Article

Generalized site occupancy models allowing for false positive and false negative errors

U.S. Geological Survey, Patuxent Wildlife Research Center, Laurel, Maryland 20708, USA.
Ecology (Impact Factor: 5). 05/2006; 87(4):835-41. DOI: 10.1890/0012-9658(2006)87[835:GSOMAF]2.0.CO;2
Source: PubMed

ABSTRACT Site occupancy models have been developed that allow for imperfect species detection or "false negative" observations. Such models have become widely adopted in surveys of many taxa. The most fundamental assumption underlying these models is that "false positive" errors are not possible. That is, one cannot detect a species where it does not occur. However, such errors are possible in many sampling situations for a number of reasons, and even low false positive error rates can induce extreme bias in estimates of site occupancy when they are not accounted for. In this paper, we develop a model for site occupancy that allows for both false negative and false positive error rates. This model can be represented as a two-component finite mixture model and can be easily fitted using freely available software. We provide an analysis of avian survey data using the proposed model and present results of a brief simulation study evaluating the performance of the maximum-likelihood estimator and the naive estimator in the presence of false positive errors.

1 Follower
 · 
111 Views
 · 
0 Downloads
  • Source
    • "However, if there are relatively few poor quality photographs in which the species is not known exactly, it may be more appropriate to ignore these photographs and consider them as non-detections. If these were true detections, ignoring photographs will lower the detection probability and likely reduce the precision of the occupancy estimates, while including them when the species was incorrectly identified will often introduce bias (Miller et al. 2011; Royle and Link 2006). Ideally, all photographs with primate detections will be identified to the species without error. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Field-based primate studies often make population inferences using count-based indices (e.g., individuals/plot) or distance sampling; the first does not account for the probability of detection and thus can be biased, while the second requires large sample sizes to obtain precise estimates, which is difficult for many primate studies. We discuss photographic sampling and occupancy modeling to correct for imperfect detection when estimating system states and dynamics at the landscape level, specifically in relation to primate ecology. We highlight the flexibility of the occupancy framework and its many applications to studying low-density primate populations or species that are difficult to detect. We discuss relevant sampling and estimation procedures with special attention to data collection via photographic sampling. To provide tangible meaning to terminology and clarify subtleties, we use illustrative examples. Photographic sampling can have many advantages over observer-based sampling, especially when studying rare or elusive species. Combining photographic sampling with an occupancy framework allows inference to larger scales than is common in primate studies, addresses uncertainty due to the observation process, and allows researchers to examine questions of how landscape-level anthropogenic changes affect primate distributions.
    International Journal of Primatology 10/2014; Online. DOI:10.1007/s10764-014-9761-9 · 1.99 Impact Factor
  • Source
    • "There are few, if any systems, where species detection is perfect [69]. Imperfect detection (false negatives) may occur because a species is rare or cryptic [108], [109], or alternatively, a false positive report or misidentification may occur [110]–[112]. We underestimated Bd occupancy of amphibian habitats within our study area by 14%, which is greater than other observations from a different geographic region that used the same filtration method [80]. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Biodiversity losses are occurring worldwide due to a combination of stressors. For example, by one estimate, 40% of amphibian species are vulnerable to extinction, and disease is one threat to amphibian populations. The emerging infectious disease chytridiomycosis, caused by the aquatic fungus Batrachochytrium dendrobatidis (Bd), is a contributor to amphibian declines worldwide. Bd research has focused on the dynamics of the pathogen in its amphibian hosts, with little emphasis on investigating the dynamics of free-living Bd. Therefore, we investigated patterns of Bd occupancy and density in amphibian habitats using occupancy models, powerful tools for estimating site occupancy and detection probability. Occupancy models have been used to investigate diseases where the focus was on pathogen occurrence in the host. We applied occupancy models to investigate free-living Bd in North American surface waters to determine Bd seasonality, relationships between Bd site occupancy and habitat attributes, and probability of detection from water samples as a function of the number of samples, sample volume, and water quality. We also report on the temporal patterns of Bd density from a 4-year case study of a Bd-positive wetland. We provide evidence that Bd occurs in the environment year-round. Bd exhibited temporal and spatial heterogeneity in density, but did not exhibit seasonality in occupancy. Bd was detected in all months, typically at less than 100 zoospores L-1. The highest density observed was ∼3 million zoospores L-1. We detected Bd in 47% of sites sampled, but estimated that Bd occupied 61% of sites, highlighting the importance of accounting for imperfect detection. When Bd was present, there was a 95% chance of detecting it with four samples of 600 ml of water or five samples of 60 mL. Our findings provide important baseline information to advance the study of Bd disease ecology, and advance our understanding of amphibian exposure to free-living Bd in aquatic habitats over time.
    PLoS ONE 09/2014; 9(9):e106790. DOI:10.1371/journal.pone.0106790 · 3.23 Impact Factor
  • Source
    • "Citizen science data, however, is collected less rigorously, making this assumption questionable. Previous work has incorporated the possibility of false positives into the OD model (Royle and Link 2006). More recent work has modeled false positives in the citizen science context by distinguishing between experts and novices in the detection process (Yu, Wong, and Hutchinson 2010). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Data quality is a common source of concern for large-scale citizen science projects like eBird. In the case of eBird, a major cause of poor quality data is the misiden-tification of bird species by inexperienced contributors. A proactive approach for improving data quality is to discover commonly misidentified bird species and to teach inexperienced birders the differences between these species. To accomplish this goal, we develop a latent variable graphical model that can identify groups of bird species that are often confused for each other by eBird participants. Our model is a multi-species extension of the classic occupancy-detection model in the ecology literature. This multi-species extension requires a structure learning step as well as a computationally expensive parameter learning stage which we make efficient through a variational approximation. We show that our model can not only discover groups of misidentified species, but by including these misidentifications in the model, it can also achieve more accurate predictions of both species occupancy and detection.
    Proceedings of the 28th Conference on Artificial Intelligence (AAAI), Quebec City, Quebec, Canada; 07/2014
Show more

Questions & Answers about this publication

  • Mark Stanaway added an answer in Mycorrhiza:
    What do you think about pseudo-absences in ecological studies?

    The objective of my study is to elucidate the relationship between individual species of mycorrhiza and environmental variables using Spearman correlation. I wonder if I should exclude absences records because not detection doesn't mean inexistence of the specie.... I think that this is true but I am not sure if this is appropriate. What do you think? Is there classical evidence of this topic?

    Mark Stanaway · Queensland University of Technology

    You should not throw good information away. You are acknowledging that your sampling is not perfect but you hope that more mycorrhiza are found if the environment is more favourable. There are a number of approaches that you could use depending on your environmental variables. A place to start could be zero-inflated models. These model both the true absences and the false absences. There should also be some links to suitable methods in the publications below.