Pseudoproxy modelling to explore ecosystem dynamics and assess uncertainties in palaeoecology

Contemporary ecosystems are experiencing increasing anthropogenic pressure and there is growing concern over how they will respond to such novel pressures. Palaeoecology offers crucial understanding of how ecosystems have responded to environmental change in the past, and provides valuable insight into how ecosystems will change in the future. However, palaeo-proxy records are associated with many uncertainties from environmental processes and process and observer error that may limit the inferences drawn from them. There is a call for palaeoecology to move from a descriptive to a quantitative discipline by quantifying uncertainties and employing rigorous statistical data analysis methods. Thus, the sources of uncertainty in the data must be assessed for their influence on statistical analyses, in terms of false positive and false negative pattern identification (e.g. of regime shifts or indicators of resilience). Virtual assessments of how uncertainty in palaeoecological data affect statistical analyses are few and do not adequately represent the underlying ecological dynamics or the multivariate data typical of proxy records. This lack of understanding of uncertainties is a key knowledge gap in the scientific literature that I seek to address in this thesis. First, I develop a model for generating multivariate proxy data based on underlying ecological dynamics. Second, I apply multivariate statistics to the simulated data under increasing levels of uncertainty introduced from core mixing, sub-sampling and proxy counting to determine which individual and combined sources of uncertainty have the greatest influence on the analyses. Results suggest that individual sources of uncertainty, such as sub-sampling frequency, have a greater influence on analyses than mixing or proxy counting resolution; however, the greatest influence is from the interaction effect of the combined uncertainties of sub-sampling with proxy counting. Third, I apply change point analyses to the multivariate statistics to assess the probability of detecting a shift in proxy data, caused by a variety of environmental conditions, when subject to proxy uncertainties. The probability of detecting a shift in proxy data at a known point in time depended on the rate of change in the driving conditions relative to the response rate of the system. Abrupt environmental shifts combined with fast ecological responses were detected more frequently than slower environmental changes and ecological response. The effects of uncertainties in the proxies themselves were inconsistent, but the probability of statistically detecting change at a known time decreased at the highest level of uncertainty. Finally, both chronological and proxy uncertainties are introduced to the data to determine how they affect the timing of change points detected at a known time in the simulated proxy record. Timing was found to be precise to a century level, but the accuracy of determining the true timing of change reduced as uncertainty increased. Overall, I conclude that for empirical palaeoecological studies where best to focus effort on reducing uncertainty depends on the question and scales of interest and knowledge of the study area. In the case of poor knowledge of a site and its environmental drivers over time, core replication and spatial coverage is likely to be necessary to increase the probability of detecting key patterns of change over focusing on laboratory methods of sub-sampling frequency. Furthermore, I recommend applying a suite of statistical methods and report chronological and proxy uncertainties.

Name thesis:degree_name

PhD

Level thesis:degree_level

Doctoral

Discipline thesis:degree_discipline

Environmental Science

Grantor dc:publisher

ResearchSpace@Auckland

Year dc:date.issued

2021

Author dc:creator

• Asena, Quinn

Advisors dc:contributor.advisor

• Perry, George
• Wilmshurst, Janet

Statement dc:rights

• Items in ResearchSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

Licence dc:rights.uri

• https://researchspace.auckland.ac.nz/docs/uoa-docs/rights.htm

Handle dc:identifier.uri

https://hdl.handle.net/2292/58282

OAI identifier oai:identifier

oai:researchspace.auckland.ac.nz:2292/58282