Maduro’s Land of Darkness Redux: Documenting A(nother) National Crisis with Remote, Alternative Data
Just before 4pm, 22 Jul 2019, the lights went out across Venezuela. That much the world knows. But getting accurate information on exactly where and when the outage hit, and what happened next is non trivial in Maduro’s Land of Darkness. Here, we document, for the public record, the spatio-temporal pattern of the electricity outage and subsequent recovery as observed by the Monash University Global IP Observatory.
Darkness Again in Venezuela
Sadly, here we are again.
Back in March, our team at the Monash University IP Observatory, started providing near real-time observations of Venezuela’s IP (internet protocol addressing) space. With rising tensions, we were concerned that the incumbent regime may be tempted to tamper with the nation’s internet infrastructure, as has been documented time and again for their blocking of various websites at certain, strategic times.
What we didn’t know then was that we would be one of several international organisations who would bear witness to the tragic, nation-wide electricity crisis that engulfed the country from around 5pm, 7 March 2019, and rumbled on for the rest of the month and into April.
Now, as this week draws to a close, we are once again sending observations to our friends in Venezuela and abroad, providing granular observations on internet activity across the country to help give some light amidst the darkness.
With the regime once again making various claims of the cause for this week’s outage, following the rule book from last time, we seek here to provide citizens, scientists, engineers, and community leaders with our observations in space and time of the outage and recovery for their expert review and analysis.
In this post we provide:
- A dynamic visualization created from millions of observations made by our global measurement platform, showing, at state level, and hourly granularity, the moment the outage occurred and then the differential recovery patterns across the country;
- A tabulation of the impact period for each state, in hours, providing an externally established, empirical baseline for comparison; and
- An open-sourced data-set of our connectivity measure (explained below) across all 13 states depicted in our dynamic visualization at hourly granularity.
The Dynamic Pattern of Failure & Recovery
The GIF below is a visualization of our data from 0am, 22 July 2019, approximately 16 hours prior to the outage.
Each location is the centroid of each of 13 states across Venezuela we observed through this period. The colour and diameter of each disc represents the degree of internet ‘connectivity’ (explained in more detail below) we observed from millions of remote measurements. The legend at right gives the interpretation of the colours in connectivity index terms.
In addition, the wonderful Kepler.gl framework we used provides the average and variance of the connectivity measure as a time-series line plot at the bottom of the figure, with the time scrubber showing the position, in local time, of the visualisation as it proceeds.
Immediately obvious is that the state of Bolívar was largely unaffected by the outage event. Venezuela’s power largely comes from the Simón Bolívar Hydroelectric Plant (Guri Dam) in this state, and so one could draw the conclusion that the plant was not at fault, nor were the local connectors. However, the adjacent state, Monagas, suffered immediately, and was one of the more impacted states (see table below), suggesting that the transmission problem was felt almost immediately beyond the state of Bolívar itself.
In addition, in the recovery phase, it is easy to pick out the spatially related, lagging states of Táchira and Barinas.
Finally, we note that (although text is overlapping), aside from Bolívar, it would seem that Miranda state, just to the East of the Distrito Capital, was the least impacted by the outage, despite neighbouring states suffering.
Tabulating the Impact
In addition to visualising the impact, we can also do some simple tabulations. In the following table, we sum up the total hours that the connectivity measure for a given state was at or below 70. Given that during pre-outage conditions, all states obtained connectivity above this level, we can use the time spent below this threshold as an indicator of the impact time for the state.
STATE IMPACT DURATION (H)
________________________________________
'Bolívar' 0
'Miranda' 10
'Anzoátegui' 20
'Carabobo' 21
'Aragua' 23
'Distrito Capital' 23
'Nueva Esparta' 23
'Lara' 24
'Mérida' 30
'Táchira' 30
'Zulia' 33
'Monagas' 36
'Barinas' 43
________________________________________
MEDIAN 23
AVERAGE 24.3It is clear from the tabulation that the ‘typical’ outage impact experience across states was of the order of 20–24 hours. This band captures 6 states.
Outside of this band are the two states mentioned above — Bolívar, with the hydro plant, and Miranda, to the East of the capital district —recording 0 or just 10 hours of impact by this measure.
At the other end, we can see three states distant from the hydro plant, Táchira, Zulia, and Barinas, and one state adjacent to the plant, Monagas.
Full data are available on our open-sourced data page housed at Git.
We leave it to the expert scientific community of Venezuela and abroad to make further insight from these observations.
Methods | Connectivity
To generate the data behind these observations, we combine a commercially available geo-located IP database with our powerful scanning technology which measures the online or offline status of millions of Internet addresses globally every hour.
Our observational methodology uses the most basic Internet messaging protocol that is widely used billions of times a day to establish routes for your email, tweet, or share. After developing a carefully selected set of Internet addresses (IPs) to measure, we periodically send them one of these tiny messages, essentially asking, ‘Are on you online?’.
Since most online devices require electricity to run, measurements of online activity serve as a remarkably good proxy for the availability of electricity: when the device has power, it is typically found to be ‘online’, without power, no answer is received.
The technology behind the IP Observatory regularly counts the number of online internet addresses such that, at any time, the level of online addresses can be compared to a long running baseline. For simplicity, we create an index which takes a value of 100 at the baseline online count for that jurisdiction. We call this index ‘connectivity’.
The mission of the Monash University IP Observatory — ‘internet insights for social good’ — is to monitor the availability and quality of the Internet during critical events such as elections, natural disasters or conflict to provide. The observatory was founded by Klaus Ackermann, lecturer in Econometrics and Business Statistics, and Simon Angus, and Paul Raschky, Associate Professors in Economics. The observatory is a project of SoDa Laboratories at the Monash Business School, and tweets @IP_Observatory.
