Dealing with vagueness in complex forest landscapes: A soft classification approach through a niche-based distribution model Valerio Amici ⁎ BIOCONNET, Biodiversity and Conservation Network, Department of Environmental Science “G. Sarfatti”, University of Siena, Via P.A. Mattioli 4, 53100 Siena, Italy abstract article info Article history: Received 11 April 2011 Received in revised form 1 July 2011 Accepted 3 July 2011 Available online 7 July 2011 Keywords: Classification uncertainty Forecasting forests Forest cover map Fuzzy set Maxent Remote sensing The increasing interest in biodiversity conservation has led to the development of new approaches to facilitate ecologically based conservation policies and management plans. In this context, the development of effective methods for the classification of forest types constitutes a crucial issue as forests represent the most widespread vegetation structure and play a key role in ecosystem functioning. In this study a maximum entropy approach (Maxent) to forest type classification in a complex Mediterranean area, has been investigated. Maxent, a niche-based model of species/habitat distribution, allowed researchers to estimate the potential distribution of four forest types: Holm oak, Mixed oak, Mixed broadleaved and Riparian forests. The Maxent model's internal tests have proved a powerful tool for estimating the model's accuracy and analyzing the effects of the most important variables in the produced models. Moreover the comparison with a spectral response-based fuzzy classification, showed a higher accuracy in the Maxent outputs, demonstrating how the use of environmental variables, combined with spectral information in the classification of natural or semi- natural land cover classes, improves map accuracies. The modeling approach followed by this study, taking into account the uncertainty proper of the natural ecosystems and the use of environmental variables in land cover classification, can represent a useful approach to making more efficient and effective field inventories and to developing effective conservation policies. © 2011 Elsevier B.V. All rights reserved. 1. Introduction In recent years, the increasing interest in biodiversity conservation has led to a renewal of efforts for developing effective methods for land cover mapping (Rocchini and Ricotta, 2007; Turner et al., 2003). Image classification and predictive distribution modeling are common approaches to facilitating development of ecologically based conser- vation policies and management plans (Felix-Locher and Campa, 2010; Roloff and Haufler, 1997; Wang et al., 2010). In particular, the development of methods for the classification of forest types is a crucial issue as forests constitute the most widespread vegetation structure and play a key role in ecosystem functioning (Oren et al., 2001; Perry, 1994; Sohngen et al., 1999). The efficiency of forest management and conservation could be improved if forest managers used thematic maps created through the use of field data and remote sensing data (Amici et al., 2010a; Butler et al., 2004; McRoberts and Tomppo, 2007; Romero-Calcerrada and Perry, 2004; Wulder, 1998). Typically, thematic maps are derived from both classification of remotely sensed images and from data analysis in geographic information system (GIS) technology (Gopal and Woodcock, 1994); the information conveyed by the maps depends on the adopted classification scheme (Rocchini and Ricotta, 2007). A variety of different classification outputs can be obtained by applying different classifiers; the classifiers have different capabilities and their performance depends of the application fields and image characteristics (Liu et al., 2002). Thematic map classifications are usually crisp: a polygon or a pixel can describe only a single land cover category applying a Boolean membership in the integer set [0, 1]; thus, the degree to which it is in reality mixed cannot be differentiated, dividing the gradual variability of a landscape into a finite number of non-overlapping classes (Rocchini and Ricotta, 2007). Then in classical land cover maps, a polygon or a pixel can be described by only one land cover category. Due to the intrinsic structure of most terrestrial landscapes, which are at the same time both spatially continuous and hierarchically organized (Woodcock and Strahler, 1987), no matter how accurately map classes are defined, the uncertainty associated with class mixtures will be never completely eliminated (Fonte and Lodwick, 2004). Fuzzy classification offers an alternative to crisp logic by evaluating pixels based on their membership of each category. Fuzzy membership is based on the “fuzzy set theory”, which assumes that membership of a given category will range from complete membership (100%) to non- membership (0%), and that pixels may be classified as partial members of two or more categories (Gopal and Woodcock, 1994). The mathematical function which defines the degree of an element's membership in a fuzzy set is called membership function. The fuzzy Ecological Informatics 6 (2011) 371–383 ⁎ Tel.: +39 0577 232864; fax: +39 0577 232896. E-mail addresses: valerio.amici@gmail.com, valerio.amici@unisi.it. 1574-9541/$ – see front matter © 2011 Elsevier B.V. All rights reserved. doi:10.1016/j.ecoinf.2011.07.001 Contents lists available at ScienceDirect Ecological Informatics journal homepage: www.elsevier.com/locate/ecolinf