Lebanon COVID-19 Lockdown Based on Bad Data — Here’s what to do.
In Lebanon, a COVID-19 risk is calculated for every town. The larger the number of cases per 100,000 people, the higher the risk. Once a town reaches high risk levels, an order is issued for a local lockdown. These are displayed on the following map created by the Lebanese COVID-19 taskforce.
This data-driven approach may seem effective. Yet we’ve discovered inaccuracies in the data used to make these analyses. In one of our first articles by OpenMapLebanon, we’d like to explain the scale of the problem and what can be done to resolve it.
The Lebanese population is overestimated by est. 50–75%. This means real risk levels are much higher than currently estimated.
When summing up the population estimate numbers used to calculate COVID-19 risks, the total is equal to 10.6 million. This shows a population estimate that is much larger than any estimate currently available.
Here is how the data looks like in comparison to the best and highest resolution population density dataset that we could access:
More information about the Facebook dataset can be found here.
The Facebook dataset used here has its limitations and it’s used for illustration purposes. It does not substitute a real census where individuals are counted directly. The Facebook population estimates are not people registered on Facebook, the total population figures are based on government and internationally reported datasets. It is based on high resolution satellite imagery, and uses state-of-the-art artificial intelligence. A population density is then estimated from the buildings detected. Then, it is calibrated by the latest census data from the Lebanese Central Administration for Statistics which you can find here. (Labour Force and Household Living Conditions Survey 2018–2019 Lebanon).
We also note that the boundaries in the map used here leave some municipalities out. Also, daily COVID-19 reports from different government agencies are using different maps. This is a separate issue that needs to be addressed. Here’s how the map looks like with the current population estimates:
If the base population estimate is flawed, this means that the risk calculations might be under or over estimated — since it’s calculated by dividing the number of COVID-19 cases by the population of an area. Comparing the Facebook population estimates with the current density map shows the scale of the problem.
The Caza level analysis shows us how many of these areas are overestimated — although some are underestimated as well. This is a tough situation as it casts doubt on the risks described in each area. This leads to seriously underestimating the risks of each region.
Your town in Lebanon might be at a much higher risk than reported.
To support the COVID-19 response, citizen volunteers need Open Data. But data is hard to reach. We want to fix this.
So far, COVID-19 data has been released as pdfs and has been hard to find and access. This means it’s expensive and difficult for scientists, academics and volunteers like us to better support the response.
This is why OpenMapLebanon has begun the work to standardise, clean and publish Lebanon COVID-19 timeseries for open access in more useful formats. This work has already started and we’re lucky to be supported by great partners.
So what does this mean?
Simply put, the risk levels displayed are flawed. This not only means that the lockdown policy is driven by bad data, but the public is misinformed. The COVID-19 situation is probably worse than described so far.
The Lebanese people must understand that the COVID-19 risk is much larger in many areas than what is reported. The chances of spreading the virus, increasing the number of cases and hospitalisation are likely higher too.
Reliable data must be the basis that drives COVID-19 measures. There are human costs due to bad data and inaccessible datasets.
The Lebanese government, agencies and municipalities must invest more in making reliable data openly accessible. This is the only way for expert citizens, volunteer scientists and Lebanese expats to better support.
We are a growing community of volunteer technologists, mappers, academics and data scientists who are supporting the relief effort in Beirut and believe that there are obvious benefits to open & safe data sharing. www.openmaplebanon.org