Developing tools for improved population and range estimation in support of extinction risk assessments for Neotropical birds

Devenish, Christian (2017) Developing tools for improved population and range estimation in support of extinction risk assessments for Neotropical birds. Doctoral thesis (PhD), Manchester Metropolitan University.

Preview

Available under License Creative Commons Attribution Non-commercial No Derivatives. Download (13MB) | Preview

Abstract

Species abundance and distribution metrics are cornerstones of conservation planning, for example, in establishing extinction risk and selecting priority areas, but abundance data are scarce and costly to obtain in comparison to those on species occurrence. Occurrence records, often from citizen science or nonsystematic surveys, are increasingly used to model species’ distributions using environmental predictors. Methods to relate occurrence models to abundance, and therefore, provide greater understanding of patterns of abundance across species’ ranges and population size estimates could bring important benefits for conservation decisions. This thesis aims to develop tools, combining different analytical techniques, field data and GIS, to provide improved estimates of species distribution and abundance in support of extinction risk assessments in threatened Neotropical bird species. To achieve this aim, a case study was implemented over the ranges of 14 dry forest birds from the Tumbesian region of Peru –an area of critical conservation importance due to high endemism and severe anthropogenic threats– with the following objectives: to model the distribution of study species (Chapter 2); to estimate local abundance of species across their ranges using covariate Distance sampling (Chapter 3); to explore range-wide variation in abundance (Chapter 4); to explore the relationship between relative probability of occurrence, derived from modelling, and bird abundance, derived from field studies (Chapter 5). First, ensemble species distribution models, using four modelling methods, were built with a median of 150 occurrence records per species, bioclimatic variables and vegetation indices. Modelled Extent of Occurrence, using a 5% omission error threshold to define presence and absence, was compared to existing range estimates used in extinction risk assessment. Additionally, field data were obtained on the local abundance of the study species and habitat characteristics along four 2.5 km transects at 26 sites over the study area. Covariate Distance sampling was used to estimate bird abundances at each site. Where sites represented discrete or delimited units (e.g. protected areas), specific population sizes were estimated. Local abundance was compared across sites and by range core versus edge; spatial autocorrelation was examined with multivariate Mantel tests; and, relationships with environmental variables were examined using Generalised Additive Models. Finally, relationships between abundance estimates, obtained from the field study, and relative probability of occurrence, obtained from distribution models, were tested using correlations, and where significant relationships were found, these were modelled using hierarchical logistic regression. Individual species distribution modelling methods performed adequately and coincided highly in terms of ranked correlation but differed in the distribution of their predicted values. Range size estimates, from thresholded models, were generally smaller than, but coincided spatially with, published ranges, with the exception of three species of conservation interest. Local abundance varied by one or two orders of magnitude across sites for almost all species, with abundance not necessarily highest at the centre of species’ ranges. Sites of maximum abundance for individual species did not coincide – nine different sites held the highest densities of at least one species. Eleven of 14 species showed significant positive correlations between their abundance and modelled occurrence for at least one modelling technique. Modelling techniques are discussed in light of complementing existing techniques to estimate Extent of Occurrence for extinction risk assessments. Abundance estimates, using methods that incorporate detectability, can be obtained for rare species over very patchy habitats with relatively low survey effort, using a suitably designed sampling protocol. The extreme variation in species' abundances and the complexity in relationships with environmental variables has conservation implications, for example, in the design of conservationmotivated surveys and regarding the need for multiple reserves to capture high local abundances of key species. The relationship between modelled species' occurrence and local abundance is a promising area of research with a view to obtaining better abundance information with less survey effort. In terms of biodiversity conservation in north Peru, critical sites are recommended for urgent protection, and updated extinction risk categories are given for threatened species.