Felix May, Itamar Giladi, Michael Ristow, Yaron Ziv, Florian Jeltsch

• Understanding the regional dynamics of plant communities is crucial for predicting the response of plant diversity to habitat fragmentation. However, for fragmented landscapes the importance of regional processes, such as seed dispersal among isolated habitat patches, has been controversially debated. Due to the stochasticity and rarity of among-patch dispersal and colonization events, we still lack a quantitative understanding of the consequences of these processes at the landscape-scale. In this study, we used extensive field data from a fragmented, semi-arid landscape in Israel to parameterize a multi-species incidence-function model. This model simulates species occupancy pattern based on patch areas and habitat configuration and explicitly considers the locations and the shapes of habitat patches for the derivation of patch connectivity. We implemented an approximate Bayesian computation approach for parameter inference and uncertainty assessment. We tested which of the three types of regional dynamics - the metacommunity, theUnderstanding the regional dynamics of plant communities is crucial for predicting the response of plant diversity to habitat fragmentation. However, for fragmented landscapes the importance of regional processes, such as seed dispersal among isolated habitat patches, has been controversially debated. Due to the stochasticity and rarity of among-patch dispersal and colonization events, we still lack a quantitative understanding of the consequences of these processes at the landscape-scale. In this study, we used extensive field data from a fragmented, semi-arid landscape in Israel to parameterize a multi-species incidence-function model. This model simulates species occupancy pattern based on patch areas and habitat configuration and explicitly considers the locations and the shapes of habitat patches for the derivation of patch connectivity. We implemented an approximate Bayesian computation approach for parameter inference and uncertainty assessment. We tested which of the three types of regional dynamics - the metacommunity, the mainland-island, or the island communities type - best represents the community dynamics in the study area and applied the simulation model to estimate the extinction debt in the investigated landscape. We found that the regional dynamics in the patch-matrix study landscape is best represented as a system of highly isolated island' communities with low rates of propagule exchange among habitat patches and consequently low colonization rates in local communities. Accordingly, the extinction rates in the local communities are the main drivers of community dynamics. Our findings indicate that the landscape carries a significant extinction debt and in model projections 33-60% of all species went extinct within 1000 yr. Our study demonstrates that the combination of dynamic simulation models with field data provides a promising approach for understanding regional community dynamics and for projecting community responses to habitat fragmentation. The approach bears the potential for efficient tests of conservation activities aimed at mitigating future losses of biodiversity.…