DEVELOPMENT AND APPLICATION OF A CENSUS-BASED REGIONAL RESIDENTIAL GROWTH MODEL FOR BIODIVERSITY RISK ASSESSMENT Stephen D. Smith, Ph.D. Cornell University 2018 The USGS National GAP Program is a biodiversity mapping program implemented at the state level via the Cooperative Fish & Wildlife Research Units (CFWRU). The New York CFWRU completed NY-GAP analysis in 2001, providing, for the first time, a statewide vertebrate species distribution dataset. A subsequent regional project, HR-GAP, documented 75% of the State’s terrestrial vertebrates as having a significant portion of their range within the Hudson River Valley region (HR). The presence of high biodiversity in conjunction with development pressures was the impetus for efforts to develop a regional residential growth prediction model, based on Block Group (BG) level Census data, with the purpose of identifying biodiversity regions at risk from future residential development. Initial efforts resulted in a regression model which predicted 77 of the 2,212 total BG in the study area to be prime candidates for a substantial percentage of the predicted new residential growth. These BGs, classified as intensive growth areas (IGA), were intersected with biodiversity data to quantify that 53% of the State’s vertebrate species are within and intensive growth BG, as well as 41% of the threatened, endangered, or special concern (TES) species. Additional model development provided a slight improvement to the predictability of the model while using only digitally available regional data. The second model explained 38% of the variance associated with the identification of IGAs and identified the top 5% of BGs showing substantial increases in residential housing units over the last decade. Of the BGs predicted to be areas of fast growth, 53% and 41% were IGAs as computed from 2000 and 2010 Census data, respectively. Of the IGAs predicted for 2000 and 2010, 16% and 8%, respectively, were also species-rich BGs. A third modeling effort was undertaken to improve upon the earlier residential housing prediction models based on regression analysis of Census-based BG data and physiographic variables aggregated to the BG level geography. It was hypothesized that increasing the spatial resolution through dasymetric mapping of the BG data would further improve model results and subsequently the identification of biodiversity areas at risk. The model results from the dasymetric mapping did not reveal significant improvement to earlier model results. Investigations of various alternative Census-based datasets yielded similar results. These efforts to model residential growth at the landscape scale support the hypothesis that the spatial distribution of residential housing growth can be modeled using Census Block Group (BG) level data and other publicly available data to provide a coarse filter for the identification of biodiversity areas at risk from projected residential growth.