STICS: a generic model for simulating crops and their water and nitrogen balances. II. Model validation for wheat and maize

An evaluation of a generic crop growth model, STICS, described in Brisson et al. [11], is presented, based on an agronomic database which combines various wheat crop and maize crop situations in France. Emphasis is placed on the need to use standard references for parameterising varieties, particularly concerning the development stages. The validation was carried out for the model's output variables, defined as being the final variables of agronomic interest (yield and components, above-ground biomass, flowering and maturity dates, nitrogen contents in the plant and grain, water and nitrogen contents in the soil) using several mathematical criteria (square errors, mean deviation, efficiency). Results indicated that the two crops behave quite similarly with square errors of 1.6 t.ha(-1) for wheat yield and 2.4 t.ha(-1) for maize yield. The two yield components, grain number and grain weight, were simulated less successfully, as was the case for the simulations concerning nitrogen both in the plant and soil, which were systematically biased. However, the water content in the soil was simulated accurately. An analysis of the dynamics of the main state variables in the system, such as leaf area index or nitrogen nutrition index, which in some cases were extracted from the database, made it possible to reveal the shortcomings in the model and propose ways of modifying it. The results we will retain include the introduction of a relationship between grain number and maximal grain weight, which makes the "grain number" variable dependent on the variety, the consideration of leaf senescence due to environmental stress, and the end of nitrogen absorption at the onset of grain filling. These modifications help to improve modelling results of yield components and soil and plant nitrogen contents. They have little effect on biomass and yield, for which errors remain at levels of approximately 15%; the impossibility of reducing the error concerning biomass, and consequently that concerning yield, illustrates the model's robustness.