What do trees do when we are not looking?

Douglas Sheil CIFOR

Getting to the root of the dos, whys and workings of trees can be an obsession for forest researchers. And for my fellow obsessed- pinpoint accuracy is our common ambition. So why is this so hard to do?

Accuracy is essential if we are to correctly understand the many details governing how forests, and their vast stores of carbon, behave in a wide range of circumstances. Such accuracy is also vital outside of research where, for example, payments to forest owners and others are based on how much carbon is stored over time.

But, however much care is invested in the tree measurements, there are still challenges in using data to make accurate assessments of forest biomass production and loss. In our recent article, my colleagues Professor Takashi Kohyama, Dr Tetsuo Kohyama and myself describe, share and illustrate how such estimates can be improved.

Accurate measurements of how forests change takes a lot of time and care. Generally, the most precise ways to assess how forests change over time is to measure large numbers of individual trees and then wait for a year of more and measure them again—and this is how most detailed science on forest change is conducted. But tree measurements are not the same as biomass measurements, and there are a lot of calculations and assumptions that go into using these data to estimate biomass change.

One specific challenge is that with intermittent observations we have to infer what happens in between. For an analogy, imagine you wanted to know the average number of students in a classroom over an eight-hour day and took a roll during two different classes: you see some pupils are the same and some are different. You would want to know what was happening, when you were not looking—what happened for example during the lunch break? Similar uncertainties arise when we count and measure trees—they don’t go out for lunch (or do they?), but we don’t know exactly what has been happening in the period between our measurements—and these uncertainties increase as time between measurements increase.

Such unknowns have consequences. Indeed, up until now there was no agreed and simple way to make comparisons between changes in one forest recorded over say two years and another over say ten. Rather, until now, various rather clunky fixes have been devised and applied. These create uncertainties over interpretation too: when values differ we cannot be sure to what extent these differences reflect the forests or the methods.

In our article, we provide worked examples from four forests (in Indonesia and Japan) to illustrate the improved accuracy that arises by using these approaches. Our examples show that the differences due to using standard methods, versus ours, are typically in the order of between one and seven percent over a five year period – with the biggest differences arising in the Sumatra forest data likely due to the broad diversity of species with different behaviours.

Do these errors matter? Seven percent doesn’t sound a lot—note that we only looked at four locations over short periods and much larger errors are possible in theory: especially over longer periods and when forests are diverse or when vegetation is changing. But for careful researchers who really want to pin down how a forest is changing as accurately as possible even a one percent discrepancy could be a concern—it would certainly be something to avoid when possible. Also, imagine you are a forest owner hoping to get paid or the buyer who wants to know what you are paying for—seven percent more or less is not negligible. Ideally all studies and payment systems would agree on the best way to make these calculations or it could be a cause of confusion and conflict. Ideally any such calculation methods would be as accurate, reliable, and consistent as possible. Our approach seems to be the best claim to such a standard, and thus provides a solution.

We have built on our previous work and intricately examined these challenges and described the theory and calculations behind our improved estimation approaches. Our revised methods will be especially valuable when accuracy is important, when there is a need to compare studies over different durations, when the forests in question includes trees with very different rates of population replacement, and where composition is changing.

We also provide the equations and code for running in the free number-crunching software ‘R’ , to allow others to apply these methods. Given the pivotal role of forests in capturing carbon from our atmosphere – we know that many should welcome our improved accuracy in assessing this role. In any case, at least some of the measurement problems that have faced forest researchers are now fixed.

Improved accuracy should permit better assessments and ultimately, better science. And to my fellow obsessed- please use this improved approach…build on it…feed it with your own data…after all, we can’t beat the ravages of climate change unless we work together.