Forest-savanna mosaics are important for biodiversity, but the savannas in these mosaics are often considered degraded forests due to low tree cover, and are thus targeted for tree planting. Yet, these mosaics may be naturally bistable systems, wherein disturbance regimes such as fire and herbivory create alternative stable states of forest and savanna. Globally, forest-savanna mosaics have been present from pre-historic times and map to regions with high biodiversity today. Here, we conduct a meta-analysis of paleo-ecological studies in Central India-a highly biodiverse forest-savanna mosaic landscape threatened by tree plantations today-to understand the spatiotemporal antiquity and dynamics of the mosaics across this region. We find that alternate states of low and high tree cover have been present in Central India since the early Holocene and that the tree cover is explained by the interaction of mean annual precipitation (MAP) and the disturbance regime of fire. We find no statistical evidence for bimodality or hysteresis-conditions that are required for alternative stable states-although patterns suggestive of alternative stable states are present. Further, in contradiction to the hypothesis of high and low tree cover states being stable, this system transitions between alternate states of high and low tree cover at time periods ranging from ~40 to 220 years. Switching back and forth between alternate states is significantly more frequent in sites with higher richness of fire-resistant tree taxa. Our historical data thus lend support to the idea that low tree cover regimes have been created or maintained through interactions between climatic conditions and disturbance regimes such as fire, and that tree cover can increase when either of these factors changes. The study further suggests that restoration should focus on maintaining the ability to switch between low and high tree cover rather than increasing tree cover in Central India.
The loss and fragmentation of natural habitats due to the intensification of agricultural land use have detrimental impacts on the biodiversity of arthropods. The reduction of natural habitats results in a decreased availability of essential resources, which may select for rapid development and phenotypes enhancing dispersal ability. We here compared replicated populations of the butterfly Coenonympha pamphilus in field-caught females and their laboratory-reared offspring across two landscape types: highly fragmented and intensified "modern" and less fragmented "traditional" agricultural landscapes. We also examined the effects of food stress and landscape parameters representing compositional and configurational landscape heterogeneity on intraspecific trait variation at different spatial scales. The differences between the two landscape types in butterfly traits were nonsignificant throughout, but both field-caught females and their offspring exhibited various responses to the measured landscape parameters. In particular, landscapes with (1) high heterogeneity of habitat patches (i.e., relatively smaller grassland patches with high boundary length), (2) higher proportion of non-crop habitats (i.e., grassland, forests, and woodland), and (3) lower proportion of crop fields seemed to select for phenotypes enhancing dispersal ability. Flight propensity of male offspring was increased under food stress, indicating plastic responses to resource scarcity. In conclusion, our findings suggest that the compositional and configurational landscape heterogeneity, namely parameters indicative of agricultural intensification, select for enhanced dispersal in C. pamphilus. As higher investment in dispersal often comes at a cost to reproduction, such trait shifts may reduce population viability, which may have important implications for insect declines in agricultural landscapes.

