ORIGINAL ARTICLE Detection of the effects of restoration on community composition in a calcareous grassland: Does scale matter? Simona Maccherini 1 , Elisa Santi 2 and Michela Marignani 3 1 Biodiversity and Conservation Network (BIOCONNET), Department of Life Sciences, University of Siena, Siena, Italy 2 IRPI-CNR, Perugia, Italy 3 Department of Life and Environmental Sciences, University of Cagliari, Cagliari, Italy Keywords before-after control-impact design; Bromus erectus; grain; principal response curves; restoration. Correspondence Simona Maccherini, BIOCONNET, Biodiversity and Conservation Network, Department of Life Sciences, University of Siena, via P.A. Mattioli 4, Siena 53100, Italy. Email: simona.maccherini@unisi.it Received 1 July 2013; accepted 12 November 2013. doi: 10.1111/grs.12036 Abstract The importance of the scale of observation in vegetation science has long been recognized. We have evaluated the effect of grain (plot dimension) on the detec- tion of changes induced by shrubs cutting on plant composition of a calcareous grassland in Southern Tuscany (Italy). We conducted a 2-year before 2-year after-control-impact (BACI) field sampling design. We collected the cover of vascular plants using nested square quadrats with 0.5, 1 and 2 m sides. Any devi- ation of the restored plots from the control was analyzed by using the principal response curve (PRC) technique. Differences between the control and the restored plots, in time, accounted for 4.5, 3.3 and 7.4 of the between-plot varia- tion in species composition, respectively for 0.25, 1 and 4 m 2 plots, but only the first PRC component of the largest scale was statistically significant. The results showed that the ability to differentiate the control from the restored plots, chan- ged with increasing grain size although we did not obtain a monotonic behavior of the variance explained by the multivariate model. Including the assessment of scale dependence in the monitoring project appears useful and necessary to cor- rectly evaluate the effects of restoration actions in a consistent way. Introduction “There is no panacea in ecological sampling…” (Kenkel et al. 1989), this was declared 20 years ago on the topic of sampling procedures in population and community ecology. Although considerable progress has already been made in sampling procedures, multivariate monitoring of communities remains a problematic approach for three main reasons: inadequacy of a single monitoring design for all species, unpredictable and often unquantifiable variability of multivariate dataset, and last but not least, the cost of data collection, management and analysis (Elzinga et al. 2001). The importance of sampling scale in vegetation science has long been recognized (Greig-Smith 1952) and the abil- ity to detect patterns in ecological studies is a function of the extent and the grain of investigation (O’Neill et al. 1986). Kenkel and Podani (1991), with regards to parame- ter estimation efficiency in multivariate ecological studies, found that an increase in sample size increases estimation efficiency and recommend the use of the largest plot size possible, taking into consideration sampling time and effort. A positive effect of increasing plot size was also detected for variance explained by the ordination method in vegetation dataset (Schlup and Wagner 2008). The use of the nested sampling design has been advo- cated for many studies in vegetation science, including vegetation change over time (see Critchley and Poulton 1998). Recently, many authors have emphasized the need for a powerful sampling design and statistical analysis to evaluate restoration experiments, considering spatial and temporal variability of habitat (Michener 1997; White and Walker 1997; Chapman 1999; Chapman and Underwood 2000; Block et al. 2001; Palmer et al. 2005; Young et al. 2005). In particular, in restoration monitoring it is strongly recommended sampling at a series of scale (Block et al. 2001; Chapman 1999; Underwood and Chapman 2003; Metlen and Fiedler 2006; Marignani et al. 2007, 2008) to address the limitation caused by scale-dependent variation in ecosystems and landscapes (Michener 1997; White and Walker 1997). In spite of this evidence, multi- scale monitoring of restoration is rare (but see Abella and © 2014 Japanese Society of Grassland Science, Grassland Science, 60, 31–35 31 Japanese Society of Grassland Science ISSN1744-6961 Japanese Society of Grassland Science