Reconstructing long-term vegetation cover along streams has become a popular practice for ecological restoration in fragmented agricultural landscapes. To maximize ecological restoration benefits with the least cost, this paper developed a GIS- based model framework for the geographical design of the location and width of riparian buffers to be reconstructed with long-term vegetation cover. The framework integrated three models： the 0-1 mathematical programming model, an economical model and an ecological model. It was tested in an agricultural landscape with a drainage system consisting of 92 stream segments with cropland banks. Each segment had five restoration options in terms of buffer width： 0, 5, 10, 15 and 20 m, thus making a total of 460 decision units in the landscape. For each unit, the economical model estimated the annual rental payment to the farmer who owned the targeted cropland margin along a stream to be converted into a vegetation buffer, whereas the ecological model predicted its corresponding habitat benefits at the landscape level. The 0-1 programming model minimized the total payments, subject to the designed habitat improvement and proposed acreages of riparian cropland set-aside annually. The model identified a cost-effective combination of riparian buffers from 460 units. The results showed that achieving the habitat improvement goal at the least cost required a spatial configuration of vegetation buffers with variable width. This finding suggests that the size and location of buffers directly affect restoration costs and environmental benefits. Thus, the best management practices of agricultural landscape restoration can only be achieved by a specific geographical design with the model framework.
