Spatial scale and sampling size affect the concordance between remotely sensed information and plant community discrimination in restoration monitoring

Remote sensing cartography and GIS are part of ordinary practice in restoration ecology in discriminating patches of habitats, defining objectives, and planning the monitoring phase, but derived information is not always consistent with field survey. We assessed the mapping process efficiency in discriminating different communities, relying on plant composition data, and considering the effect of sample size and plot dimension (grain), in a heterogeneous environment in Tuscany (central Italy). We identified four land cover classes on a land cover map produced with object-oriented technique; hence we conducted a sampling of 64 plots (4 zones x 4 classes x 4 plots), estimating vascular plant cover using a point-quadrant method. Plots were nested squares with side lengths of 0.50 m, 1 m and, limited to a sub-sample, 2 m. We evaluated the effect of sample size and grain using permutational multivariate analysis of variance (PERMANOVA), testing the simultaneous response of species composition compared to land cover classes. Results demonstrated that for a sample size of 64 plots, grain does not influence the ability of discriminating among the habitat types investigated, while for a smaller sub-sample the effect of grain is significant and communities cannot be distinguished at all plot dimensions. Outcomes corroborate the hypothesis that sampling at a series of scales of observations and an adequate sample size can improve monitoring efficiency in restoration ecology.