New investigation i've been working on for a while for
@bellingcat
, identifying almost 180 ghost ships loading grain at a sanctioned port in occupied Ukraine. This potentially ammounts to millions of tons of illicitly exported grain.
New: Bellingcat has been given access to over 140 satellite images showing activity at a Crimean shipping terminal that some reports have claimed has been enabling the transportation of stolen Ukrainian grain by Russia.
Turns out lots of missile defense radars interfere with open source satellite imagery when they're turned on. I built a tool that lets anyone monitor when and where these radars are deployed.
Ships are required by law to broadcast their poisition and identity via the Automatic Identification System. In the image above, only four ships had their AIS turned on. To get around this, i built a tool that uses radar satellite imagery to identify ships
I built a tool to conduct damage assessment in Turkey:
Worked with an amazing team the Economist to provide analysis for the story below. It uses an algorithm i've been developing to detect persistent changes in Sentinel-1 imagery.
Just draw a box over an area of interest, and it will show you the dates a ship was present (and even estimate the length). Here's a fun one where you can spot the sinking of the Russian cruiser Moskva, which used to be docked in Sevastopol
Did some more analysis for
@TheEconomist
, this time doing damage detection across Ukraine on the anniversary of the war.
Again, here is an interactive tool allowing users to explore destroyed buildings across the entire country:
Want to learn how to carry out Bellingcat-style research? We're hosting an in-person workshop in London on November 20th-24th. Find the details and how to sign up here:
How has
#Lockdown
in
#London
impacted mobility? I made some timelapses using anonymized mobile phone data from
@Mapbox
showing the massive drop in movement as the first lockdown was put in place:
just finished my second week of teaching online not realizing that i have a bookmark called "chicks" with a webcam icon next to it. How do i subtly explain next class that this is a live stream of my chickens?
A project I’ve been working on for a while is out now as a CSAE working paper! It explores the relationship between droughts, irrigation, and insurgency in Southeastern Turkey. 1/11
Been working on this one with
@wammezz
for a while now, finally out in
@bellingcat
! We use open source data and tools to identify major sources of pollution in Iraq. Spoiler: there are a lot of them.
Iraq has huge oil reserves, but it has also been blighted by rapidly declining air quality & oil pollution in recent years
How can open source tools help us monitor these issues & pinpoint who is responsible for pollution? New from
@wammezz
&
@oballinger
Just got back from
@glastonbury
, completely awestruck by the fact that they just build a city of 210,000 people in a field over the course of a month. Slept, showered, and headed over to
@planet
to make a timelapse.
If anyone's bored and wants to spot Russian military hardware on the Ukrainian border there are a bunch of interesting signals in 2021 (red dotted lines):
Very good to hear BP is looking into carbon offsetting at
#COP26
! Let's have a look at what they're already doing for the planet in one of their largest production countries, Iraq
With
#Election2020
tomorrow, we're about to be bombarded with very bad maps. Here's a thread on why most election maps are misleading, and how satellite images of the earth taken at night can help fix them.
New paper out with
@wammezz
in the International Review of the Red Cross! We look at how the growth in computational power and open-access data can be leveraged to protect the enviornment in conflict zones.
this is a reAL FIGURE PUBLISHED IN NATURE. The article is actually very good but my god i didnt know we were allowed to use the chad/virgin meme template for research purposes
What the heck is this beam on May 3rd right next to the Azovstal plant in Mariupol? (47.114804, 37.59192) visible in
@planet
imagery taken at 08:20 (but not at 08:13).
40 cases of blood clots in 17 million people who received the vaccine corresponds to odds of 1 in 425,000. The image below has 424,999 blue pixels and one red pixel. Before worrying about the odds of randomly picking the red pixel, can you even find it?
A reminder that ALL
@googleearth
engine app source code is publicly accessible, whether or not you publish the code itself. Good to keep in mind as people build OSINT tools/projects on Ukraine.
@JackDetsch
@EliotHiggins
will be fun to spot it when it arrives. Looks like Poland has turned on a patriot battery near the border with Ukraine, my best guess is that it's at the Rzeszów-Jasionka Airport
;
I urge everyone to go through the state department’s bizarre Instagram feed, which has been posting extremely combative content about Iran non-stop for the past three years
Made a sick gif for the course i'm teaching this fall to illustrate the The Central Limit Theorem. I use individual-level U.S. income data as the "population" (i know. shh).
Everyone knows that regular
#election
maps are full of huge empty precincts that visually overstate rural vote shares. Here's how the 2016 map would change if we only displayed the vote shares of populated areas:
Planes appear as red, green, and blue dots in satellite imagery because of slight differences in the timing of the imaging of each band. The plane is heading around 70 degrees north.
One thing to watch for tomorrow is how
#BidenHarris2020
fare relative to Hillary in these suburban areas early on.
I wrote full sentences about this here:
You can play around with the maps yourself here:
Interesting find of the day: the Kanarya neighbourhood of
#Istanbul
is conspicuously missing from google street view at exactly the polygon boundary for the neighbourhood. Wonder why?
Journalists saw the precinct maps (blue coasts, red heartland) and assumed that Trump won thanks to rural voters. It took years for this to be debunked and for people to realize that his main base of support is suburban voters. This is clear in Populated Area Voteshare maps:
Methodology paper is currently under blind peer review for a conference but will share ASAP! Also caveats: accuracy will improve as more imagery becomes available; accuracy for areas with less coverage (northern syria) probably lower than e.g. kahramanmaras
Lets take a closer look at what's going on using Dallas as an example. The precinct map makes it look like
#Trump
won on the basis of rural votes. When we filter voteshare using nighttime lights, we see that Trump's "rural base" is actually suburban.
WHAT THE HOLY F*CK?! Steve Bannon calls for the ***beheading*** of Dr Fauci and FBI Director Wray as a warning to federal workers. This video is real. Captured by
@peltzmadeline
.
Some gems: the last map of the Soviet Union (March 1990), and the first map of “Russia and the Newly Independent Nations of the Former Soviet Union” (March 1993). Gorgeous use of transverse polyconic projection
Geopolitical analysis from my Turkish mom (her side of the family lives on the Bosporus). We used to wait for Russian navy ships/subs to pass by before jumping in to swim, they made the biggest waves!
Similar trend in Nevada, where the precinct map makes it look like a Trump landslide even though Hillary actually won the state. Populated Area Voteshare shows that the whole state is mostly empty, and the real battlegrounds were Reno and Vegas
Spooky find of the day-- the defacto standard benchmark dataset for facial recognition studies is full of junk. Working on a paper involving clustering faces across images and was getting weird accuracy statistics...
Want to monitor air pollution from gas flaring in iraq? Detect rare earth mining in Myanmar using AI? Identify oil spills in Northern Syria from space? We run through the data sources and methods you'll need. All open-source.
To answer this question, I collect data on topographical and climatic factors, nighttime lights, infrastructure, tribal control, >3m ballot-box level election results, conflict incidence and insurgent recruitment, and measure irrigation using satellite imagery: 4/11
Likely geolocation of last week's
#Russia
-
#Belarus
training exercises: RIT picked up two radar signatures in the West of Belarus, both on Feb 14. Offensive military radar systems were a key component of the 2014 war in
#Ukraine
. May be worth keeping an eye on Belarus. 1/8
For each histogram, I take 10,000 samples of size n from the "population" and plot the distribution of sample means. Even though the underlying population distribution is very skewed, as sample sizes increase the distribution of sample means tends towards the normal distribution!
The exact coordinate for the Pyramid of Giza is 29.9792458°N and the speed of light is exactly 299,792,458 meters per second.
Aliens or time travelers or coincidence?
Whats going on here? Using the fact that populated areas are lit up at night, I isolated these areas using nighttime satellite imagery of the earth. More people=more light. This is similar to
@kennethfield
's dasymetric map in
@NYMag
, but retains important spatial info
If
#Beirut
explosion was caused by 2750 tons of ammonium nitrate, it was probably equivalent to between 1.2 and 2.2 kilotons of TNT. ~7-14% the size of Hiroshima
Stunning picture from the August 1992 issue on Kurds in which Christopher Hitchens spends three months in the mountains interviewing Öcalan, Barzani, and Talabani.
With another round of
@ucu
strikes coming up next week, a reminder from the ghost of Christmas future about what happens when academic labor isn’t organised and casualisation is allowed to go unchecked
How is this real?
"The Department of Chemistry and Biochemistry at UCLA seeks applications for an Assistant Adjunct Professor on a without salary basis. Applicants must understand there will be no compensation for this position."
An important caveat: this effect is wiped out in areas with highly unequal land tenure. Areas such as Batman actually see *increased* conflict following irrigation, as rival Kurdish clans compete over a new source of rents 10/11
big shoutouts to
@EoghanMacguire
for somehow turning this into a coherent article,
@gena_d
for the awesome date counter module in
@googleearth
engine, and
@sdbernard
for documenting his
@FT
wildfire smoke map which inspired the SO2 figure!
Are you an experienced researcher in spatial data science (broadly defined) with exceptional organisational skills?
We have a grade 8 fixed term lecturer opening
@CASAUCL
to run the dissertation module with guidance from me.
Apply by 26 Nov/start ASAP:
lmfao i used the twitter API to download tweets containing the N-word in the past 10 days and surprise surprise!! Number of N-word containing tweets per hour increased >10x since elon tweeted "the bird is freed"
More caveats: Is industrial monoculture farming sustainable? Is flooding villages and heritage sites good? Can the PKK simply be irrigated away? No. Is shielding Kurdish farmers from drought a better strategy to win hearts and minds than setting villages on fire? Probably. 11/11
@wammezz
I made a GEE app a while back that puts VIIRS/S2 IR/S2 NC together to spot/verify burning wells: . may be of interest! Here's the full blog post about IS attacks on oil infrastructure:
I estimate the impact of irrigation on conflict using two sets of models; a cross-sectional instrumental variables approach exploiting exogenous topographic variation in the placement of irrigation schemes, and spatial panel models that rely on the timing of irrigation. 5/11
@MapMakinMeyers
been looking through your open data catalogue, amazing work! I've got some village-level shapefiles for Pakistan kicking about if you want em: