Title

Determining Directly-connected Impervious Areas in Residential Land Uses

Date of Completion

January 2012

Keywords

Land Use Planning|Environmental Sciences|Remote Sensing|Urban and Regional Planning

Degree

Ph.D.

Abstract

Urban growth, suburban sprawl, and land cover changes have occurred rapidly in the United States and other large countries in recent years. The associated site development features and transportation infrastructure typically convert pervious land cover types such as forest to anthropogenic land cover called impervious cover (IC)—also called total impervious area (TIA). Many environmental problems have been associated with the conversion of the landscape to IC including alterations to the local hydrologic cycle, changes in receiving water channel morphology, degraded water quality, and impacts to the stream ecology. Planners have used TIA to evaluate the environmental impacts to communities with some degree of success but this metric is usually applied using coarse data and doesn't link the impervious cover and receiving water impacts with any process. ^ This research explores an alternative metric for residential land-use areas that may have more predictive value and more inherent process information than TIA. This metric called Directly-connected Impervious Area (DCIA) evaluates IC area by their hydrologic connectivity to local receiving waters and is an innovative method in IC assessment research. The emphasis of this project was initially to create an inductive geospatial model to predict DCIA in residential areas at the parcel scale but a hydrologic response unit conception of DCIA worked better. Data for the model creation and statistical analysis were derived from a GIS and field work, for site development, slope, drainage, and land cover variables, and were collected at the parcel scale for four field sites in Connecticut. ^ Data collected in this research showed a bimodal distribution for the aggregation of within parcel DCIA compared to within parcel IC. One peak was at the middle of a normal distribution while the other peak was approximately 0 sq ft (0 m2) and was caused by parcels that were disconnected. The bimodal distribution is non-parametric which eliminates the use of many statistical methods and required the use of subsetting and machine learning analytic tools. In addition, IC% and DCIA% were not related at the parcel scale. ^ This research resulted in the generation of a statistically significant equation to predict parcel level DCIA, and is the first research project known to have done so. The important variables from this equation are total IC within the parcel, the absence/presence of driveway drains, and slope direction of the driveway. The three-variable model intended for parcels with street inlets only, predicted 65% of variance. In addition, a rapid assessment protocol is described for determining within parcel DCIA in residential land use areas. The connectedness of parcel level IC areas in neighborhoods with low housing densities are impacted by the edge containment, or lack of it, for roads adjacent to the parcel. ^

COinS