Green = Healthy? Think again. Using data on scenic values as an indicator for human health

Seresinhe, C. I., Preis, T. and Moat, H. S. (2015) “Quantifying the impact of scenic environments on health”, Scientific Reports, 5, 16899

Seresinhe, C. I., Preis, T. and Moat, H. S. (2017) “Using deep learning to quantify the beauty of outdoor places”, Royal Society Open Science, 4, 7, 170170

 

Is greenness really a good indicator for health, or is there a missing link?

Often people assume that living in urban areas is not good for their health. This is true from the standpoint of pollution levels and risk factors associated with traffic. In contrast, people living in rural areas can enjoy cleaner air, less traffic, and wide-open, green spaces. In Great Britain, researchers set out to ask whether these open natural, green spaces are related to human well-being in terms of health. To do so, they obtained nationally available datasets on self-perceived health status from the 2011 Census for England and Wales. The researchers mapped this data along with landcover data from a Generalised Land Use Database, classified into how “green” areas are (Figure 1).

 

The missing link: “Scenicness”

The researchers made use of an online game galled “Scenic-or-Not” that presents people with geotagged photos that can be rated on a scale from 1 (not scenic) to 10 (very scenic). The photos used in the study were obtained from Geograph, a website that collected geographically representative photos of every square kilometer in Great Britain.

With over 1.5 million ratings on over 200,000 photos, the authors mapped “scenicness” and included the finding, along with greenness, in computer simulations to predict health. Surprisingly, scenicness was a really important factor in predicting health in all areas, whether they were urban, rural, or suburban (Figure 2)!

To make sure that pollution in urban areas or variables such as income did not influence how healthy people perceived themselves to be, the researchers included these factors in their simulation. Even then, it was confirmed that scenicness was the best predictor of health across Great Britain.

 

 

So, what makes an area scenic?

It turns out that not all areas that are “green” are always scenic. The researchers wanted to know why that was and identified features in the photos, giving them clues as to what would be rated as scenic or not. People seem to have a preference for areas where there are a variety of colors (blue, brown, green), such as areas with mountains or water features, whereas large open fields of green (think athletic fields, or grass) do not rank highly on Scenic-or-Not. In cities, large proportions of roads or buildings in a photo, usually represented as grey, are also associated with less scenic ratings. This is even the case when some green (trees) is in the photo. In contrast, urban areas can also be rated as scenic when certain types of buildings are in the photo. This can include cottages, aqueducts, churches, etc. (Figure 3).

Having discovered this phenomenon using data simulation, the researchers went on to use something called a neural network. Essentially, a neural network is a kind of computer learning that is built on the human brain. In this case, the neural network learns to identify how scenic a photo is without having to have people rate the photo on Scenic-or-Not. This neural network can identify features in a photo (mountains, lakes, rivers, ponds, coasts, parking garage, grass, office building, etc.) based on “training” with over 8 million photos. As with humans, practice makes perfect – or pretty close, anyways! After that, the researchers were able to use the neural network on over 200,000 photos taken in London. They were able to identify what exactly makes up scenic photos and rate them without having to rely on crowd-sourcing (Figure 4).

Relating this to policy and environmental protection

It seems that scenicness motivates us to go outside more, exercise more, and walk around more. Knowing this and having the ability to use these computer simulations that learn how to interpret photos gives us a chance to predict which areas are beneficial to humans in terms of health. These simulations are more powerful than traditional methods in many ways, as they allow for a much greater amount of data to be processed than would be possible by human encoders, but are also significantly cheaper than sending field crews to remote locations to survey scenicness. These large data analyses are important as they reflect national trends and can easily be picked up by policymakers to decide which parts of a country are valuable and need to be protected. This research also helps to identify areas that are not perceived as scenic (and might also have inhabitants that perceive their health to be poorer). With this information, we can put efforts towards improving the scenicness of less scenic areas so that local residents can enjoy them.

So, go out there, take pictures of your surroundings and do your part in environmental conservation!