Innovation Brief: Interview with the team behind 'StormSense Rises to the Challenge: Using Voice and Data Science to Strengthen Community Preparedness' and Upcoming Virtual Event

Featuring Dr. J. Derek Loftis, Assistant Research Scientist at College of William & Mary and Sridhar Katragadda, Systems Analyst at the City of Virginia Beach, VA

ARTICLE | Sep 6, 2018

Looking ahead to the live virtual session on September 18!

One of the great successes coming out of StormSense was the regional collaboration.  There’s been a lot of discussion around ‘smart cities’, but perhaps it’s more appropriate to call them ‘smart regions’, since collaboration is so critical. Could you share how this partnership was achieved and what’s next for the program?

The modern Smart City of today is one in which innovation is not simply dreamed of, but is proactively employed. StormSense brings innovation to the smart cities of Hampton Roads, VA, in six ways: 

  1. Testing of multiple sensor technologies, their communication protocols, energy requirements, and accuracy, 
  2. Leveraging a high-resolution hydrodynamic model results for street-level flood forecasting, 
  3. Integration architecture for disparate data sources in near real-time, 
  4. Leveraging multi-cloud environments using Amazon Web Services (AWS) for data, AI, and voice-response query integration, 
  5. Data presentation and big data support using AWS, Microsoft Cloud and Power BI interactive visualization, and 
  6. Spatial mapping of inundation vulnerability scenarios for the short and long-term

StormSense is a collaborative effort between the Virginia Institute of Marine Science (VIMS) and a legion of city governments including; Newport News, Hampton, Norfolk, Virginia Beach, Chesapeake, Portsmouth, Williamsburg, and York County in the Greater Hampton Roads Region of Tidewater Virginia. This effort started as a small collaborative footprint with passionate engagement by cities driven to be innovative and proactive by engaging in shark-tank like challenge, leveraging the latest inundation modeling research from VIMS to strive to bring awareness to flooding vulnerability. StormSense will provide long-term sensor data for modeling and provide new insights into the flow patterns in different rainfall and hurricane events to the decision makers and citizens. Thus, StormSense endeavors to produce an iteratively predictive tool in development for emergency managers to continually visualize flooding from storm surge, rain, and tides up to 36 hours in advance of a flood event. Since decisions made by one locality related to flood mitigation tend to impact those on the other side of the water, StormSense is a functional collaboration among resilient communities, so the cooperation wasn’t particularly difficult to foster.

Our goal is to continue to develop the platform by adding 16 more new sensors by the year’s end, and publicly release 36-hr tidal forecasts for existing sensors in the program, and add a function to the StormSense Alexa App to use a smart device’s location to tell people what the water level is near them. Chatbots will soon provide real-time information through several access points. In the future, StormSense will use AWS DeepLens camera to delineate flood boundaries, and machine learning techniques to refine prediction from 25-year reanalysis data from NOAA’s National Water Model data history. This process may provide data to enhance the predictability of such models.

The StormSense program has so many great aspects, but one that stands out is its scalability and affordability. In your 2018 case study, it was noted that the sensors went from ‘$3k/ vs. 38k/sensor’. Could you detail how you achieved the unit price drop and do you foresee further price reductions?

The objective of StormSense is to enhance the capability of communities to prepare and respond to the disastrous impacts of sea level rise and coastal flooding in ways that are replicablescalablemeasurable, and make a comparable difference (see figure below). In pursuit of this, the project's mantra is to advance the field of emergency preparedness by advancing research to ultimately help better predict flooding resulting from storm surge, rain, and tides. Thus, cost differences between sensors affect its replicability and scalability in other coastal environments around the world. In many cases, these are the result of differences in hardware, communication, and software as different sensors for different applications. The cost of water level sensors and their operation varies depending upon the user’s need for accuracy.

The National Oceanic and Atmospheric Administration (NOAA) has the highest data standards for their water level data, and their sensors can cost between ~$30-50k/station depending upon whether it is just a water level sensor or a full weather and/or water quality station. Their water level sensors use top-of-the-line Ka-band radar sensors and have a requirement of measuring water level of within 1 mm during typical tidal conditions. Their sensors communicate their data through the GOES satellite network. The US Geological Survey (USGS) use similar style water level gauges, but use slightly less expensive hardware (and are typically accurate within a couple mm) and communicate their data through the Iridium Satellite network for roughly ~$38k/sensor. These costs are variable depending upon whether a full weather station is attached to that site ~$38k, or if there is just a water level sensor deployed (~$30k), and maintenance costs vary based upon location of each sensor and have a variable baseline of ~$5000/sensor/year.

StormSense uses two types of IoT sensors in its network: 1) Sensors from Valarm, based in Los Angeles, California. These communicate through cellular broadband (and can be customized by desired carrier average cost ~$8/sensor/mo) to Valarm’s tools cloud environment (costs ~$7/sensor/mo). These permanent water level sensors range from $3-4k/sensor depending upon the sensor type (Ultrasonic Sonar (cheaper - $550-780/each) to Ka-Band Radar (more expensive - ~$1200/each)) and size of a proportionally-sized solar panel, adjusting for power output needs (radar uses more power). Most of these sensors are mounted to city-maintained bridges over open waterways. An article StormSense recently published in the Journal of the Marine Technology Society in April 2018 reported accuracy on 4 ultrasonic StormSense sensors that were temporarily mounted to the same bridges as USGS sensors to be within ~1.5 cm over 4 mo.

2) Sensors from Green Stream based in Norfolk, VA. These sensors are even cheaper at approximately ~$400/sensor and can pretty much be installed anywhere. The city of Norfolk has installed them over open waterways and frequently flooded intersections, while Virginia Beach uses them near retention ponds with past history of storm water drainage concerns during heavy rainfall events to closely monitor developing weather events in real time. These sensors can communicate via Long Range WiFi of up to 1 mi. to a gateway device (~$1200/gateway), but in urban environments, transmission corridors can become problematic with buildings potentially blocking gateways if they are not positioned on top of a tall building. These can also be configured to communicate via cellular broadband but have the same /device monthly costs and the sensors from Valarm noted above. Since each integrated sensor device from Green Stream costs less than the actual sensor components used in the Valarm setup, it is worth noting that its accuracy is less ~4.5 cm.

Cloud services and data architecture. IoT and sensors. Data presentation and visualization. Climate science. For those aspiring transformational leaders in local government and smart cities, please name three skills (and briefly why) you see these are needed in the public sphere & innovation.

StormSense is transformational in its innovative approach to making water level data more accessible to modern citizens to aid in raising awareness and informed decision-making, while broadcasting the flood observations and tidal forecasts up to 36-hours in advance of impending inundation events. This enables people to make short-term and long term personal decisions for themselves. This can be little things, from route planning to avoid inundated intersections, to recognizing flood risk that may not be represented in today’s FEMA flood risk inundation maps and optionally obtaining a flood insurance policy even when it isn’t legally required outside of the floodplain. Meanwhile their communities use StormSense’s information to make planning decisions. Drawing from Dr. Loftis’ undergraduate leadership minor, from Christopher Newport University, James MacGregor Burns paints a picture of transformational leaders as those who functionally maintain a feedback loop with followers to cooperatively achieve goals in business and in our communities:

  • [Engage Your Community]. This has been accomplished with StormSense by engaging the community to help proactively map an inundation event through the Sea Level Rise Mobile App, which helped calibrate the new StormSense water level sensors during the king tide in November 2017, while simultaneously educating citizens regarding how to access the new sensor and flood model data. This helps from a leadership perspective by intrinsically investing the community in a collaborative resilience solution by leveraging modern technology in monitoring an ongoing problem.
  • [Be Accessible]. Find multiple ways for you or your resource to be available for people to use and be responsive to feedback. StormSense’s water level data (among other storm-relevant data such as wind speed or rainfall) can be voice queried via the "Storm Sense" Alexa skill from any smart phone or compatible smart home device via the Reverb App and Amazon Alexa, respectively. Chatbots and cloud-based call center such as Amazon Connect will soon provide real-time information through several access points. This functionality is powered by Amazon Web Services (AWS) in collaboration with data scientists in Virginia Beach, VA. This mobile functionality, in addition to being able to provide real-time water levels on the dynamic web map at, 36-hr water level forecasts provide time series forecasts available on VIMS' Tidewatch site to provide unique and immersive new ways for anyone from key decision-makers to citizens to access flood information and become more informed about their own flood vulnerability. By publicly sharing the sensors' and predictive model's web services, decision-making can be informed in near-real time, with the goal of informing adaptive route planning services, such as Google Maps and Waze, (which is already being partially implemented in Norfolk, VA).
  • [Meaningful Data Integration]. To remain transformative, StormSense aims to integrate data from different sensors and sources, automate flood control structures and tide gates alongside guiding traffic patterns to keep citizens of the smart cities of Hampton Roads out of the flood path, as the best decision is often the one that you don't have to make (most of the time). As one might imagine, this has already become an increasingly ambitious goal as sea levels continue to rise, but one that is not impossible as we hope to bring our solution to - and learn from - other coastal communities facing similar inundation issues.

And finally, considering your experience now with voice integrations, IoT, and sensors, could you speculate about other areas or departments that would benefit from this technology?

Coming from a background of marine research, I saw many ways that sensor systems could be employed to validate my flood modeling research. Sridhar in Virginia Beach has a science background and was the one who initially pointed me to applications in riverine environments where IoT sensors had been used to calculate flood impacts and predictions based upon up-river water level observations on tributaries of the Mississippi River. These sensors had been tested in Manila in the Philippines and widely deployed in their coastal environment, but we are the first to use them as a mesh network for interpolating for geospatial water level forecasts.

Since StormSense itself is a technological approach adapted to use level sensors to remotely sense for still fluids in chemical vats and water in storage reservoirs to observe open environment conditions in riverine and coastal flooding applications, there are presumably many possibilities.

The sensor technology is being used in self-driving cars already to determine how close another vehicle or moving object is and adjust speed accordingly. The automated flood alerts that are set up through valarm and through the cities’ own systems can be used for other dangers if integrated through other sensor systems. These could be related to air quality, crime, open web cameras, etc. The Alexa skill could be tailored to address other key functions beyond just those related to flooding. Currently StormSense uses water levels, rainfall, wind speed, air pressure, and stream flow data, but others from similar systems could benefit from these technologies.

Have questions? Contact Ryan Spillers, Program Coordinator at Sign up for the live session today  StormSense Rises to the Challenge: Using Voice and Data Science to Strengthen Community Preparedness!

You may also be interested in