitivity Test; RUE, radiation use effi- ciency; Topt , optimum temperature for growth; RipL0, partition coefficient to leaves at emergence; k, extinction coefficient for solar radiation; LHS, Latin hypercube sam- pling; SLA till , specific leaf area at tillering; SA, sensitivity analysis; TDCC, top-down concordance coefficient. ∗ Corresponding author. Tel.: +39 02 50316515; fax: +39 02 50316575. E-mail address: roberto.confalonieri@unimi.it (R. Confalonieri). of dynamic modelling of water and chemical budgets (e.g., De Willigen, 1991; Hopmans and Bristow, 2002; Gervois et al., 2004; Tixier et al., 2007). For this reason, crop models are important tools to assess the impact of the agricultural practices on the environ- ment, also in relation to greenhouse gases emission and climate change (e.g., Tubiello et al., 2007). Since crop models are simplifi- cations of the agricultural systems under study, assumptions and inputs can be inaccurate to different degrees, depending on the variability associated to both input variables and parameters. Sensitivity analysis (SA) is a fundamental tool in the build- ing, use and understanding of mathematical models of all forms (Tarantola and Saltelli, 2003). SA provides information regard- ing the behaviour of the simulation model being evaluated. This information ranges from the identification of relevant model inputs (parameters or variables), to information on model balance 0304-3800/$ – see front matter © 2010 Elsevier B.V. All rights reserved. doi:10.1016/j.ecolmodel.2010.04.021