Des chercheurs de l’université de la Pampa (Argentine) ont proposé une approche intéressante pour expliquer la variation du rendement au niveau intra-parcellaire : la mesure précise de l’altitude et le calcul d’un indice topographique (Topographic precision index, TPI) associé. Les résultats mettent en évidence une corrélation significative entre le TPI et le rendement de blé mesuré au niveau intra-parcellaire sur une parcelle de 100 ha. Cette approche montre l’importance de la topographie et de la micro-topographie pour expliquer la variabilité spatiale des paramètres de production en agriculture. Dans le cas de ce travail, le modèle d’élévation a été déterminé par un relevé effectué au GPS différentiel. Cette étude met en évidence tout l’intérêt de la photogrammétrie précise tel que les drones permettent de la faire.
résumé : In this work we propose the use of the topographic position index (TPI), which takes into account the local topography for a given neighborhood, to delineate management units (MU) for site-specific systems. This study was performed in the province of La Pampa, in central Argentina, an area with sandy soils where the main limiting condition for crops is soil moisture. Usually, multi-annual yield maps are used for the delineation of MU. However, those are strongly influenced by issues that could be related to un-calibrated data and previous agronomical practices. Thus, there was a need for a methodology based on stable and unbiased parameters. The methodology was developed for a representative 100 ha field. The average size and orientation of the topographic structures were characterized applying the autocorrelation function on the topographic data, which was then used to determine an optimum neighborhood size for the TPI. TPI performed better than the topographic map to characterize the variability of the field. The correlation between yield and TPI was higher (r = 0.74) than that between yield and topography (r = 0.54). The resulting management units were delineated using an unsupervised classification approach on the TPI maps. From the confusion matrices, the overall accuracy was higher for the TPI derived maps than for the topography derived maps (62% against 47%) when compared to a yield map used as reference. We estimate that this methodology could be used for operational applications, the only requirement being topographic data for a given field, since it is simple, the algorithms used are unbiased and it could be performed using free software.
références : Mieza, M. S., Cravero, W. R., Kovac, F. D., & Bargiano, P. G. (2016). Delineation of site-specific management units for operational applications using the topographic position index in La Pampa, Argentina. Computers and Electronics in Agriculture, 127, 158-167.