Testing trends in global N cycling

Some of the best evidence that N availability has been declining globally.
Standardized patterns of wood d15N for forests across the US. From McLauchlan et al. 2017.

Because we monitor, weather all around the world, we can detect changes in global temperatures and precipitation patterns.

Because we monitor gas concentrations throughout the world, we can quantify the increases in CO2 concentrations as well as other gases like methane and N2O.

But what about the N cycle? Has N availability to plants been increasing or decreasing throughout the world?

We really have no idea. There is no global N monitoring network. We cannot tell if plants have been experiencing increased or decreased N availability over the past, say, 100 years.

Considering how crucial the N cycle is to plant productivity worldwide, it seems important to have some index of whether N availability has been going up or down.

Through different projects over the last decade or so, I've been involved in trying to reconstruct N availability over time. Most of these have involved N isotopes in one way or the other, examining patterns in herbarium samples, tree rings, and sediments.

Most of these papers have shown declines in N availability over time.

There's one study left that I never got to, though. And it's been bugging me for about 5 years.

That's seeing if we can see a trend over time in foliar N isotopes at the global scale.

Most simply, if N availability has been increasing globally, all other things equal, d15N should be increasing. If N availability has been decreasing globally, all other things equal, d15N should be declining.

The last time we synthesized global foliar 15N patterns, we stopped data collection in 2006 and had about 12,000 data points. In order to update the database, I've read through about 500 papers that were flagged as potentially having appropriate data. About 250 did. So, Andrew Elmore and I started sending out emails to assemble the new global database.

With the new synthesis, we're now over 40,000 data points assembled so far with a couple dozen more datasets left to incorporate.



Given all the variability that exists at the global scale, we'll likely need all 40,000 data points to detect any trend that might be there (or to be sure there isn't a trend).

As Andrew and I talked about how we wanted to conduct this study, we wanted to make sure that the analyses were trusted. We wanted to preclude any criticisms that, for example, we left out data that didn't fit a given desired outcome**.

**Although most of the other patterns we've uncovered support declining to stable N availability, any of the three outcomes are equally interesting and important, as long as we have strong enough analyses to be certain about the patterns.

To make this project as trusted as possible, we have pre-registered our data analysis plan with the open science framework. See https://osf.io/thnyf/#. This means all of our hypotheses and analysis plans have been declared ahead of data collection. The R code that we will use to analyze the data has also been written and posted. This pre-registration is important in that it locks in our approach ahead of time and eliminates many forms of scientific bias that can misrepresent results. This transparency will be important to creating trust in whatever results we find.

We should be finishing up data acquisition by the end of the year, given the current pace of people getting back to us (which has been pretty good--over 75% of the requests have been answered).

One neat thing about writing all the R code ahead of time is that once we have all the data, we just push the big red button, and we'll see our answer.

Something to look forward to in 2018.

McLauchlan, K. K. et al. 2017. Centennial-scale reductions in nitrogen availability in temperate forests of the United States. Scientific Reports 7: 7856.