FloodFinder

To resolve the lack of real time information during or after a flood, FloodFinder applies deep machine learning algorithms on the traffic camera feeds to calculate water surface area.

Early alert systems are an adaptive measure for climate change and an important component of disaster risk management strategies. To resolve this lack of real time data/information issue, we designed and implemented FloodFinder. It applies deep machine learning algorithms to the traffic and private camera(Ring Doorbell) feeds to calculate water surface area in each frame to determine the difference between puddles and water accumulations and provide a map layer that can be utilized by citizens and emergency personnel during and after flash flooding.

San Jose city has 1200+ city-owned traffic cameras distributed across the downtown area which covers 46% of the area. By leveraging the city's owner volunteer program to share the homeowner's external camera feed to the Police department for quick response, we will increase the camera feeds for better coverage.  Leveraging an existing technology (traffic camera) will help the city adapt the early. A map layer will help end users to drive away from the city by avoiding flooded streets.

It seeks to develop Urban Water Resilience in San Jose and work toward achieving Sustainable Development Goals: Industry, Innovation, and Infrastructure, Sustainable cities and communities,  and Climate actions.

As humans continue to pour greenhouse gases into the atmosphere, oceans have tempered the effect. The world's seas have absorbed more than 90 percent of the heat from these gases, but it’s taking a toll on our oceans. Average sea levels have swelled over 8 inches since 1880, with about three of those inches gained in the last 25 years. In a new study, Stanford researchers report that intensifying precipitation contributed one-third of the financial costs of flooding in the United States over the past three decades, totaling almost $75 billion of the estimated $199 billion in flood damages from 1988 to 2017.

In 2017, San Jose was hit with heavy rain that resulted in creeks and reservoirs overflowing that overwhelmed the area. The flood resulted in $100 million in economic losses and 14,000 people displaced. The biggest issue during the 2017 flood was the city’s failure to notify its residents that their neighborhoods were under imminent danger of flooding. By the time San Jose put an evacuation order into place, many residents were already standing waist-deep in water. Since 1950, the city has dealt with 13 floods, and the absence of any warning has remained a consistent issue due to the lack of real-time flood data. 

More than half of flood-related drownings occur when someone drives into hazardous water, according to the Centers for Disease Control and Prevention and the National Weather Service. Even water that’s 12 inches deep can move a small car, and 2 feet of raging water can dislodge and carry most vehicles, the NWS says. And getting stuck can put others in peril, especially emergency workers who may need to come to your aid. Beyond personal risk, driving into floodwater also can leave you with a car that’s totaled. AAA recommends that any vehicle that has been damaged by floods be inspected by a professional mechanic before driving. Flood finder will notify San Jose residents  any street in 5 miles radius is flooded more then 30%. They can view the map with live traffic camera feeds to gauge the situation on ground and drive away from the city by avoiding flooded streets. 

City needs to be ready for disasters such as flooding. Being prepared for a crisis can mean the difference between loss and survival. FloodFinder provide city officials a visual dashboard with manholes locations and direction of storm pipes. This is a key information for city officials to redirect the water in the storm pipes away from the flooded areas and streets using storm pumps.

To better understand our community we interviewed people from different walks of life and they all had witness 2017 floods. Among them were State Hazard Mitigation Officer, Emergency manager from CA Department of water resource, fire department chef, a 911 dispatcher, and 2 residents that were impacted during 2017. They were send an email before the zoom call to be ready with data and statistics. The number one issue was the lack real time data at the street level was important to both the city officials and the San Jose residents. the documentation on interviews can be found here.

There is a pilot joint project underway with DWR, FEMA, and CalOES ( as per the article).  It uses existing bond funds for floodplain management  to put gauges on some ungated spillways of dams across the state. This will focus on those that have disadvantaged communities with some degree of hazard exposure downstream.  The issue is without a gauge on the spillway, there’s no way to get real-time flow rate coming from that dam or facility. Installing and maintaining these devises is expensive. 

FloodFinder is at a prototype phase. We have

1. Finished creating personas.
2. Interviewed Personas over Zoom with a set of questions
3. Designed and implemented Panoptics machine learning training model to predicts the water flooding in each of the acquired images and updates the database with the predicted results. The algorithm is trained to identify water segments and calculate the percentage area covered in the image using the trained model.
4. Designed and implemented monitoring service which observes the prediction results from the previous step to raise an alert. Currently If the calculated surface area is more than 30% then the push notification is generated.
5. Designed and implemented Dashboard, an interactive geographic information system (GIS) web UI  built with Leaflet.js map layers. It shows city camera locations with color coded, city manhole locations with details, and storm hole water pipes  and the direction of water flow in those pipes. This holistic visualization brings all these disparate information in one visual to help first responders make quick and correct decisions. The camera icons exhibit different colors based on the flooding area coverage in the image, allowing easy consumption of the information.
         

FloodFinder is collection of services written in Python and hosted in Azure cloud. 

Collector service: It is a Python based service in node that is scheduled to collect all San Jose city owned camera feeds every 5 minutes and store the images to MySQL databases. 

Model training : Use Panoptic FPN, to generate semantic and instance segmentations. It significantly outperforms the individual networks on the COCO and Cityscapes image recognition benchmarks. 

Processor service: Predicts the water flooding in each image. If the aspect ratio of water to other objects in an image is less than 10% than that image is given a value green. If it is between 10% to 35% then that image is given a value yellow. If it is between 35% to 60% then that image is given a value orange. And finally, if it is more than 60% that image is given a value red. 

Monitoring service: A monitoring service use Amazon Simple Notification Service (SNS). 

Dashboard: An interactive geographic information system (GIS) web UI  built with Leaflet.js map layers. It shows city camera locations with color coded, city manhole locations with details, and storm hole water pipes  and the direction of water flow in those pipes. This holistic visualization brings all these disparate information in one visual to help first citizens make correct decisions. The camera icons exhibit different colors based on the flooding area coverage in the image.

We have spent 5 months interviewing San Jose, California stakeholders to get their requirement before implementing the product.  We have pitched IT department including Mr. Jordan Sun in April of this year. We are going to implement our first version by Q4. Currently there are 1,013,240 residence who live in San Jose. We plan to pilot with area overage for 10,000 residence (1%) of residents before targeting to the larger audience.

FloodFinder seeks to achieve the following goals as it deploys initial pilots:

• Develop and expand to serve more users in city of San Jose.
• Refine business and revenue model from a prototype to a replicable service.
• Cultivate a marketing and social media presence to reach new partners and clients.
• Support marketing and communications that clearly distribute data
• Build connections to investors and grantors to progress their fundraising strategy

FloodFinder measure the progress :

• Integrate Floodfinder.org with city data and city servers.
• Create Facebook and twitter social media account.
• Increase the number of covered cities to 5-9 in coming year.
• Add new features like depth of flooded water to assess the risk and damage cost.
• Explore integrations with insurance sector.
• Hire more human resources to help grow the technology and product in the sustainability area.

FloodFinder needs

• A technical team to work on improving the model accuracy and enhancements
• A marketing team to help advertise to other cities.
• A financial help for operational cost. 

1. The Flood-Finder was part of the Open17 Water Challenge, a 5-week online coaching program, to help global innovators shape ideas tackling Urban Water Resilience into viable innovation projects. The Open17 Water Challenge is supported by the Crowd4SDG Project and co-supported by the Geneva-Tsinghua Initiative (GTI) at the University of Geneva. Flood Finder is one of the projects that got selected for further development as part of the SDG Accelerator program at the GTI. 
2. In August 2020 I joined Code of San Jose, a nonprofit volunteer group. Their mission is to make community services more transparent, accessible, and equitable by collaborating with local government and community-based organizations on civic projects to improve San José and the wider South Bay for everyone. I worked as a data engineer to create web scrapper program to fetch free school meal from county websites. 
3. In August I took AP computer science A with Data structures at my school that help me expose to object oriented programming.
4. In November 2020 I took a online course in Machine learning from Stanford to help with the basic concepts.
5. I have been using YouTube videos to self learn AWS concepts.

FloodFinder currently has two groups of partners for technical and project management help and initial piloting:

• The Harker School Programming club.
• The Code for San Jose,  a nonprofit volunteer group. Their mission is to make community services more transparent, accessible, and equitable by collaborating with local government and community-based organizations on civic projects to improve San José and the wider South Bay for everyone.

It has became more apparent to me that in our quest to find answers and solutions, our society has gravitated toward blaming each other. I strongly feel we should channel our efforts to tackle bigger issues like climate change and sustainability. And to combat big problems we need innovation in technologies and new policies. So I started a Harker Civic Tech club at my school. The aim of the club is to  develop engagement in citizens to solve civic literacy issues by identifying issues in the government and inspiring creative tech and policy making solutions.

My efforts have been:
1. Coordinate with outreach officer to invite City Council members to discuss real world problem, and then divide the team member in a group to brain storm ideas to find  technology-based or policy based solutions.
2. Work with Vice President to help publish Op-Eds/ blogs on technology awareness and its impact on society
3. Work with outreach officers to collaborate with non-profit organizations to make code open-sourced, and use technology to promote civics.
4. Host a Civic Hackathon at the end of the year to showcase the work

I am not in Boston area, so I don't qualify