Mining flood height data from social media

I think we can all agree that social media is a gold mine of data; one particularly underutilized area is extracting hazards-related data and making it useful for analysis.

Last September 12, two typhoons almost simultaneously crossed the Philippines bringing heavy rains in many parts of the country. This caused floods in many different areas. Naturally, a lot of people shared photos and descriptions of floods in social media. I noted that they did it mainly to warn others not to pass through a particular road, to tell people that they’re safe, or to ask others for help.

There were official reports from government agencies and traditional media, but I found that it was hard to pinpoint the geographic locations of the reports. They mostly had vague accounts on the location, such as the barangay (village), sometimes even incorrect geographic descriptions. Other times, they’d mention which streets are flooded, but fail to mention which parts of that street. Unless it’s catastrophic flooding, you’d expect only sections of streets to be flooded, and that’s the key information that we needs to be known,  more so if we have to commute to or from work or school.

When studying impacts of hazards like floods, the three most important pieces of information (at least for me) that needs to be taken into account are the location, flood height, and time. And when people post stuff in social media, oftentimes these three are already present.

So, I gathered photos uploaded by people in social media, mostly Facebook and Twitter, analyzed the photos to extract the location, get the approximate flood height, and time of flood.

Google Street View was invaluable in determining the location of the photo. Maybe I would have been able to do this without Street View, but it would have been much, much harder. And I’d probably get less data. I would have to rely on photos having street signs and really prominent landmarks, or be lucky that I know the place.


This one, for instance, in the famously flood-prone street España Blvd. in Manila City. The left photo was posted in Facebook, while one on the right is a screenshot from Google Street View. Although the Street View photo was taken three years ago, there were still enough clues in the photos to narrow down at which location the photo was taken.

After determining where the photo was taken, I took the geographic coordinates from Google Maps. If there are people around or objects that gives me a good sense of scale, I take note of the approximate flood depth. To make it easier, I divide this into five categories: ankle-deep, knee-deep, waist-deep, neck-deep, and higher than neck. Taking previous photo as an example, we can see that the flood reaches up to roughly a quarter of the car’s wheel (barely got into the photo), which would be more or less ankle-deep. It’d be nearly impossible for me to measure the depth in actual units, so this should do.


Here’s the map I produced from a whole day of collecting these data online. I was able to get data as far away as Atimonan and Lopez in Quezon province, southeast of Metro Manila. It wasn’t a lot though because there wasn’t much that showed up in my social media network. You can download a higher resolution version of the map here and the kmz file here. I saved the raw data in a spreadsheet, just comment below if you have any corrections or if you want to contribute.


Note: Image and data are distributed under CC BY 4.0 and ODC-BY 1.0 licenses, respectively.



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s