Inspiration
We have collected challenge ideas from organizations across the globe to provide further inspiration for your projects. These organizations either provide direct disaster response and relief services, or focus on research to improve disaster response.
Global focus
- NASA Earth Science Data Systems
How can we help society respond to natural disasters by using Earth remote sensing observations and complementary tools to map their impacts? - HOT - Humanitarian OpenStreetMap Team
How might we identify (and/or accurately outline) features such as buildings, roads, schools, and bridges from aerial and/or street level imagery? Note that these features may look different in different parts of the world.
Latin American focus
- NASA Disaster Response Latin America Program & AmeriGEO
How can we use machine learning to create landslide management strategies at community level?
North America focus
- ITDRC - Information Technology Disaster Resource Center
How might we more quickly understand the impacts of natural disasters on internet connectivity in a given area? - SBP
How might we accurately and efficiently determine the extent of damage to individual homes in a given disaster-impacted area?
Asia / Pacific focus
- ADB - Asian Development Bank
How might we quickly and accurately determine the condition of critical infrastructure (such as roads, bridges, water sources, health clinics, and schools) and prioritize recovery needs after a disaster impacts a given area?
How can we help society respond to natural disasters by using Earth remote sensing observations and complementary tools to map their impacts?
Shared by: NASA Earth Science Data Systems https://earthdata.nasa.gov/esds
How It’s Relevant: Natural hazards in our environment intersect with human activity, often leading to disasters of long-lasting impact to infrastructure, agriculture, and the environment. While some are well-predicted, such as hurricanes and severe weather, their significant impacts are still felt by many and can be large enough in scope that satellite remote sensing, machine learning, and other tools are helpful in mapping and estimating the full extent of their impact and longer-term recovery.
For example, there are opportunities to use remote sensing to address a number of questions following a disaster event:
- Where has infrastructure damage occurred, with obstructed roads and impacted buildings limiting transportation and shelter?
- How far inland has a storm surge traveled, leaving significant flooding, in combination with flooding rains as a hurricane moves inland?
- How can we map these impacts and combine them with other socioeconomic data to answer questions about economic loss, the speed of recovery, and whether that recovery is shared equitably across the entire population?
As events recur over time, we also have an opportunity to map changes in their frequency, severity, and extent such as what we may experience as the result of climate change, or, perhaps see the impacts lessen over time as strategic investments are made to mitigate disaster impacts.
Suggested Ideas:
- Use optical or SAR imagery to identify the extent of water and flooding from a major tropical cyclone
- Use machine learning to classify features in aerial and other imagery as damaged buildings, water, or other topics of interest
- Use machine learning to classify smoke features in satellite imagery for early warning processes.
Suggested Resources:
- https://appliedsciences.nasa.gov/what-we-do/disasters/inside-disasters
- https://registry.opendata.aws/usgs-landsat/
- https://registry.opendata.aws/sentinel-2-l2a-cogs/
- https://registry.opendata.aws/sentinel-1/
- https://registry.opendata.aws/osm/
- https://registry.opendata.aws/spacenet/
- https://registry.opendata.aws/ladi/
- https://registry.opendata.aws/noaa-goes/
- https://registry.opendata.aws/noaa-himawari/
- https://registry.opendata.aws/sentinel-s2-l2a-mosaic-120/
How might we identify (and/or accurately outline) features such as buildings, roads, schools, and bridges from aerial and/or street level imagery? Note that these features may look different in different parts of the world.
Shared by: HOT - Humanitarian OpenStreetMap Team https://www.hotosm.org/
How It’s Relevant: As humanitarian organizations prepare for and respond to natural disasters, having updated information on the location of key infrastructure can significantly improve both the speed and effectiveness of their planning and response. The Humanitarian OpenStreetMap Team (HOT) is an international team dedicated to humanitarian action and community development through open mapping. As part of this work, HOT mobilizes volunteers from all over the world to create open map data that enables decision-makers to better allocate resources and reach communities in need. HOT also develops open source tools for collaborative mapping and geospatial data collection. These tools are free for all to use and are leveraged by global nonprofits, international organizations, government agencies, and local community organizations. In this way, HOT and a growing number of volunteers are working to map an area home to over one billion people, who are living in disaster prone areas or experiencing high levels of poverty, to achieve a world where missing maps are no longer a factor in human suffering or loss of life in humanitarian crises.
Suggested Ideas: Geospatial data, satellite and drone imagery, as well as social media, may be useful in solving this challenge.
Suggested Resources:
- https://registry.opendata.aws/osm/
- https://registry.opendata.aws/spacenet/
- https://registry.opendata.aws/ladi/
- https://registry.opendata.aws/sentinel-2-l2a-cogs/
- https://registry.opendata.aws/sentinel-1/
- https://openaerialmap.org/
Note: Closed Captions (CC) are available in Spanish and English via the Youtube player controls.
How can we use machine learning to create landslide management strategies at community level?
Shared by: NASA Disaster Response Latin America Program & AmeriGEO https://www.amerigeoss.org/
How It’s Relevant: Latin America and the Caribbean is the second most disaster-prone region in the world. According to the United Nations Office for the Coordination of Humanitarian Affairs, during the first 2 decades of the 20th century, from 2000 to 2019, in Latin America and the Caribbean, 152 million people were affected.
In recent years, the landslide in Guatemala in 2015, which caused 350 deaths, and Colombia in 2017, which caused 349 deaths and affected more than 45,000 people, stand out as particularly destructive landslide events in the region. There are many different types of landslides and they almost always have multiple causes, including rainfall, changes in water level, stream erosion, earthquakes and volcanic activity. Human activity can also be a contributing factor in causing landslides.
Currently, the inputs used to determine the risk of landslides all over the world are limited by large-scale maps that do not allow actions to mitigate the risk at the local level. Therefore, there is an opportunity for researchers and scientists to develop dynamic mapping tools to determine the risk of landslides more accurately.
The challenge is to create an application for the use of satellite data and information locally provided by citizens to determine the risk for landslides in towns and rural areas to give a forecast of the level of risk. This will inform governments decision-making and contribute to saving lives and building the resilience of communities.
Suggested Ideas:
There is an opportunity to create prototypes and methodologies to integrate satellite data and local open data, provided by national entities and scientific institutes. Additionally, to improve the precision of the analysis, it will be valuable if the general public could contribute by capturing data in their local territories. Finally, it will be ideal if any tool or prototype that meets the overall objective of this challenge must potentially be implemented and executed at a low cost by local governments.
Goals:
To solve this challenge, we believe there may be ways to leverage machine learning to determine risk zoning and to assess levels of risk for landslides. We also believe that a tool that has a graphical interface will allow communities and local governments to view, interpret, include local information, and analyze decision-making results.
To address this problem, you must take into account the geology, the historical databases of landslides in Latin America, the elevation and surface information of the territory, the Normalized Difference Vegetation Index (NDVI), the current land use, demographics, and other variables that you consider helpful in your methodology. Included variables must be processed under a model that allows the method to be standardized and replicable in different municipalities. Remember that incorporating data capture by the community will considerably enrich your analysis of both detection and prediction of risks.
As you develop your solution, you may (but are not required to) consider the following:
- Presenting information in a straightforward, intuitive, and easy process by communities and local governments is critical.
- Develop a prototype that is scalable but also operable at a low cost.
- Build or increase models that allow quantifying potential human losses and the cost of impacts from landslides.
- Develop machine learning training data sets to predict landslides so that other researchers can easily incorporate their training data into their processes.
- Create a code repository for your project so other people can review and take advantage of your efforts.
Suggested Resources:
- https://science.nasa.gov/citizenscience
- IMERG: Integrated Multi-satellitE Retrievals for GPM | NASA Global Precipitation Measurement Mission NASA IMERG
- https://gpm.nasa.gov/landslides/ NASA landslides resources
- https://pmm.nasa.gov/applications/global-landslide-model NASA Landslides model
- Landslide Hazards - Data & Tools (usgs.gov) USGS Landslides data
- https://ceos.org/ourwork/workinggroups/disasters/landslide-pilot/ CEOS Landslides demo.
- Landslides - Emergency Management - Thematic Areas - Sentinel Online - Sentinel Online (esa.int)
- Earthdata (nasa.gov)
- https://www.meteorologicaltechnologyinternational.com/news/nowcasting/nasa-machine-learning-model-doubles-accuracy-of-global-landslide-nowcasts.html
- https://registry.opendata.aws/usgs-landsat/
- https://registry.opendata.aws/sentinel-2-l2a-cogs/
- https://registry.opendata.aws/sentinel-1/
- https://registry.opendata.aws/osm/
- https://registry.opendata.aws/noaa-goes/
- https://registry.opendata.aws/noaa-himawari/
How might we more quickly understand the impacts of natural disasters on internet connectivity in a given area?
Shared by: ITDRC - Information Technology Disaster Resource Center https://www.itdrc.org/
How It’s Relevant: Disasters inherently disrupt terrestrial communications, leaving responders and survivors without reliable means to communicate with emergency crews and loved ones. The lack of communications can delay emergency aid to victims, and also increases anxiety and insecurity of those who have just experienced the traumatic event.
Having the ability to immediately understand the extent of the communications impact after a catastrophic event enables first responders, mass care NGOs, and emergency management agencies to better understand the impact on a community, and respond accordingly.
While private sector communications companies have the ability to understand impacts to their own infrastructure, accurate outage data is rarely shared with responders; and doesn’t always convey the full impact of damage.
As a trusted resource to both industry and emergency management agencies, ITDRC provides temporary communications and technical resources to Communities in Crisis™. Being able to more quickly understand the impacts of natural disasters on internet connectivity would enable us to mobilize teams and resources faster to targeted areas within communities to provide them with temporary connectivity until the public infrastructure can be repaired.
Suggested Ideas: Internet speed data and social media information may be useful in solving this challenge.
Suggested Resources:
How might we accurately and efficiently determine the extent of damage to individual homes in a given disaster-impacted area?
Shared by: SBP https://sbpusa.org/
How It’s Relevant: As state, local, tribal, and territorial governments seek to determine the magnitude of damage from a natural disaster, they routinely conduct damage assessments that focus on determining the extent of damage to the homes of their citizens. This is often a manual, time-intensive, error-prone process – particularly in areas where infrastructure is also damaged or where these citizens are unable to immediately return to their communities. In addition, studies have shown that low-income homeowners are often under-compensated resulting in them receiving fewer resources to repair their damaged homes.
Having the ability to quickly understand the extent of damage to individual homes enables a more rapid and equitable start to the recovery process, which can save millions – if not billions – of dollars in economic impact. When residents are able to rebuild timely, they can get back to work and school, have better health and mental health outcomes and can fully participate in their communities thereby contributing to a more efficient recovery. This would ensure more communities are aligned with the United Nations Sustainable Development Goals goal to “substantially decrease the direct economic losses” caused by disasters, “with a focus on protecting the poor and people in vulnerable situations.”
SBP is a social impact organization that helps under-resourced communities prepare for and recover from disasters via three strategic initiatives: 1) SBP prepares individuals, communities, and organizations to mitigate risk and speed recovery; 2) SBP shapes national policy and system change and local disaster recovery programs to be more efficient and effective; and 3) SBP builds resilient communities efficiently and effectively and shares our proven model. Being able to more quickly understand the impacts of natural disasters on homes would enable us, as well as other nonprofits and government entities, to jumpstart the rebuilding process.
Suggested Ideas: Geospatial data, satellite and drone imagery may be useful in solving this challenge.
Suggested Resources:
- https://registry.opendata.aws/osm/
- https://registry.opendata.aws/spacenet/
- https://registry.opendata.aws/ladi/
- https://registry.opendata.aws/digital-globe-open-data/
- https://openaerialmap.org/
Please note: As an additional frame of reference, current standards within the United States define damage in four categories, defined below:
- Affected: a home is considered affected if the damage to the home is mostly cosmetic.
- Minor: a home with repairable non-structural damage.
- Major: a home with structural damage or other significant damage that requires extensive repairs.
- Destroyed: the home is a total loss.
How might we quickly and accurately determine the condition of critical infrastructure (such as roads, bridges, water sources, health clinics, and schools) and prioritize recovery needs after a disaster impacts a given area?
Shared by: ADB - Asian Development Bank https://www.adb.org/
How It’s Relevant: After a disaster, those responding and planning the recovery must determine how to prioritize limited resources to rebuild damaged infrastructure and meet critical gaps. Across Asia and throughout the world, this is often a manual process of collating field reports, making educated guesses, and extrapolating data that takes time to consolidate and reach decision makers. This can delay recovery efforts by responding organizations such as governments, nonprofits, and international organizations, and may decrease the effectiveness of the response.
Having the ability to quickly understand the extent of damage to critical infrastructure after a disaster enables those responding to better understand the impact on affected communities, and to steer resources to the places where they are most needed.
Post-disaster, ADB supports its developing member countries to assess damage and loss and quantify recovery needs over the short, medium, and long term. As governments lead the post-disaster needs assessment and recovery planning process, partners like ADB can provide sector specific expertise, including on roads, agriculture, health and education, to inform and contribute finance to the recovery. The quality of post-disaster assessments and recovery plans are determined by the accuracy of data.
Suggested Ideas: Geospatial data, imagery, and social media information may be useful in solving this challenge.
Suggested Resources: