A Sensitivity Analysis of the SPACSYS Model

Agronomie, 2002

This study was aimed at developing and using a method for analysing the sensitivity of a crop simulation model to its internal parameters, in order to provide information on the relative effect of parameters on intermediate or output variables. Because the total number of parameters is very large in a crop simulation model such as the STICS model, we divided our analysis into two steps. The first step evaluates the most sensitive parameters acting in each specific module on intermediate variables. The second step compares sensitivity produced by each module on overall model outputs. Analyses are conducted using analysis of variance and surface response tools. Results showed that grouping together situations to estimate the sensitivity in multiple growth conditions is correct as long as the effect of the conditions is not preponderant. If it is, grouping may conceal the different significant effects of the parameters according to the conditions. Analysing the effect of several parameters simultaneously makes it possible both to show their possible interactions (or correlations) and to rank them according to their importance in a module. We found that some of the parameters only have an effect depending on growing conditions: calibration must be made in these conditions. The effect of a parameter also depends on the considered output variable: we must therefore choose the parameters to be fitted according to the variable of interest. In case, parameters have little influence in all circumstances, fine calibration is unnecessary. The intermodule study showed that the parameters of shoot production modules all have an effect on yield and water uptake, but not on the quantity of water drained and nitrogen leached. On the contrary, the parameters of the soil dependent modules act on all the output variables and particularly on the drained water and leached nitrogen. Among the studied parameters, ones which have a systematic effect are the field capacity and parameters of the rooting module. sensitivity analysis / crop model / crop production, water use, drained water, leached N / wheat, maize / parameter estimation Communicated by Daniel Wallach (Castanet-Tolosan, France)

Ecological Modelling, 2014

Process-based crop models use many cultivar parameters to simulate crop growth. Usually, these parameters cannot be directly measured and need to be calibrated when the crop model is applied to a new environment or a new cultivar. Determining the relative importance of the cultivar parameters to the specific outputs could streamline the calibration of crop models for new cultivars. Sensitivity analysis can quantify the influence of model input parameters on model outputs. We applied the variance-based global sensitivity analysis to the wheat module of the Agricultural Production Systems sIMulator (APSIM) for the first time and calculated the sensitivity of four outputs including yield, biomass, flowering day, and maturity day to ten cultivar parameters including both the main and total effects sensitivity indices. We explored the effects of changing climate, soil, and management practices on parameter sensitivity by analyzing two fertilization rates (0 and 100 kg N ha −1 ), across five sites in Australia's cereal-growing regions. Uncertainties for the four outputs with varying cultivar parameters, climate-soil conditions and management practices were evaluated. We found that yield was most sensitive to the cultivar parameters that determine the yield component (grains per gram stem, max grain size, and potential grain filling rate) and the phenology parameters that determine length of the reproductive stages (thermal time from floral initiation to flowing and thermal time from start grain filling to maturity). All ten cultivar parameters affected biomass, amongst which the parameters of vernalization sensitivity and thermal time from floral initiation to flowering were the most influential. Fertilization influenced the rank order of parameter sensitivities more strongly than climate-soil conditions for yield and biomass outputs. Under 0 kg N ha −1 , with the variation of cultivar parameters simulated yield varied from 64 to 3559 kg ha −1 (minimum and maximum), biomass from 693 to 12,864 kg ha −1 . Fertilization of 100 kg N ha −1 increased the maximum yield to 9157 kg ha −1 and biomass to 22,057 kg ha −1 . We conclude that to minimize cultivar-related uncertainty, cultivar parameters should be carefully calibrated when applying the APSIM-wheat model to a new cultivar in a new environment. By targeting the most influential phenological parameters for calibration first and then the yield component parameters, the calibration of APSIM can be streamlined. (G. Zhao). such as climate change, and food and energy security . Different cultivars of the same crop may behave distinctly under varying growth environments. Process-based crop models mimic these cultivar-related behaviours by using a set of cultivar parameters which adjust crop development, phenology, partitioning, and reproduction performance. However, it is often difficult to obtain accurate values for these parameters. A common practice to estimate them is fitting the simulated results to measured data using methods such as regression or http://dx.

Field Crops Research, 2018

Process-oriented agro-ecosystem models are increasingly applied to assess crop management options or impacts of climate change on agricultural production, food security and ecosystem services. Thereby, the aggregation of initial soil and climate information is a widely used approach for performing simulations at larger scales such as regions, nations or even globally. In this context, the ability of models to respond to different site conditions is essential for high quality impact assessment through the use of modelling tools. As part of a model inter-comparison the present study investigated models' yield response on variable site conditions using data sets from two well-documented fields, one located in Germany and one in Italy. The fields were sampled at 60 and 100 grid points, respectively, and soil and crop variables were recorded at varying intensity for the entire simulation period covering three growing seasons. The data was provided successively to the participating modelling groups in three calibration steps (a, b, and c) and the first growing season was considered for calibration. Model validation was based on these steps and each growing season as well as on the entire simulation period considering the soil state variables mineral nitrogen and water content (N, WC) as well as crop yield, biomass, and leaf area index (LAI). The WC was best depicted by the models, resulting in high correlation coefficients (r) up to 0.81 between simulated and observed values. The root mean square error (RMSE) of simulated N ranged from 20 kg ha -1 to 1072 kg ha -1 regarding all steps and growing seasons. The annual within-field variability of yields was better simulated by the models when observed subsoil information was provided. However, the RMSE ranged from 0.5 t ha -1 to 3.5 t ha -1 at the German field, and from 0.6 t ha -1 to 5.9 t ha -1 at the Italian field, respectively. It was found that intensified calibration did not necessarily lead to improved model output. Furthermore, single models showed specific inconsistencies in their algorithms when, for example, underestimated WC was associated with overestimated yields. In total, the sensitivity of models to spatially variable site conditions differed considerably. The importance of quality-assured soil and yield information for model improvement was highlighted.

. A comprehensive sensitivity analysis for an ecological simulation model.

The present article shows the results of a study on the soil module of the STICS (Simulateur Multidisciplinaire des Cultures Standards, developed by INRA, France) crop model. Simulation models are often applied to regions where conditions are substantially different from the ones which the model was originally developed for and validated against. This was the reason to study the sensitivity of the STICS soil module and to analyze model behavior with regard to spatial transferability. The model was parameterized with data collected from an area close to the German city of Trier. Using this parameterization as a baseline, an initial study was carried out on the sensitivity of the soil parameters. This was followed by an analysis of model behavior concerning parameters which also in the real system are responsible for successful or poor plant growth. This provided some improvements over the initial simulation results. However, the model failed to match the real system's behavior concerning yield, biomass development, and root growth. From various approaches to parameterization it has become clear that a high level of abstraction is required to produce a satisfactory model of the soil-plant-atmosphere continuum and in particular the soil water dynamics. This applies especially to extreme locations as well as to relatively extreme climatic years.

Environmental Pollution, 2011

Modelling complex systems such as farms often requires quantification of a large number of input factors. Sensitivity analyses are useful to reduce the number of input factors that are required to be measured or estimated accurately. Three methods of sensitivity analysis (the Morris method, the rank regression and correlation method and the Extended Fourier Amplitude Sensitivity Test method) were compared in the case of the CERES-EGC model applied to crops of a dairy farm. The qualitative Morris method provided a screening of the input factors. The two other quantitative methods were used to investigate more thoroughly the effects of input factors on output variables. Despite differences in terms of concepts and assumptions, the three methods provided similar results. Among the 44 factors under study, N 2 O emissions were mainly sensitive to the fraction of N 2 O emitted during denitrification, the maximum rate of nitrification, the soil bulk density and the cropland area.

Ecological Modelling, 2010

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 relationships. 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 computationally 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 Morris 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. Rankings 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 calculating 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.

Environmental Modelling & Software, 2010

Sensitivity analysis studies how the variation in model outputs can be due to different sources of variation. This issue is addressed, in this study, as an application of sensitivity analysis techniques to a crop model in the Mediterranean region. In particular, an application of Morris and Sobol' sensitivity analysis methods to the rice model WARM is presented. The output considered is aboveground biomass at maturity, simulated at five rice districts of different countries (France, Greece, Italy, Portugal, and Spain) for years characterized by low, intermediate, and high continentality. The total effect index of Sobol' (that accounts for the total contribution to the output variation due a given parameter) and two Morris indices (mean m and standard deviation s of the ratios output changes/parameter variations) were used as sensitivity metrics. Radiation use efficiency (RUE), optimum temperature (T opt ), and leaf area index at emergence (LAI ini ) ranked in most of the combinations site  year as first, second and third most relevant parameters. Exceptions were observed, depending on the sensitivity method (e.g. LAI ini resulted not relevant by the Morris method), or site-continentality pattern (e.g. with intermediate continentality in Spain, LAI ini and T opt were second and third ranked; with low continentality in Portugal, RUE was outranked by T opt ). Low s values associated with the most relevant parameters indicated limited parameter interactions. The importance of sensitivity analyses by exploring site  climate combinations is discussed as pre-requisite to evaluate either novel crop-modelling approaches or the application of known modelling solutions to conditions not explored previously. The need of developing tools for sensitivity analysis within the modelling environment is also emphasized.

1998

J.G.Wesseling, J.G. Kroes and K. Metselaar. Global sensitivity analysis of the Soil-Water-Atmosphere-Plant (Swap) model. Wageningen (The Netherlands), DLO Winand Staring Centre. Report 160. 62 pp.; 5 Figs; 13 To gain insight in the sensitivity of the results of the one-dimensional simulation model for transient unsaturated/saturamodel Swap to changes on some of its input parameters a sensitivity analysis was performed with this model. Generation of parameter values and the analysis were carried out with the statistical package Usage for different crop-soil combinations. The large influence of the bottom boundary condition is shown. The influence of input parameter strongly varies with the chosen crop/soil combination. It is recommended to perform a more extensive research on all input parameters.

2017

Site conditions and soil properties have a strong influence on impacts of climate change on crop production. Vulnerability of crop production to changing climate conditions is highly determined by the ability of the site to buffer periods of adverse climatic situations like water scarcity or excessive rainfall. Therefore, the capability of models to reflect crop responses and water and nutrient dynamics under different site conditions is essential to assess climate impact even on a regional scale. To test and improve sensitivity of models to various site properties such as soil variability and hydrological boundary conditions, spatial variable data sets from precision farming of two fields in Germany and Italy were provided to modellers. For the German 20 ha field soil and management data for 60 grid points for 3 years (2 years wheat, 1 year triticale) were provided. For the Italian field (12 ha) information for 100 grid points were available for three growing seasons of durum whea...