Summary
The past decade has seen an explosion in the development and application of models aimed at estimating species occurrence and occupancy dynamics while accounting for possible non-detection or species misidentification.
We discuss some recent occupancy estimation methods and the biological systems that motivated their development. Collectively, these models offer tremendous flexibility, but simultaneously place added demands on the investigator.
Unlike many mark–recapture scenarios, investigators utilizing occupancy models have the ability, and responsibility, to define their sample units (i.e. sites), replicate sampling occasions, time period over which species occurrence is assumed to be static and even the criteria that constitute ‘detection’ of a target species. Subsequent biological inference and interpretation of model parameters depend on these definitions and the ability to meet model assumptions.
We demonstrate the relevance of these definitions by highlighting applications from a single biological system (an amphibian–pathogen system) and discuss situations where the use of occupancy models has been criticized. Finally, we use these applications to suggest future research and model development.

https://besjournals.onlinelibrary.wiley.com/doi/pdf/10.1111/2041-210X.12100

Advances and applications of occupancy models

Large-scale presence-absence monitoring programs have great promise for many conservation applications. Their value can be limited by potential incorrect inferences owing to observational errors, especially when data are collected by the public. To combat this, previous analytical methods have focused on addressing non-detection from public survey data. Misclassification errors have received less attention but are also likely to be a common component of public surveys, as well as many other data types. We derive estimators for dynamic occupancy parameters (extinction and colonization), focusing on the case where certainty can be assumed for a subset of detections. We demonstrate how to simultaneously account for non-detection (false negatives) and misclassification (false positives) when estimating occurrence parameters for gray wolves in northern Montana from 2007–2010. Our primary data source for the analysis was observations by deer and elk hunters, reported as part of the state’s annual hunter survey. This data was supplemented with data from known locations of radio-collared wolves. We found that occupancy was relatively stable during the years of the study and wolves were largely restricted to the highest quality habitats in the study area. Transitions in the occupancy status of sites were rare, as occupied sites almost always remained occupied and unoccupied sites remained unoccupied. Failing to account for false positives led to over estimation of both the area inhabited by wolves and the frequency of turnover. The ability to properly account for both false negatives and false positives is an important step to improve inferences for conservation from large-scale public surveys. The approach we propose will improve our understanding of the status of wolf populations and is relevant to many other data types where false positives are a component of observations.

/pdf/determining-occurrence-dynamics-when-false-positives-occur-527kaf0o7k.pdf

Determining Occurrence Dynamics when False Positives Occur: Estimating the Range Dynamics of Wolves from Public Survey Data

False positive errors are a significant component of many ecological data sets, which in combination with false negative errors, can lead to severe biases in conclusions about ecological systems. We present results of a field experiment where observers recorded observations for known combinations of electronically broadcast calling anurans under conditions mimicking field surveys to determine species occurrence. Our objectives were to characterize false positive error probabilities for auditory methods based on a large number of observers, to determine if targeted instruction could be used to reduce false positive error rates, and to establish useful predictors of among-observer and among-species differences in error rates. We recruited 31 observers, ranging in abilities from novice to expert, who recorded detections for 12 species during 180 calling trials (66 960 total observations). All observers made multiple false positive errors, and on average 8.1% of recorded detections in the experiment were false positive errors. Additional instruction had only minor effects on error rates. After instruction, false positive error probabilities decreased by 16% for treatment individuals compared to controls with broad confidence interval overlap of 0 (95% CI: � 46 to 30%). This coincided with an increase in false negative errors due to the treatment (26%; � 3t o 61%). Differences among observers in false positive and in false negative error rates were best predicted by scores from an online test and a self-assessment of observer ability completed prior to the field experiment. In contrast, years of experience conducting call surveys was a weak predictor of error rates. False positive errors were also more common for species that were played more frequently but were not related to the dominant spectral frequency of the call. Our results corroborate other work that demonstrates false positives are a significant component of species occurrence data collected by auditory methods. Instructing observers to only report detections they are completely certain are correct is not sufficient to eliminate errors. As a result, analytical methods that account for false positive errors will be needed, and independent testing of observer ability is a useful predictor for among-observer variation in observation error rates.

https://www.researchgate.net/profile/David_Miller27/publication/230712781_Experimental_investigation_of_false_positive_errors_in_auditory_species_occurrence_surveys/links/0deec51bf6add03532000000.pdf

Experimental investigation of false positive errors in auditory species occurrence surveys.

Opportunistic citizen science (CS) programs allow volunteers to report species observations from anywhere, at any time, and can assemble large volumes of historic and current data at faster rates than more coordinated programs with standardized data collection. This can quickly provide large amounts of species distributional data, but whether this focus on participation comes at a cost in data quality is not clear. Although automated and expert vetting can increase data reliability, there is no guarantee that opportunistic data will do anything more than confirm information from professional surveys. Here, we use eButterfly, an opportunistic CS program, and a comparable dataset of professionally collected observations, to measure the amount of new distributional species information that opportunistic CS generates. We also test how well opportunistic CS can estimate regional species richness for a large group of taxa (>300 butterfly species) across a broad area. We find that eButterfly contributes new distributional information for >80% of species, and that opportunistically submitting observations allowed volunteers to spot species ~35 days earlier than professionals. Although eButterfly did a relatively poor job at predicting regional species richness by itself (detecting only about 35-57% of species per region), it significantly contributed to regional species richness when used with the professional dataset (adding ~3 species that had gone undetected in professional surveys per region). Overall, we find that the opportunistic CS model can provide substantial complementary species information when used alongside professional survey data. Our results suggest that data from opportunistic CS programs in conjunction with professional datasets can strongly increase the capacity of researchers to estimate species richness, and provide unique information on species distributions and phenologies that are relevant to the detection of the biological consequences of global change.

Opportunistic citizen science data transform understanding of species distributions, phenology, and diversity gradients for global change research.

Summary
Biological monitoring programmes are increasingly relying upon large volumes of citizen-science data to improve the scope and spatial coverage of information, challenging the scientific community to develop design and model-based approaches to improve inference.
Recent statistical models in ecology have been developed to accommodate false-negative errors, although current work points to false-positive errors as equally important sources of bias. This is of particular concern for the success of any monitoring programme given that rates as small as 3% could lead to the overestimation of the occurrence of rare events by as much as 50%, and even small false-positive rates can severely bias estimates of occurrence dynamics.
We present an integrated, computationally efficient Bayesian hierarchical model to correct for false-positive and false-negative errors in detection/non-detection data. Our model combines independent, auxiliary data sources with field observations to improve the estimation of false-positive rates, when a subset of field observations cannot be validated a posteriori or assumed as perfect. We evaluated the performance of the model across a range of occurrence rates, false-positive and false-negative errors, and quantity of auxiliary data.
The model performed well under all simulated scenarios, and we were able to identify critical auxiliary data characteristics which resulted in improved inference. We applied our false-positive model to a large-scale, citizen-science monitoring programme for anurans in the north-eastern United States, using auxiliary data from an experiment designed to estimate false-positive error rates. Not correcting for false-positive rates resulted in biased estimates of occupancy in 4 of the 10 anuran species we analysed, leading to an overestimation of the average number of occupied survey routes by as much as 70%.
The framework we present for data collection and analysis is able to efficiently provide reliable inference for occurrence patterns using data from a citizen-science monitoring programme. However, our approach is applicable to data generated by any type of research and monitoring programme, independent of skill level or scale, when effort is placed on obtaining auxiliary information on false-positive rates.

https://besjournals.onlinelibrary.wiley.com/doi/pdf/10.1111/2041-210X.12542

Uncertainty in biological monitoring: a framework for data collection and analysis to account for multiple sources of sampling bias

Inferring the distribution and abundance of a species from field records must deal with false-negative and false-positive errors. False-negative errors occur if a species present goes undetected, while false-positive errors are typically a consequence of species misidentification. False-positive observations in studies of rare species may cause an overestimation of the distribution or abundance of the species and distort trend indices. We illustrate this issue with the monitoring of the Eurasian lynx in the Alps. We developed a three-level classification of field records according to their reliability as inferred from whether they were validated or not. The first category (C1) represents ‘hard fact’ data (e.g. dead lynx); the second category (C2) includes confirmed data (e.g. tracks verified by an expert); and the third category (C3) are unconfirmed data (e.g. any kind of direct visual observation). For lynx, which is a comparatively well-known species in the Alps, we use site-occupancy modelling to estimate its distribution and show that the inferred lynx distribution is highly sensitive to presence sign category: it is larger if based on C3 records compared with the more reliable C1 and C2 records. We believe that the reason for this is a fairly high frequency of false-positive errors among C3 records. This suggests that distribution records for many lesser-known species may be similarly unreliable, because they are mostly or exclusively based on unconfirmed and thus soft data. Nevertheless, such soft data form a considerable part of species assessments as presented, for example in the International Union for Conservation of Nature Red List. However, C3 records can often not be discarded because they may be the only information available. When inferring the distribution of rare carnivores, especially for species with an expanding or shrinking range, we recommend a rigorous discrimination between fully reliable and un- or only partly reliable data, in order to identify possible methodological problems in the distribution maps related to false-positive records.

Monitoring in the presence of species misidentification: the case of the Eurasian lynx in the Alps

The Eurasian lynx is of special conservation concern based on the European Union's Habitat Directive and its populations need to be maintained or restored at favourable conservation status. To evaluate lynx population status, appropriate monitoring needs to be in place. We modelled the distribution dynamics of lynx in the Alps (200 000 km2) during 1994–2014 at a resolution of 100 km2. Lynx distribution and detection probability varied by year, country, forest cover, elevation and distance to the nearest release site. Occupancy of neighbouring quadrats had a strong positive effect on colonization and persistence rates. Our analyses demonstrate the importance of accounting for imperfect detection: the raw data underestimated the lynx range by 55% on average, depending on country and winter. Over the past 20 years the Alpine lynx range has expanded at an average rate of 4% per year, which was partly due to the lynx translocations to new areas. Our approach to large-scale distribution modelling and analysing trends using site occupancy models can be applied retrospectively and is useful in many cases where a network of trained people is established to report the presence of target species, for example, in Europe where member states of the European Union have to report conservation status of species of community interest. Hence, dynamic occupancy models are an appealing framework for inference about the large-scale range dynamics based on opportunistic data and a useful tool for large-scale management and conservation programmes.

https://zslpublications.onlinelibrary.wiley.com/doi/pdf/10.1111/acv.12369

M. Wölfl

Papers

Monitoring in the presence of species misidentification: the case of the Eurasian lynx in the Alps

Mapping range dynamics from opportunistic data: spatiotemporal modelling of the lynx distribution in the Alps over 21 years