PAPER: Pseudoproxy modelling of uncertainties

Is the past recoverable from the data? Pseudoproxy modelling of uncertainties in palaeoecological data

May 2024

Asena Q, Perry GL, Wilmshurst JM. Is the past recoverable from the data? Pseudoproxy modelling of uncertainties in palaeoecological data. The Holocene 0(0):

ABSTRACT

There is growing concern about the response of contemporary ecosystems to increasing and novel anthropogenic pressures and environmental conditions. Palaeoecology is crucial to understanding how ecosystems have responded to past environmental changes and can inform management of contemporary ecosystems and contribute to forecasts of ecosystem responses to change. However, palaeoecological data are subject to uncertainties that arise from environmental processes, field and laboratory methods, and data processing, and that affects inferences drawn from them. Understanding how different sources of uncertainty affect the analyses of proxy records remains limited, and records are often interpreted solely qualitatively. We present a virtual ecology approach for assessing how uncertainties inherent in empirical proxy data influence statistical analyses and the inferences drawn from them. In the virtual ecology approach, both the data and the observational process are recreated in simulation to assess sampling and analytical methods. We demonstrate results from a new model for simulating core-type samples of pseudoproxies comparable to empirical proxy data but not subject to the same sources of proxy and chronological uncertainties. These ‘error-free’ pseudoproxies generated under known driving conditions have uncertainties (e.g. core mixing, sub-sampling, and proxy quantification) systematically introduced to them to assess how individual and combined sources of uncertainty influence analytical methods. Results indicate that inferences drawn from statistical analysis, such as the stability of a system, or the rate of ecological turnover, can change substantially between the ‘error-free’ pseudoproxies, and degraded and sub-sampled data. We show how our approach can advance understanding of uncertainties in palaeoecological data and how it can help shape research questions by quantifying of their influence on proxy data.

KEYWORDS

ecological modelling; palaeoecology; proxy system modelling; pseudoproxy; virtual ecology