Skip to Content

Impervious Surface Mapping of NS

Example of Nova Scotia Impervious Surface data on a map

Impervious Surface Mapping AGRG March 2014 Report (pdf)

Impervious Surface Mapping of NS Poster (pdf - right click to save)

Download Data: Raw Nova Scotia Impervious Surface Raster (tif - compressed)

Description

Many different datasets were utilized in the production of this map which represents the percentage of impervious surface area (%ISA). High resolution imagery (1 metre) were classified into developed or non-developed areas. A 29 m x 29 m focal sum kernal was applied to the 1 m classified image to produce an image that represents the count of impervious cells per 29 m x 29 m pixel window. This focal sum image was resampled to represent a training area for the Landsat 30 m imagery. Training areas, a 30 m road raster and visible, near-infrared and mid-infrared Landsat bands were prepared and input into the Cubist classification and regression tree (CART) software. Cubist determines the rules which best describe the impervious percentages from the training data, roads and Landsat bands. These rules are translated from Cubist into ERDAS IMAGINE for classification of the entire Landsat scene. The classification is inspected and modified by removing pixel that represent agricultural lands, clear cuts, bare soil, areas which have a similar spectral signatures as impervious areas in the Landsat scene. The final classification is a map of the percentage of impervious surface for each 30 metre Landsat pixel. Eight Landsat scenes were classifed for %ISA and merged to produce this province wide map. The methods are decribed on the top left and the details of the large map and insets are described below the methods. Many different datasets were utilized in the production of this map which represents the percentage of impervious surface area (%ISA). High resolution imagery (1 metre) were classified into developed or non-developed areas. A 29 m x 29 m focal sum kernal was applied to the 1 m classified image to produce an image that represents the count of impervious cells per 29 m x 29 m pixel window. This focal sum image was resampled to represent a training area for the Landsat 30 m imagery. Training areas, a 30 m road raster and visible, near-infrared and mid-infrared Landsat bands were prepared and input into the Cubist classification and regression tree (CART) software. Cubist determines the rules which best describe the impervious percentages from the training data, roads and Landsat bands. These rules are translated from Cubist into ERDAS IMAGINE for classification of the entire Landsat scene. The classification is inspected and modified by removing pixel that represent agricultural lands, clear cuts, bare soil, areas which have a similar spectral signatures as impervious areas in the Landsat scene. The final classification is a map of the percentage of impervious surface for each 30 metre Landsat pixel. Eight Landsat scenes were classifed for %ISA and merged to produce this province wide map. The methods are decribed on the top left and the details of the large map and insets are described below the methods.

Conditions of Download: Acknowledgement & Reference

You are free to use these data for your research and work. However if graphics are generated using these data for reports or presentations, we request that the source of the information be acknowledged as “Impervious Surface layer generated by the Applied Geomatics Research Group, Nova Scotia Community College”. Also, please use the reference below when these data are used in a report or for any analysis. Thank you.

Citation:

Webster, T. MacDonald, C. Mouland, C. 2014. Impervious Surface Mapping for Nova Scotia. Study conducted for Environment Canada.

Tagged in: