The View from the Crowds

A UC Berkeley scholar uses Flickr to study what the public considers scenic.

By Lydia W. LeeLAM-Jan2013-Interview-HalfDome

Even though Alexander Dunkel, Student ASLA, has never visited the High Line in New York City, he can tell you exactly what part of the park is the most popular: the 10th Avenue Square. How? He spent a year analyzing Flickr, the popular image web site, and seeing where people take the most photos. Because many of the images in Flickr collections are tagged with their precise geographic location as well as a descriptor (“Golden Gate Bridge,” for example), Dunkel was able to generate maps of an area’s most frequently photographed subjects. From his home in Dresden, Germany, he spoke about his research at the University of California, Berkeley, which won a 2012 ASLA Student Honor Award.

What inspired you to study Flickr?

Flickr is a unique source of data that shows how people interact with the landscape. Some people take pictures all the time, some people only take a picture of things that are really important to them, but if you look at the whole data set, you see what the majority opinion is.

I was inspired by Eric Fischer, who works at Google and in his free time creates maps based on crowdsourced data. He had done amazing maps based on Flickr, and he helped me a lot.

How did you get the information out of Flickr?

Flickr has a very well-documented API (application programming interface), so it was easy to start writing scripts to pull information out of it. I’m a landscape architect, not a programmer, so I used Microsoft Visual Basic. There are a lot of parts that you can use from other people and put together based on your needs.

Why is this data of interest to you?

In Germany, geographic data is supplied by the government and has a lot of licensing problems. We have to pay for the data. And there’s data that you can’t get from the government, including “soft” data such as personal opinions, up-to-date data, and data that doesn’t exist on a very local scale. Flickr is free, and it contains information that you can’t get anywhere else. I’ve also started to look at, which is a sort of Wikipedia of geographic data.

What are some of the interesting observations that emerged from the specific areas you investigated?

When you look at my map of Treasure Island, you can see that people most often stop at one point and take a picture of the San Francisco skyline. Then some people step over to the eastern side of the island and take a picture of the Bay Bridge or a reflection in the water. So Flickr is telling you a lot of specific things about how people experience a certain place. Likewise, if you look at the map of the Berkeley Marina and the pier, you see that when people approach the pier, the pier itself is the most dominant image, but once people step onto the pier, they see the Golden Gate Bridge, the western hills—other things start to influence their perception.

It’s also possible to generate historic maps, like my map of the Chelsea district in New York before and after the opening of the High Line. The difference is really amazing. By slowing down the pace from driving to walking, the High Line has given people the opportunity to explore the area and experience the entire neighborhood in a whole different way.

A skeptic might say that you’re just confirming what people find the most stereotypically photogenic—true?

Ninety-nine percent of the time, if you know a place, the data will display what you already expected. But you wouldn’t normally have data to back up your personal experience. This way you can visualize your thoughts and show proof of what you already know.

Can you provide an example of how this real-world data could be useful for landscape architects?

In downtown Dresden, my hometown, there is a famous historic church. I was walking by there two days ago and there were 20 people all standing in the same place taking a picture. From the Flickr data, I can prove that most people take pictures of the church from the same angle. If I were to make a master plan for Dresden, I would normally use GIS viewshed analysis to determine the important sight lines to this church. But digital viewshed analysis does not reflect the true sightings of the church, and it doesn’t take into account trees or other buildings; sight lines based on Flickr photos would.

More generally, as a designer, there is a great possibility to react to the data in your design. If I know there is a really beautiful spot and it’s getting ignored—I see that 90 percent of the photographers are taking a photo here but it isn’t the best spot for taking a picture—I can design better access to the spot over there. It’s a base layer for making decisions. Once you make changes, you could look again at the data and see if photo taking has changed based on your improvements.

How might this data affect your own practice?

In my research at Dresden University, I work on projects for the government and the European Union. The exact equivalent of my job doesn’t exist in the United States, but it’s comparable to regional planner. Right now I am working on the cultural development plan for all of Germany; a master plan for how to combine new energy sources like wind and solar with landscape development on a national level over the next 10 years. While my thesis work on Flickr isn’t directly connected to this research project, you can imagine how Flickr could help us identify important spots that many people take photos of, so we would avoid building energy lines there.

Where would you like to take your research next?

My future research will focus more on how to communicate with people. I’d like to bring people into a design project and have them participate actively. I want to combine different data streams like Twitter and Flickr and allow people to connect through their favorite platform to make design a more collaborative process. At the start of creating a new master plan for a harbor, I can imagine putting up a web site and asking people to submit pictures of their favorite spots in the area, to tweet about them, and to check in at Foursquare venues, and then combining all this information into a map. Technologies are developing so fast, I think it will be easy to create these kinds of project sites in the near future.

Lydia W. Lee is an architecture and design writer based in the San Francisco Bay Area.

Leave a Reply