Planning restoration in a cultural landscape in Italy using an object-based approach and historical analysis

Abstract

We present a proposal for a standardized method to develop restoration practices capable of increasing the efficacy of landscape management and create the necessary bridge between restoration planning and landscape ecology. This methodology was developed in order to identify the reference landscape and to define areas within that landscape that possess different degrees of potential for restoration purposes in a cultural landscape. We utilized retrospective data to compare former ecosystem arrangements, taking into account ecological, spatial and temporal issues, such as historical information on changes in land use, in addition to diachronically analyzed aerial photos taken between 1954 and 2002, using an object-based approach. The test area is a Nature Reserve in Tuscany (Italy) that preserves the cultural landscape of biancane badlands – erosion forms generated on Plio-Pleistocene marine clay outcrops – which is characterized by a high erosion rate. In the first step, a land cover map was obtained by image segmentation on the 1954 photographs and the patches classified as “target habitats” were used as a selection mask on the 2002 image. As a second step, a more detailed land cover map was created for the areas selected as masks in the previous step. Hence, the target habitats that showed stability (persistence) between the two dates were excluded from the analysis, as well as the land cover classes not suitable for restoration (broad-leaved forests, arable land, artificial and other agricultural areas). The selected sites, covered by four vegetation types in the 2002 land cover map, accounted for approximately 91 ha. The method focuses on selecting sites for restoration in order to reduce efforts and negative impact and to maximize the restoration results.

Introduction

From a European perspective on nature conservation, cultural landscapes created by low-intensity farming contain a mosaic of significant wildlife habitats (Antrop, 2005, Moreira et al., 2006). The recent abandonment of traditional management techniques in favour of the intensification of agriculture is quickly reducing the amount of traditional cultural landscape units within the landscape matrix, producing an overall homogenization of the landscape (Vos and Stortelder, 1992, Debussche et al., 1999, Stoate et al., 2001, Van Eetvelde and Antrop, 2004, Antrop, 2005). The loss of the traditional state of dynamic equilibrium between human intervention and natural dynamics (Mazzoleni et al., 2004, Tatoni et al., 2004, Romermann et al., 2005), accompanied by the regeneration of natural systems (e.g. shrub encroachment), has direct implications on biodiversity (Svenning, 2002, Mac Dougall et al., 2004, Naveh, 2005, Vandvik et al., 2005, Rocchini et al., 2006). As a result, both scientists and the public are calling for restoration action.

For cultural landscapes the choice of the reference landscape can be highly arbitrary (Aronson and Vallejo, 2005). We define reference landscape as the reference system that helps identify restoration targets (Moreira et al., 2006). White and Walker (1997) identify different types of reference for ecosystems combining space and time, which are also applicable to landscapes: in this paper we refer to a “same place, different time” type, where integrative research on land use changes can inform landscape restoration (Moreira et al., 2006). Hence, temporal transformation and environmental heterogeneity need to be considered to define the reference landscape and identify sites in which different management techniques can be tested (Bell et al., 1997, Briske et al., 2003, Freckleton, 2004, Lane and LeJeune, 2005, Moreira et al., 2006). To establish a reference landscape it is crucial to locate its corresponding period of origin: historical information such as early cadastral documents and aerial photographs can be used to characterize the local landscape's historical range of variability (Antrop, 2005, Käyhkö and Skånes, 2006). A comparison between the current state and the range of variability can then be used to prioritize restoration objectives (Swetnam et al., 1999, Mac Dougall et al., 2004) and guide restoration efforts.

In restoration ecology, the selection and prioritization of sites for restoration is a major issue (Chapman and Underwood, 2000; Society for Ecological Restoration International Science and Policy Working Group, 2004). The prioritization of sites for the restoration of cultural landscapes is particularly sensitive because of the restricted availability of land due to agriculture and urban expansion. When dealing with cultural landscapes, a strict ecosystem restoration approach has limited application (Aronson and Vallejo, 2005, Moreira et al., 2006). Moreover, restoration of the disturbance regime may be critical, considering that management techniques that fight against successional trends (i.e. shrub encroachment) are far less likely to succeed than those that work with them (Bakker and Berendse, 1999, Young et al., 2005). Despite the fact that an inadequate site selection can result in restoration failure and wasting limited financial support (Russell et al., 1997, Moerke and Lamberti, 2004), few papers specifically address this issue (Palik et al., 2000, Burnside et al., 2002), except perhaps for wetland restoration (Kentula, 1997, Russell et al., 1997, Pieterse et al., 2002). The objective of our study is to identify a method for applying restoration principles to a cultural landscape, defining the reference landscape and delimiting areas with different degrees of potential for restoration purposes, taking into account spatial and temporal issues.

As Geographic Information Systems (GIS) and landscape pattern analysis are increasingly being incorporated into ecological management to support decision making (Freckleton, 2004, Romero-Calcerrada and Perry, 2004), the need for rapid mapping operations is increasing continuously. Object-oriented approaches (e.g. image segmentation) can assist the development of methods for selecting areas to be prioritized for restoration purposes, since they act on images using a pre-defined, and thus repeatable, algorithm (Blaschke and Strobl, 2001, Burnett and Blaschke, 2003, Laliberte et al., 2004, Schiewe and Ehlers, 2005, Langanke et al., 2007). The objects obtained, namely the “patches” in landscape ecology (Devereux et al., 2004), are readily available at multiple scale, allowing consistent and fast definition of sites, minimizing time and costs.

We present a case study on the use of historical information, present day observation and an object-oriented approach to select sites where non-recurring disturbances (e.g. bush clearing and small-scale fires) are expected to promote a controlled increase of the erosion rate, hampering the natural vegetation growth. The test areas are biancane badlands ecosystems, a traditional cultural landscape in Tuscany, central Italy.