Previous PageTable Of ContentsNext Page

CERES-Rape, a model for the production and environmental impact of rapeseed

Benoît GABRIELLE, Romain ROCHE, and Ghislain GOSSE

Institut National de la Recherche Agronomique, Unité de Recherche en Bioclimatologie,
F-78850 Thiverval-Grignon E-mail:


CERES-Rape is a simple, process-based model simulating the growth and development of winter rapeseed, over time, as a function of crop microclimate and of soil water and nitrogen dynamics. It has been developed within the framework of the CERES family of soil-crop models, and designed to predict physiological and agronomic variables as well as mass fluxes of environmental interest. Simulated variables include leaf area index, crop net photosynthesis and nitrogen uptake, partitioning of carbon and nitrogen among the main plant parts, nitrate leaching and denitrification, and ammonia volatilization from soil.

Here we evaluated the ability of CERES-Rape to predict the above variables, under several sets of pedo-climatic conditions in Europe. In most cases the model was fairly successful as regards the dynamics of crop growth, soil water and nitrogen, but the prediction of gaseous losses proved more difficult. It also appeared that model accuracy was highly dependent on the amount of information available on soil functional properties, such as hydraulic conductivity or mineralisation potential.

Overall, these comparisons allowed us to characterize the uncertainties associated with model simulations, and to issue advice for its use in agronomic and environmental assessment. Although some of its components need to be refined, and its domain of validity extended, the CERES-Rape model demonstrated a good potential to fulfill the latter objective.

KEYWORDS: modelling, environmental impact, nitrogen cycle, yield prediction


Mechanistic soil-crop models are a powerful and sometimes exclusive tool to investigate the effect of management practices on the productivity or environmental impacts of arable crops. Because they simulate the major physical and biological processes occurring in these systems, they may in principle be extrapolated to climatic conditions, soil types or cropping systems other than those on which they have been developed and tested. As a consequence, they may be used to approach the effects of climate variability, when run on long sequences of past or generated weather data, or to provide recommendations as regards fertiliser use or irrigation.

The CERES models (e.g., Jones and Kiniry, 1986) provide a simple and coherent framework for the simulation of the water, carbon and nitrogen cycles in soil-plant systems. Over the past fifteen years, they have been widely used and tested in applications ranging from decision-aid in irrigation to global assessment of crop productivity (Rosenzweig and Parry, 1994). Although they now exist for a number of crops (including wheat, maize, peanuts, sorghum and sunflower), no version for rapeseed have been reported to date. More generally speaking, efforts on simulating the growth of this crop have focused on its relation to micro-climate, notwithstanding the influence of soil water and nitrogen status (e.g., Habekotté, 1996). This may be due to the relatively minor importance of winter rapeseed (Brassica napus L.), although in Europe the acreage cropped to rape is on the increase, but also to the particular challenges posed by its functioning. It features in particular an indeterminate flowering, prior to which heavy losses of leaves, and thus of dry matter and nitrogen may occur as a result of leaf senescence and freeze.

In this paper we present a rapeseed model that incorporates the effect of solar radiation and availability of soil nitrogen on the dynamics of leaf area and crop net photosynthesis and yield components. It was recently developed based on a detailed data set from France (Gabrielle et al., 1998) and we also show how it may be extrapolated to other data sets with examples from Denmark and Germany.


CERES-Rape was developed within the framework of the CERES models (Jones and Kiniry, 1986), with improvements for the soil water and C-N turnover modules recently proposed and tested under French conditions (Gabrielle and Kengni, 1996). It runs on a daily time step, and we focus here on the plant components, which were adapted to rapeseed from CERES-Maize. They comprise the following modules:

• leaf and pod area: potential growth in area is calcultated according to air temperature; senescence terms may then apply for lack of radiation or nitrogen

• net photosynthesis: it is a function of intercepted solar radiation, through area indices, for leaves and pods

• nitrogen uptake module: a plant demand is calculated form current crop biomass and potential daily

• nitrogen stress: the latter affects both leaf area growth and, to a lesser extent, net photosynthesis. It is based on the Nitrogen Nutrition Index, which is the ratio of actual N concentration in aerial biomass to a critical N content which is optimal for biomass growth (Colnenne et al., 1998)


Experimental data

Data from three field-trials in Europe were used to test the ability of CERES-Rape to predict environmental- and agronomic-related outputs. The Table lists general characteristics about these experiments, whose locations present a 10° range in latitude, from Central to Northern Europe. Monitored variables included: crop leaf area index, biomass, and N content, and the dynamics of soil water and inorganic nitrogen. All three experiments investigated the effect of various doses of fertiliser nitrogen on the latter, with either detailed observations over a single growing season or more sparse measurements over a number of treatments. The experiment in Châlons, France, featured a detailed monitoring of the nitrogen cycle and included field-measurements of ammonia volatilization and denitrification. It served to develop and calibrate the plant components of CERES-Rape, namely its equations for LAI growth, radiation interception, nitrogen uptake, and the partitioning of assimilated carbon and nitrogen among the main plant compartements (Gabrielle et al., 1998a,b). The other two experiments were used as independent validation data sets.

Selected characteristics of the three experiments used for calibration and testing of the CERES-Rape model. Fertiliser nitrogen was supplied in mineral form, and crops received adequate protection from pests and pathogens.

Name and


Soil type

Fertiliser N doses

(1U = 1 kg N ha-1)




Châlons, F

(41.2 N; 6.1 E)

Rendzina over chalk

0 U;+135 U;

+272 U



Gosse et al., 1999

Kiel, DE

(54.3 N; 10.0 E)

Brown earth,


0 U; + 120 U;

+ 240 U


Rialto, Bandit


Jyndevad, DK

(54.3 N; 12.3 E)

Coarse sand

0 U; + 155 U;

+ 260 U



Gabrielle et al., 1998

Model parameterization

In all sites, the soil components of CERES-Rape were parameterized based on a standard procedure that converts routine soil physico-chemical characteristics into functional properties by means of pedo-transfer functions. Because varietal parameters have not been established yet, the temperature sum that determines the date of flowering was adjusted to match the observed date.


The Figure shows sample results from the CERES simulations in the three locations, for various doses of fertiliser N. Whether for the plant- or for the soil-related variables, the fit is obviously remarkably good for the calibration site (Chalons), but deteriorates in the extrapolation sites. Overall, crop biomass is still reasonably well simulated, whereas crop nitrogen tended to be under-estimated in Jyndevad and Chalons for the N-limited crops. This appeared to be a consequence of the simulation of crop N losses associated with leaf senescence in winter and spring, with the model over-estimating the concentration of N in the dead leaves. In the calibration phase, a correct simulation of the dynamics of senesced leaf area, biomass and nitrogen appeared impossible with the current framework of the model, and additionnal data seem to be required to fine-tune this critical component.

The outcome of the soil nitrate simulation was highly sensitive on the parameters of the soil organic matter and N turnover module, and more specifically on the sequence of immobilisation and mineralisation of freshly-incorporated residues from the preceding crop. For instance a better fit to the surface nitrate data at Kiel could be achieved by tuning the amount of residue present upon sowing. However we considered such fitting beyond the scope of our extrapolation exercise, and so left this parameter unchanged. It is also unrealistic to calibrate the parameters relative to such short-range processes based on only two data points, which may only provide a check that the order of magnitude predicted by the model for surface nitrate is correct over the season.

On the other hand the Châlons data provided a quite stringent test for the soil components of CERES-Rape, and the nitrate leaching chart shows that the model was quite successful at predicting this flux, as well as water drainage and net mineralisation (not shown). Again this could be achieved by using a standard parameterization from basic soil properties and management data.

Simulated (lines) and observed (symbols) dynamics of crop leaf area index, biomass and nitrogen content, surface soil nitrate and moisture content, and nitrate leaching in the various locations (1U = 1 kg N ha-1).


In the comparisons we could make so far against experimental data, CERES-Rape demonstrated a good potential for simulating the growth and development of rape-seed, as well as the dynamics of water and nitrogen in the soil profile. Further tests are scheduled against field experiments in central France, southern Italy, and southern England. These are expected to underline the extent to which CERES-Rape may be extrapolated across Europe, as well as to provide insight into the improvement of weak components of the model.

So far it appears that the prediction of crop phenology, and more specifically of the onset of stem elongation and flowering need to be better addressed. Likewise, the simulation of potential leaf area growth and its subsequent senescence is being revised based on field trials involving very early plantings.

Lastly we expect to integrate the fate of pesticides into this model, so that it may be used to yield a more complete picture of the environmental balance of rapeseed, in response to management practices such as sowing date, application of fertiliser N and pesticides, and soil tillage.


The authors would like to thank M. Langensiepen (University of Kiel, Germany) for providing the Kiel data set, and M.N. Andersen (Foulum Centre, Denmark) for providing the Jyndevad data set.


1. C. Colnenne, J.M. Meynard, R. Reau, and A. Merrien, Determination of a critical nitrogen curve for winter oilseed rape, Annals of Botany 81: 311-317, 1998.

2. B. Gabrielle and L. Kengni, Analysis and field-evaluation of the CERES models' soil components: Nitrogen transfer and transformation, Soil Sci. Soc. Am. J. 60:142-149, 1996.

3. B. Gabrielle, P. Denoroy, G. Gosse, E. Justes and M. Andetersen, Development and evaluation of a CERES -type model for winter oilseed rape, Field Crops Res. 57: 95--111, 1998.

4. G. Gosse P. Cellier, P. Denoroy, B. Gabrielle, P. Laville,E. Justes, B. Nicolardot, B. Mary, S. Recous, J.C. Germon, C. Hénault and P.K. Leech, Carbon and Nitrogen cycling in a rendzina soil cropped with winter oilseed rape: the Châlons Oilseed Rape Database, Agronomie (in the press) 1999.

5. B. Habekotté, Winter oilseed rape: analysis of yield formation and crop type design for higher yield potential, PhD Dissertation, Wageningen Agricultural University, NL, 1996.

6. C.A. Jones and J.R. Kiniry, Ceres-N Maize: a simulation model of maize growth and development,

7. Texas A&M University Press, College Station, Temple, TX, 1986.

8. C. Rosenzweig and M.L. Parry, Potential impact of climate change on world food supply, Nature 367: 133-138, 1994

Previous PageTop Of PageNext Page