Is Ecosystem-Atmosphere Observation in Long-Term Networks actually Science?

Abstract:

Science uses observations to build knowledge by testable explanations and predictions. The “scientific method” requires controlled systematic observation to examine questions, hypotheses and predictions. Thus, enquiry along the scientific method responds to questions of the type “what if …?” In contrast, long-term observation programs follow a different strategy: we commonly take great care to minimize our influence on the environment of our measurements, with the aim to maximize their external validity. We observe what we think are key variables for ecosystem-atmosphere exchange and ask questions such as “what happens next?” or “how did this happen?” This apparent deviation from the scientific method begs the question whether any explanations we come up with for the phenomena we observe are actually contributing to testable knowledge, or whether their value remains purely anecdotal.

Here, we present examples to argue that, under certain conditions, data from long-term observations and observation networks can have equivalent or even higher scientific validity than controlled experiments. Internal validity is particularly enhanced if observations are combined with modeling.

Long-term observations of ecosystem-atmosphere fluxes identify trends and temporal scales of variability. Observation networks reveal spatial patterns and variations, and long-term observation networks combine both aspects. A necessary condition for such observations to gain validity beyond the anecdotal is the requirement that the data are comparable: a comparison of two measured values, separated in time or space, must inform us objectively whether (e.g.) one value is larger than the other. In turn, a necessary condition for the comparability of data is the compatibility of the sensors and procedures used to generate them. Compatibility ensures that we compare "apples to apples": that measurements conducted in identical conditions give the same values (within suitable uncertainty intervals). In principle, a useful tool to achieve comparability and compatibility is the standardization of sensors and methods. However, due to the diversity of ecosystems and settings, standardization in ecosystem-atmosphere exchange is difficult. We discuss some of the challenges and pitfalls of standardization across networks.