In the context of precision nitrogen (N) management, to ensure that the yield potential could be reached each year, farmers have too often applied quantities of fertilizers much larger than what was strictly required. However, since 2002, the Belgian Government transposed the European Nitrate Directive 91/676/EEC in the Belgian law, with the aim to maintain the productivity and the revenue of Belgian's farmers while reducing the environmental impact of excessive N management strategies.
On the one hand, in a context where increasing frequencies of weather extremes are observed, much effort have been put in recent years to quantify climatic uncertainty impacts on cropping system productivity. On the other hand, at the plot level, crop models are powerful tools to assess the impacts of different cropping systems inputs, such as agro-environmental conditions (e.g. soil conditions, selected cultivar), management practices (e.g. sowing date, fertilisation level and calendar), and climatic conditions (e.g. extreme weather sequences) on the crop harvestable organs.
A feasible approach to cope with climatic uncertainty analysis in crop modelling is to quantify the risk associated to historical climate records. However, because historical climatic records are usually not numerous, the yield distributions obtained using such an approach may appear highly discontinuous. Therefore, to support the decision-making process, it would be highly relevant to have at one's disposal a methodology which would allow finely‑discretised yields distribution to be studied, as an answer to highly contrasted climatic variations and in interaction with the N management practices. This study is an effort that way !
The Belgian's farmer historical and current practice consists to fertilize a total 180kgN.ha-1, split in three equal fractions (60kgN.ha-1) respectively applied at tillering, stem elongation and flag‑leaf stages. This study will analyse in depth the relevance of such treatment, comparing this latter to similar practices where the N rates applied at the flag-leaf stage will be modified.
In particular, three types of farmer behaviours were analysed. The basic decision aimed to find the N strategy that maximise the yields. The second decision rule would be to look for the highest marginal net revenue. Finally, the third behaviour would aim to reduce the environmental impact linked to potential N leaching, for which taxes may be imputed to farmer under inappropriate applied N rates.
For the purpose of this research, we made use of data originating for a five years experiment designed to study the answer of a winter wheat crop (Triticum aestivum L.) under different nitrogen fertilisation level and over multiple seasons. The experiment aimed to characterise the dynamic of the soil-plant-atmosphere system. The LAI development, the biomass growth, the N level exported by the plants and remaining in the soil were measured along the seasons. The whole climatic data set of explanatory variables and necessary to run a crop model was collected at a daily time step (min. and max. temperature, vapour pressure, solar radiation, wind speed, rain amount and frequency).
The STICS soil-crop model was used to simulate crop growth. As a preliminary requirement, the model was calibrated and validated on the different seasons of measurements. Parameters of the model involved in the different formalism of yield elaboration were optimised to match the wheat growth under the Belgian's growing conditions and wheat cultivar.
A 30-yers weather database, acquired four kilometers from the experimental field, was then analysed using the LARS-Weather Generator (WG). It computed a set of parameters representative of the experimental site. On the basis of these characteristic values, the LARS-WG options were used to generate a set of stochastic weather time-series representative of the climatic conditions in the area. Not less than 300 synthetic climates were so used as inputs of the STICS model.
Our preliminary results showed that the so-obtained yield distribution discontinuities were clearly reduced. In a general way, the modulation of N level around the farmer current practice showed high level of asymmetry. In other word, these practices maximised the probability to achieve yields that were at least superior to the mean of the distributions, which decreased the risk for the farmers.
The highest practice assessed (60-60-100kgN.ha-1) was the one offering the highest yield distribution. When a simple economical criteria was computed, a 60-60-80 kgN.ha-1 protocol was found optimal between 80 and 90% of the time. However, no statistical differences were evidenced between this practice and the farmer current practice. Finally, when taxes linked to high level of potentially leachable N remaining in the soil after harvest were considered, our methodology clearly demonstrated that three years out of four, 30kgN.ha-1 could systematically be saved in comparison to what is usually done.
