Species distribution models (SDMs) are an effective tool in ecological restoration, but the application of SDMs in wetland restoration is rare owing to limited survey data. Based on field survey of the iconic, widespread wetland species Phragmites australis (Poaceae), we built maximum entropy (MaxEnt) models to study the current and future delimitation of wetlands in the Songnen Plain in Northeast China. In building the model, we evaluated the selection of environmental predictors and data resolutions. The evaluation showed the addition of soil and topohydrological (e.g. slope, Topographic Wetness Index, water table depth) variables significantly improved model performance in comparison with climate-only models, and the best model performance was under spatial resolution of 250 m. This study found that wetlands of the Songnen Plain are facing threats from climate change and anthropogenic expansion. The selected models predicted approximately a 15% net loss in suitable habitats, as well as expansion trends towards more humid and higher altitude areas in the future. Based on the modeling results, we delimited 1.51 x 10(3) km(2) and 1.48 x 10(3) km(2) of potential wetland restoration areas under RCP 4.5 and RCP 8.5 scenarios, respectively. Our study highlighted the importance of optimizing the selection of variables and spatial resolution in the application of SDMs in wetlands, and illuminated priority areas of the Songnen Plain for consideration in future wetland restoration efforts.

Species distribution models (SDMs) are an effective tool in ecological restoration, but the application of SDMs in wetland restoration is rare owing to limited survey data. Based on field survey of the iconic, widespread wetland species Phragmites australis (Poaceae), we built maximum entropy (MaxEnt) models to study the current and future delimitation of wetlands in the Songnen Plain in Northeast China. In building the model, we evaluated the selection of environmental predictors and data resolutions. The evaluation showed the addition of soil and topohydrological (e.g. slope, Topographic Wetness Index, water table depth) variables significantly improved model performance in comparison with climate-only models, and the best model performance was under spatial resolution of 250 m. This study found that wetlands of the Songnen Plain are facing threats from climate change and anthropogenic expansion. The selected models predicted approximately a 15% net loss in suitable habitats, as well as expansion trends towards more humid and higher altitude areas in the future. Based on the modeling results, we delimited 1.51 x 10(3) km(2) and 1.48 x 10(3) km(2) of potential wetland restoration areas under RCP 4.5 and RCP 8.5 scenarios, respectively. Our study highlighted the importance of optimizing the selection of variables and spatial resolution in the application of SDMs in wetlands, and illuminated priority areas of the Songnen Plain for consideration in future wetland restoration efforts.