All over ecological research, inferences are being made from related yet indirect measures.
Some of it is explicit, like in applied research where surrogate measures are used to represent a true attribute of interest that is too difficult or costly to measure directly. However, some is implicit, like in fundamental research where ‘productivity’ is almost exclusively inferred from other measures, such as chlorophyll A, leaf litter biomass, or remotely sensed ‘greenness’ metrics.
‘But hold on’ I hear you saying….. Those are the well-established measures demonstrated to accurately represent productivity.
Sure. But to generalise that there is no context-specificity, variability, uncertainty or error in how well your measure represents your target (1) mischaracterises the risks associated with indirect measures, (2) does not consider that the inference could be wrong, and (3) may result in over-simplification and a false-confidence of understanding.
Even for the measures used to represent productivity.
In an essay just published in BioScience, my co-authors and I discuss how many standard approaches for measuring ecosystem properties and processes are in fact surrogates, in order to highlight the broader usefulness of a whole body of literature for evaluating variability, error and context-dependency in surrogate-target relationships.Ecological surrogacy is typically only considered as something for applied research on conservation monitoring and management. This is where time and resources are most limiting, both in terms of the capacity to collect information and use that information to inform management decisions. As a result, you can’t always collect the most ideal measures. And therefore, evaluating how accurately one measure relates to a different target – along with acknowledging trade-offs, uncertainty and risk – is of critical importance.
However, thinking this way is also important in fundamental research. Evaluating surrogacy establishes the conditional boundaries under which relationships between variables exist. In our paper, we discuss how this may be under-appreciated when it comes to many ecological relationships that are considered well-established and therefore generalisable across all situations.
We know why this occurs – there is a need to simplify complexity. And of course all associations in ecology have variability, error and context specificity. The key point of our paper is that no one benefits from simply ignoring that.
Instead, improved recognition of surrogate use in fundamental research probably has some benefits. Surrogate research offers a whole literature of frameworks and approaches for accounting for uncertainty and ensuring accuracy in inferences being made.
We contend that greater recognition of surrogacy could represent a significant knowledge transfer from applied to fundamental research. Adopting the tools for evaluating surrogates will allow researchers to quantitatively formalize trade-offs between accuracy and cost-effectiveness to more clearly justify the use of surrogates, whatever the context.
Hopefully this paper will get you thinking about the measures in your own research that you may be implicitly using as surrogates. These could be almost anything. For example, are you really measuring fire severity, or are you measuring scorch-height on tree trunks and simply inferring that that accurately represents fire severity?
The important question that you should hopefully come away from our paper asking is: do I need to contextually evaluate the measure I’m using to make this inference, to ensure it accurately represents what I’m saying it represents?
Ultimately this comes down to whether you are making generalisations or context-specific evaluations and what level of accuracy and certainty is acceptable in either. It’s not always going to be the same.