Sampling Bias Correction Research Articles

Bias correction in species distribution models based on geographic and environmental characteristics

Correcting sampling bias in species distribution models (SDMs) is challenging. The difficulty lies in accurately identifying and quantifying bias and the scarcity of samples, which greatly impedes the implementation of bias correction. Current methods often adjust the distribution of presence or background points within geographic or environmental spaces to correct the sampling bias in probability estimation within SDMs. However, these methods may lead to information loss, rely on subjective assumptions, and often separate geography and environment when correcting for bias. This study proposes a novel and easily implementable method termed “aggregation background.” This method selects background data based on the aggregation degree of presence points in the geographic and environmental feature space, thereby approximating the representation and correction of sampling bias in the presence samples. We compared this new method with other prevalent sampling bias correction methods in the existing literature by analyzing ecological authenticity. Under varying biases and sample sizes, the aggregation background and geographic filtering methods achieved more accurate species distribution predictions compared to the target group background and other methods. Notably, when the sample size was small (≤70), the aggregation background was superior to that obtained using the geographic filtering method. These findings underscore the effectiveness of the aggregation background in improving bias correction using limited available presence sample data, without relying on assumptions about sampling bias. Our method provides a new approach for correcting complex unknown biases in SDMs.

A practical approach to making use of uncertain species presence-only data in ecology: Reclassification, regularization methods and observer bias

Various statistical models and software platforms aim to produce species distribution models to better predict where species occur as a function of the environment. However, there are many practical challenges that arise with observations coming from opportunistic surveys. Such data may be of low quality with respect to accuracy and may also exhibit sampling bias. Here, we explore three main challenges. First, species identification can be misleading with the changes in taxonomy where the identification of species has changed for some genus, rendering older records confounded with respect to species identity. Second, the observers' sampled pattern may not reflect the true species distribution as some observers may favor some areas where the species is found. Furthermore, ecological knowledge of environmental drivers of a species distribution may be limited, which presents challenges in selecting appropriate covariates to include in species distribution models. In this paper, we extend two algorithms we recently developed which make use of misidentified observations in order to predict species distributions using spatial point processes. In particular, these algorithms incorporate sampling bias correction and address potential overfitting of the model via lasso-type penalties. We compare the performance of these algorithms to models which do not make use of the confounded species data, and explore the effects of the lasso penalty and bias correction on model performance. We apply the best performing methods to a real dataset of eastern Australian frogs for which taxonomy recently changed. Including confounded observations in the models is particularly relevant for informing management decisions regarding endangered species and species in remote areas.

Evaluating sampling bias correction methods for invasive species distribution modeling in Maxent

As globalization continues, the spread of invasive species is accelerating, posing a severe threat to native biodiversity. To manage such species, reduce their negative impact on native biota and utilize management costs efficiently, a profound understanding of their geographical distribution pattern is mandatory. In this study, the species distribution model Maxent was used to predict the potential spatial distribution of U. europaeus. To account for sampling bias, three bias correction methods were applied, including a novel approach to increase the number of presence points by sampling occurrences based on satellite images. Furthermore, a decision structured process was used to evaluate and select optimal Maxent parameterization and account for limitations of single evaluation criteria. The currently suitable area of U. europaeus is primarily distributed in the coastal and central regions of Chilean natural region Zona Sur in south-central Chile. Annual mean temperature (bio1), annual precipitation (bio12), and precipitation seasonality (bio15) were the most important environmental variables that affected the distribution of U. europaeus. The sampling of additional presence points could effectively correct for sampling bias in species occurrence data. The use of a decision structured process for model evaluation proved to be useful in determining optimal model parameterization for decreased model complexity. This study highlights the importance of optimized Maxent calibrations to yield results as accurately as possible. The predicted suitable habitats can inform nature conservation planners and landscape managers to guide and prioritize conservation measures.

Assessing the effect of sample bias correction in species distribution models

1. Open-source biodiversity databases contain a large number of species occurrence records but are often spatially biased; which affects the reliability of species distribution models based on these records. Sample bias correction techniques require data filtering which comes at the cost of record numbers, or require considerable additional sampling effort. Since independent data is rarely available, assessment of the correction technique often relies solely on performance metrics computed using subsets of the available – biased – data, which may prove misleading.2. Here, we assess the extent to which an acknowledged sample bias correction technique is likely to improve models’ ability to predict species distributions in the absence of independent data. We assessed variation in model predictions induced by the aforementioned correction and model stochasticity; the variability between model replicates related to a random component (pseudo-absences sets and cross-validation subsets). We present, then, an index of the effect of correction relative to model stochasticity; the Relative Overlap Index (ROI). We investigated whether the ROI better represented the effect of correction than classic performance metrics (Boyce index, cAUC, AUC and TSS) and absolute overlap metrics (Schoener’s D, Pearson’s and Spearman’s correlation coefficients) when considering data related to 64 vertebrate species and 21 virtual species with a generated sample bias.3. When based on absolute overlaps and cross-validation performance metrics, we found that correction produced no significant effects. When considering its effect relative to model stochasticity, the effect of correction was strong for most species at one of the three sites. The use of virtual species enabled us to verify that the correction technique improved both distribution predictions and the biological relevance of the selected variables at the specific site, when these were not correlated with sample bias patterns.4. In the absence of additional independent data, the assessment of sample bias correction based on subsample data may be misleading. We propose to investigate both the biological relevance of environmental variables selected, and, the effect of sample bias correction based on its effect relative to model stochasticity.

Fin whale (Balaenoptera physalus) distribution modeling on their Nordic and Barents Seas feeding grounds

AbstractUnderstanding cetacean distribution is essential for conservation planning and decision‐making, particularly in regions subject to rapid environmental changes. Nevertheless, information on their spatiotemporal distribution is commonly limited, especially from remote areas. Species distribution models (SDMs) are powerful tools, relating species occurrences to environmental variables to predict the species' potential distribution. This study aims at using presence‐only SDMs (MaxEnt) to identify suitable habitats for fin whales (Balaenoptera physalus) on their Nordic and Barents Seas feeding grounds. We used spatial‐block cross‐validation to tune MaxEnt parameters and evaluate model performance using spatially independent testing data. We considered spatial sampling bias correction using four methods. Important environmental variables were distance to shore and sea ice edge, variability of sea surface temperature and sea surface salinity, and depth. Suitable fin whale habitats were predicted along the west coast of Svalbard, between Svalbard and the eastern Norwegian Sea, coastal areas off Iceland and southern East Greenland, and along the Knipovich Ridge to Jan Mayen. Results support that presence‐only SDMs are effective tools to predict cetacean habitat suitability, particularly in remote areas like the Arctic Ocean. SDMs constitute a cost‐effective method for targeting future surveys and identifying top priority sites for conservation measures.

Flexsdm: An r package for supporting a comprehensive and flexible species distribution modelling workflow

Abstract Species distribution models (SDM) are widely used in diverse research areas because of their simple data requirements and application versatility. However, SDM outcomes are sensitive to data input and methodological choices. Such sensitivity and diverse applications mean that flexibility is necessary to create SDMs with tailored protocols for a given set of data and model use. We introduce the r package flexsdm for supporting flexible species distribution modelling workflows. flexsdm functions and their arguments serve as building blocks to construct a specific modelling protocol for user's needs. The main flexsdm features are modelling flexibility, integration with other modelling tools, simplicity of the objects returned and function speed. As an illustration, we used flexsdm to define a complete workflow for California red fir Abies magnifica. This package provides modelling flexibility by incorporating comprehensive tools structured in three steps: (a) The Pre‐modelling functions that prepare input, for example, sampling bias correction, sampling pseudo‐absences and background points, data partitioning, and reducing collinearity in predictors. (b) The Modelling functions allow fitting and evaluating different modelling approaches, including individual algorithms, tuned models, ensembles of small models and ensemble models. (c) The Post‐modelling functions include tools related to models' predictions, interpolation and overprediction correction. Because flexsdm comprises a large part of the SDM process, from outlier detection to overprediction correction, flexsdm users can delineate partial or complete workflows based on the combination functions to meet specific modelling needs.

Here be dragons: important spatial uncertainty driven by climate data in forecasted distribution of an endangered insular reptile

AbstractThe effect of future climate change is poorly studied in the tropics, especially in mountainous areas, yet species living in these environments are predicted to be strongly affected. Newly available high‐resolution environmental data and statistical methods enable the development of forecasting models, but the uncertainty related to climate models can be strong, which can lead to ineffective conservation actions. Predictive studies aimed at providing conservation guidelines often account for a range of future climate predictions (climate scenarios and global circulation models). However, very few studies consider potential differences related to the source of climate data and/or do not account for spatial information (overlap) in uncertainty assessments. We modelled the environmental suitability for Phelsuma borbonica, an endangered reptile native to Reunion Island. Using two metrics of species range change (difference in overall suitability and spatial overlap), we quantified the uncertainty related to the modelling technique (n = 10), sample bias correction, climate change scenario, global circulation models (GCM) and data source (CHELSA vs. Worldclim). Uncertainty was mainly driven by GCMs when considering overall suitability, while for spatial overlap, the uncertainty related to data source became more important than that of GCMs. The uncertainty driven by sample bias correction and variable selection was much higher when assessed based on the spatial overlap. The modelling technique was a strong driver of uncertainty in both cases. We provide a consensus ensemble prediction map of the environmental suitability of P. borbonica to identify the areas predicted to be the most suitable in the future with the highest certainty. Predictive studies aimed at identifying priority areas for conservation in the face of climate change need to account for a wide panel of modelling techniques, GCMs and data sources. We recommend the use of multiple approaches, including spatial overlap when assessing uncertainty in species distribution models.

Is geographic sampling bias representative of environmental space?

Species occurrence data from public repositories are widely used in biogeography, and conservation research. However, these data are prone to several sampling biases limiting their usefulness in biodiversity studies, and particularly in species distribution models (SDMs). Specifically, a geographic sampling bias can lead to overfitted SDMs while an environmental sampling bias can affect the estimate of the true environmental space occupied by a species. Several methods can be used to correct for sampling biases, but most of these are focused on the geographic component only. We aim to assess if sampling bias correction based on geographical factors can represent a viable solution also for environmental bias. We focused our analyses on plants and invertebrates in Africa. To model sampling bias in the spatial dimension we estimated sampling rates as a function of distance from a set of potential bias factors (roads, cities, waterbodies, and coastline) in a Bayesian framework. To model sampling bias in the environmental dimension we calculated the environmental multivariate distance between the average climate in the species' occurrences and the rest of the study area by using a set of bioclimatic variables. We quantified the spatial relationship between geographic and environmental bias calculating a Local Indicator of Multivariate Spatial Association (LISA), highlighting local clusters of spatial correlation. For both plants and invertebrates, almost 50% of Africa showed a non-significant relationship between the geographical and the environmental bias components, indicating the absence of a direct correspondence between the two types of biases. Furthermore, from 20.4% to 24.4% of the study area showed an opposite pattern between the two biases, clearly indicating that a correction in the spatial component would be detrimental in the environmental component. Our findings showed that geographical factors cannot be considered as good proxies for environmental bias and, consequently, we suggest considering both geographic and environmental corrections when sampling bias is a problem.

Static species distribution models in the marine realm: The case of baleen whales in the Southern Ocean

AbstractAimInformation on the spatio‐temporal distribution of marine species is essential for developing proactive management strategies. However, sufficient information is seldom available at large spatial scales, particularly in polar areas. The Southern Ocean (SO) represents a critical habitat for various species, particularly migratory baleen whales. Still, the SO’s remoteness and sea ice coverage disallow obtaining sufficient information on baleen whale distribution and niche preference. Here, we used presence‐only species distribution models to predict the circumantarctic habitat suitability of baleen whales and identify important predictors affecting their distribution.LocationThe Southern Ocean (SO).MethodsWe used Maxent to model habitat suitability for Antarctic minke, Antarctic blue, fin and humpback whales. Our models employ extensive circumantarctic data and carefully prepared predictors describing the SO’s environment and two spatial sampling bias correction options. Species‐specific spatial‐block cross‐validation was used to optimize model complexity and for spatially independent model evaluation.ResultsModel performance was high on cross‐validation, with generally little predicted uncertainty. The most important predictors were derived from sea ice, particularly seasonal mean and variability of sea ice concentration and distance to the sea ice edge.Main conclusionsOur models support the usefulness of presence‐only models as a cost‐effective tool in the marine realm, particularly for studying the migratory whales’ distribution. However, we found discrepancies between our results and (within) results of similar studies, mainly due to using different species data quality and quantity, different study area extent and methodological reasons. We further highlight the limitations of implementing static distribution models in the highly dynamic marine realm. Dynamic models, which relate species information to environmental conditions contemporaneous to species occurrences, can predict near‐real‐time habitat suitability, necessary for dynamic management. Nevertheless, obtaining sufficient species and environmental predictors at high spatio‐temporal resolution, necessary for dynamic models, can be challenging from polar regions.

Comparing sample bias correction methods for species distribution modeling using virtual species

AbstractA key assumption in species distribution modeling (SDM) with presence‐background (PB) methods is that sampling of occurrence localities is unbiased and that any sampling bias is proportional to the background distribution of environmental covariates. This assumption is rarely met when SDM practitioners rely on federated museum records from natural history collections for geo‐located occurrences due to inherent sampling bias found in these collections. We use a simulation approach to explore the effectiveness of three methods developed to account for sampling bias in SDM with PB frameworks. Two of the methods rely on careful filtering of observation data—geographic thinning (G‐Filter) and environmental thinning (E‐Filter)—while a third, FactorBiasOut, creates selection weights for background data to bias locations toward areas where the observation dataset was sampled. While these methods have been assessed previously, evaluation has emphasized spatial predictions of habitat potential. Here, we dig deeper into the effectiveness of these methods by exploring how sampling bias not only affects predictions of habitat potential, but also our understanding of niche characteristics such as which explanatory variables and response curves best represent species–environment relationships. We simulate 100 virtual species ranging from generalist to specialist in their habitat preferences and introduce geographic and environmental bias at three intensity levels to measure the effectiveness of each correction method to (1) predict true probability of occurrence across a study area, (2) recover true species–environment relationships, and (3) identify true explanatory variables. We find that the FactorBiasOut most often showed the greatest improvement in recreating known distributions but did no better at correctly identifying environmental covariates or recreating species–environment relationships than G‐Filter or E‐Filter methods. Narrow niche species are most problematic for biased calibration datasets, such that correction methods can, in some cases, make predictions worse.

Spatial distribution of the pine marten (Martes martes) and stone marten (Martes foina) in the Italian Alps

The pine marten (Martes martes) and stone marten (Martes foina) are considered similar carnivores in terms of morphology, habitat requirements, diet and distribution. Despite their range overlap widely throughout Europe, few studies have analysed their spatial distribution in sympatry. With the aim of investigating the environmental factors which may enhance their coexistence, we focused on the Italian sector of the Alpine biogeographical region, an area offering homogeneous environmental conditions and resources. Recent, presence-only data were collected through a collaboration network and ad hoc surveys for faecal DNA, yielding a total of 1658 and 329 records for stone marten and pine marten, respectively. Species distribution models (SDM) were run using MaxEnt with sampling bias correction. Three detailed models were fitted, each considering different landscape descriptors (land use, topography and climate) and, finally, an overall model was developed for each species. The pine marten selected forested areas, particularly coniferous forests, at intermediate altitude, while lower-altitude belts, partially including cultivated and urban areas, were more suitable for the stone marten. Both martens avoided open areas. Moreover, the thermophilic stone marten preferred warmer and drier areas than the pine marten. About 50% of the Alpine area was filed as suitable for each species, with 34.6% of cells resulting suitable for both martens. Differential use of space, probably associated with segregation along with other dimensions of their niches (e.g. the temporal niche), lowers resource competition and enhances the coexistence of these strictly-related species.

A simple correction for COVID-19 sampling bias

COVID-19 testing has become a standard approach for estimating prevalence which then assist in public health decision making to contain and mitigate the spread of the disease. The sampling designs used are often biased in that they do not reflect the true underlying populations. For instance, individuals with strong symptoms are more likely to be tested than those with no symptoms. This results in biased estimates of prevalence (too high). Typical post-sampling corrections are not always possible. Here we present a simple bias correction methodology derived and adapted from a correction for publication bias in meta analysis studies. The methodology is general enough to allow a wide variety of customization making it more useful in practice. Implementation is easily done using already collected information. Via a simulation and two real datasets, we show that the bias corrections can provide dramatic reductions in estimation error.

Meta-replication, sampling bias, and multi-scale model selection: A case study on snow leopard (Panthera uncia) in western China.

Replicated multiple scale species distribution models (SDMs) have become increasingly important to identify the correct variables determining species distribution and their influences on ecological responses. This study explores multi‐scale habitat relationships of the snow leopard (Panthera uncia) in two study areas on the Qinghai–Tibetan Plateau of western China. Our primary objectives were to evaluate the degree to which snow leopard habitat relationships, expressed by predictors, scales of response, and magnitude of effects, were consistent across study areas or locally landcape‐specific. We coupled univariate scale optimization and the maximum entropy algorithm to produce multivariate SDMs, inferring the relative suitability for the species by ensembling top performing models. We optimized the SDMs based on average omission rate across the top models and ensembles’ overlap with a simulated reference model. Comparison of SDMs in the two study areas highlighted landscape‐specific responses to limiting factors. These were dependent on the effects of the hydrological network, anthropogenic features, topographic complexity, and the heterogeneity of the landcover patch mosaic. Overall, even accounting for specific local differences, we found general landscape attributes associated with snow leopard ecological requirements, consisting of a positive association with uplands and ridges, aggregated low‐contrast landscapes, and large extents of grassy and herbaceous vegetation. As a means to evaluate the performance of two bias correction methods, we explored their effects on three datasets showing a range of bias intensities. The performance of corrections depends on the bias intensity; however, density kernels offered a reliable correction strategy under all circumstances. This study reveals the multi‐scale response of snow leopards to environmental attributes and confirms the role of meta‐replicated study designs for the identification of spatially varying limiting factors. Furthermore, this study makes important contributions to the ongoing discussion about the best approaches for sampling bias correction.

Spatial conservation prioritisation in data-poor countries: a quantitative sensitivity analysis using multiple taxa

BackgroundSpatial conservation prioritisation (SCP) is a set of computational tools designed to support the efficient spatial allocation of priority areas for conservation actions, but it is subject to many sources of uncertainty which should be accounted for during the prioritisation process. We quantified the sensitivity of an SCP application (using software Zonation) to possible sources of uncertainty in data-poor situations, including the use of different surrogate options; correction for sampling bias; how to integrate connectivity; the choice of species distribution modelling (SDM) algorithm; how cells are removed from the landscape; and two methods of assigning weights to species (red-list status or prediction uncertainty). Further, we evaluated the effectiveness of the Egyptian protected areas for conservation, and spatially allocated the top priority sites for further on-the-ground evaluation as potential areas for protected areas expansion.ResultsFocal taxon (butterflies, reptiles, and mammals), sampling bias, connectivity and the choice of SDM algorithm were the most sensitive parameters; collectively these reflect data quality issues. In contrast, cell removal rule and species weights contributed much less to overall variability. Using currently available species data, we found the current effectiveness of Egypt’s protected areas for conserving fauna was low.ConclusionsFor SCP to be useful, there is a lower limit on data quality, requiring data-poor countries to improve sampling strategies and data quality to obtain unbiased data for as many taxa as possible. Since our sensitivity analysis may not generalise, conservation planners should use sensitivity analyses more routinely, particularly relying on more than one combination of SDM algorithm and surrogate group, consider correction for sampling bias, and compare the spatial patterns of predicted priority sites using a variety of settings. The sensitivity of SCP to connectivity parameters means that the responses of each species to habitat loss are important knowledge gaps.

Identification of homogeneous region boundaries of fractured rock masses in candidate sites for Chinese HLW repository

Identification of homogeneous region boundaries in a fractured rock mass is the basis of statistics and modeling of discontinuities. In engineering practice, an objective division can be obtained by adopting the major influential factors as indicators and the optimal approach as a tool. For discontinuities in Beishan, a main candidate site for Chinese high-level radioactive waste (HLW) repository, pretreatment techniques (e.g., sampling window truncation, sampling bias correction and block-net variation correction) were used to deal with field data. Programming was then applied to realize homogeneous region division via several methods (e.g., the improved Miller’s method, Mahtab and Yegulalp’s method, and correlation coefficient method). This study preliminarily examined these methods’ distinguishing capability, applicability, and limitations. Results showed that the applicability of the correlation coefficient method as well as the Mahtab and Yegulalp’s method was weak; the improved Miller’s method appeared most satisfactory, especially with a large-area strategy (34 blocks). Finally, two statistical homogeneous regions were obtained by applying the optimal approach to the Jijicao block. Findings can offer guidance for subsequent research on discrete fracture network (DFN) modeling and seepage path simulation, which is important for the prediction and evaluation of radionuclide migration in rock masses.

Correcting the effect of sampling bias in species distribution modeling – A new method in the case of a low number of presence data

Species distribution models that only require presence data provide potentially inaccurate results due to sampling bias and presence data scarcity. Methods have been proposed in the literature to minimize the effects of sampling bias, but without explicitly considering the issue of sample size.A new method developed to better take into account environmental biases in a context of data scarcity is proposed here. It is compared to other sampling bias correction methods primarily used in the literature by analyzing their absolute and relative impacts on model performances.Results showed that the number of presence sites is critical for selecting the applicable method. The method proposed was regularly placed in the first or second rank and tends to be more proficient than other methods in the context of presence site scarcity (〈100). It tends to improve results regarding environment-based performance indexes. Eventually, its parametrization, requiring background knowledge on species bio-ecology, appears to be more robust and convenient to perform than those based on geographical criteria.

Evaluating Private Equity Performance Using Stochastic Discount Factors

We examine the performance of 2,790 private equity (PE) funds incepted during 1979-2008 using Stochastic Discount Factors (SDFs) implied by the two leading consumption-based asset pricing models (CBAPMs) — external habit and long-run risks — as their assumptions appear consistent with investment objectives of avid PE investors. In contrast to CAPM-based inference, venture funds did not destroy value under these CBAPMs in post-2000 vintages and may even have outperformed buyouts and generalists in the full sample. We find that 2007-08 venture vintages provide a better hedge for post-crises consumption shocks than other types of PE, and that the temporal variation in PE excess returns is significantly smaller under CBAPMs. Our contribution is also methodological. We extend the realized risk premia matching insight of Korteweg and Nagel (2016) to a more general class of SDFs, namely portfolio-specific discount factors that reflect non-tradeable assets unspanned by standard benchmarks. To this end, we propose a more efficient estimation of SDF parameters in this context and develop a finite sample bias correction for NPV-based inference with long-duration assets.

Sampling bias correction in species distribution models by quasi-linear Poisson point process

Species distribution modeling has an essential role in ecology to investigate habitat suitability based on the relationship between species occurrences and environmental conditions. The presence-only data for organisms is usually assumed to be obtained randomly from the region of interest; however, it is often the case that it is biased toward the areas easily accessed and adversely affects prediction accuracy. To address this sampling bias problem relevant to the prediction accuracy and model fitting of habitat distributions, we propose a new Poisson point process (PPP) model named as a quasi PPP, where environmental effect and sampling bias are explicitly modeled in separate clusters in a framework of quasi-linear modeling. The quasi-linear modeling is designed for capturing homogeneity within clusters and heterogeneity between clusters to improve the estimation accuracy of species distribution. The proposed model includes conventional models such as thinned and superposed PPPs as special cases. We have found that the quasi PPP outperforms the other existing methods in terms of goodness of model fitting. A statistical index based on the quasi-linear modeling is proposed to measure how the presence-only data used for the estimation of the species habitat distribution is affected by the sampling bias. The utility of the quasi PPP has been illustrated using simulation studies as well as the comprehensive vascular plant data in Japan. Our proposed model flexibly incorporates the effect of sampling biases to improve the estimation accuracy of species distributions. The results of data analysis are easily reproducible and applications to other data sets are also easily implementable by a package of qPPP of a statistical software R.

Quantifying international human mobility patterns using Facebook Network data.

Quantifying global international mobility patterns can improve migration governance. Despite decades of calls by the international community to improve international migration statistics, the availability of timely and disaggregated data about long-term and short-term migration at the global level is still very limited. In this study, we investigate the feasibility of using non-traditional data sources to fill existing gaps in migration statistics. To this end, we use anonymised and publicly available data provided by Facebook’s advertising platform. Facebook’s advertising platform classifies its users as “lived in country X” if they previously lived in country X, and now live in a different country. Drawing on statistics about Facebook Network users (Facebook, Instagram, Messenger, and the Audience Network) who have lived abroad and applying a sample bias correction method, we estimate the number of Facebook Network (FN) “migrants” in 119 countries of residence and in two time periods by age, gender, and country of previous residence. The correction method estimates the probability of a person being a FN user based on age, sex, and country of current and previous residence. We further estimate the correlation between FN-derived migration estimates and reference official migration statistics. By comparing FN-derived migration estimates in two different time periods, January-February and August-September 2018, we successfully capture the increase in Venezuelan migrants in Colombia and Spain in 2018. FN-derived migration estimates cannot replace official migration statistics, as they are not representative, and the exact methods the FN uses for classifying its users are not known, and might change over time. However, after carefully assessing the validity of the FN-derived estimates by comparing them with data from reliable sources, we conclude that these estimates can be used for trend analysis and early-warning purposes.

Lung Density Changes Following SBRT: Association with Lung Dose and Clinical Pneumonitis

Radiation pneumonitis is an inflammation of the lungs that can develop following radiation therapy, the cause of which is not fully understood. The hypothesis of this study is that radiation induced lung changes are associated with radiation dose to the lung and pneumonitis. This work investigates the relationship between lung dose and the development of radiographic lung changes following stereotactic body radiotherapy (SBRT). In addition, the lung dose and radiographic changes were investigated for their ability to predict the development of clinical grade ≥2 radiation pneumonitis (RP≥2). Regions of consolidation (dense) and ground-glass opacities (hazy) were semi-automatically contoured on 262 3- and 6-month post treatment CT-scans for 145 subjects who underwent SBRT between 2008 and 2018 receiving 40-60 Gy in 3-10 fractions. 84% of the subjects received 48Gy in 4 fractions. RP≥2 was diagnosed in 9 patients (6%). The PTV volume, relative and absolute lung volumes irradiated within defined dose bins were extracted from the treatment plans. Volumes of dense and hazy radiographic changes, both absolute volumes in mL and volumes normalized by the PTV volume were assessed at 3- and 6-months. Correlations for all 3- and 6-month hazy and dense volumes with PTV volumes, relative and absolute lung volumes in defined dose bins, and treatment technique (3D conformal radiotherapy (3DCRT), static intensity modulated radiotherapy (IMRT), or volumetrically modulated arc therapy (VMAT)) were established through Spearman correlation using Bonferroni correction. Univariate logistic models, using the Firth small sample bias correction, were developed for predicting RP≥2. PTV volume and relative lung dose bins between 25 and 40 Gy were significantly correlated with the hazy and dense volumes on both the 3 and 6 month images (p=<0.0001-0.0019). There was no significant difference in the hazy or dense volumes at either 3- or 6-months by the treatment technique. Subjects developing RP≥2 on average demonstrated a 2- to 4-fold increase in the hazy and dense absolute and normalized volumes at both 3- and 6-months (n.s.). The 3-month dense (p=0.03), 6-month dense (p=0.02), 6-month dense normalized (p=0.04), and 6-month hazy volumes (p=0.03) were significantly correlated with RP≥2 in univariate logistic models (AUC 0.64-0.67). Lung V20-40Gyappears to be correlated with radiographic changes following lung SBRT. While irradiated lung volumes were not correlated with RP≥2, the amount of dense volume present at 3 and 6 months in particular appears to be related to the development of RP≥2. This observation indicates that in addition to dose, demographic and clinical factors may influence the development of RP and need to be further investigated.