Team Flood Riders on improving flood impact maps
The following article covers the work of team Floor Riders, the winners of the “3D mapping for inundation” challenge part of the third Mothership Mission titled The Big Blue. For more info on the program visit the website.
In the midst of the Corona pandemic in mid- 2020, our attention turned to another form of a natural disaster — floods. Like a disease, floods have accompanied Mankind since the bud, and are featured in ancient tales and legends. These days they are still the most common weather-related disaster, and global warming suggests their frequency may increase further yet.
Source: https://asf.alaska.edu/how-to/data-recipes/create-a-dem-using-sentinel-1-data/
The five of us participated in a challenge that could hopefully alleviate the impact of flood events on vulnerable communities. Relief and aid efforts can greatly benefit from ‘flood risk maps’, a form of a map which can portray in advance the impact of a flood in a given area. A flood risk map is based on the topographical data for a specific area, given in the form of a Digital Elevation Model (DEM).
But open-access global DEMs generally have a low resolution, particularly in remote areas, making them insufficient for relief planning. Our task was to find whether global satellite imagery available by the European Space Agency (ESA) could be used to generate a high-resolution publicly-available DEM for the benefit of communities and relief initiatives.
This challenge was invited on behalf of ‘510’, an initiative of the Netherland Red Cross. They would be our end users. ‘510’ had a specific test case in mind for us: the coastal city of Dar es Salaam, Tanzania. A better DEM would improve the Red Cross ‘510’ flood impact maps for the area. The program was facilitated under the joint auspices of ‘space4good’, and The Hague-based ‘World Startup Factory’ and AI.Lab’s program called ‘Mothership Mission’.
As it turned out, we’re grateful for these numerous helping hands, since this was not a trivial task … To our advantage, we found there is existing academic research on using satellite images to produce DEMs, as well as ‘user guides’ available from the scientific community. These referred us to the Sentinel Application Platform (SNAP). This refers to the ‘Sentinels’, the Earth-Observation satellites of the European Space Agency (ESA). OK, enough name-dropping. Readers are welcome to Google further as they please.
For us, the availability of SNAP, with its user community and forums, essentially meant that there was a program ‘out there’ capable of performing the calculation feat with all its tweaks and tunes. It still needed to be evoked and configured to the flow we found. But more good news — SNAP could successfully interface with Python code. And we likewise had access to some heavyweight virtual machines. So we did it!
Not that it was easy. But we managed to produce a DEM model for Dar es Salaam which is at 10m resolution, three times better than the publicly-available DEMs of 30m resolution. Not bad! Add to that, our vertical accuracy (so the relative height of objects) was 4m versus 6m for the prior models. A job well done — on the technical side.
There were also a number of business-related problems. For once, the process of DEM generation is quite demanding in terms of computational resources and storage capacity. On the other hand, a requirement of our challenge was to make this service open-source and the data publicly-available. So how do we combine the two requirements?
Our solution was ‘fee in, free out’. The first client to request a DEM for new area funds it. One that it completed, the data is available on-line, ready for immediate access through our partners’ websites. This service allows DEMs to be generated when needed, and be updated on a frequent basis. This will make sure that risk models are based on the most up-to-date datasets.
So much for the technical and business side. What was our personal experience as individuals and as a team in this hackathon?
The most striking experience is that not all obstacles were foreseeable from the get-go. Some surprises which popped up along the way included the extent of domain knowledge required or the trade-offs in the selection of image sources. Even technical issues that had been flagged as potential problems, such as high computational load and time-demands of the model generation, punched with their full weight when we hit them along the way.
That is precisely why professional help mattered. We had consultation sessions with users, technical experts, and the program organizers. It helped. There was a wealth of topic information as well as specific tool know-how that had to be digested by the team, and experts’ advice helped us find our way.
Notwithstanding, there was tremendous value to prior knowledge held by team members, whether it was in breaking down a computing task to multi-threads or in evaluating satellite image sets and tools.
Of course, to meet a challenge of this scope, contributions were made by every member of the team. Varied as we were, we still successfully found a constructive use for each of the team member’s skills, whether it was in domain knowledge, business development, coding, or pitching.
This was our first time participating in a highly specialized challenge. This is quite different from a typical event when you are encouraged to come up with your own business idea or technology application, with only a loose theme occasionally in place.
Lastly, as a side-note, let’s actually do away with the term hackathon, which is not so proper here. This was not a one-sitting development sprint. Rather this program ran eight weeks long. As such it may better be called a ‘challenge’. And a challenge it was.
What would be the ultimate outcome? Ideally, we would like to empower communities and relief efforts. This can mean making a change from long before a flood (e.g. optimal city or infrastructure planning), during an emergency (safer evacuation routes) to after the event (e.g. forecasting when homes are accessible again).
The satellites’ eye in the sky can make a difference on the ground. We have proven that it can be done.
Written By: Roy Birnholtz, Rene da Costa, Veronika Heidegger, Kishaan Sutharsan and Kotryna Valečkaitė
Team Flood Riders won the prize for their challenge co-organized with 510 Red Cross.
The Mothership is an open innovation program helping teams to develop a proof of concept and business model for solutions related to the Sustainable Development Goals. The program is co-organized by AI Lab One, Space4Good, and WorldStartup.
