The Decline of Acid Rain in the United States
The inspiration for this graphic came from reading Bill Bryson's description of the threat of acid rain in his (otherwise amusing) account of his hike of the Appalachian Trail, A Walk in the Woods: Rediscovering America on the Appalachian Trail. As a nature lover, I wondered why one rarely hears about the effects of acid rain.
It turns out that the problem of acid rain has been largely solved by eliminating the pollution that gives rise to it. The National Atmospheric Deposition Program (NADP) has roughly thirty years of data collected through its National Trends Network (NTN). As can be seen from the graphic, the pH3 of rain in the U.S. has become far less acidic, on average, over the past 2 decades.
Weaknesses of the VisualizationThere are three primary remaining problems with this graphic in my view:
- There is more data (such as precipitation data) in the source data that I could not find a good way to include.
Initially, I planned to map precipitation quantity to the size of the circles and not color the states, but I felt the resulting graphic was misleading. One tends to interpret the amount of "ink" as the amount of precipitation, but the amount of ink is as much a function of the number of detection stations in that scenario as it is a function of precipitation. I think one could include such data as another dimension by having a uniform grid of dots whose radius corresponds to a kernel density estimate of precipitation, but I did not try this.
- The effect of the measurement stations is somewhat distracting in the animation. It reminds me of camera flashes.
I think it's best to include the measurement station data because it gives the viewer some idea of the variability within states (particularly large states), even if it is a little distracting when the values are rapidly changing.
- I'm not sure I have mapped pH to color in an optimal way.
Even after experimenting with a lot of options, I'm not totally happy with the way color works in this plot. It's natural to use a diverging scale for something like pH. But pH is a log scale so I feel the current coloring under-emphasizes big differences in pH. However, if you map the color directly to ion concentration, everything but a few extreme measurement stations will be washed out. Even on the log scale, the measurement stations report much more extreme values than the state averages, so the chloropleth part of the map doesn't take full advantage of the color scale and the overall message is somewhat obscured. Perhaps the answer to the measurement station/color issue is a kernel density plot, but real kernel densities don't play nicely with vector graphics.
I am curious about what's giving rise to the handful of basic pHs measured. Seawater is somewhat basic, but I don't know what's making the rainwater basic.
If you found this graphic interesting, check out the recommended reads.
AcknowledgmentsThanks to the NADP for making their data public. This visualization was created using R and D3. The following packages were particularly helpful:
Notes1 Alternatively, one could weight upon precipitation, or take the minimum pH. I don't know enough about the biological effects of acid rain to make an informed choice, so I went with the simplest option. I don't think it makes much of a difference for visualization purposes.↩
2 Charlson, R. J., and Henning Rodhe. "Factors controlling the acidity of natural rainwater." Nature 295.5851 (1982): 683.↩
3 pH is the negative base-10 logarithm of hydrogen ion concentration. The key word here is logarithm, meaning a drop of 1 pH unit corresponds to a 10-fold increase in acidity.↩