Looking to the past to conserve present-day diversity

Thesis

This thesis analyses, or develops tools for, the investigation of macroecological patterns of change across varied temporal and spatial scales. I have sought to develop new and innovative methods for the gathering and analysis of a diverse range of ecological data, alongside previously unimplemented strategies within the field of macroecology.

In the first study I developed and then tested the Historical Occurrence Georeferencing System (HOGS). This exploits historical distributional data in printed literature to generate previously untapped distributional occurrence information. HOGS was able to detect and classify large volumes of map-marker occurrence data in a fraction of the time taken for previous manual methods. It provides a rapidly deployable method for the generation of historical data, vital to accurate definition of baseline species conditions.

The second study determined the degree and factorial underpinnings of avian community shift within southern Africa by utilising community weighted mean temperature and precipitation metrics. I found that avian species responded strongly to these climatic changes and the multi-metric approach allowed for a more in-depth analysis than traditionally temperature-only strategies. Indices to measure the rate of change for avian communities employed alongside derivation of important predictive factors is highly useful when designing management strategies going forward.

The third study was conducted in partnership with Kew. I created a voronoi tessellation scheme for the assignment of areas of control in the analysis of socioeconomic and wellbeing indicators surrounding Madagascan protected areas. This allowed for a more nuanced understanding of fine scale heterogeneity in poverty indicators that can inform the successful delivery of conservation and sustainable development interventions.

In the final study I developed a modular machine learning pipeline for the automated analysis of mollusc field images. It combines object detection and classification techniques to detect molluscs and drill holes, identify species, and enable biologically meaningful measurements. The system demonstrates strong performance across tasks, achieving high accuracy in mollusc detection, robust species classification across a range of dataset thresholds and reliable localisation and classification of drill holes.

Together, these studies add new methodologies for modern and historical macroecological assessment, which will be crucial in a world experiencing accelerating change.

Authors: Ravenscroft, H

• DOI: 10.5287/ora-pm4gmp9qa
• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Subjects: Macroecology