The abandonment of traditional anthropogenic activities is an important driver shaping landscape patterns. Therefore, multi-scale pattern analysis over time is needed to identify appropriate scales for biodiversity conservation and monitoring of abandoned landscapes. We compared spatial and temporal changes in a pair of alpine watersheds in Italy (Cajada and Tovanella), which are similar in size, geo-climatic conditions, and land-use histories; but have had divergent anthropogenic abandonment processes since the 1950s. We hypothesize that this divergence has led to corresponding dissimilarities in multi-scale patterns of landscape change. To examine this hypothesis, we analyzed land cover maps from three years (1954, 1980/83, 2006) and described the changes using transition matrices. For each year and watershed, landscape heterogeneity and a set of class-level metrics (i.e. percentage of the landscape, area-weighted mean patch size, patch density, area-weighted mean shape index, edge density, and aggregation index) were also measured at different scales using random sampling techniques, and the results were summarized by using scalograms. Woodland expansion occurred mainly at the expenses of grasslands, meadows, and shrublands. These changes were greater during the first time-period (1954-80/83) than in the more recent period (1980/83-2006), with a mean annual value that decreased from +5.18 to +1.33ha/year and from +4.08 to +1.96ha/year in the abandoned and managed watersheds, respectively. Landscape heterogeneity decreased over time with a similar pattern in both watersheds, which indicates a general process of homogenization. Management regime affected the spatial-scale response of class-level metrics; these metrics showed a variety of multi-scalar responses, which were not always consistent over time and under different management regimes. When considering the response of the indices across spatial-scales for both watersheds, certain historical curves showed a scale break, representing a significant change in the shape and slope of the curve (i.e. scale divergence). The presence of scale breaks in the scalograms can potentially reveal important thresholds for biodiversity. For example, grassland and meadow patch density at small spatial scales (<200m radius), which was found to be important for protected butterfly species, had a greater reduction over time in the managed watershed when compared to the abandoned watershed. In conclusion, the findings of this study indicate that there is good potential for understanding changes in landscape patterns under different management abandonment regimes by combining spatial and temporal analysis of class-level metrics.