Risk of forest loss in Maryland's Green Infrastructure, based on 1997-2000 patterns of development

Spatial development risk model
(view larger map)


The objective of this study was to create a development risk model for Green Infrastructure (GI) forest land, based on conversion of Green Infrastructure forest to development between 1997 and 2000, and comparing to a wide range of predictive variables. We found that six counties (Cecil, Garrett, Howard, Montgomery, St. Mary's, and Washington) were significantly more likely to be developed than the statewide average. Five counties (Allegany, Dorchester, Somerset, Wicomico, and Worcester) were significantly less likely to be developed than the statewide average. Four of these five were the counties of the Lower Eastern Shore. Discounting regional landscape differences, sites with less protection, less wetland area, within a Priority Funding Area (PFA) or with existing or planned sewer service, with higher population growth between 1990 and 2000, closer to Washington DC and Baltimore, with higher market land value, closer to primary roads, and closer to other development were more likely to be developed. Of these, distance to previous development, market land value, population growth, and percent area of wetlands were significant regressors in a multivariate model. This model correctly classified independent validation data 76.3% of the time. We applied this model spatially to each 0.3 acre grid cell within unprotected GI forest, and used it to display relative risk of development there. Averaged within parcels at least 20 acres in size, areas with higher mean cell scores are more likely to be developed than those with lower scores, given the limitations of the model and source data. The model does not attempt to predict risk of development outside of GI forest. In these areas, a 2001 risk model can be used, at least until a better statewide model is developed. The most significant model limitation is the time period considered, 1997 to 2000. The model should be tested against newer land use data as it becomes available. Although it was 76% accurate in predicting trends between 1997 and 2000, changing economic conditions, real estate preferences, zoning, permit requirements, and other factors may make the model less accurate the further in time one projects. Use of development modeling, including the model described in this paper, can help planners focus on protecting those areas most at risk, as well as creating and implementing policies that minimize the harmful effects of sprawl development.

Download the Document (354 KB PDF file)

( Note: Documents are in PDF format and require Adobe Acrobat Reader which can be downloaded free from http://www.adobe.com)

Return to Green Infrastructure main page
DNR Home

2004 Maryland Department of Natural Resources

DNR logo