Comparative Land Evaluation approaches: an itinerary from FAO framework to simulation modelling.

Abstract The FAO framework for Land Evaluation (LE) has been the primary procedure used worldwide to address local, regional, and national land use planning. Despite its widespread and long-term application, the process has been criticised by the scientific community for its qualitative and empirical base, which is not effective to address many new agro-environmental challenges where the dynamic characterisation of the interrelated physical and chemical processes taking place in the soil landscape is a must. In today's environment, the LE expert is asked to choose the best LE methodology considering costs, complexity of the procedure and benefits in handling a specific land evaluation. Unfortunately very little scientific literature supports this choice especially in terms of the comparison of different LE approaches. In this scenario, we performed a forage maize land suitability study by comparing different methods having increasing complexity and costs. The study area was comprised of approximately 2000 ha and was located on the Lodi Plain (Po Valley) of North Italy. The primary land use in the region is forage maize and this study was designed to assess forage maize biomass. Methods were developed and applied to compare the efficacy of procedures of increasing complexity. The range of the 9 employed methods ranged from standard LE approaches to the extensive use of simulation modelling (SWAP and CropSyst), using as data input pre-existing soil information (soil map 1:50,000) and also hydraulic properties measured as well estimated by PTF(pedotransfer functions). The different methods were compared based on both predictive ability and cost. Independent estimates of forage maize biomass were obtained by locally tested remote sensing measurements, and predictive ability was estimated using statistical indexes including correlation, relative variance, and ANOVA to evaluate the mean differences in maize biomass. The level of expertise required to apply a specific methodology was a factor in the cost of the LE. In addition, an increase in method complexity corresponded to a higher quality/quantity of input parameters and as a consequence higher costs. Generally, more complex methods gave better results in terms of their predictive ability but this occurred in much discontinuous steps; this finding contradicts a simplified view on the better performance of more complex and mechanicistic methods. Methods operating on measurements resulted in increased performance relative to those operating on PTF. Furthermore, methods operating on “true” soils without aggregations (i.e. soil-mapping units) exhibited higher performance.