Improved inference in point-transect models: application to Hawaiian forest birds

Mary Woodcock Kroble
Friday 10 November 2017
Date: 14 March 2018
Time: 2:00 pm - 3:00 pm

Speaker: Richard Camp (CREEM)

Abstract

Hawaii is the extinction capital of the world, and unfortunately birds are at the forefront of this calamity. We know this because of work by early naturalists and long-term monitoring. In this presentation I will briefly introduce the Hawaiian avifauna, impacts of Polynesian colonization and Western contact, major monitoring efforts, and progress on my PhD research focusing on improving inference in point-transect models.

Distance sampling is widely used to estimate wildlife population density. Point-transect distance sampling is particularly useful for surveying birds, and over the last four decades it is the standard method employed in Hawaiian forest bird monitoring. Accurate and precise estimates are essential to understanding how populations change across space and through time, particularly in response to conservation efforts and management actions. For widespread, common species distance sampling model assumptions are generally met and recommended numbers of detections for reliable detection probability modelling obtained. Achieving adequate detections is problematic however for rare, endangered and range-restricted species resulting in uncertainty estimates that are biologically unrealistic and useless for conservation and management purposes. Population densities can vary considerably over small areas, which further inflates estimate uncertainty. We explore two methods to decrease variance in density and trend estimates.

Trend in densities across a time series is usually evaluated at the stratum (i.e., forest) level. Point-transect sampling, however, involves recording distances to individually detected birds at samplers located along transects. Capitalizing on the sampler-level data we calculated density estimates and evaluated the trend in densities at each sampler for the endangered Hawai?i ?Akepa (Loxops coccineus). Accounting for the sampler-level trends better predicts spatially restricted regional or local trends than what can be observed at the stratum-wide averaged trend, which reduces trend variance.

Recently developed and demonstrated is a protocol to estimate population density from single automatic sound recorders for Hawai`i `amakihi (Chlorodrepanis virens). The protocol uses cue rates from the target species and an estimate of the distance from individuals to the recorder based on the power of the sound. Measurement error in the estimated distances, however, precluded applying the protocol to other bird species. We account for measurement error in `Oma`o (Myadestes obscurus) vocalizations to minimize estimator bias and correctly account for estimator uncertainty. Incorporating methods to correct for measurement error makes this protocol useful for collecting large scale and long-term information on animal populations, particularly those that are rare or that live in remote areas that are difficult to access.

Share this story