CONICET | Buscador de Institutos y Recursos Humanos

Ecological sites are the basic entity used in rangeland health assessment. This study evaluates the syn-ergistic use of multi- and hyper-spectral satellite imagery for sub-pixel classification of ecological sites in semiarid rangelands. Hyperion and Landsat enhanced thematic mapper (ETM) data are included in atwo-step procedure to mapping ecological sites in Patagonian rangelands of Argentina. Firstly, mixturetuned matched filtering and logistic regression analyses are used for Hyperion data processing to obtainecological site probability images in the area covered by hyperspectral imagery. Secondly, artificial neu-ral networks are applied to model the relationships between the spectral response patterns of Landsatand the probability images from Hyperion, and used to map ecological sites over the entire study area.Overall classification accuracy was 81% (kappa = 0.77) with relatively high accuracies for all ecologicalsites demonstrating that their spectral signatures are sufficiently distinct to be detectable. Better accu-racies were obtained for shrub steppes with desert pavement (producers and users accuracies of 89% and 84%, respectively), and shrub-grass steppes associated to tertiary calcareous outcrops (producers and users accuracies of 100% and 86%, respectively), while poorer accuracies resulted for shrub-grasssteppes on old alluvial plains (producers and users accuracies of 75% and 56%, respectively). Fuzzy mapsof ecological sites as presented in this research can provide rangeland managers with a tool to stratify thelandscape and organize ecological information for rangeland health assessment and monitoring, priori-tizing and selecting appropriate management actions, and promoting the recovery of areas degraded inthese environments.

enviar mensaje