October 5, 2022

Geo For Good

Capacity building with SERVIR’s service on HYDrologic Remote Sensing Analysis for Floods (HYDRAFloods)

Tim Mayer

Emil Cherrington

Biplov Bhandari

Capacity Building

Pre-training survey

Session Outline

CONNECTING SPACE TO VILLAGE

SERVIR is a joint initiative of NASA, USAID, and leading geospatial organizations in Asia, Africa, and Latin America that partners with countries and organizations to address challenges in climate change, food security, water and related disasters, land use, and air quality.

Using satellite data and geospatial technology, SERVIR co-develops innovative solutions through a network of regional hubs to improve resilience and sustainable resource management at local, national and regional scales.

​

SERVIR Focuses on Countries in Asia, Africa, & the Americas

FOCUS COUNTRIES

ADDITIONAL BENEFITTING COUNTRIES

International Research Institute for Climate & Society

Lamont Doherty Earth Observatory

Columbia University

SERVIR Connects US Science to Global Development Challenges

IFDC

NASA Marshall Space Flight Center

NASA SPoRT Center/ENSCO

Universities Space Research Association

University of Alabama in Huntsville

Geospatial Collaborative LLC

Columbus Technologies

ESRI

Google

Mapbox

NASA Jet Propulsion Laboratory

Spatial Informatics Group, LLC

University Of California, Los Angeles

University Of California, Santa Barbara

University of California, Berkeley

University of California, Irvine

University of San Francisco

University of Florida, Gainesville

George Mason University

US Agency for International Development

USGS, Eastern Geographic Science Center

University of Richmond

Michigan State University

Oregon State University

US Forest Service, PNW Research Station

Amazon

University of Washington

New Mexico State University

University Of Houston

Texas Agricultural Experiment Station

​

NOAA/National Severe Storms Laboratory

University of Oklahoma, Norman

University of South Carolina

DevSeed

Resources For The Future, Inc.

World Wildlife Fund

University of Minnesota

Desert Research Institute

Boston University

Clark University

Earth Big Data

Univ. of Massachusetts, Amherst

Woods Hole Research Center

​

Johns Hopkins University

NASA Goddard Space Flight Center

University of Maryland

Univ. of Maryland, Baltimore County

Univ. of Maryland, College Park

USRA, Columbia

Alaska Satellite Facility

University of Alaska Fairbanks

Maxar Technologies

Univ. of Colorado, Boulder

Brigham Young University

US Forest Service, Rocky Mountain Research Station

US Geological Survey, Salt Lake City

University of Wisconsin, Madison

• Poverty reduction & resilience
• Data-dependent issues in data-scarce places
• International field presence

• 30+ Earth observing satellite missions, free & open data
• Major research portfolio
• Societal benefit from space

Hub Consortium Members:

Regional Hub Host Institutions:

Private sector collaborators:

Who Is SERVIR?

​

Research collaborators: 20+ US universities & research centers through the SERVIR Applied Sciences Team; ITC, in-region university networks

USG collaborators:

Intergovernmental, NGO collaborators:

Geospatial Co-Development

Weather & Climate

Agriculture & Food Security

Water & Water- Related Disasters

Land Cover, Land Use Change & Ecosystems

CONNECTING SPACE TO VILLAGE

Colab Notebook

***Copy a version to your drive***

Colab Set-up

Copy the notebook to your drive

-

File: “Save a copy to my Drive”

Colab Set-up: Restart runtime

Restart runtime

-

Circle back

-

• pip install hydrafloods geemap: Cell #1

Colab Set-up: Authentication “Road Map”

Step 1

Step 2

Step 3

Step 4

Step 5

Step 6

Copy Authorization Code

Step 7

Paste Authorization Code

Colab Set-up: Authentication “Road Map”

Step 1

Step 2

Colab Set-up: Authentication “Road Map”

​

Step 3

Step 4

Copy Authorization Code

Colab Set-up: Authentication “Road Map”

​

5

6

Paste Authorization Code

Colab Set-up: Authentication “Road Map”

Step 7

Lets run the code and talk application

How familiar are you with SAR and specifically Sentinel-1 ?

-

Resources: SAR Handbook

Knowledge Check-in

HYDrologic Remote Sensing Analysis for Floods

HYDRAFloods

• The publicly available, web-based service delivering near real-time information for improved flood monitoring.
• Designed to provide information on flood location and extent to assist with flood preparedness, emergency response and relief efforts.

Provide information to users on:

• The extent of flooded areas
• Possible severity of flood (e.g., depth, duration, age)

HYDRAFloods

What is HYDRAFloods?

Open Source - anyone can use and modify for FREE!!!

• Documented to increase transparency

What is HYDRAFloods?

Documented to increase transparency

​

What is HYDRAFloods?

• Cloud-based - overcome big data challenges
• Built on top of Google Earth Engine and Google Cloud ecosystem
• End-to-End processing - users have all the tools needed to create their own high quality surface water map
• QA masking
• SAR speckle filters
• Terrain correction (SAR and Optical)
• Time series processing
• Machine learning workflows
• Multi-sensor water mapping algorithms

What makes HYDRAFloods powerful?

• Flood Maps can be generated from an array of sensors
• Making it powerful to combat temporal gaps
• Is built in GEE and Python for end-user ease of use

Background: Science

• Single sensor (flood) maps
• Sentinel-1
• Sentinel-2
• Landsat 8
• Daily data fused (flood) maps

Otsu

QA UNet

based on Donchyts et al. (2016), Application: Markert et al. (2020)

based on Hughes et al. (2019), Application: Mayer et al. (2021)

all

optical

Thresholding Algorithm

Otsu’s Thresholding

• Automated histogram-based thresholding approach
• Maximizes inter-class variance between two classes
• Assumes there are only two classes, a background and foreground

Markert et al. (2020)

HYDRAFloods S-1 Flood detection workflow

On November 3, Eta made landfall on the Caribbean coast of Nicaragua as a category 4 hurricane, followed by Iota on November 14 as a category 5 hurricane.

The compounding storms’ reach extended throughout Central America, with heavy flood impacts in Nicaragua, Honduras, and Guatemala.

An estimated 7.5 million people were affected across Central America

Source: Relief Web

SERVIR Method Transfer

Photo: European Union 2020, D. Membreño on Flikr

Central America Example

Hurricanes Eta & Iota

Central America example of flood detected from Sentinel-1 using HYDRAFloods

Let's revisit the code

Explore in GEE

Google Earth Engine Example

One Stop Shop

​

HYDRAFloods

​

Open Science

​

Open Source

​

Web portal with analytics

​

Dedicated data streams

​

Capacity building

HF Operational Flowchart

HYDRAFloods SERVIR-Mekong Website

Applications of HYDRAFloods

WFP Cambodia

• Currently providing merged water classification layers from multiple sensors
• Provide multi sensor surface water maps
• Uses machine learning and deep learning methods

Explore SERVIR-Mekong HYDRAFloods Website

Explore the interface

HYDRAFloods SERVIR-Mekong Website

Use Date: 9/27/2022

Zoom to

Preah Netr Preah Bântéay Méanchey, Cambodia

Conclusions

What this provides

• Daily wall-to-wall maps as high as 10m resolution of estimated surface water
• Ability to support surface water as daily basis

​

What this cannot provide

• Will not capture flash-floods or dam breaks as instantaneous events will not be in time series
• These are estimated water maps so may not align exactly with observations

Interested in more Capacity Building?

TensorFlow Working Group

​

Capacity Building

Post-training survey

Thank you to the rest of the SERVIR Science Coordination Office, SERVIR Hubs, and Applied Science Team that make all of these services possible.

​

​

Tim Mayer

Emil Cherrington

Biplov Bhandari

Thank You

For further questions, please contact:

timothy.j.mayer@nasa.gov

biplov.bhandari@nasa.gov

emil.cherrington@nasa.gov

SERVIRglobal.net

Methods, Algorithms, and Science Co-Development

Daily Surface Water Fusion Process (DSWFP)

Flood Water Depth Estimation Tool

-

Cohen et al. 2019 | Peter et al. 2020

Deep Learning: Visual Geometry Group (VGG) 19 model architecture

SAR Speckle Filtering Algorithms

Generic Water Mapping Algorithms

Etc.

HF used in Central America

Set-up in Colab

Algorithms: Change Detection

• Pre-event and post-event change detection using logarithmic amplitude ratio (LAR).

Sentinel-2 QA Model

Additional Example Uses

Deltares Flood Impact Analysis on Road Networks

• Used to understand how floods can potentially affect road networks and peoples access to critical services

Methods, Algorithms, and Science Co-Development

Leverage multiple sensors easily with common syntax and data fusion workflows.