A WORLD OF POSSIBILITIES
Lassoing satellite data
for global and local
investigations
Access this slideshow at
Introduction
Carl Churchill, Wall Street Journal
Deborah Nelson, University of Maryland
Laura Kurtzberg, Florida International University
RASTER DATA
Akhund, Sadig. (2022). Analysis of Spatial Big Data for Geographical Information Systems. 10.13140/RG.2.2.20522.70080/2.
EXAMPLES
SATELLITE
DATA
PROJECT
EXAMPLE
The Financial Times | Steve Bernard
September 2020
SATELLITE
DATA
PROJECT
EXAMPLE
ABC News Australia
How heat and drought turned Australia into a tinderbox
By Colin Gourlay, Tim Leslie, Matt Martino and Ben Spraggon
Story Lab
8 Feb 2020
RASTER
DATA
PROJECT
EXAMPLE
ProPublica
The Most Detailed Map of Cancer-Causing Industrial Air Pollution in the U.S.
by Al Shaw and Lylla Younes, November 2, 2021, Updated August 28, 2023
Additional reporting by Ava Kofman
Not All Raster Data is Satellite Data
Aquazônia
MAPPING IMPACTS ON WATER
Darker, deep blue areas represent low combined impact
Light yellow areas represent extreme impact on freshwater
MAPPING IMPACTS ON WATER
I calculated 9 different impacts per river microbasin using GEE
Cecília Gontijo Leal
Scientific Consultant,
Biologist &
Researcher
IIAA= 3*URB + 3*GAR + 3*HDL + 2*AGP + 2*MIN + 2*PRE + EST + HID + DEG
MAPPING IMPACTS ON WATER
📋 Data Sources
AWII:
Amazon Water Impact Index
Constructing the site
THANK YOU!
Check out the tools, sources, and tutorials at the end of the slideshow →
Carl Churchill
The Wall Street Journal
Graphics Reporter
carl.churchill@wsj.com
Preparing and Downloading Satellite Image Stacks with GEE*
A world of possibilities: Lassoing satellite data for global and local investigations
NICAR 2024
*and some other stuff
Landsat 7 and 8 produce 1 terabyte of data every day.
Give me the stupid data
I want this on my computer
Where is it
This story is pubbing in 4 hours i have a liberal arts degree what am i doing
[And others]
Y
X
Z (time)
Bands
Sentinel-2 (actually twin satellites)
C a r r i z o P l a i n
Satellite image
1,100 official datasets
1,352 community-uploaded datasets
Filters
Dataset
(already in GEE)
Vegetative health (EVI) 16-d comp
(day of year)
Vegetative health (EVI) 16-d comp
(forward from Jan 1st 2020)
EVI (Sentinel 2) = 2.5 * ((B8 – B4) / (B8 + 6 * B4 – 7.5 * B2 + 1)).
Inundated farmland after flooding
GPM
Global Precipitation Measurement
Earth Engine Alternatives
Not everyone has time to learn coding
The EO Browser
Climate Engine
Giovanni
NASA EO
Tutorials!
Carl Churchill
The Wall Street Journal
Graphics Reporter
carl.churchill@wsj.com
The code for the hurricane GIF earlier:
https://code.earthengine.google.com/b8ff6cc38ab556fc3598f9483feb721d
(its a old photo)
Questions?
Deborah Nelson,
Professor of Investigative Journalism
Philip Merrill College of Journalism, University of Maryland
“Ground-truthing” satellite data
Satellite data can help document the what, where and how.
But you need to come back to earth
for the causes
and consequences
Ground-truth the data
to report on local conditions and dynamics
that can’t be seen from the satellites
or measured in the data.
FEET ON THE GROUND
Your own, or through collaboration with a local journalist or connection with trusted sources.
VIRTUAL FEET ON THE GROUND
The usual suspects: Clips, satellite imagery, street view mapping, social media, etc.
Key resource: Google Scholar. Type the location of interest along with your keywords into Google Scholar. You may be surprised at how often you get geographically relevant results.
- Mine the top for authors with scientific and geographic expertise.
- Mine the middle for geographic coordinates, location descriptions, methodology that may inform your own modeling and clues about local political/social/environmental dynamics.
- Mine the bottom -- footnotes and supplements -- for more sources, details & data
Most human viruses initially came from animals
-- an event known as zoonotic spillover.
A growing body of scientific evidence has linked spillover
to environmental factors….
Case study…
Some are natural -- like precipitation and temperature.
Others are tied to human activity -- tree loss, urbanization, intensive farming --
that increase our contact with wildlife and disturb habitats in ways thought
to make some animals more susceptible to developing and shedding viruses.
While doing some deep reading on the subject, I found a reference to a database of zoonotic outbreaks over the past 20 years.
in a footnote of a report. It is maintained by Gingko, a public benefit company, and includes location information.
He thought we might be able to predict where and why future outbreaks could occur by:
I mentioned all this to my longtime reporting partner at Reuters:
Ryan McNeill, deputy data editor in London
We decided to focus our analysis on viruses linked to bats because there were so many spillovers -- at least 95 since 2002.
And they represented some of the deadliest new diseases to emerge in the last half century: SARS, Ebola, Marburg and Nipah.
Although we don’t know how people were first infected by SARS-CoV-2, we do know it’s related to a family of bat viruses responsible for the 2003 SARS pandemic out of China.
DONT BLAME THE BATS! THE EARTH NEEDS THEM!
When a bat virus jumps to humans, it’s a warning that the ecosystem is way out of whack.
The deets:
Ryan recruited Reuters data journalist Allison Martell to help geolocate
95 documented spillovers 2002-2020 from that database. Researchers at the company provided us with a list of outbreaks of diseases in which bats are the reservoirs. Ryan and Allison mined studies, public health records and other documents to vet the list, expand on it and identify the locations of first known victims.
They then developed a machine learning model -- in consultation with scientists -- that analyzed 8 billion data points related to conditions around those spillovers, most derived from satellite data imagery drawn from Google Earth Engine.
The model included 56 covariates at a 25-square-kilometer resolution, including environmental, social and economic conditions such as land surface temperatures, precipitation, the estimated number of bat species present in an area and measures of deforestation and urban development, that scientific research has linked to spillover risk. (Bat species richness was calculated from IUCN Red List data.)
Based on the machine-learning analysis of past spillovers, we identified other places where similar conditions exist. We took those that scored in the 95th percentile and called them JUMP ZONES.
Jump zones = the highest risk areas areas on earth for spillover and potential starting points for future outbreaks, epidemics and pandemics.
For more info: tinyurl.com/Batlandsmethod
WHAT WE FOUND:
9 million sq km of high-risk zones in 113 countries.
A 57% increase in people living in areas at highest risk for spillover over 2 decades -- Nearly 1.8 billion people -- 1 of 5 on the planet -- now live in these areas.
99% of highest-risk areas globally are in lower- and middle-income countries -- mostly tropical locales undergoing rapid urbanization, much of it driven by wealthy countries’ demand for raw materials.
Since 2020, the last year of the analysis:
…all in geographic areas made up almost entirely of hotspots identified in our analysis -- including in countries with no prior known occurrences of the virus.
…all pathogens with potential to develop into an epidemic.
“GROUND-TRUTHING”* THE DATA
We used the analysis to zero-in on specific locations to better understand the changes that turned them into hotspots…. and to help to decide where to put reporters’ feet on the ground…
…or drones in the air…
…in order to report on local dynamics that couldn’t be seen from the satellites or measured in the data.
We collaborated with a dozen journalists to conduct field reporting on six continents.
The view from the ground in Ghana, where Marburg virus made its first known appearance in 2022.
Bat-eaten fruit and pulp underfoot in a deforeseted area where the virus infected a farmer and his baby.
IN INDIA
The data identified a dramatic increase in risk in the state of Kerala in the years leading up to Nipah’s first known appearance there in 2018.
The view from the sky on Google Earth Pro, showed rapid urbanization of once heavily wooded areas.
On the ground, Sreekanth and I found backyards full of fruit trees and a type of fruit bat known to carry Nipah roosting nearby in small patches of woods. They’d raid the gardens at night. Residents would pick up the discarded fruit to toss away or eat. Scientists told us the virus from an infected bat’s saliva could survive hours on a piece of fruit.
Throughout West Africa we found governments are granting mining concessions in ecologically risky areas without requiring companies to assess and address the dangers.
On the ground, Reuters reporters visited the site where scientists had found evidence of Ebola in bats near where a 2013 epidemic that killed 11,000 people began. Ryan used their phone GPS signal to locate the site on a major iron ore concession.
Biggest reporting challenge: Laos.
Jump zones more than doubled to 73%
-- a bigger % increase than any other country
That was accompanied by a 19% increase in tree loss.
To figure out what was driving those changes and to understand the risks they were creating, we needed to know what was happening on the ground.
The challenge: Laos is not a historically open country.
We mined every source we could find….
… to document how China’s demand for rubber and food crops were driving deforestation and spillover risk.
…allowing us to geolocate them on our risk map
A 2017 study found coronaviruses
in bats for sale at markets in Laos
and included a dataset with their locations…
…and show what risk looks like on the ground…
MORE
EXAMPLES
& TUTORIALS