Centre de documentation
scientifique

European Forest Types: toward an automated classification / Francesca Giannetti in Annals of Forest Science, vol 75 n° 1 (March 2018)  

Public [article]  inAnnals of Forest Science > vol 75 n° 1 (March 2018) Titre : European Forest Types: toward an automated classification Type de document : Article/Communication Auteurs : Francesca Giannetti, Auteur ; Anna Barbati, Auteur ; Leone Davide Mancini, Auteur ; et al., Auteur Année de publication : 2018 Note générale : Bibliographie Langues : Anglais (eng) Descripteur : [Termes IGN] algorithme de tri [Termes IGN] base de règles [Termes IGN] classification par forêts d'arbres décisionnels [Termes IGN] classification pixellaire [Termes IGN] Europe (géographie physique) [Termes IGN] Fagus (genre) [Termes IGN] milieu naturel [Termes IGN] système d'information forestier [Termes IGN] système d'information géographique [Termes IGN] système expert [Vedettes matières IGN] Inventaire forestier Résumé : (Auteur) Key message: The outcome of the present study leads to the application of a spatially explicit rule-based expert system (RBES) algorithm aimed at automatically classifying forest areas according to the European Forest Types (EFT) system of nomenclature at pan-European scale level. With the RBES, the EFT system of nomenclature can be now easily implemented for objective, replicable, and automatic classification of field plots for forest inventories or spatial units (pixels or polygons) for thematic mapping. Context: Forest Types classification systems are aimed at stratifying forest habitats. Since 2006, a common scheme for classifying European forests into 14 categories and 78 types (European Forest Types, EFT) exists. Aims: This work presents an innovative method and automated classification system that, in an objective and replicable way, can accurately classify a given forest habitat according to the EFT system of nomenclature. Methods: A rule-based expert system (RBES) was adopted as a transparent approach after comparison with the well-known Random Forest (RF) classification system. The experiment was carried out based on the information acquired in the field in 2010 ICP level I plots in 17 European countries. The accuracy of the automated classification is evaluated by comparison with an independent classification of the ICP plots into EFT carried out during the BioSoil project field survey. Finally, the RBES automated classifier was tested also for a pixel-based classification of a pan-European distribution map of beech-dominated forests. Results: The RBES successfully classified 94% of the plots, against a 92% obtained with RF. When applied to the mapped domain, the accuracy obtained with the RBES for the beech forest map classification was equal to 95%. Conclusion: The RBES algorithm successfully automatically classified field plots and map pixels on the basis of the EFT system of nomenclature. The EFT system of nomenclature can be now easily and objectively implemented in operative transnational European forest monitoring programs. Numéro de notice : A2018-318 Affiliation des auteurs : non IGN Thématique : FORET/GEOMATIQUE Nature : Article nature-HAL : ArtAvecCL-RevueIntern DOI : 10.1007/s13595-017-0674-6 Date de publication en ligne : 03/01/2018 En ligne : https://doi.org/10.1007/s13595-017-0674-6 Format de la ressource électronique : URL article Permalink : https://documentation.ensg.eu/index.php?lvl=notice_display&id=90450 [article]