Page 44 - GCN, October/November 2018

40
GCN OCTOBER/NOVEMBER 2018 • GCN.COM
Public Sector Innovations
PROJECT: StormSense
Commonwealth Center for Recurrent Flooding Resiliency
Better warnings as Virginia waters rise
In 2016, the Virginia Institute of Marine Science launched the StormSense program to track water level rise in Virginia’s Hampton Roads area. Today, there is a network of 28 sensors in Norfolk, Virginia Beach and Newport News, made available with funding from the National Institute of Standards and Technology.
The sensors track flooding conditions and report back to VIMS, which has created an Alexa skills app that informs residents about water levels throughout the region. The information is also displayed via a cloud-based platform that incorporates Esri’s mapping and visualization tools.
“We continue to add functionality to the app to report the water levels for the StormSense gauges, [U.S. Geological Survey] gauges and now the National Oceanic and Atmospheric Administration gauges as well,” said Derek Loftis, program manager for StormSense. “We’ve been adding to our Amazon Alexa chatbot to answer inquiries.”
There is also an ongoing process with the local branch of the National Weather Service to generate alerts based on the gauge readings. Modeled on VIMS’ Tidewatch network, which predicts flooding 36 hours in advance, the alerts could help residents better prepare for flooding in the region.
In addition, Loftis said he hopes to incorporate data from 75 to 80 sensors maintained by a regional sanitation district by the end of the year.
“The data architecture is there to add the extra sensors, but there are not public [application programming interfaces] to link on an individual gauge basis,” he added. “We are extremely interested in those measuring devices so it is not just water-level sensors in the network.”
By using analytics and machine learning to assess real property values, Wake County, N.C., is removing much of the human subjectivity from the process, resulting in more accurate numbers and cost savings.
Machine learning software from SAS crunches data on the county’s 400,000 properties. That data includes more than 140 variables such as square footage, exterior finish, neighborhood and number of bathrooms. The software uses a sophisticated algorithm to turn around estimations in minutes.
“It’s important to note that the method we are utilizing to assess properties hasn’t changed,” said Marcus Kinrade, the county’s revenue director. All counties must still comply with the state’s General Statutes and use a uniform schedule of values, standards and rules.
“That being said,” he added, “there is
a lot of subjective analysis performed by our appraisers when applying the schedule of values, particularly in a market where property values are increasing rapidly. The SAS machine learning models only consider the data, and there is no emotion or human judgment involved. So as a tool providing
a check and balance to the work our appraisers are doing, it’s invaluable.”
The department has also used the software to review the boundaries of valuation control areas, or groups of similar properties. “The real benefit is we have a completely objective tool to either validate or possibly invalidate the historical appraisal techniques our office performs,” Kinrade said.
The software is saving the county money because appraisers can work from the office rather than in the field, and the department has had to hire fewer contractors because its employees are more productive.
PROJECT: Wake County Tax Assessment Model
Wake County, N.C., Revenue Department
More science, less art in
property assessments