COS 16-8
Patterns of land-cover change in New Hampshire: Geographic drivers of logging and clearcuts in recent years
Regional-scale ecosystem services are influenced by the geographic distribution of land-cover and land use types. Here, we analyze current trends in land use change as a starting point for modeling future land use. We used regression tree analysis (RTA) to characterize spatial patterns of anthropogenic land cover change (development and logging) for the state of New Hampshire. As input for our RTAs, we selected variables suitable for future dynamic modeling because they were either unlikely to change (elevation, slope, whether land is classified as prime farmland, soil drainage classification, flood zone classification, and distances from water, conserved land, interstates, and cities of various sizes) or variables that can be directly derived from dynamic land use change models (e.g., land cover categories, forest type, and distance from developed land). All input variables as well as land cover change data were obtained from publicly available datasets (NOAA C-CAP, USGS Digital Elevation Models, SSURGO, flood insurance maps, U.S. Census data, New Hampshire Land Cover Assessment, and the NH Public Roads layer from the GRANIT database), and RTAs were performed on the likelihood of development or logging as a function of these input variables for randomly distributed points within New Hampshire.
Results/Conclusions
According to our analysis, variables spatially correlated with development probabilities included: elevation, distance from previously developed land, current land cover classes, distance from interstates, and soil drainage classification. Of these, elevation was the most important driver, with more development in areas below 90 meters, reflecting more intense development in the low-lying river-valleys of the southern portion of the state, and in a region closer to Boston. Higher probabilities of development also existed for land within 70 meters of already developed land and non-forest land (grassland or shrubland). Variables correlated with logging included distance from larger municipalities, forest type, distance from water, soil hydrologic group, and percent slope. The most important driver for the spatial distribution of logging was distance from the two largest cities, reflecting heavy logging in the northernmost and least populated portion of the state. The RTA for logging also quantified a higher probability of clearcut logging for oak/beech and other hardwood forests compared to evergreen and mixed forests, and for terrain with slope less than 6.5%, providing quantitative estimates of the geographic distribution of logging. These results provide a baseline for alternative development scenarios.