Understanding how metapopulations persist in dynamic working landscapes requires assessing the behaviors of key actors that change patches as well as intrinsic factors driving turnover. Coupled human and natural systems (CHANS) research uses a multidisciplinary approach to identify the key actors, processes, and feedbacks that drive metapopulation and landscape dynamics. We describe a framework for modeling metapopulations in CHANS that integrates ecological and social data by coupling stochastic patch occupancy models of metapopulation dynamics with agentâbased models of landâuse change. We then apply this framework to metapopulations of the threatened black rail (Laterallus jamaicensis) and widespread Virginia rail (Rallus limicola) that inhabit patchy, irrigationâfed wetlands in the rangelands of the California Sierra Nevada foothills. We collected data from five diverse sources (rail occupancy surveys, landâuse change mapping, a survey of landowner decision making, climate and reservoir databases, and mosquito trapping and West Nile virus testing) and integrated them into an agentâbased stochastic patch occupancy model. We used the model to (1) quantify the drivers of metapopulation dynamics, and the potential interactions and feedbacks among them; (2) test predictions of the behavior of metapopulations in dynamic working landscapes; and (3) evaluate the impact of three policy options on metapopulation persistence (irrigation district water cutbacks during drought, incentives for landowners to create wetlands, and incentives for landowners to protect wetlands). Complex metapopulation dynamics emerged when landscapes functioned as CHANS, highlighting the importance of integrating human activities and other ecological processes into metapopulation models. Rail metapopulations were strongly topâdown regulated by precipitation, and the black rail's decadeâlong decline was caused by the combination of West Nile virus and drought. Theoretical predictions of the two metapopulationsâ responses to dynamic landscapes and incentive programs were complicated by heterogeneity in patch quality and CHANS couplings, respectively. Irrigation cutbacks during drought posed a serious extinction risk that neither incentive policy effectively ameliorated.