Dealing with imperfect detectability in biological surveys for native grassland management

Garrard, G 2009, Dealing with imperfect detectability in biological surveys for native grassland management, Doctor of Philosophy (PhD), Global Studies, Social Science and Planning, RMIT University.


Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Garrard.pdf Thesis application/pdf 9.88MB
Title Dealing with imperfect detectability in biological surveys for native grassland management
Author(s) Garrard, G
Year 2009
Abstract The default assumption of many environmental impact assessments is that a species that is present at a site will be detected during a survey of that site. However there is now evidence that a range of species have detection probabilities of less than one during biological surveys, and the consequences of failing to detect threatened and invasive species may be severe. A number of methods exist for characterising detectability and determining the number of repeat visits to a site necessary to ensure an acceptably high probability that a resident species will be detected. However, no such method exists for determining the survey duration necessary in a single visit to ensure a reasonable probability of detecting a plant species that is present at a site. In this thesis a new method, based on failure time techniques, is proposed for estimating the time required to detect a plant species during a flora survey. The exponential detection time model presented in this study is demonstrated to be a reliable estimator of the average time to detection of a plant species under a range of simulated scenarios.

Using the exponential detection time model introduced in this thesis, the detectability of two threatened plant species (Pimelea spinescens and Dianella amoena) and two invasive weeds (Nassella neesiana and N. trichotoma) in a threatened native grassland community are investigated. Imperfect detection is demonstrated for all four species and a number of observer and environmental factors that influence detectability are identified. In particular, it is demonstrated that experienced observers detect each species more quickly than their less experienced counterparts. Other variables that affect detectability include the search method and cover of the dominant grass species. Under favourable conditions, predicted average detection times range from 26 (P. spinescens) to 41 (D. amoena) minutes per hectare, and estimates of the survey effort required to achieve a probability of 0.80 that the target species will be detected if it is present are between 42 and 66 minutes per hectare. As with other detectability studies, the findings of this research demonstrate that the survey effort required to detect these species increases substantially under suboptimal survey conditions.

A multi-species detection time model that characterises the average time to detection according to plant traits and characteristics is also introduced. Variables shown to influence detectability in this model include the lifeform and rarity of the plant, and flower and leaf characteristics. While not able to provide precise and accurate estimates of individual species’ detection probabilities, this model may be used to bound survey effort requirements where no other information exists.

The findings of this thesis have important implications for conservation policy and practice in Australia and internationally. In particular, they may be used to inform survey recommendations, including minimum survey effort requirements, for environmental impact assessments and weed surveillance. A number of recommendations for a more effective handling of imperfect detection in conservation policy are made.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Global Studies, Social Science and Planning
Keyword(s) detectability
false absence
survey effort
surveillance
impact assessment
Versions
Version Filter Type
Access Statistics: 299 Abstract Views, 959 File Downloads  -  Detailed Statistics
Created: Thu, 20 Jan 2011, 14:24:47 EST
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us