Effects Of Preferential Sampling Research Articles

The Peregrine population study in the French Jura mountains 1964–2016: use of occupancy modeling to estimate population size and analyze site persistence and colonization rates

AbstractWe summarize key results of the first 53 years of one of the longest-running avian population studies in the world, on the Peregrine Falcon(Falco peregrinus),in the French Jura mountains (12,714 km2), launched in 1964. A total of 449 cliff sites in 338 potential Peregrine territories were surveyed: 287 (85%) of these territories were occupied by an adult pair at least once, while in 51 (15%) we never detected an adult pair. Most sites were visited several times during a breeding season to survey occupancy and later fecundity, but the proportion of sites visited was highly variable over the years. We highlight the power of the Bayesian implementation of site-occupancy models (MacKenzieet al.2002, 2003) to analyze data from raptor population studies: to correct population size estimates for sites not visited in a given year and for the biasing effects of preferential sampling (when better sites are more likely to be checked). In addition, these models allow estimation and modeling of the site-level persistence and colonization rates, which can provide important clues about drivers of population dynamics, even without individually marking any birds. Changes in the dynamics rates may serve as early-warning signals for subsequent population declines.Since 1964, the observed number of adult pairs varied between 17 in 1972 and 196 in 2008, but the proportion of sites visited increased from 43% in 1964 to 80–90% after 2002. Hence, this raw population total must be an underestimate. We found strong evidence for preferential sampling in our study. Correcting for this, we estimated 56 pairs in 1964, after which the population dropped to a minimum of 18 in 1972, but then recovered rapidly, leveling off somewhat around 1995 and reaching a maximum of 200–210 adult pairs during 2000–2012. This was then followed by a decline to 170–190 pairs. In any one year, the raw counts underestimated the true population size by 5–39% (mean 11%), due to sites not being visited (this correction ignores imperfect detection though). Site persistence rates declined from 78% to less than 60% during 1967–1972, and then increased rapidly to over 90% during 1980–1990, suggesting that once pesticide effects vanished, individual survival probability increased rapidly and as a consequence also site persistence. Since the 1990s, persistence has declined slowly, which may indicate decreasing adult survival. In contrast, colonization rates increased steadily from about 3% in the early years to maxima of 46–49% during 1994–2001, but declined thereafter and currently reach about 33%. Taller cliffs had greater persistence and colonization rates than medium or small cliffs.Both the decline in colonization and in persistence rates during the last 15 years may reflect density-dependence, predation by the expanding European Eagle Owl(Bubo bubo)population, human persecution or any as yet unknown factors. Importantly, we note that both persistence and colonization rates began to decline many years before the recent population decline became apparent. Thus, analysis of population studies using dynamic occupancy models can provide early-warning signals for future population declines. Our study demonstrates the benefits of modern analytical methods that can correct for several key deficiencies in probably all raptor population studies: incomplete coverage of sites and imperfect detection (though we only dealt with the former here). Occupancy models, possibly accounting for preferential sampling, appear to represent the logical analytical framework for abundance in raptor population studies.

Breaking down the lithification bias: the effect of preferential sampling of larger specimens on the estimate of species richness, evenness, and average specimen size

AbstractLithification, the transition of unconsolidated sediments to fully indurated rocks, can potentially bias estimates of species richness, evenness, and body size distribution derived from fossil assemblages. Fossil collections made from well-indurated rocks consistently exhibit lower species richness, lower evenness, and larger average specimen size relative to collections made from unconsolidated sediments, even when collections are drawn from the same assemblage. This phenomenon is known as “lithification bias.” While the bias itself has been demonstrated empirically, much less attention has been paid to its causes. Proposed causes include taphonomic processes (e.g., destruction of small specimens during early diagenesis) and methodological differences (e.g., sieving vs. counting specimens on outcrops, bedding surfaces, or mechanically split surfaces). Here we investigate the potential effects of preferential intersection that could also result in a methodologically related bias: the preferential sampling of larger specimens relative to smaller ones when fossils are counted on rock surfaces. We used an analogue model to simulate preferential intersection (fossil collection via splitting fossiliferous rocks) and compare the results with a random-draw model that approximates the effects of sieving. The model was parameterized using nine different combinations of species abundance and species size distributions. The results show that, with rare exceptions, species richness is 5–23% lower, evenness 5–25% lower, and average specimen size 24–150% larger in preferential-intersection than in random-draw simulations. We conclude that preferential intersection can impose a significant bias independent of other mechanisms (e.g., preferential destruction of smaller specimens during diagenetic or sampling processes), that the magnitude of this bias is partially dependent on the species abundance and size distributions, and that this bias alone does not fully account for empirically observed lithification bias on species richness (i.e., other sources of bias are also at work).

Confronting preferential sampling when analysing population distributions: diagnosis and model‐based triage

SummaryPopulation surveys are often used to estimate the density, abundance, or distribution of natural populations. Recently, model‐based approaches to analyzing survey data have become popular because one can more readily accommodate departures from pre‐planned survey routes and construct more detailed maps than one can with design‐based procedures.Spatial models for population distributions (SMPDs) often make the implicit assumption that locations chosen for sampling and animal abundance at those locations are conditionally independent given modelled covariates. However, this assumption may be violated when survey effort is non‐randomized, leading to preferential sampling.We develop a hierarchical statistical modelling framework for detecting and alleviating the biasing effects of preferential sampling in spatial distribution models fitted to count data. The approach works by specifying a joint model for population density and the locations selected for sampling, and specifying a dependent correlation structure between the two processes.Using simulation, we show that moderate levels of preferential sampling can lead to large (e.g. 40%) bias in estimates of animal density and that our modelling approach can considerably reduce this bias. In contrast, preferential sampling did not appear to bias inferences about parameters informing species–habitat relationships (i.e. slope parameters).We apply our approach to aerial survey counts of bearded seals (Erignathus barbatus) in the eastern Bering Sea. As expected, models with a preferential sampling effect led to lower abundance than those without. However, several lines of reasoning (better predictive performance, higher biological realism) led us to prefer models without a preferential sampling effect for this dataset.When population surveys break from traditional scientific survey design principles, ecologists should recognize the potentially biasing effects of preferential sampling when estimating population density or occurrence. Joint models, such as those described in this paper, can be used to test and correct for such biases. However, such models can be unstable; ultimately the best way to avoid preferential sampling bias is to incorporate design‐based principles such as randomization and/or systematic sampling into survey design.

The effect of Reynolds number on inertial particle dynamics in isotropic turbulence. Part 1. Simulations without gravitational effects

In this study, we analyse the statistics of both individual inertial particles and inertial particle pairs in direct numerical simulations of homogeneous isotropic turbulence in the absence of gravity. The effect of the Taylor microscale Reynolds number, $R_{{\it\lambda}}$, on the particle statistics is examined over the largest range to date (from $R_{{\it\lambda}}=88$ to 597), at small, intermediate and large Kolmogorov-scale Stokes numbers $St$. We first explore the effect of preferential sampling on the single-particle statistics and find that low-$St$ inertial particles are ejected from both vortex tubes and vortex sheets (the latter becoming increasingly prevalent at higher Reynolds numbers) and preferentially accumulate in regions of irrotational dissipation. We use this understanding of preferential sampling to provide a physical explanation for many of the trends in the particle velocity gradients, kinetic energies and accelerations at low $St$, which are well represented by the model of Chun et al. (J. Fluid Mech., vol. 536, 2005, pp. 219–251). As $St$ increases, inertial filtering effects become more important, causing the particle kinetic energies and accelerations to decrease. The effect of inertial filtering on the particle kinetic energies and accelerations diminishes with increasing Reynolds number and is well captured by the models of Abrahamson (Chem. Engng Sci., vol. 30, 1975, pp. 1371–1379) and Zaichik & Alipchenkov (Intl J. Multiphase Flow, vol. 34 (9), 2008, pp. 865–868), respectively. We then consider particle-pair statistics, and focus our attention on the relative velocities and radial distribution functions (RDFs) of the particles, with the aim of understanding the underlying physical mechanisms contributing to particle collisions. The relative velocity statistics indicate that preferential sampling effects are important for $St\lesssim 0.1$ and that path-history/non-local effects become increasingly important for $St\gtrsim 0.2$. While higher-order relative velocity statistics are influenced by the increased intermittency of the turbulence at high Reynolds numbers, the lower-order relative velocity statistics are only weakly sensitive to changes in Reynolds number at low $St$. The Reynolds-number trends in these quantities at intermediate and large $St$ are explained based on the influence of the available flow scales on the path-history and inertial filtering effects. We find that the RDFs peak near $St$ of order unity, that they exhibit power-law scaling for low and intermediate $St$ and that they are largely independent of Reynolds number for low and intermediate $St$. We use the model of Zaichik & Alipchenkov (New J. Phys., vol. 11, 2009, 103018) to explain the physical mechanisms responsible for these trends, and find that this model is able to capture the quantitative behaviour of the RDFs extremely well when direct numerical simulation data for the structure functions are specified, in agreement with Bragg & Collins (New J. Phys., vol. 16, 2014a, 055013). We also observe that at large $St$, changes in the RDF are related to changes in the scaling exponents of the relative velocity variances. The particle collision kernel closely matches that computed by Rosa et al. (New J. Phys., vol. 15, 2013, 045032) and is found to be largely insensitive to the flow Reynolds number. This suggests that relatively low-Reynolds-number simulations may be able to capture much of the relevant physics of droplet collisions and growth in the adiabatic cores of atmospheric clouds.

Mitigating the effects of preferentially selected monitoring sites for environmental policy and health risk analysis

The potential effects of air pollution are a major concern both in terms of the environment and in relation to human health. In order to support both environmental and health policy there is a need for accurate estimates of the exposures that populations might experience. The information for this typically comes from environmental monitoring networks but often the locations of monitoring sites are preferentially located in order to detect high levels of pollution. Using the information from such networks has the potential to seriously affect the estimates of pollution that are obtained and that might be used in health risk analyses. In this context, we explore the topic of preferential sampling within a long–standing network in the UK that monitored black smoke due to concerns about its effect on public health, the extent of which came to prominence during the famous London fog of 1952. Abatement measures led to a decline in the levels of black smoke and a subsequent reduction in the number of monitoring locations that were thought necessary to provide the information required for policy support. There is evidence of selection bias during this process with sites being kept in the most polluted areas. We assess the potential for this to affect the estimates of risk associated air pollution and show how using Bayesian spatio–temporal exposure models may be used to attempt to mitigate the effects of preferential sampling in this case.

Quantifying and Mitigating the Effect of Preferential Sampling on Phylodynamic Inference.

Phylodynamics seeks to estimate effective population size fluctuations from molecular sequences of individuals sampled from a population of interest. One way to accomplish this task formulates an observed sequence data likelihood exploiting a coalescent model for the sampled individuals’ genealogy and then integrating over all possible genealogies via Monte Carlo or, less efficiently, by conditioning on one genealogy estimated from the sequence data. However, when analyzing sequences sampled serially through time, current methods implicitly assume either that sampling times are fixed deterministically by the data collection protocol or that their distribution does not depend on the size of the population. Through simulation, we first show that, when sampling times do probabilistically depend on effective population size, estimation methods may be systematically biased. To correct for this deficiency, we propose a new model that explicitly accounts for preferential sampling by modeling the sampling times as an inhomogeneous Poisson process dependent on effective population size. We demonstrate that in the presence of preferential sampling our new model not only reduces bias, but also improves estimation precision. Finally, we compare the performance of the currently used phylodynamic methods with our proposed model through clinically-relevant, seasonal human influenza examples.

Optimal Design in Geostatistics under Preferential Sampling

This paper analyses the effect of preferential sampling in Geostatistics when the choice of new sampling locations is the main interest of the researcher. A Bayesian criterion based on maximizing utility functions is used. Simulated studies are presented and highlight the strong influence of preferential sampling in the decisions. The computational complexity is faced by treating the new local sampling locations as a model parameter and the optimal choice is then made by analysing its posterior distribution. Finally, an application is presented using rainfall data collected during spring in Rio de Janeiro. The results showed that the optimal design is substantially changed under preferential sampling effects. Furthermore, it was possible to identify other interesting aspects related to preferential sampling effects in estimation and prediction in Geostatistics.

Impact of preferential sampling on exposure prediction and health effect inference in the context of air pollution epidemiology.

Preferential sampling has been defined in the context of geostatistical modeling as the dependence between the sampling locations and the process that describes the spatial structure of the data. It can occur when networks are designed to find high values. For example, in networks based on the U.S. Clean Air Act monitors are sited to determine whether air quality standards are exceeded. We study the impact of the design of monitor networks in the context of air pollution epidemiology studies. The effect of preferential sampling has been illustrated in the literature by highlighting its impact on spatial predictions. In this paper, we use these predictions as input in a second stage analysis, and we assess how they affect health effect inference. Our work is motivated by data from two United States regulatory networks and health data from the Multi-Ethnic Study of Atherosclerosis and Air Pollution. The two networks were designed to monitor air pollution in urban and rural areas respectively, and we found that the health analysis results based on the two networks can lead to different scientific conclusions. We use preferential sampling to gain insight into these differences. We designed a simulation study, and found that the validity and reliability of the health effect estimate can be greatly affected by how we sample the monitor locations. To better understand its effect on second stage inference, we identify two components of preferential sampling that shed light on how preferential sampling alters the properties of the health effect estimate.

Catchments as heterogeneous and multi-species reactors: An integral approach for identifying biogeochemical hot-spots at the catchment scale

Summary From a biogeochemical perspective, catchments can be regarded as reactors that transform the input of various substances via precipitation or deposition as they pass through soils and aquifers towards draining streams. Understanding and modeling the variability of solute concentrations in catchment waters require the identification of the prevailing processes, determining their respective contribution to the observed transformation of substances, and the localization of “hot spots”, that is, the most reactive areas of catchments. For this study, we applied a non-linear variant of the Principle Component Analysis, the Isometric Feature Mapping (Isomap), to a data set composed of 1686 soil solution, groundwater and stream water samples and 16 variables (Al, Ca, Cl, Fe, K, Mg, Mn, Na, NH4, NO3, SO4, total S, Si, DOC, electric conductivity and pH values) from the Lehstenbach catchment in Germany. The aim was (i) to assess the contribution of the prevailing biogeochemical processes to the variability of solute concentrations in water samples taken from soils, in groundwater and in stream water in a catchment and (ii) to identify hot spots at the catchment scale with respect to 16 solutes along different flow paths. The first three dimensions of the Isomap analysis explained 48%, 30% and 11%, respectively, i.e. 89% of the variance in the data set. Scores of the first three dimensions could be ascribed to three predominating bundles of biogeochemical processes: (i) redox processes, (ii) acid-induced podzolization, and (iii) weathering processes. In general, the upper 1 m topsoil layer could be considered as hot spots along flow paths from upslope soils and in the wetland, although with varying extents for the different prevailing biogeochemical processes. Nearly 67% and 97% of the variance with respect to redox processes and acid induced podzolization could be traced back to hot spots, respectively, representing less than 2% of the total spatial volume of the catchment. In contrast, biogeochemical processes in deep groundwater had only minor effects on the biogeochemical turnover in catchment waters. Hot spots with respect to weathering processes along upslope and wetland flow paths could not be quantified due to effects of preferential sampling in soil solution samples. Predominating flow paths and biogeochemical processes crucial for the variability of stream water chemistry differed substantially for three streams but were consistent with presumed mixing ratios.

Reducing estimation bias in adaptively changing monitoring networks with preferential site selection

This paper explores the topic of preferential sampling, specifically situations where monitoring sites in environmental networks are preferentially located by the designers. This means the data arising from such networks may not accurately characterize the spatio-temporal field they intend to monitor. Approaches that have been developed to mitigate the effects of preferential sampling in various contexts are reviewed and, building on these approaches, a general framework for dealing with the effects of preferential sampling in environmental monitoring is proposed. Strategies for implementation are proposed, leading to a method for improving the accuracy of official statistics used to report trends and inform regulatory policy. An essential feature of the method is its capacity to learn the preferential selection process over time and hence to reduce bias in these statistics. Simulation studies suggest dramatic reductions in bias are possible. A case study demonstrates use of the method in assessing the levels of air pollution due to black smoke in the UK over an extended period (1970-1996). In particular, dramatic reductions in the estimates of the number of sites out of compliance are observed.

On the effect of preferential sampling in spatial prediction

The choice of the sampling locations in a spatial network is often guided by practical demands. In particular, many locations are preferentially chosen to capture high values of a response, for example, air pollution levels in environmental monitoring. Then, model estimation and prediction of the exposure surface become biased due to the selective sampling. Since prediction is often the main utility of the modeling, we suggest that the effect of preferential sampling lies more importantly in the resulting predictive surface than in parameter estimation. Our contribution is to offer a direct simulation-based approach to assessing the effects of preferential sampling. We compare two predictive surfaces over the study region, one originating from the notion of an 'operating' intensity driving the selection of monitoring sites, the other under complete spatial randomness. We can consider a range of response models. They may reflect the operating intensity, introduce alternative informative covariates, or just propose a flexible spatial model. Then, we can generate data under the given model. Upon fitting the model and interpolating (kriging), we will obtain two predictive surfaces to compare. It is important to note that we need suitable metrics to compare the surfaces and that the predictive surfaces are random, so we need to make expected comparisons.

Null models reveal preferential sampling, spatial autocorrelation and overfitting in habitat suitability modelling

Nowadays, species are driven to extinction at a high rate. To reduce this rate it is important to delineate suitable habitats for these species in such a way that these areas can be suggested as conservation areas. The use of habitat suitability models (HSMs) can be of great importance for the delineation of such areas. In this study MaxEnt, a presence-only modelling technique, is used to develop HSMs for 223 nematode species of the Southern Bight of the North Sea. However, it is essential that these models are beyond discussion and they should be checked for potential errors. In this study we focused on two categories (1) errors which can be attributed to the database such as preferential sampling and spatial autocorrelation and (2) errors induced by the modelling technique such as overfitting, In order to quantify these adverse effects thousands of nulls models were created. The effect of preferential sampling (i.e. some areas where visited more frequenty than others) was investigated by comparing model outcomes based from null models sampling the actual sampling stations and null models sampling the entire mapping area ( Raes and ter Steege, 2007). Overfitting is exposed by a fivefold cross-validation and the influence of spatial autocorrelation is assessed by separating test and training sets in space. Our results clearly show that all these effects are present: preferential sampling has a strong effect on the selection of non-random species models. Crossvalidation seems to have less influence on the model selection and spatial autocorrelation is also strongly present. It is clear from this study that predefined thresholds are not readily applicable to all datasets and additional tests are needed in model selection.