“Not everything that counts can be counted, and not everything that can be counted counts” – Albert Einstein
In this episode of the Core Ideas podcast, we conclude our introduction to paleolimnology with a discussion of how lake sediments are used to reconstruct past environments. We examine some reasons for developing environmental reconstructions, and how calibration sets can produce transfer functions to apply knowledge of modern environments to a dated sediment core and infer the history of an environmental variable from biological, chemical or physical indicators preserved in the sediments. Finally, we cover some finer details to keep in mind when using these techniques.
Typically, environmental reconstructions are used to answer questions like “How have lakes changed through time with respect to a particular environmental variable?” These can be very complex questions, as direct measurements of relevant data are rare for any substantial timespan. The lack of data may be due to either a simple absence of measurements, but often the problem of interest can be older than the relevant measurement techniques (e.g. acid rain was recognized in the late-1800s, but the concept of pH was only developed in 1909). Indirect approaches can be used to work around missing data, using methods such as calibrating the indicators preserved in lake sediments to specific environmental variables.
For most environmental variables, the majority of taxa have an optimum value and well-defined tolerances. A calibration (or training) set of lakes, can be selected to span an environmental gradient of interest. By examining the composition of various indicator communities (e.g. diatoms) contained within surface sediment samples (representative of recent history) across a full calibration set (typically 40-70 lakes), the community composition can be calibrated against measured values of the environmental variable. A transfer function can then be produced to infer values for that variable from a given community, allowing knowledge of modern conditions to be transferred to deeper sediment intervals.
After our quick description of calibration sets/transfer functions we spent some time on:
- How well do calibration sets age?
- What are the implications of using a calibration set as the underlying conditions change (i.e climate warming)?
- How often are new calibration sets produced for specific variables when a transfer function already exists?
- Difficulty in building a well-performing model that others will want to use?