Loading…
Fall NEARC 2018 has ended
Welcome to the interactive web schedule for the 2018 Fall NEARC Conference! To return to the NEARC website, go to: https://www.northeastarc.org/fall-nearc.html

Filter the schedule to view only the presentations with slides available.

View and download a PDF of the final program.

Tuesday, October 30 • 2:30pm - 3:00pm
Remote Sensing Track. Estimating Percent Impervious Cover from Landsat-based Land Cover: An Evaluation of a Simple and Transferable Regression Model

Sign up or log in to save this to your schedule and see who's attending!

AUTHORS: Jason R. Parent, Qian Lei – University of Connecticut

ABSTRACT: Percent impervious cover (PIC) is often estimated from moderate-resolution satellite data which is known to overestimate PIC in urban areas and underestimate PIC in rural areas. Regression-based models (e.g. ISAT, ETIS) have been developed to calibrate Landsat-based PIC estimates to improve accuracy. However, it is unknown how these models perform if they are used outside of the geographic area for which the models were developed or if the size of the analysis units (e.g. watershed) affects model performance. Furthermore, these models tend to be applicable only for specific land cover datasets and may require ancillary data such as population estimates. This study evaluated the robustness of a simple regression model, based solely on Landsat-based impervious land cover, to estimate PIC for different geographic areas, land cover datasets, and analysis units.

We tested the model for analysis units ranging in size from 2 to 100+ ha for four locations in Connecticut, Massachusetts, and Ohio. The model was developed in southwestern CT and validated in the three other locations. Model RMSE values ranged from 1.5% to 10.0% with the performance improving as the analysis unit size increased. The model had slightly lower performance (0.0 to 2.7% higher RMSE) when applied outside the area in which it was developed. Overall, this study showed that a simple PIC estimation model, based only on the impervious cover classes of Landsat-based land cover datasets, can be effective for a variety of analysis unit sizes and for locations outside of the model calibration areas.


Tuesday October 30, 2018 2:30pm - 3:00pm
Saratoga 1/2

Attendees (2)