Air Quality Trends from Lichen Sampling

This first map shows US Forest Service lichen sampling plot locations (red dots) and wilderness areas (green shaded areas). There's a wilderness or some under those dense clusters way out west. The one out in the Atlantic is probably an error.

Lichen monitoring is an active and expensive process, and is not immune to regional and national budget priorities. Although resulting data are used for long-term air quality analysis, plots can go unvisited for years before we have a second sample. The series of maps below show plot visitation from 1980 to now. Nine results from 1973 to 1978 and one historical observation from 1900 are omitted here.

Because a single sample isn't a trend, we revisit plots to collect a series of observations over the decades. This second graphic shows all nitrogen airscore values by collection year (tiny dots), with lines connecting repeated visits to a plot.

It's a busy graphic, but we can see repeat sampling picks up right around 1993, and that database updates kind of stall out after 2017.

Now that we have shown everything at once, let's take a closer look at lichen around one wilderness area. Let's!

This map is the same as the first up above, but zoomed in on some arbitrary Forest Service wilderness area (the Stikine-LeConte). Lichen plots are still red dots/circles, with larger dots showing plots associated with the wilderness by USFS air quality specialists.

This area seems to have pretty good coverage, but for our measure trend we are looking for plots that have been sampled at least twice -- preferably once around the time the wilderness area was designated and once again recently. We'll skip the minimaps here and go straight to the graph.

The much less busy graph here shows the data for the lichen plots assigned to the Stikine-Leconte. As above, dots show the nitrogen airscore by year, and lines connect paired observations across sample rounds.

This wilderness area was designated in 1980, and we would prefer a trend from 1980. Sampling seems to have started in 1989, and there are two 1980s plots (TNF030 and TNF031) with repeat visits in subsequent decades (2004 and 2012). We'll use these as inputs to our trend test.

We compare our 1980s sample group to our 2010s group using a paired two-tailed t-test. The first graphic is a box plot that is only useful for more samples than we have here, if ever.

The second graph here is a quantile-quantile plot comparing the two sample rounds. More points would make this a more exciting visual, but we will never have more than two, because our protocol specifies matched pairs and that's what we had in the 1980s.

Below the graphs we have the trend and the p-value.