International audienceThe diagnosis of agronomic problems in cropping systems is often limited to surveys among farmers, in which they are asked about the problems they encounter. This paper describes an alternative method of diagnosis to understand the variations in crop yield on a regional scale. It is based on multi-annual on-farm surveys, where data are collected on the cropping system, environment and crop-yield build-up. The different steps of the method: choice of study area and of farmers' fields; choice of methods; design of further experiments; and diagnosis itself, plus the way they are linked to each other, are developed and discussed

No abstract

On cherche à tester le modèle de HÉ NIN-Du P uis et l'ensemble des références numériques qui lui sont associées à partir d'une enquête sur des parcelles supposées au voisinage de l'équilibre organique et correspondant à une large gamme de sols et de systèmes de culture. Pour cela, on confronte les teneurs d'équilibre calculées à partir des informations sur le sol et le régime d'entretien organique aux teneurs observées. Les deux séries de valeurs ainsi obtenues sont corrélées mais les valeurs calculées sont, en moyenne, inférieures aux valeurs observées surtout si elles sont exprimées en carbone plutôt qu'en matière organique. Différentes hypothèses, susceptibles de rendre compte de ce biais systématique et de la dispersion résiduelle autour de la régression, sont discutées. Vis-à-vis du biais, l'hypothèse la plus plausible est celle d'une surestimation de la minéralisation ; quant à la dispersion, elle est probablement encore liée à celle des statuts organiques antérieurs. En l'état actuel, le modèle de HË N tN-Du p Uts peut être considéré comme un bon instrument de classement des systèmes de culture vis-à-vis de leurs conséquences sur l'évolution de l'état organique du sol dans une région donnée, mais, par rapport à la description d'une évolution, son caractère monocompartimental est sans doute trop simpliste. Mots clés additionnels : Matière organique, résidus de culture, fumier. SUMMARY Cropping system and organic status of soils : application of the 7 V c / ! M-Z) M/ ?MM model. We tested the H EN) N-Du p uis model and the numerical references associated with it. This was done by means of a survey on 89 farm plots, supposed to be close to organic equilibrium and corresponding to a wide range of soils (fig. 1) and cropping systems in a restricted area. For each plot, the expected value of organic content (calculated from information collected on soil and on organic supply) was compared with the observed content. Expected and observed values proved to be correlated (fig. 2 and 3), but expected ones were systematically lower than those observed, especially if the content was expressed as carbon instead of organic matter. Different hypotheses were examined to explain this bias and the residual dispersion around the linear regression (expected content versus observed). It was concluded that the most plausible origins were, for the bias, an overestimation of the annual rate of mineralization k 2 , and, for the residual dispersion, the influence of unknown former organic status. In addition, the collection of numerous results concerning plots where farmyard manure was used allowed us to estimate the isohumic coefficient (k l) of this product (0.34). In the present state of knowledge, the Ht NIN-Du p uis model with associated numerical references can be considered as a good tool for grading cropping systems from the point of view of organic management. However in order to predict the change in organic content of a given soil with time, this tool seems inadequate, due to its monocompartmental character.

In a small area of the Paris basin in France, 71 pea (Pisum sativum) crops, cv. Solara, were examined to elucidate the determinants of their yield over three years.

In a first approach, we note that the observed yields were often different from the objectives ( fig. 1). However, using grain number per unit area as a criterion of nitrogen efficiency, allowed us to analyse the cases of bad prevision. 4 types of response curves to nitrogen are distinguished (tabl. 5 and fig. 3 We conclude in favour of the validity of the tested method, which allowed to obtain, in 8 cases out of 10, when the yield objective was not over-estimated and the soil structure favourable, at least the previous yield at the previous fertilizer rate.