Department of Natural Resources and the Environment, University of New Hampshire, 114 James Hall, Durham, NH, 03824, USA.
Carsey School of Public Policy, University of New Hampshire, Huddleston Hall, 73 Main Street, Durham, NH, 03824, USA.
Environ Manage. 2018 Dec;62(6):1089-1107. doi: 10.1007/s00267-018-1102-x. Epub 2018 Sep 26.
The Great Lakes region contains productive agricultural and forest lands, but it is also highly urbanized, with 32 of its 52 million residents living in nine large metropolitan areas. Urbanization of undeveloped areas may adversely affect the productivity of agricultural and forest lands, and the provision of ecosystem services. We combine demographic and remote sensing data to evaluate land cover change in the region using a two-phase statistical modeling approach that predicts the incidence and extent of land cover change for each of the region's 10,579 county subdivisions. Observed patterns are spatially uneven, and the probability of land cover change is influenced by current land use, human habitation, industry, and demographic change. Pseudo R values varied from 0.053 to 0.338 for the first-phase logistic models predicting the presence of land cover change; second-stage beta models predicting the rate of change were more reliable, with pseudo R ranging from 0.225 to 0.675. Overall, changes from agriculture or greenspace to development were much more predictable than changes from agriculture to greenspace or vice versa, and demographic variables were much more important in models predicting change to development. Although models successfully predicted the general location of land cover change, and models from before the Great Recession were useful for predicting the location but not the amount of change during the recession, fine-grained prediction remained challenging. Understanding where future changes are most probable can inform planning and policy-making, which may reduce the impact of development on resource production, environmental health, and ecosystem services.
五大湖地区拥有富饶的农田和林地，但城市化程度也很高，其 5200 万居民中有 3200 万人居住在 9 个大都市区。未开发地区的城市化可能会对农田和林地的生产力以及生态系统服务产生不利影响。我们结合人口统计和遥感数据，使用两阶段统计建模方法评估该地区的土地覆盖变化，该方法预测了该地区 10579 个县分区中每一个的土地覆盖变化的发生率和程度。观测到的模式在空间上不均匀，土地覆盖变化的概率受到当前土地利用、人类居住、工业和人口变化的影响。预测土地覆盖变化存在的第一阶段逻辑模型的伪 R 值从 0.053 到 0.338 不等；预测变化率的第二阶段 beta 模型更可靠，伪 R 值从 0.225 到 0.675 不等。总体而言，从农业或绿地向发展的转变比从农业向绿地或反之的转变更具可预测性，人口变量在预测向发展转变的模型中更为重要。虽然模型成功地预测了土地覆盖变化的大致位置，而且大衰退前的模型对于预测衰退期间变化的位置而非变化的数量很有用，但精细的预测仍然具有挑战性。了解未来变化最可能发生的地方可以为规划和决策提供信息，这可能会减少发展对资源生产、环境健康和生态系统服务的影响。
