Résumé
It was hypothesized that a fuzzy logic inference system (FIS) would be able to combine critical plant and soil based spatial information to provide nitrogen (N) fertilization recommendations adapted to conditions prevailing at different locations within fields. Identification of key soil and plant parameters was based on experiments with N rates ranging from 0 to 250 kg/ha conducted over three seasons (2005-2007) on fields of contrasting edaphic and topographic features. For soils, the following parameters were selected: apparent soil electrical conductivity (ECa), elevation (ELE) and slope (SLP). The NSI (Nitrogen Sufficiency Index; the ratio of NDVI status to the one of a fully fertilized reference plot) was used as the plant-based parameter of interest. N rate optimization by the FIS was defined against maximal corn growth in the weeks following in-season N application. This mid-season growth was assessed through remotely sensed imagery by a CASI (Compact Airborne Spectrographic Imager) at the 1 m2 pixel resolution. Best mid-season growth was experienced in areas of low ECa, low SLP and high ELE. Expert knowledge was built up into a set of rules involving three defined soil parameters levels (low, medium and high) and two NSI levels (low and high). For favourable soil conditions, maximal mid-season growth was obtained without in-season N. Recommendations for N were high where soil conditions were naturally unfavourable to growth. FIS appeared as a useful and handy tool to incorporate expert knowledge into spatially variable N recommendations.