College of Agriculture and Natural Resources
Center for Land Use Education and Research


ETIS calculates percent of each land cover class per selected unit of analysis (e.g. Census tract, block, watershed, etc.), and then uses these values in linear regression along with known population density (PopDen) data and built-in set of coefficients to calculate the percentage imperviousness (%IS):                          

%IS = b1+b2 PopDen+Sum(bi %Ai)

where b1 is the IS-axis intercept, b2 is the coefficient for population density expressed in persons per square mile, and bi are the coefficients for the percentage of the land cover category area (%Ai) within the unit area.

Two regression models were created in the JMP Statistical Discover Software 5.0.1 (SAS Institute, Inc. 2002) for developing two sets of impervious coefficients, one for use on the NLCD land cover and the other for use on the CCL land cover, to be used with ETIS. The first model relied on NLCD 2001 land cover as well as Census tract polygons with associated population values. Stepwise regression analysis was performed on 15 land cover classes from NLCD and 13 from CCL. The percentage of each land cover class with a given census tract, along with its population density, formed the independent variable set. Actual percent imperviousness, the dependent variable, was calculated from the Union of the planimetric data layer with the tract boundaries, and the amount of each land cover class present in the study area was derived using the Tabulate Areas command from the ArcToolbox Spatial Analyst Tool for each tract. Planimetric-based percent imperviousness, population density and area values for each land cover class were exported into the JMP input table. Values of the percent imperviousness per tract were selected as role (dependent) variables population density and area values are added as Model Effects (independent variables).

Available Coefficients Sets

There are several Coefficients Sets provided with the ETIS Toolbox: some for Connecticut Changing Landscape (CCL) and National Land Cover Data (NLCD) 2001 datasets. These datasets were calculated using calibration planimetric data from three separate states (CT, MA, and NY) as well as the combinations of them.


NLCD 2001: CT, MA, NY, CT & NY, CT & MA & NY