Beth Tellman

Addressing flood justice and equity impacts of adaptation and urban expansion with satellite observations

NSF CAREER Award

To expand urban development, preserve property values, or avoid paying for flood insurance, real estate developers in the US may request to remove properties from the Federal Emergency Management Agency (FEMA) regulatory floodplain, often by building flood mitigation infrastructure like levees and improved drainage. However, this flood mitigation infrastructure could be draining floodwaters to others downstream, an effect has not been measured. This Faculty Early Career Development CAREER project will advance knowledge by generating observed flood data to measure if and where flood mitigation infrastructure changes flooding and who is affected. The project will build capacity for environmental justice organizations to use satellite-based flood maps to challenge changes to FEMA flood maps and advocate for adaptation funding for their communities. Read more here

FLUJOS: Flood Justice using Satellite Imagery - Rio Grande Valley

Google Environmental Justice Data Fund, Climate Change AI, U Arizona Community Engagement Grant. With Texas Rio Grande Legal Aid, Lucas Belury, Zhijie Zhang

This project will leverage satellite-based flood maps in four Rio Grande Valley (RGV) counties (Willacy, Cameron, Hidalgo and Starr) to co-produce a data tool, designed and evaluated by local community-based organizations, in one of the most flood vulnerable regions of the US. We will: i) build an open access history of satellite-observed inundation in the RGV, ii) advance legal advocacy opportunities by identifying disparate impact, inverse condemnation, or discriminatory intent in flood patterns and infrastructure, iii) overlay satellite-based flood maps with the service areas of existing community-based organizations to identify currently unknown flood vulnerable areas or populations and, iv) co-develop, build, and evaluate a prototype data tool to further flood justice in the RGV. To build a library of flood maps, we will use machine learning approaches to extract surface water imagery from five satellites (Landsat 8, Sentinel-1, Sentinel-2, MODIS, and PlanetScope) at various resolutions. The maps will be assessed and ground truthed in field visits and community workshops to improve existing algorithms and assess the limits of satellite-based flood data in this region. The ultimate tool and the data created will be made publicly available.

2022-2023 International Working Group to Assess Flood Recovery, Adaptation, and Resilience from Space

Funded by The Airzona Institute for Resilient Environment and Societies led by Hannah Friedrich

The working group assembled a network of scientists and practitioners to develop a conceptual framework and research agenda for using satellite imagery to monitor flood resilience, adaptation, and recovery. This working group unified flood researchers and practitioners to achieve three outcomes to foster innovation in measuring flood recovery and adaptation from satellites. We organized a workshop at the Global Flood Partnership 2022 meeting and the public 2023 Flood Justice Sympsium at University of Arizona. A recording of that event is available here. We are currently exploring funding to keep the network alive!

Key papers:

Earth Observation to Address Inequities in Post-Flood Recovery

Five Key Needs for Addressing Flood Injustice

Enhancing the Evidence for Humanitarian Action (EEHA) in the Face of Climate Change

Bureau of Humanitarian Affairs, PI Zack Guido. Co-I Tellman, Mukherjee, and Sharma from Social Pixel Lab.

The Enhancing Evidence for Humanitarian Action (EEHA) will identify, develop, and test novel indicators and methods for measuring the results and effectiveness of humanitarian investments in three sub-sets of USAID DRR and related programming; local early warning systems, DRR governance (action, contingency, and respond planning), and diversified livelihood strategies that reduce disaster risk and enable households and communities to adapt and become more resilient to climate variability and change.

Feasibility of Satellite Based Flood Extent Monitoring in Pima County

Funded by the Pima County Flood Control District, 2022

This project explored different satellite data sources and modeling techniques in order to assess the potential for remote sensing of monsoon flood events for Pima County.

NASA funded flood mapping efforts

NASA Commercial SmallSat Data Analysis. 20-CSDA20-0121. 2021-23. PI with collaborators Iskha Gurung, Andrew Molthan (NASA), Kobus Barnard, and Upmanu Lall. High resolution imagery to train and validate deep learning models of inundation extent for multiple satellite sensors

Our dataset, called FloodPlanet, will include 25 major flood events, labeled with co-located Planet, Maxar, HLS (Harmonized Sentinel-2 and Landsat), and Sentinel-1 data to promote cross-sensor fusion across public and commercial sensors and deep learning modeling efforts in the remote sensing and computer vision fields. We will develop our own deep learning models for high resolution urban flood detection, and release the the high quality benchmark dataset we will produce in this proposal. As an extension of this project, we are also testing the potential for water mapping with the commercial Black Sky satellite

NASA Terrestrial Hydrology. 19-THP19-0042. 2021-2024. PI with Co-Is Upmanu Lall, Pierre Gentine, and Venkat Lakshmi. Mapping flood impacts using multi-sensor satellite data fusion in urban areas

In this research, we are developing novel sensor fusion techniques to take advantage of the expanding availability of sensors with different spatial and temporal resolutions to monitor urban flooding. To do so, we combine public optical (MODIS, Landsat, Sentinel-2), radar (Sentinel-1), passive microwave (AMSR, SMAP) imagery and commercial data from PlanetScope with the eventual goal of generating 30-meter per-pixel probabilities of inundation in urban areas globally every 3 days.

Validation and machine learning approaches for surface water mapping with Planetscope. 2022-2023.

This collaboration with Fritz Policelli aims to develop a benchmark dataset for surface water detection sampling a diversity of rivers around the world and labeling water using Planetscope. All images are co-located with Sentinel-1 imagery to validate Dr. Policelli's gaussian distribution approach to Sentinel-1 water mapping aimed at extracting river widths over time. The labeled dataset will be released upon completion.

Flood Mapping and Insurance in Bangladesh

This project is funded by the NASA New Investigators Program in Earth Sciences, Tellman PI with Collaborators Upmanu Lall, Saiful Islam, Mohammed Bhuyan, Mehadi Hasan, and Sarder Raihan. Understanding flood risk in human altered landscapes from cities to farms: inferences from satellites and machine learning, with initial funding from the Columbia University ActToday Program.

This project leverages inundation observations to understand extreme flood risk in human-modified landscapes, specifically in urban and agricultural areas in Bangladesh. We will use commercial and public sensors to i) map maximum inundated area for extreme flood events, ii) quantify flood risk at watershed scales, and iii) estimate of flood return periods at household scales (30m resolution). \ These objectives will be reached by using deep learning approaches to map floods and damage, and estiamte return periods with Bayesian Hierarchical Models. We will use optical and radar satellites, including MODIS, Sentinel-1, Planetscope, and the HLS (Harmonized Landsat Sentinel-2 dataset), comparing satellite-based flood maps and return period predictions for Bangladesh to those produced by the Bangladesh Flood Forecasting and Warning Centre (from a physically-based model). In May 2023, we discussed applications for insurance and climate risk financing with collaborators in Bangladesh and ran a training on how to make a flood map in Google Earth Engine and with Machine Learning. Training materials from our 2 days course (code plus slides) are available here. We have developed a 500m weekly inundation data set for Bangladesh from 2001-2022 that fuses MODIS with Sentinel-1, see our datasets page for access.

Key papers:

Inferring the past: a combined CNN-LSTM deep learning framework to fuse satellites for historical inundation mapping

A framework to assess remote sensing algorithms for satellite-based flood index insurance.

Regional Index Insurance using Satellite-based Fractional Flooded Area.

Selected media:

Understanding flood risk from space: opportunities to adapt to changing risk with improved monitoring at The University of Coloroado, Boulder, Geography Colloquium

Mapping floods for insurance applications in Bangladesh at WKCR Radio Interview

Climate Insurance, (Mal)adaptation, delaying inevitable retreat or migration? 2021 Columbia Managed Retreat Session

Illicit activity and land change: Narcotrafficking in Central America

Illicit economic activity shapes land use around the world, but inadequate data makes it difficult to quantify relationships. We've developed methods to leverage news media and satellite data to study how illicit activity shapes landscapes in Central America, the region where the vast majority of cocaine that arrives in the USA travels through. As the US war on drugs attempts to (unsuccessfully) intercept drug traffickers, they move into remote forested regions. This arrival of drugs, and the capital that comes with it, causes changes in land cover, use, ownership, and control with adverse impacts on people, forests, and biodiveristy in Central America. In collaboration with a larger research team, we seek to better understand how narcotrafficking transforms landscapes and what policies could reduce violence and support sustainable development in the region. See opportunities for our recent NASA grant to continue this work.

Selected media:

To preserve tropical forests, empower local communities: Phys.org ; Los guardianes de la selva maya desafían a narcos e inversores: ElPais ; Beyond 'business-as-usual': using news media and satellites to connect drug trafficking to forest loss: Behind the paper, Nature Sustainability ; Paths to Sustaining Forests and Communities in Guatemala's Maya Reserve: Sustain What? podcast with Andy Revkin The impact of cocaine trafficking on tropical forests in Central America Reddit Science AMA (Ask Me Anything)

Key papers:

Illicit Drivers of Land Use Change: Narcotrafficking and Forest Loss in Central America

Understanding the role of illicit transactions in land-change dynamics

Informal urbanization: Mexico City

Urban expansion is growing fast in the Global South, much of which occured on land unplanned for formal urban development. Informal urban settlements provide needed housing for growing populations, with subsequent needs for services like water, electricity, and property titles. We've been studying what (and who!) governs informal urban expansion in Mexico City, and the social and environmental consequences.

Key papers:

The role of institutional entrepreneurs and informal land transactions in Mexico City’s urban expansion.

Governing the gaps in water governance and land-use planning in a megacity: The example of hydrological risk in Mexico City.

Transferencia de riesgos sociohidrológicos y vulnerabilidad emergente en el Valle de México: un modelo conceptual basado en agentes. Las paradojas de la megalopolis: un debate actual a distintas voces

Infrastructure, water vulnerability, and adaptation: Rainwater Harvesting in Mexico City

As megacities grow, infrastructure choices may either reduce or further entrench vulnerability to water security. One decentralized adaptation to water stress is household rainwater harvesting, which can improve water access in the rainy season. We worked with Isla Urbana, an NGO in Mexico City, to estimate how their harvesting system could improve water security across the city if implemented on every home. Oxfam funded a study for us to build a large geospatial model to integrate satellite imagery, census data, climate data, and surveys to identify the neighborhoods where rainwater harvesting systems would reduce water stress and how many systems should be installed. The study was the basis for the city government's plan to stall 100,000 systems where we indicate they could most improve water access.

Key papers:

Adaptive Pathways and Coupled Infrastructure: adaptations to water risk, and the production of vulnerability in Mexico City since 1325

Opportunities for natural infrastructure to improve urban water security in Latin America

Selected media:

Study Paves Way for Rainwater Harvesting in Mexico City: Earth Institute, Columbia University Viven 38 mil en precariedad hídrica alta en CDMX: La Reforma Presentación de la investigación: Captación de lluvia en la CDMX: with Isla Urbana and Oxfam Seminario Captación y aprovechamiento de agua de lluvia y desigualdad espacial en la Ciudad de México con WATERLAT-GOBACIT.