Observer bias in bird census: in forest assemblages, richness estimates appear the most sensitive metrics – a cautionary note with educational implications

Aim We examined the relative influence of geographical location, habitat structure (physiognomy), and dominant plant species composition (floristics) on avian habitat relationships over a large spatial extent. Although it has been predicted that avian distributions are more likely to covary with physiognomy than with floristics at coarse scales, we sought to determine, more specifically, whether there remained a significant association between gradients in assemblages of bird species and dominant plant species within a general biome type, after statistically controlling for structural variation and geographical location of sampling sites.Location Our sample consisted of a subset of North American Breeding Bird Census survey sites that covered most of the range of eastern forests, from Florida to Nova Scotia, and west to Minnesota and North Dakota (up toc.2500 km between sites).Methods We restricted our analyses to the single year (1981) that provided the largest sample of sites (47) for which vegetation data were available within ± 2 years of the avian surveys. We examined the relationship between avian community composition and tree species composition over this series of forested plots. Data were divided into four sets: (1) bird species abundances, (2) tree species abundances, (3) physiognomic or structural variables and (4) geographical location (latitude and longitude). We performed separate detrended correspondence analysis ordinations of birds and trees, before and after statistically partialling out covariation associated with structural variables and geographical location. To gauge the relationship between the two sets of species we correlated site scores resulting from separate ordinations. We also compared continental‐scale patterns of variation in bird and tree assemblages to understand possible mechanisms controlling species distribution at that scale.Results Both bird and tree communities yielded strong gradients, with first‐axis eigenvalues from 0.75 to 0.97. All gradients were relatively long (> 4.0), implying complete turnover in species composition. However, geographical location accounted for < 10% of the total variation associated with any ordination. Prior to partialling out covariation resulting from location and physiognomy, bird species ordinations were strongly correlated with tree species ordinations. The strength of association was reduced after partialling, but one bird and one tree axis remained significantly correlated. There was a significant species–area effect for birds, but not for trees.Main conclusions There was a significant relationship between bird species assemblages and tree species assemblages in the eastern forests of North America. Even after partialling out covariation associated with spatial location and forest physiognomy, there remained a significant correlation between major axes from bird and tree ordinations, consistent with the hypothesis that floristic variation is likely to be important in organizing assemblages of birds within a general biome type, albeit over a much larger spatial extent than originally predicted. Forest tree species ordinations differed from bird species ordinations in several ways: trees had a higher rate of turnover along underlying environmental gradients; trees appeared more patchily distributed than birds at this scale; and tree species were more spaced out along the underlying ecological gradients, with less overlap. By understanding the relationship between bird assemblages and forest floristics, we might better understand how avian communities are likely to change if tree species distributions are altered as a result of climatic changes.

Patterns of phylogenetic structure of assemblages are increasingly used to gain insight into the ecological and evolutionary processes involved in the assembly of co-occurring species. Metrics of phylogenetic structure can be sensitive to scaling issues and data availability. Here we empirically assess the sensitivity of four metrics of phylogenetic structure of assemblages to changes in (i) the source of data, (ii) the spatial grain at which assemblages are defined, and (iii) the definition of species pools using hummingbird (Trochilidae) assemblages along an elevational gradient in Colombia. We also discuss some of the implications in terms of the potential mechanisms driving these patterns. To explore how source of data influence phylogenetic structure we defined assemblages using three sources of data: field inventories, museum specimens, and range maps. Assemblages were defined at two spatial grains: coarse-grained (elevational bands of 800-m width) and fine-grained (1-km2 plots). We used three different species pools: all species contained in assemblages, all species within half-degree quadrats, and all species either above or below 2000 m elevation. Metrics considering phylogenetic relationships among all species within assemblages showed phylogenetic clustering at high elevations and phylogenetic evenness in the lowlands, whereas those metrics considering only the closest co-occurring relatives showed the opposite trend. This result suggests that using multiple metrics of phylogenetic structure should provide greater insight into the mechanisms shaping assemblage structure. The source and spatial grain of data had important influences on estimates of both richness and phylogenetic structure. Metrics considering the co-occurrence of close relatives were particularly sensitive to changes in the spatial grain. Assemblages based on range maps included more species and showed less phylogenetic structure than assemblages based on museum or field inventories. Coarse-grained assemblages included more distantly related species and thus showed a more even phylogenetic structure than fine-grained assemblages. Our results emphasize the importance of carefully selecting the scale, source of data and metric used in analysis of the phylogenetic structure of assemblages.

The consequences of biodiversity decline in intensified agricultural landscapes hinge on surviving biotic assemblages. Maintaining crucial ecosystem processes and services requires resilience to natural and anthropogenic disturbances. However, the resilience and stability of surviving biological communities remain poorly quantified. From a 10-y dataset comprising 2,880 bird censuses across a land-use gradient, we present three key findings concerning the resilience and stability of Costa Rican bird communities. First, seed dispersing, insect eating, and pollinating guilds were more resilient to low-intensity land use than high-intensity land use. Compared with forest assemblages, bird abundance, species richness, and diversity were all ~15% lower in low-intensity land use and ~50% lower in high-intensity land use. Second, patterns in species richness generally correlated with patterns in stability: guilds exhibited less variation in abundance in low-intensity land use than in high-intensity land use. Finally, interspecific differences in reaction to environmental change (response diversity) and possibly the portfolio effect, but not negative covariance of species abundances, conferred resilience and stability. These findings point to the changes needed in agricultural production practices in the tropics to better sustain bird communities and, possibly, the functional and service roles that they play.

Standardized data on large‐scale and long‐term patterns of species richness are critical for understanding the consequences of natural and anthropogenic changes in the environment. The North American Breeding Bird Survey (BBS) is one of the largest and most widely used sources of such data, but so far, little is known about the degree to which BBS data provide accurate estimates of regional richness. Here, we test this question by comparing estimates of regional richness based on BBS data with spatially and temporally matched estimates based on state Breeding Bird Atlases (BBA). We expected that estimates based on BBA data would provide a more complete (and therefore, more accurate) representation of regional richness due to their larger number of observation units and higher sampling effort within the observation units. Our results were only partially consistent with these predictions: while estimates of regional richness based on BBA data were higher than those based on BBS data, estimates of local richness (number of species per observation unit) were higher in BBS data. The latter result is attributed to higher land‐cover heterogeneity in BBS units and higher effectiveness of bird detection (more species are detected per unit time). Interestingly, estimates of regional richness based on BBA blocks were higher than those based on BBS data even when differences in the number of observation units were controlled for. Our analysis indicates that this difference was due to higher compositional turnover between BBA units, probably due to larger differences in habitat conditions between BBA units and a higher likelihood of observing geographically restricted species. Our overall results indicate that estimates of regional richness based on BBS data suffer from incomplete detection of a large number of rare species, and that corrections of these estimates based on standard extrapolation techniques are not sufficient to remove this bias. Future applications of BBS data in ecology and conservation, and in particular, applications in which the representation of rare species is important (e.g., those focusing on biodiversity conservation), should be aware of this bias, and should integrate BBA data whenever possible.

Termites are major decomposers in tropical ecosystems. To characterize their assemblages in terms of taxonomical and functional composition, Jones and Eggleton (2000, Journal of Applied Ecology 37, 191–203) recently proposed a standardized sampling protocol based on belt transects of 100 m × 2 m. We evaluated the representativeness of samples obtained by this protocol, and its suitability to calculate diversity statistics, by replicating it in an area of naturally fragmented subtropical forest. We sampled six 100 m transects in separate small forest islets, and one transect extended to 500 m in a large islet, recording presence/absence data (occurrences) of termite species in successive quadrats of 5 m × 2 m. In the large islet, strips of 100 m within the 500 m transect produced extremely variable species richness figures. This variability was primarily due to heterogeneity in the spatial distribution of soil‐dwelling termites. Combining non‐contiguous quadrats allowed us to span a broader diversity of microhabitats for an equal effort, providing less variable results and faster species accumulation. Individual transects of 100 m in small forest islets yielded too few samples to allow reliable estimations of total species richness, although these transects when pooled constituted a useful data set for comparison with other sites. In the focal habitat, a single 100 m transect appeared therefore inadequate to allow a reliable characterization of the termite assemblage, even at the level of a single forest islet. To improve the rate of species accumulation and to obtain diversity statistics allowing intersite comparisons, we suggest the use of smaller, non‐contiguous quadrats, and that sampling be continued until stable diversity estimates are obtained. In the habitat studied, such an alternative protocol could be adequately combined with a standardized protocol for collecting ground‐dwelling ants.

To characterize the structure and species density of termite assemblages in tropical forests, we developed a sampling protocol based on a single large quadrat (belt transect) that standardises sampling effort and area. Emphasizing the estimation of species richness, Roisin and Leponce (2004, Austral Ecology, 29: 637–46) recommend sampling numerous small quadrats, and varying the number of quadrats depending on local species density. However, their method would result in the loss of standardisation and speed, taking up to four times longer to collect and process the samples. We show that even for a subtropical forest with low species density, our protocol produces robust results.

Documenting and estimating species richness at regional or landscape scales has been a major emphasis for conservation efforts, as well as for the development and testing of evolutionary and ecological theory. Rarely, however, are sampling efforts assessed on how they affect detection and estimates of species richness and rarity. In this study, vascular plant richness was sampled in 356 quarter hectare time-unlimited survey plots in the boreal region of northeast Alberta. These surveys consisted of 15,856 observations of 499 vascular plant species (97 considered to be regionally rare) collected by 12 observers over a 2 year period. Average survey time for each quarter-hectare plot was 82 minutes, ranging from 20 to 194 minutes, with a positive relationship between total survey time and total plant richness. When survey time was limited to a 20-minute search, as in other Alberta biodiversity methods, 61 species were missed. Extending the survey time to 60 minutes, reduced the number of missed species to 20, while a 90-minute cut-off time resulted in the loss of 8 species. When surveys were separated by habitat type, 60 minutes of search effort sampled nearly 90% of total observed richness for all habitats. Relative to rare species, time-unlimited surveys had ∼65% higher rare plant detections post-20 minutes than during the first 20 minutes of the survey. Although exhaustive sampling was attempted, observer bias was noted among observers when a subsample of plots was re-surveyed by different observers. Our findings suggest that sampling time, combined with sample size and observer effects, should be considered in landscape-scale plant biodiversity surveys.

Summary1.Diverse sites have long‐attracted ecologists, yet the overwhelming variety of species can confound attempts to enumerate species richness. Various predictive methods estimate species richness by comparing the rate at which species are first detected with the rate at which they are detected again, yielding richness estimates of known precision without exhaustive sampling.2.While frequently used for arthropods, predictive methods are rarely applied to vertebrate surveys where species identity is often a priority. Expressing observed richness as a function of estimated richness, an estimate of survey completeness can be derived, offering the potential for inventories of standardized precision for comparison and further analysis.3.To realize this potential, I conducted 402 h of bird surveys on Barro Colorado Island (Panama) and performed a series of retrospective analyses to address three questions: (i) How much effort is required to achieve complete inventories (maximum completeness)? (ii) What is the least amount of effort required to yield robust richness estimates (maximum efficiency)? and (iii) How much effort is required to optimize sampling, balancing completeness and efficiency?4.Whereas the richness estimate for all species required thirty 6‐h samples to attain maximum completeness, once migrants, waterbirds and non‐forest‐dependent species were excluded, the richness of forest‐dependent residents could be estimated to the same precision with fifteen samples and to 80% completeness with four samples.5.Of the 186 bird species detected, 70 represented unique or duplicate records, seen in only one or two sampling periods. These low detectability species were dominated by migrants (28) and raptors (14) and also included seven waterbirds, five nocturnal species and four aerial foragers, justifying the widespread practice of excluding these groups from surveys of forest assemblages.6.In addition to demonstrating the reliability of predictive approaches, this study demonstrates the practicality of results‐based stopping rules for sampling diverse sites, especially for targeted groups of species. Combining predictive methods with targeted sampling represents an efficient and rigorous design, increasing the number of sites that can be sampled and enhancing the overall power and value of the study.

Estimates of species richness for a given area require that repeat surveys be taken, so that the statistical robustness of the estimate can be assessed. But how should these repeat surveys be organized in time? Here we present a case study of Australian woodland birds, surveyed using the ‘active timed area search’ method, which has become the standard unit for the Australian Bird Atlas, a continental‐scale bird survey. To date, there has been no assessment of how estimates of species richness derived from this method are affected by the temporal organization of the repeat surveys. For instance, can conducting the repeat surveys in sequence on the same day effectively capture richness, or will additional species be obtained by repeating the surveys on different days within a season? If so, does the spacing of the repeat visits throughout the season have an effect? To answer these questions, we surveyed woodland birds in the Mount Lofty Ranges, South Australia, during late spring–summer 1999–2000, and compared the performance of two different temporal configurations of repeat visits to sites: (i) six repeat surveys performed on the same day; and (ii) three repeat surveys on different days. For both, we calculated the average number of species actually sighted and also estimated total species richness. The data supported our hypothesis that the same‐day surveys would yield fewer species and underestimate total species richness. The different‐day repeats captured significantly more species per unit of survey effort, and yielded a higher richness estimate. However, the timespan over which different‐day surveys were conducted within a season did not have a significant influence on species richness estimates, evincing a qualitative advantage to surveying on different days, regardless of the spacing of repeat visits. These results may be of assistance to conservation managers when planning cost‐efficient monitoring regimes.

Aim To improve predictions of spatial and temporal patterns of species richness it is important to consider how species presence at a site is defined. This is because this definition affects our estimate of species richness, which should be aligned with the aims of the study, e.g. estimating richness of the breeding community. Here we explore the sensitivity of species richness estimates to criteria for defining presence of species (e.g. in relation to number of days present during the breeding season) at 107 wetlands. Innovation We use opportunistic citizen science data of high density (a total of 151,817 observations of 77 wetland bird species; i.e. about 16 observations per day) to build site-occupancy models calculating occupancy probabilities at a high temporal resolution (e.g. daily occupancies) to derive probabilistic estimates of seasonal site use of each species. We introduce a new way for defining species presence by using different criteria related to the number of days the species are required to be present at local sites. We compared patterns of species richness when using these different criteria of species inclusions. Main conclusion While estimates of local species richness derived from high temporal resolution occupancy models are robust to observational bias, these estimates are sensitive to restrictions concerning the number of days of presence required during the breeding season. Unlike complete local species lists, summaries of seasonal site use and different presence criteria allow identifying differences between sites and amplifying the variability in species richness among sites. Thus, this approach allows filtering out species according to their phenology and migration behaviour (e.g. passer-by species) and could improve the explanatory power of environmental variables on predictive models.

Useful sampling guidelines have recently been provided to improve the precision of bird population parameter estimates from point count data. Less attention has been given, however, to the issue of accuracy or completeness of counts at each sampling point. While it may not be critical in studies that compare average counts among study plots, completeness of counts at the level of individual points may pose several problems in studies that assess relationships between response variables (species' occurrence or abundance) and explanatory variables (vegetation structure, size and isolation of the fragment, proportion of interior habitat) on a point by point basis. Reliability of models in such studies is greatly dependent on the accuracy of counts at each point. We examined how bias of incomplete counts could be minimised through refinements of the point-count method. Six forest types of the northeastern deciduous forest in Quebec (Canada) were intensively surveyed during an entire breeding season. Census performances of the point-count method for measuring species richness at individual points were assessed by resampling the original data set for a simulated census protocol of two visits at a point. Effects of daily variation, time of season, time of day, and count duration were tested. Census performances of counts conducted the same day were significantly poorer than counts from different days. Census performances were influenced by the time of season during which counts were conducted. When pooled across habitats, census performances for counts that combined visits at the beginning and at the end of the breeding season were significantly better than counts that were concentrated in either one of these periods. The 05:00 to 08:30 morning period yielded the best results for five of the six habitats. Census performances in open field were, however, significantly better during the first sampling period (04:30), suggesting that bird censuses in open habitats could be initiated before sunrise. Longer counts (25 and 20 min) greatly improved species richness estimates at a point. For studies that focus on comparisons at the scale of individual points, longer counts are preferable to short counts (15 and 10 min). We stress the importance of the scale (study plots or individual points) at which statistical analyses of point count data are conducted in the decision-making process of point count sampling procedures.

Although indigenous farmers and other traditional peoples have disturbed Amazonian forests for centuries, few studies investigate the effects of these disturbances on biodiversity. This short-term study investigates how bird assemblages are affected by agricultural practices adopted by the residents of a national park in the Brazilian Amazon. Twelve sites in secondary forest (four sites in three age categories) and 12 sites in primary forest were selected for bird sampling. Audio-visual censuses of birds were conducted in small plots (1 ha) of young secondary growth (4–5 y), middle-aged (7–15 y), and old (20–35 y) secondary forests. Each site in secondary forest had a corresponding control site in adjacent primary forest. Young secondary growth had fewer species than old secondary and primary forests. Bird species richness was similar in middle-aged, old secondary and primary forests. Bird communities of young secondary habitat were dominated by species normally found in open habitats, such as agricultural fields and igapó flooded forests. The bird species composition of young secondary growth was more similar to that of middle-aged forest, which in turn was more similar to that of old secondary forests. Bird species composition differed between young secondary growth and old secondary forests and primary forests and between middle-aged secondary and primary forests. Nectarivore/insectivores and frugivore/insectivores of monospecific flocks are significantly associated with secondary forests, especially in initial stages of forest regeneration. In contrast, specialized insectivorous birds that join mixed-flocks show a significant association with primary forest.

Sampling understory birds in different habitat types using point counts and camera traps

Visual assessments of topographic habitat structure and benthos on coral reefs were appraised using quantitative data collected from 16 replicate surveys within each of 21 sites on Seychelles reefs. Results from visual assessments of reef benthos were similar to those obtained using techniques frequently used to assess benthic complexity and composition. Visual estimates of habitat topography were correlated with rugosity, reef height and holes of 10–70 cm diameter, whilst visual estimates of benthic composition were very similar to those obtained from line intercept transects. Visual estimates of topography correlated strongly with species richness of fish communities and explained 42% of the variation in these data. The relationship between visual estimates of topography and species richness is strongest with fish 10–30 cm total length (TL), abundance of fish within this size category also correlating positively with topographic visual assessments. Visual techniques are prone to observer bias, however with regular training they can be used to quickly provide a reliable and effective means of assessing habitat complexity and benthos on coral reefs.

Point counts are the most frequently used technique for sampling bird populations and communities, but have well-known limitations such as inter- and intraobserver errors and limited availability of expert field observers. The use of acoustic recordings to survey birds offers solutions to these limitations. We designed a Soundscape Recording System (SRS) that combines a four-channel, discrete microphone system with a quadraphonic playback system for surveying bird communities. We compared the effectiveness of SRS and point counts for estimating species abundance, richness, and composition of riparian breeding birds in California by comparing data collected simultaneously using both methods. We used the temporal-removal method to estimate individual bird detection probabilities and species abundances using the program MARK. Akaike's Information Criterion provided strong evidence that detection probabilities differed between the two survey methods and among the 10 most common species. The probability of detecting birds was higher when listening to SRS recordings in the laboratory than during the field survey. Additionally, SRS data demonstrated a better fit to the temporal-removal model assumptions and yielded more reliable estimates of detection probability and abundance than point-count data. Our results demonstrate how the perceptual constraints of observers can affect temporal detection patterns during point counts and thus influence abundance estimates derived from time-of-detection approaches. We used a closed-population capture–recapture approach to calculate jackknife estimates of species richness and average species detection probabilities for SRS and point counts using the program CAPTURE. SRS and point counts had similar species richness and detection probabilities. However, the methods differed in the composition of species detected based on Jaccard's similarity index. Most individuals (83%) detected during point counts vocalized at least once during the survey period and were available for detection using a purely acoustic technique, such as SRS. SRS provides an effective method for surveying bird communities, particularly when most species are detected by sound. SRS can eliminate or minimize observer biases, produce permanent records of surveys, and resolve problems associated with the limited availability of expert field observers. Los conteos por punto son la técnica más utilizada para muestrear poblaciones y comunidades de aves, pero tienen limitaciones tales como los errores intra- e inter-observador y la escasa disponibilidad de observadores expertos. Una solución a estas limitaciones es el uso de grabaciones en el campo para censar aves. Diseñamos un sistema de grabación de paisajes sonoros, SRS (por sus siglas en inglés), que combina un sistema de grabación de cuatro canales discretos con un sistema de reproducción cuadrafónica para el censado de comunidades de aves. Comparamos la efectividad del SRS y la de los conteos por punto para estimar la abundancia, riqueza y composición de comunidades de aves en hábitats riparios en California, comparando los datos colectados simultáneamente con ambas técnicas. Utilizamos el método de remoción-temporal para estimar la probabilidad de detección de individuos y abundancia de especies utilizando el programa MARK. El Criterio de Información Akaike proveyó evidencia sustancial de que existen diferencias en la probabilidad de detección entre ambos métodos y entre las 10 especies más comunes de aves. La probabilidad de detección de aves resultó mayor cuando se escucharon las grabaciones del SRS en el laboratorio que durante los conteos por punto en el campo. Además, los datos tomados con la técnica SRS demostraron estar mejor adaptados a los supuestos del modelo de remoción-temporal y ofrecieron estimados más confiables de abundancia y de probabilidad de detección que los datos de conteos por puntos. Nuestros resultados demuestran cómo la percepción de los observadores puede afectar los patrones de detección temporal durante los conteos por punto, influyendo en las estimaciones de abundancia derivadas de procedimientos de tiempo de detección. Utilizamos el método de captura-recaptura de población cerrada para calcular las estimaciones de jackknife de riqueza de especies y la probabilidad de detección de especies para SRS y conteos por punto, utilizando el programa CAPTURE. Ambos tuvieron índices de riqueza de especies y probabilidades de detección similares. Sin embargo, los métodos difirieron en la composición de especies detectadas basados en el índice de similitud de Jaccard. La mayoría de los individuos detectados durante los conteos por punto (83%) vocalizaron al menos una vez durante el periodo de censado, y estuvieron disponibles para ser detectados utilizando una técnica acústica, como SRS. La técnica SRS provee un método efectivo para censar comunidades de aves, particularmente cuando la mayoría de las especies son detectadas por sus sonidos. El SRS puede eliminar o minimizar los sesgos del observador, producir registros permanentes de los censos, y resolver problemas asociados a la limitada disponibilidad de expertos en el campo.