Halifax Water Uses ENVI and IDL to Overhaul Billing Structure
Anonym
In 2007, Halifax Regional Municipality (HRM) transferred wastewater and stormwater assets to Halifax Water, establishing the first regulated and integrated water, wastewater and stormwater utility in Canada. Even prior to the transfer, HRM realized that the methodology for stormwater revenues had to be reviewed. Halifax Water embarked on a formal cost of service study to change the way stormwater charges were historically collected in an effort to make it more equitable to their customers, and more cost-effective to administer. The first thing this study concluded was that a new rate structure needed to be established to separate costs for wastewater and stormwater services.
This was different from the current billing structure that allocated revenue to stormwater as a proportion of wastewater revenue that was billed based on volumetric water consumption. An added benefit of separate stormwater charges was to hopefully influence customers to reduce the amount of stormwater that was getting into the wastewater system. This would be advantageous because inflow and infiltration accounted for a significant amount of the wastewater treatment capacity.

With the understanding that a larger amount of impervious surface on a property increases the amount of stormwater runoff, Halifax Water decided to use a pilot program to determine the feasibility of using remote sensing technology to map impervious surfaces. The goal was to demonstrate a cost-effective approach to assess stormwater usage without relying on costly and labor-intensive manual surveying. The use of remote sensing not only offered the potential of lower cost, but also improved accuracy, and provided regular monitoring on a regional scale. Basing the billing structure on impervious surface also offered a higher correlation to cost causation than a structure based on water consumption or a per-lot charge. Today Halifax Water has 98,000 customers which include 15,000 stormwater only customers.
PILOT PROJECT FOR IMPERVIOUS SURFACE MAPPING
The pilot project was launched with three different software providers to not only determine the feasibility of the project, but to evaluate the strengths and weaknesses of different software to determine which would be best equipped to complete the larger scope of work. NV5 (formerly Exelis VIS) completed the pilot study for Halifax Water using pansharpened 8-band multispectral Worldview-2 imagery. NV5 developed semi-automated methods to map impervious surfaces of the three small study areas identified by Halifax Water using its commercial ENVI and IDL software.
The pilot study not only demonstrated the effectiveness of the approach, but the products NV5 created for Halifax water were 96% accurate. Using ENVI and IDL, NV5 demonstrated that the project could be completed in a cost-effective manner without compromising accuracy or detail, and was therefore awarded the contract to produce GIS-ready impervious surface mapping products for the entire stormwater boundary established by the HRM.

GETTING DOWN TO BUSINESS
The mapping approach that NV5 employed for the project was based on the methods developed during the pilot study, with a few exceptions. During the pilot study it became apparent that that the map precision of the Worldview-2 imagery was not sufficient when compared to base imagery already contained in Halifax Water’s GIS. Even after Digital Globe provided samples of their highest precision orthorectified map product that didn’t require customer-provided ground control, Halifax Water accuracy requirements still were not met. NV5 then ordered imagery in a format that was suitable for further georectification: Level 2A “ortho-ready” format. Because the ultimate use of the data was for spectral analysis, NV5 also requested the imagery be produced using only nearest neighbor resampling – a further effort to retain as much radiometric precision as possible.
With the Worldview-2 data calibrated, orthorectified, and pansharpened, the enhanced resolution multispectral data were then used to identify impervious surfaces through a combination of multiple processes. A vegetation index was used to take advantage of the strong, and fairly unique signature of vegetation in the 8-band multispectral data. Because it’s rare to see a strong vegetation signature on man-made impervious surfaces, this index image is used as an initial, rough estimate of the pervious surfaces. Other processes including vegetation masking, soil mapping, unsupervised (auto-clustering), and supervised classification were also employed. N-Dimensional spectral angle metrics were used for measures of spectral similarity, and an anomaly detection algorithm was used to reduce false positives caused by spectral artifacts.
TRACKING THE SOURCE OF ERRORS
Visual interpretations of the mapping results (without ground truth) showed that the largest errors in classification occurred for soils that occasionally get categorized as an impervious surface. In some cases these soils have unusually bright signatures, a feature that is characteristic of urban materials, and erroneously get filtered out in post-classification processing. This may be more common for altered soils, such as those mixed with sand, or those compacted due to construction activity. To help minimize this source of error, NV5 made adjustments to the post-processing to ensure that flagged spectra were secondarily assessed for the likelihood of being a soil ground cover type.
In the final analysis, selected parcels were ground-truthed with an actual survey of impervious area, the variance between the survey results and satellite imagery was 0.16%, indicating 99.84% accuracy for the project.
SOLUTION ACHIEVED
Based on the impervious surface mapping results, Halifax Water determined there should be two classes of rates: one for residential customers; one for Commercial, Industrial, and Institutional customers (ICI Customers). This decision was based on analysis that concluded it was possible to develop a residential average. Residential lot sizes differ but the ratio of impervious area to pervious area does not change much. ICI customers, on the other hand, have much more variances in lot size, and nature of the property use. As it was not possible to develop an average impervious area, it was decided to charge based on the impervious area of the actual lot.
Using impervious area as a billing determinant has provided a transparent and objective system because it is measurable and does not change a lot over time. There are some on-going challenges with respect to what gets identified as impervious area, however NV5 and Halifax Water are continuing to work together to improve the system.