- Email: [email protected]
Preservation of wild bird species in northern Portugal - Effects of anthropogenic pressures in wild bird populations (2008–2017)
Science of the Total Environment 650 (2019) 2996–3006
Contents lists available at ScienceDirect
Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv
Preservation of wild bird species in northern Portugal - Effects of anthropogenic pressures in wild bird populations (2008–2017) Andreia Garcês a, Isabel Pires b,c, Fernando A.L. Pacheco d, Luís F. Sanches Fernandes a, Vanessa Soeiro e, Sara Lóio e, Justina Prada b,c, Rui Cortes a, Felisbina L. Queiroga a,b,⁎ a
Centre for the Research and Technology of Agro-Environment and Biological Sciences, University of Trás-os-Montes and Alto Douro, Vila Real, Portugal Department of Veterinary Science of the University of Trás-os-Montes and Alto Douro, Vila Real, Portugal CECAV - University of Trás-os-Montes and Alto Douro, Vila Real, Portugal d Chemistry Research Centre, University of Trás-os-Montes and Alto Douro, Vila Real, Portugal e Wildlife Rehabilitation Centre in Parque Biológico de Gaia, R. Cunha, Avintes, Portugal b c
H I G H L I G H T S
G R A P H I C A L
A B S T R A C T
• Wild birds in northern Portugal are affected by several anthropogenic factors. • Increased number of medium companies contributed to increased mortality of wild birds. • Habitat destruction and fragmentation is linked to wild bird mortality.
a r t i c l e
i n f o
Article history: Received 6 August 2018 Received in revised form 2 October 2018 Accepted 2 October 2018 Available online 03 October 2018 Editor: Damia Barcelo Keywords: Anthropogenic pressures Mortality North of Portugal Partial least squares-path modelling Wild birds
a b s t r a c t In this study we aim to obtain a holistic view over the consequences of human-induced threats to the wild bird populations in the study area, based on data collected from a Wildlife Rehabilitation Centre (WRC) through the method of Partial Least Squares-Path Modelling (PLS-PM). The study area comprised 76 rural and urban municipalities located in northern Portugal. Within PLS-PM the threats (“anthropogenic pressures”) are termed exogenous latent variables, while the final environmental consequence (wild bird mortality) is termed endogenous latent variable. Latent variables are concepts assessed by numerical parameters. The PLS-PM results identified as most significant pressures, the number of small and medium companies in the municipalities, both in traumatic and nontraumatic deaths. Although the pattern of weights is similar regardless of the general cause of death, traumatic causes seem to play a more prominent role given the larger weights in the relevant specific causes (number of companies). The high coefficients of determination (R2 N 0,8) reveal that the variance of wild bird mortality is largely explained by the variance of the pressures, which indicates a cause-effect relationship between the independent (pressures) and dependent (mortality) variables. The ample coverage of northern Portugal with a huge dataset suggests that this cause-effect relationship is typical from this region. The use of a sophisticated statistical method PLSPM and its incorporation into a Geographic Information System (GIS) revealed to be an important tool for analysing wildlife impacts of environmental and human factors. The results obtained with this model indicate that there is a substantial negative impact of human activity on wild bird mortality. © 2018 Published by Elsevier B.V.
⁎ Corresponding author at: CITAB, University of Trás-os-Montes and Alto Douro, Vila Real, Portugal. E-mail address: [email protected] (F.L. Queiroga).
https://doi.org/10.1016/j.scitotenv.2018.10.022 0048-9697/© 2018 Published by Elsevier B.V.
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
1. Introduction In the last decades, the Earth has undergone profound and irreversible changes in the landscape. The intensification of crop and livestock production, the spatial increase in density of infrastructures such as highways, dams, buildings or power lines are some of these alterations (Santos et al., 2016). These human induced changes on the natural ecosystems have a detrimental effect on the wild populations. A major consequence of these changes is the loss and fragmentation of habitats, which in a short time span may trigger population decline and species extinction due to the decrease of resources and overpopulation in small areas (Ewers and Didham, 2007). There are numerous hazards for birds in urban centres. Collision with infrastructures (Avery, 1979; Bayne et al., 2012), collision with vehicles (Hernandez, 1988), electrocution (Bevanger, 1998; Tintó et al., 2010), traps (Rodríguez et al., 2010), hunting (Desmarchelier et al., 2010), climate change, pollution, pesticides or introduction of exotic species are some threats ultimately resulting from human activity (Acevedo-Whitehouse and Duffus, 2009; Erickson et al., 2005; Gross, 2015; Santos et al., 2018). The diversity and intensity of anthropogenic pressures on the ecosystems increases the stress level in wild animals (Orshay et al., 2008), forcing them to modify their behaviour to avoid or to mitigate stressors. Along these lines, alterations have already been observed in the patterns of movement, activity, diet or reproduction (Birnie-Gauvin et al., 2017; Ditchkoff et al., 2006; Sol et al., 2013) in several species. Although there is a tendency for the loss of biodiversity in urban centres, occasionally new ecological niches arise leading some species to prosper (Bradley and Altizer, 2007). One success story is that of peregrine falcon (Falco peregrinus), which successfully reproduced and increased its population in recent decades in several cities in North America and Europe (Ditchkoff et al., 2006; Luniak, 2004). The structural equation model (SEM) is an algorithm that facilitates estimation of causal relationships between two or more complex latent variables. Latent variables are concepts (e.g., “anthropogenic pressure”) measured by a number of observable indicators (e.g., “population density”, “road density”). The method of Partial Least Squares-Path Modelling (PLS-PM) is the Partial Least Squares approach to Structural Equation Modelling. It involves the design of a network composed of connections among latent variables, as well as among latent variables and measurable indicators, which are assumed to represent effect-cause relationship. Having defined the network, the method finds a linear regression model by projecting the predicted variables and the measurable variables onto a new common space and quantifies the connections in the sequel (Garson, 2016; Kumar Gorai et al., 2015; Sanches Fernandes et al., 2018). The PLS-PM model is composed of two sets of linear equations: 1) the structural or inner model, which specifies the relationship between latent variables; 2) the measurable or outer model, which defines the relationship between latent and associated measured variables. This model is very attractive because of its minimum demand on the sample size, and because it can achieve high levels of statistical significance as it is based on a series of ordinary least squares (OLS) regressions (Reinatz et al., 2009). The PLS-PM is widely used by researches from different disciplines in information research, strategic management, ecology and others (Hair et al., 2011; Kumar Gorai et al., 2015; Sanches Fernandes et al., 2018). However, to our knowledge the use of PLS-PM in environmental assessments involving wild birds has not been attempted. The purpose of this study is to take that step forward and use the Partial Least Squares-Path Modelling (PLS-PM) method to investigate the impact of multiple and diverse anthropogenic pressures in the mortality of wild bird during a period of 10 years (2008–2017) in northern Portugal. The notable flexibility of PLS-PM to efficiently handle large and complex sets of databases (Garson, 2016; Kumar Gorai et al., 2015; Sanches Fernandes et al., 2018), coupled with the dimension, spatial coverage and quality (detail on causes of death and ample coverage of anthropogenic pressures), is expected to produce a unique view on the impacts on human activity on wild bird populations found in
2997
urban and rural centres as well as their spatial incidence in northern Portugal. 2. Materials and methods A total of 6058 free-living birds were admitted into the Wildlife Rehabilitation Centre (WRC) in Parque Biológico de Gaia, located in Avintes, Portugal (41° 05′ 48.50″N–8° 33′ 21.34″W), from January 2008 to December 2017, and their admission records were analysed. The data collected comprises individuals that died during treatment and those that were euthanized, in compliance with its unfavourable prognosis for release into the wild. The data was separated according to the bird's taxonomic order and age classes (adult N 1 years, juvenile ˂ 1 year), based on external characteristics. Gender information was not included because it was absent in the majority of the medical records. The birds were collected from municipalities exclusively located in northern Portugal. 2.1. Definition of latent and measured variables for the PLS-PM model The PLS-PM exercise was based on two latent variables: “mortality” and “anthropogenic pressures”. The measured variables used to form “mortality” were first divided for each animal into two groups: nontraumatic and traumatic, in agreement with previous studies (MolinaLópez et al., 2011; Morishita et al., 1998; Rodríguez et al., 2010; Wendell et al., 2002). In brief, non-traumatic causes included infectious diseases, parasitic infections, captivity (animals kept in captivity illegally that presented dehydration, stress, muscular atrophy, feather problems, injuries, parasites or diet related metabolic problems), nutritional, enterotoxaemia (a paretic syndrome associated with enteritis, frequently observed in seagulls), toxics, and unknown origin. Traumatic causes included fish hooks and nets, collision with vehicles, collision with buildings, electrocution, predation, shooting and trauma of unknown origin. 2.2. Dataset preparation and model implementation The XLStat software (Addinsoft, 2014) was used to run PLS-PM with one exogenous (independent) and one endogenous (dependent) latent variable (LV), termed “anthropogenic pressures” and “mortality”, respectively. The model was run twice to investigate non-traumatic and traumatic “mortality”. The LVs were connected with a path and described by a number of measured variables (MV), namely the aforementioned death causes (LV “mortality”) and a list of anthropogenic pressures: density of burned areas, annual precipitation, number of water reservoirs, number of domestic landfills, wind farms (number and total power), road density, human population density, residential buildings density, non-residential buildings density, small companies (with ˂10 employees, 91.1% of all companies), medium companies (with 10–250 employees, 8.6%), large companies (with ˃250 employees, 0.3%), number of farms (Pordata, 2018). According to PORDATA portal (Pordata, 2018), the small companies are distributed in a similar way by the various economic activities, while the medium companies are mostly related to extraction, treatment and distribution of water (27.2%), quarries (21.1%), manufacturing industry (17.4%), construction (7.7%) and transport (7.4%). The large companies relate to distribution of water (53.1%) and electricity (18.8%). The degree of multicolinearity between the measured variables was tested using Variance Inflation Factors (VIF). In general, VIFs N10 are regarded by many practitioners as a sign of severe or serious multi-collinearity (Hair Jr. et al., 1995). However, Values of the VIF of 10, 20, 40, or even higher do not, by themselves, discount the results of regression analyses (O'Brien, 2007). The block comprising the LV and the associated MVs is termed construct. Constructs may be unidimensional, in case all measured variables are on the same scale or multidimensional otherwise. The construct linking mortality to death causes is unidimensional because the quantification of each cause is based on countings of animals dying from a non-traumatic or traumatic event. The construct “anthropogenic
2998
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
pressures” is multidimensional because the associated parameters are measured in a diversity of different scales. Robustness tests are available for unidimensional constructs, namely the Cronbach's alpha and the Dillon-Goldstein's Rho. According to Tenenhaus et al. (2005), unidimensional constructs are internally consistent when the alpha and Rho statistics are both N0,7. The links between the two LV were quantified through path coefficients while the links between LV and MV were quantified through weights, using the complex combination of multivariate statistical analyses embedded in PLS-PM. The coefficient of determination (R2) measures the inner model goodness of fit. The path coefficients account for the influence of the pressures on mortality. The weights represent the contribution of each measured variable to its latent variable. The “goodness of fit” parameter (GoF; e.g., Tenenhaus et al., 2005), complemented with the bootstrap procedure (Hair et al., 2005), can be used to test the outer model robustness. In this study, the PLS model was designed as formative model, where the MV were viewed as causes of the LV (Sanches Fernandes et al., 2018). In the XLStat software the data used to source the PLS-PM algorithm were compiled in a worksheet comprising n rows and p columns and representing the sample size and characterization, respectively. The sample size indicates the number of municipalities from northern Portugal for which data were available on the causes of death of wild birds. The sample characterization encompassed the number of measured parameters, which includes the causes of mortality and the anthropogenic pressures. The data and results of PLS-PM were spatially represented in a Geographic Information System (GIS), namely the ArcMap software of ESRI (2010) widely used in environmental studies
(see supplementary data) (e.g., Bellu et al., 2015; Sanches Fernandes et al., 2015; Terêncio et al., 2017, 2018).
3. Results 3.1. Study population The 6058 animals examined were categorized in 22 orders, as follows: Order Accipitriformes (n = 496, 8.2%), Order Anserifomes (n = 90, 1.5%), Order Apodiformes (n = 451, 7.4%), Order Bucerotiformes (n = 17, 0.3%), Order Caprimulgiformes (n = 74, 1.3%), Order Charadriiformes (n = 2055, 33.9%), Order Ciconiformes (n = 47, 0.8%), Order Columbiformes (n = 302, 5%), Order Coraciiformes (n = 17, 0.3%), Order Cuculiformes (n = 10, 0.2%), Order Falconiformes (n = 79, 1.3%), Order Galliformes (n = 12, 0.2%), Order Gaviidae (n = 1, 0.01%), Order Gruiformes (n = 15, 0.2%), Order Otidiformes (n = 2; 0%), Order Passeriformes (n = 1739, 28.7%), Order Pelecaniformes (n = 38, 0.6%), Order Phoenicopteriformes (n = 2, 0%), Order Piciformes (n = 46, 0.8%), Order Podicipediformes (n = 2, 0%), Order Procellariiformes (n = 5, 0.1%), Order Strigiformes (n = 439, 7.2%) and Order Suliformes (n = 119, 2%). Regarding the age, 4844 (80%) were adult and 1214 (20%) juveniles. Gender information was unavailable because it was absent from the majority of the medical records as described in material and methods section. Concerning the causes of death, 2606 (43%) had a non-traumatic origin and in 3452 (57%) cases, death was related to trauma. Additional information is available as supplementary data.
Fig. 1. Map with the spatial distribution of wild birds admitted to the Wildlife Rehabilitation Centre of Parque Biológico de Gaia (from 2008 to 2017). The wild birds were collected from distinct municipalities of northern of Portugal (number of events per municipality is indicated in the colour scale). (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
2999
3.2. Spatial distribution of wild bird mortality and causes of death
3.3. Spatial distribution of anthropogenic pressures
Fig. 1 shows the spatial distribution of the 6058 animals collected from 76 municipalities of northern Portugal covered in this study. The areas shaded as red are the municipalities where a larger number of animals were collected (250–2000 animals). The municipalities that contributed the most to the wild bird mortality database were Vila Nova de Gaia (n = 1875), Porto (n = 1540) and Matosinhos (n = 659). In 43% of the cases, the death was due to a non-traumatic origin. Fig. 2 shows the concomitant spatial distribution in the studied region. Independently of the cause of death, the municipalities with the greatest number of cases are located near the Atlantic coast (west), being coloured in red and orange in the figure. When looking at the specific cause of death (Fig. 2a–g), the municipalities ranked in the top position are: a) captivity-Porto (n = 246); b) nutritional-Vila Nova de Gaia (n = 164); c) parasitism-Matosinhos (n = 4); d) poisoning-Porto (n = 3); e) enterotoxaemia- Porto (n = 147); f)Infectious disease-Vila Nova de Gaia (n = 90) and g)unknown non-traumatic-Vila Nova de Gaia (n = 435). In 57% of the cases, the cause of death was due to trauma. Fig. 3 shows the spatial distribution of the main causes per municipality. As in the previous model, it becomes clear that the municipalities with the greatest number of cases are located near the coast, (Fig. 3a–g): a) collision with vehicles-Vila Nova de Gaia (n = 26); b) predation-Vila Nova de Gaia (n = 35); c) Collision with building-Vila Nova of Gaia (n = 24); d) gunshot-Santo Tirso (n = 3); e) electrocution-Santo Tirso (n = 11); f) fish hooks and netsVila Nova de Gaia (n = 33) and g) unknown trauma-Vila Nova de Gaia (n = 868).
Fig. 4 represents the spatial distribution of anthropogenic factors analysed in the municipalities from where the wild bird data were collected. The factors of population density, residential and non-residential buildings, medium and large companies and annual precipitation increase from the inland to the coastal areas. Regarding precipitation, the result exposes the influence of a reduction in the annual rainfall, related to the location of Gerês and Barroso mountains in the northwest of Portugal. For the other parameters, the results reflect the long standing asymmetry between the rural inland and the urban coast, which is consequence of governmental policies that lacked the integrated development of territories. The inland-coast trend is reversed for the factors burnt areas, water reservoirs, small companies, and farms, which are predominant in the inland areas. This result is not surprising considering the prevalence of a rural environment in the inland areas. For example, the largest forested regions are concentrated in these zones justifying the larger burnt areas. The irrigation of a larger number of farms combined with a lower annual rainfall justifies the larger number of water reservoirs. The sparse population contributes to the dominance of small companies. The factors related to wind farms are larger at the central area, where various mountain ridges allow for effective wind speeds. Finally, the factors roads and domestic landfills do not show any marked trend, which probably means that these public services are evenly distributed through the entire region. 3.4. Relation between mortality and anthropogenic pressures Fig. 5 characterizes the spatial distribution of trauma and nontrauma mortality based on the measured number of cases (Fig. 5a and
Fig. 2. a) to g): Maps with the spatial distribution of wild birds' mortality of non-traumatic origin in wild birds admitted to the Wildlife Rehabilitation Centre of Parque Biológico de Gaia (from 2008 to 2017). The wild birds were collected from distinct municipalities of northern of Portugal (number of events per municipality is indicated in the colour scale). (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)
3000
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
Fig. 3. a) to g): Maps with the spatial distribution of wild birds mortality of traumatic mortality in wild birds admitted to the Wildlife Rehabilitation Centre of Parque Biológico de Gaia (from 2008 to 2017). The wild birds were collected from distinct municipalities of northern of Portugal (number of events per municipality is indicated in the colour scale).
b) as well as based on the results of PLS-PM (Fig. 5c, d). The correspondence between the measured and modelled distributions is noticeable, which favours the applicability of PLS-PM in this study. We must remark that modelled anthropogenic pressures (Fig. 5e and f) most resemble the spatial distribution of medium companies (Fig. 4l). The diagrammatic representation of PLS-PM results in Fig. 6 illustrates the relationship between anthropogenic pressures and nontraumatic/traumatic causes of mortality. The corresponding path coefficients are 0.927 (non-trauma case) and 0.908 (trauma-case). The VIF values of anthropogenic pressures were b10 in most cases. Therefore, the vast majority of these variables are not multi-collinear. The exceptions were the number of small and medium companies. The VIF of non-traumatic causes were N10 for captivity, enterotoxaemia, toxics, and unknown origin, and b10 otherwise. The traumatic causes of death were ascribed VIF N 10 for predation, collision with vehicles, drowning and unknown origin, and b10 otherwise. These results would indicate the need to remove collinearity using standardization, principal component analysis or other methods, before running the PLS-PM model. It should be noted, however, that the collinear anthropogenic pressures are indicator (dummy) variables that represent a categorical variable with three or more categories (i.e., percentages of small, medium and large companies, which always sum 100%). The same applies to the traumatic and non-traumatic causes of death because the overall mortality is always the sum of deaths per cause. In these cases, interpretation of regression results will not be affected by the high VIFs (Allison, 1999), the reason why no attempts were made to remove multi collinearity from the independent variables. The PLS-PM model (panel A) identified the mortality associated with captivity and nutritional disorders as the main causes of death of nontraumatic origin associated with the anthropogenic pressures under study. According to the results, the main anthropogenic pressures
responsible for the wild bird mortality were small (w = 1.54) and medium (w = 1.56) companies. In a second line of influence, human population density (w = −0.77) and non-residential buildings (0.54) played also a role. Some other pressures were apparently less important considering their smaller absolute weights: residential buildings (w = −0.12), roads (w = −0.20), burnt areas (w = 0.16) and precipitation (w = 0.08). The remaining pressures (water reservoirs, annual precipitation, farms, large companies and wind farms) had insignificant or no impact on the mortality. It is worth mentioning that weights are positive for most measured variables and negative for a few others while the negative weights are usually small (≤0.2) or very small (≤0.1). In general, negative signs in the context of PLS-PM are considered an occurrence of Simpson's paradox related to the existence of improbable causal relationships (Pearl, 2009), or with instability caused by collinearity (Kock and Lynn, 2012), among other problems. In this study, a reason for the occurrence of negative signs may be the “reverse coding” of regionalized variables (i.e., variables distributed in space). For example, the number of farms is a variable that increases towards the inland area (Fig. 4n) whereas population density (Fig. 4h) or the number of medium companies (Fig. 4l) increase in the opposite direction. When these variables are assembled in a construct of a PLS-PM model (Fig. 6a and b), the reflex on the latent variable (anthropogenic pressures) of this negative spatial correlation between farms and the other two variables should be that weight signs of population density and number of medium companies are the same (e.g. +0.67 and +2.86 in the trauma mortality model; Fig. 6b) and differ from the farms sign (−0.07), regardless of the potential impact of these three anthropogenic pressures on bird mortality. Besides reverse coding of regionalized variables, the results suggest other causes for the negative signs, because variables such as density of roads (−0.2 in both models) do not show any marked spatial trend (Fig. 4g). A deeper inspection of such
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
3001
Fig. 4. a) to n): Maps with the spatial distribution of the selected anthropogenic pressures in the municipalities from where animals were originated and further admitted to the Wildlife Rehabilitation Centre of Parque Biológico de Gaia (from 2008 to 2017).
3002
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
Fig. 5. a) to f): Comparative spatial distribution of measured and modelled total trauma and non-trauma mortality model in the animals admitted to the Wildlife Rehabilitation Centre of Parque Biológico de Gaia from municipalities of northern Portugal (from 2008 to 2017).
causes would be justified if negative weights were large. Because they are small or very small, no further explanation for negative signs was explored. The PLS-PM model identified the mortality due to predation and trauma of unknown origin as the main causes of death of traumatic origin associated with the anthropogenic pressures under study. The main anthropogenic pressures identified were also small companies (w = 3.23) and medium companies (w = 2.86). Combined with the small companies' pressure, the human population density (w = 0.67), residential buildings density (w = 0.74), large companies (w = 0.21), non-residential buildings (w = −0.14) and roads density (w = −0.20) were also considered impacting pressures. A coupled representation of weights associated to traumatic and non-traumatic death causes is depicted in the radial diagram of Fig. 7. The plots are generally similar, but it seems that traumatic mortality is
more sensitive to the impact of pressures given the larger weights in the most relevant pressures (small and medium companies). It is recognized that the magnitude of weighs depends on a number of circumstances, namely sample size, number of measured variables per latent variable, among others (e.g., Jakobowicz, 2006). However, our interpretation is supported by the fact that non-trauma and trauma models were based on the same sample and anthropogenic pressures. The blue and red lines describe the impacts of pressures on traumatic and non-traumatic mortality, respectively. As regards the small and medium companies the model sensitivity is higher for traumatic mortality. The unidimensional construct “mortality” + death cases is internally consistent, characterized by Cronbach's alpha and Dillon-Goldstein's Rho always N0.9. The inner models of PLS-PM (relation between the two latent variables) are reliable because their coefficients of
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
3003
Fig. 6. (A and B): Diagrammatic illustrates the relationship between anthropogenic pressures and traumatic/non-traumatic causes of mortality. Panel A refers to non-traumatic model and panel B to traumatic model. The coloured rectangles represent the measurable variables, the coloured circles the corresponding anthropogenic pressures. The arrows represent the link between latent variables. The blue coloured rectangles represent the death causes and anthropogenic pressures with a bigger impact on the wild bird mortality, followed by the green and orange coloured rectangles. The rectangles shaded as yellow have quite small impact. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
determination (R2) are high in the non-traumatic (R2 = 0.86) and traumatic (R2 = 0.82) cases. The outer models (relationship between latent and measured variables) are satisfactory because their GoF and corresponding bootstrap values are moderate (N0.7 for the non-trauma case and N0.8 for the trauma case). 4. Discussion Nearly three quarters of the land without ice is now dominated by anthropogenic biomes with altered ecosystems, leading to their destruction or fragmentation (Ellis and Ramankutty, 2008). These changes have enormous negative consequences on the wild populations, both directly and indirectly. (Fleming and Bateman, 2018). Our data show that areas with the highest density of dead animals are located near the coast, particularly in urban areas such as Vila Nova de Gaia and Porto. In the maps depicting the spatial distributions of anthropogenic pressures (Fig. 4), it is possible to observe that the
majority of the selected anthropogenic pressures are also located in this area. The dominant ones are: burnt areas, annual precipitation, landfills, roads, population density, residential and non-residential buildings. Nevertheless, medium-sized and large companies predominate. Consequently, as expected the greater number of injured animals was found in these areas. The high density of the human population in urban areas also increases the probability of finding injured animals when compared with more rural ones. Another reason for finding a greater concentration of dead animals in urban areas is the benefits provided by urban biomes. The higher concentration of prey, the increase in the success of catches provided by artificial light and infrastructures or the use of man-made nesting structures are some of the attractions for some species to adopt this biome as a habitat (Fleming and Bateman, 2018; Molina-López et al., 2011). Nevertheless, the negative consequences for wildlife populations living in close proximity to humans are substantial, as showed by the results of our work. The use of PLS-PM evidenced the existence of a high
3004
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
Fig. 7. Radial diagram coupled representation of PLS-PM weights associated to anthropogenic pressures. The red line is associated to the weight of pressure in non-traumatic mortality and the blue line to the weight of pressure in traumatic mortality. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
relationship between mortality and anthropogenic pressures, supported by the high coefficients of determination (R2) obtained (close to 1). For the first time, to the authors' best knowledge, a spatial correlation is made between the mortality of a great variety of wild bird species and potentially threatening anthropogenic pressures, during an extensive period of time (2008–2017). The PLS-PM method reveals that small and medium companies represent the main anthropogenic pressures responsible for the mortality of free-living birds, both from traumatic and non-traumatic origin. Two thirds of these companies are related to commerce (20%), agriculture and animal production (8%), restoration (7%) or construction (7%) These are located mostly nearby the coast where the human density population is higher, and the larger metropolitan areas are located (Pordata, 2018). The reason why small and medium companies are associated with mortality can be justified by the presence of specific infrastructures that favour the occurrence of accidents (e.g., large transparent glass windows, liquid reservoirs, chimneys), pollution (smoke, gas, contamination of water with chemicals, landfills) and a great movement of people and vehicles. Small and medium industries may also contribute to a large concentration of species in the surrounding areas, since their characteristic infrastructures can provide nesting and shelter areas. Moreover, the accumulation of trash and wastes favour a high concentration of prey such as rats and insects facilitating the hunting task. The residential and non-residential buildings also had an important impact in the mortality of wild birds and can be linked to collision with windows (Avery, 1979; Bolshakov et al., 2013; Dunn, 1990; Klem, 1990). Small and medium-sized companies, despite the reduced number of workers and less complex facilities, seem to have a greater impact on birds' mortality than large companies. Although our data does not allow us to identify with precision the cause for this association, we believe that this may be associated with the dispersal of infrastructure within a given area leading to a greater fragmentation and habitat destruction (Zambrano and Martins, 2004). Habitat fragmentation and destruction has been described as one of the main threats linked to the decline of wild populations overall (Destro et al., 2018; Grimm et al., 2000; Hunter, 2007). In light of our data, small and medium-
companies should be targeted for specific monitoring programs in order to clarify strategies to mitigate their apparent negative impact on wild birds' mortality in northern Portugal. Some studies from the last decades have shown that urban areas are rendering the animals more susceptible to be infected by an ample variety of agents due to the warmer conditions, the stress induced by human presence and activity, pollutants, specific interactions with human and pets, landfills and outbreaks of zoonotic diseases (Acevedo-Whitehouse and Duffus, 2009; Bradley and Altizer, 2007; Brearley et al., 2013; Daszak et al., 2001). All these factors can be related to the negative impact of anthropogenic pressures over the mortality of these wild animals. One particular case is the enterotoxaemia syndrome, the second most common cause of non-traumatic death in our study. This syndrome was mainly observed in aquatic birds, particularly in Larus michaellis and other seagulls. It is characterized by the paresis of the pelvic limbs and enteritis. Even though frequently observed, the cause of this syndrom is not well established. In previous works carried out by our team (Garcês et al., 2018) and by others in the southern part of the country (Soares, 2014), a connexion with environmental toxics or viral origin was hypothesised, but a definitive conclusion was not reached. The stress induced by the excess of animals in urban areas in search for food together with human activities, pollution and climate changes, may be among the main contributors for the animals' enhanced susceptibility to toxics and pathogens contributing to their mortality (Acevedo-Whitehouse and Duffus, 2009). The results obtained in relation to death by trauma confirm previous results of other teams (Burton and Doblar, 2004; Erickson et al., 2005; Garcês et al., 2018). Around the globe, every year, billions of birds die due to anthropogenic causes where collision with windows and vehicles, power lines, wind turbines, cat predation or commercial fishing can be included (Banks, 1979; Erickson et al., 2005; Fowler, 1988; Gottdenker et al., 2008; Kalpakis et al., 2009; Loss et al., 2013; MolinaLópez et al., 2011; Morishita et al., 1998). Based on the investigated database, it can be concluded that human activity has a significant impact on the mortality of wild birds in northern Portugal These results confirmed the outcomes of earlier studies related to the mortality of wild birds all around the globe where the
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
predominant cause of death was associated with the direct or indirect action of humans. In the future, it would be useful introduce new variables to better explain LVs such as livestock and pet density, electricity distribution density, indicators of territory fragmentation (e.g. fractal measurements, dispersion, shape and aggregation of patches, etc.), indicators of climate change in order to have a better cause-effect relationship. In addition, because the PLS-PM involve several multidimensional components it would be appropriate to encompass other spheres such as the atmosphere, hydrosphere and lithosphere in order to enlighten our current understanding about human and wildlife interactions.
Acknowledgments This research was funded by the INTERACT project – “Integrated Research in Environment, Agro-Chain and Technology”, no. NORTE-010145-FEDER-000017, in its line of research entitled BEST - Bioeconomy and Sustainability, co-financed by the European Regional Development Fund (ERDF) (NORTE-01-0145-FEDER-000017) through NORTE 2020 (North Regional Operational Program 2014/2020). Moreover, this study was also supported by the Research Project UID/AGR/04033/2013 financed by the Portuguese Foundation for Science and Technology (FCT). Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi. org/10.1016/j.scitotenv.2018.10.022.
References Acevedo-Whitehouse, K., Duffus, A.L.J., 2009. Effects of environmental change on wildlife health. Philos. Trans. R. Soc. Lond. Ser. B Biol. Sci. 364, 3429–3438. https://doi.org/ 10.1098/rstb.2009.0128. Allison, P.D., 1999. Multiple Regression: A Premier. Pine Forge Press, Thousand Oaks, California (202p). Avery, M., 1979. Review of avian mortality due to collisions with manmade structures. Bird Control Semin. Proc. 2–11. Banks, B.R.C., 1979. Human Related Mortality of Birds in the United States. U.S. Fish and Wildlife Service, Washington, D. C. Bayne, E.M., Scobie, C.A., Rawson-clark, M., 2012. Factors in fluencing the annual risk of bird – window collisions at residential structures in Alberta, Canada. J. Compil. 39, 583–592. https://doi.org/10.1071/WR11179. Bellu, A., Sanches Fernandes, L.F., Cortes, R.M.V., Pacheco, F.A.L., 2015. A framework model for the dimensioning and allocation of a detention basin system: the case of a floodprone mountainous watershed. J. Hydrol. 533, 567–580. Bevanger, K., 1998. Biological and conservation aspects of bird mortality caused by electricity power lines: a review. Biol. Conserv. 86, 67–76. https://doi.org/10.1016/ S0006-3207(97)00176-6. Birnie-Gauvin, K., Peiman, K.S., Raubenheimer, D., Cooke, S.J., 2017. Nutritional physiology and ecology of wildlife in a changing world. Conserv. Physiol. 5, 1–18. https://doi.org/ 10.1093/conphys/cox030. Bolshakov, C.V., Bulyuk, V.N., Sinelschikova, A., Vorotkov, M., 2013. Influence of the vertical light beam on numbers and flight trajectories of night-migrating songbirds. Avian Ecol. Behav. 24, 35–49. Bradley, C.A., Altizer, S., 2007. Urbanization and the ecology of wildlife diseases. Trends Ecol. Evol. 22, 95–102. https://doi.org/10.1016/j.tree.2006.11.001. Brearley, G., Rhodes, J., Bradley, A., Baxter, G., Seabrook, L., Lunney, D., Liu, Y., McAlpine, C., 2013. Wildlife disease prevalence in human-modified landscapes. Biol. Rev. 88, 427–442. https://doi.org/10.1111/brv.12009. Burton, D.L., Doblar, K a., 2004. Morbidity and mortality of urban wildlife in the midwestern United States. Proc 4th Int Urban Wildl Symp, pp. 171–181. Daszak, P., Cunningham, A.A., Hyatt, A.D., 2001. Anthropogenic environmental change and the emergence of infectious diseases in wildlife. Acta Trop. 78, 103–116. https://doi.org/10.1016/S0001-706X(00)00179-0. Desmarchelier, M., Santamaria-Bouvier, A., Fitzgérald, G., Lair, S., 2010. Mortality and morbidity associated with gunshot in raptorial birds from the province of Quebec: 1986 to 2007. Can. Vet. J. 51, 70–74. Destro, G.F.G., De Marco, P., Terribile, L.C., 2018. Threats for bird population restoration: a systematic review. Perspect. Ecol. Conserv. 16, 68–73. Ditchkoff, S.S., Saalfeld, S.T., Gibson, C.J., 2006. Animal behavior in urban ecosystems: modifications due to human-induced stress. Urban Ecosyst. 9, 5–12. https://doi.org/ 10.1007/s11252-006-3262-3. Dunn, E., 1990. Bird mortality from striking residential windows in winter. J. F. Ornithol. 64, 302–309. Ellis, E., Ramankutty, N., 2008. Putting people in the map: anthropogenic biomes of the world. Ecol. Environ. 6, 439–447. https://doi.org/10.1890/070062.
3005
Erickson, W.P., Johnson Jr., G.D., Y., D.P., 2005. A Summary and Comparison of Bird Mortality from Anthropogenic Causes with an Emphasis on Collisions 1 Fatality Rates 1029–1042. Ewers, R.M., Didham, R.K., 2007. The effect of fragment shape and species' sensitivity to habitat edges on animal population size: contributed papers. Conserv. Biol. 21, 926–936. https://doi.org/10.1111/j.1523-1739.2007.00720.x. Fleming, P.A., Bateman, P.W., 2018. Novel predation opportunities in anthropogenic landscapes. Anim. Behav. 138, 145–155. https://doi.org/10.1016/j.anbehav.2018.02.011. Fowler, P.A., 1988. Seasonal endocrine cycles in the European hedgehog, Erinaceus europaeus. J. Reprod. Fertil. 84, 259–272. Garcês, A., Soeiro, V., Lóio, S., Prada, J., Silva, F., Pires, I., 2018. Necropsy findings and causes of mortality in wild birds in a Center for Rehabilitation of Wild Animals in the North of Portugal (Hallazgos de Necropsia y causas de mortalidad en aves silvestres en un Centro de Rehabilitación de Animales Silvestres en el). Rev. Electrónica Veterinária 19, 1–20. Garson, G.D., 2016. Partial Least Squares: Regression and Structural Equation Models. Statistical Associates Publishers, Asheboro, NC. Gottdenker, N.L., Walsh, T., Jimenez-Uzcategui, G., Betancourt, F., Cruz, M., Soos, C., Miller, E.R., Parker, P.G., 2008. Causes of mortality of wild birds submitted to the Charles Darwin Research Station, Santa Cruz, Galápagos, Ecuador from 2002–2004. J. Wildl. Dis. 44, 1024–1031. https://doi.org/10.7589/0090-3558-44.4.1024. Grimm, N., Grove, J., Pickett, S., Redman, C., 2000. Integrated approaches to long-term studies of urban ecological systems. Bioscience 50, 571–584. Gross, M., 2015. Europe's bird populations in decline. Curr. Biol. Mag. 25, R483–R485. https://doi.org/10.1016/j.cub.2015.05.057. Hair Jr., J.F., Anderson, R.E., Tatham, R.L., Black, W.C., 1995. Multivariate Data Analysis. 3rd edn. Macmillan, New York. Hair, J., Anderson, R., Tatham, R., Black, W., 2005. Análise Multivariada De Dados. Bookman, Porto Alegre. Hair, J., Ringle, C., Sarstedt, M., 2011. PLS-SEM: indeed a silver bullet. J. Mark. Theory Pract. 18, 139–152. Hernandez, M., 1988. Road mortality of the little owl (Athene noctua) in Spain. J. Raptor Res. 22, 81–84. Hunter, P., 2007. About sustainability a comprehensive foundation. EMBO Rep. 8, 316–318. https://doi.org/10.1038/sj.embor.7400951. Jakobowicz, J., 2006. Understanding PLS path modeling parameters estimates: a study based on Monte Carlo simulation and customer satisfaction surveys. COMPSTAT'06, 17th Symposium on Computational Statistics, held in Rome, Italy, in September 2006. Kalpakis, S., Mazaris, A.D., Mamakis, Y., Poulopoulos, Y., 2009. A retrospective study of mortality and morbidity factors for common buzzards Buteo buteo and long-legged buzzards Buteo rufinus in Greece: 1996–2005. Bird Conserv. Int. 19, 15–21. https:// doi.org/10.1017/S095927090800806X. Klem, D., 1990. From collisions with windows. J. F. Ornithol. 61, 115–119. Kock, N., Lynn, G.S., 2012. Lateral collinearity and misleading results in variance-based SEM: an illustration and recommendations. J. Assoc. Inf. Syst. 13 (7), 546–580. Kumar Gorai, A., Tuluri, F., Tchounwou, P.B., 2015. Development of PLS–path model for understanding the role of precursors on ground level ozone concentration in Gulfport, Mississippi, USA. Atmos. Pollut. Res. 6, 389–397. https://doi.org/10.5094/ APR.2015.043. Loss, S.R., Will, T., Marra, P.P., 2013. The impact of free-ranging domestic cats on wildlife of the United States. Nat. Commun. 4, 1–7. https://doi.org/10.1038/ncomms2380. Luniak, M., 2004. Synurbization: adaptation of animal wildlife to urban development. Proc. 4th Int. Urban Wildl. Symp. 50–55. Molina-López, R.A., Casal, J., Darwich, L., 2011. Causes of morbidity in wild raptor populations admitted at a wildlife rehabilitation centre in Spain from 1995–2007: a long term retrospective study. PLoS One 6. https://doi.org/10.1371/journal.pone.0024603. Morishita, T.Y., Fullerton, A.T., Lowenstine, L.J., Gardner, I. a, Brooks, D.L., Dale, L., Brooks, D.L., 1998. Morbidity and mortality in free-living raptorial birds of northern California: a retrospective study, 1983–1994. Medicine (Baltimore) 12, 78–81. O'Brien, R.M., 2007. A caution regarding rules of thumb for variance inflation factors. Qual. Quant. 41, 673–690. Orshay, K.E.J.F., Ohnson, P.I.T.J.J., Tock, M.E.S., En, C.A.P., 2008. Multiple anthropogenic stressors and the structural properties of food webs. Ecology 89, 2692–2699. https://doi.org/10.1002/ecy.2064. Pearl, J., 2009. Causality: Models, Reasoning, and Inference. Cambridge University Press, Cambridge, England. Pordata, 2018. Pordata (Base de Dados de Portugal Contemporâneo). [WWW Document]. URL. https://www.pordata.pt/, Accessed date: 20 July 2018. Reinatz, W., Haenlein, M., Henseler, J., 2009. An empirical comparation of the efficacy of covariance-based and variance-based SEM. Int. J. Res. Mark. 26 (332–244). Rodríguez, B., Rodríguez, A., Siverio, F., Siverio, M., 2010. Causes of raptor admissions to a Wildlife Rehabilitation Center in Tenerife (Canary Islands). J. Raptor Res. 44, 30–39. https://doi.org/10.3356/JRR-09-40.1. Sanches Fernandes, L.F., Terêncio, D.P.S., Pacheco, F.A.L., 2015. Rainwater harvesting systems for low demanding applications. Sci. Total Environ. 529, 91–100. Sanches Fernandes, L.F., Fernandes, A.C.P., Ferreira, A.R.L., Cortes, R.M.V., Pacheco, F.A.L., 2018. A partial least squares – path modeling analysis for the understanding of biodiversity loss in rural and urban watersheds in Portugal. Sci. Total Environ. 626, 1069–1085. https://doi.org/10.1016/j.scitotenv.2018.01.127. Santos, M., Bessa, R., Cabral, J.A., Leal-Pacheco, F., Leitão, D., Moreira, F., Pinto, M., Lecoq, M., Silva, J.P., 2016. Impacts of land use and infrastructural changes on threatened Little Bustard Tetrax tetrax breeding populations: quantitative assessments using a recently developed spatially explicit dynamic modelling framework. Bird Conserv. Int., 1–18 https://doi.org/10.1017/S0959270915000258. Santos, R.M.B., Fernandes, L.F.S., Varandas, S.G.P., Pereira, M.G., Sousa, R., Teixeira, A., Cortes, R.M.V., Pacheco, F.A.L., 2018. Impacts of climate change and land-use scenarios on
3006
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
Margaritifera margaritifera, an environmental indicator and endangered species. Sci. Total Environ. 511, 477–488. https://doi.org/10.1016/j.scitotenv.2014.12.090. Soares, S.P.V., 2014. Paretic Syndrome in Gulls (LARIDAE) in The South of Portugal. University of Lisbon. Sol, D., Lapiedra, O., González-Lagos, C., 2013. Behavioural adjustments for a life in the city. Anim. Behav., 1–12 https://doi.org/10.1016/j.anbehav.2013.01.023 Special is. Tenenhaus, M., Vinzi, V., Chatelin, Y., Lauro, C., 2005. PLS path modeling. Comput. Statist. Data Anal. 48, 159–205 Available at:. https://studies2.hec.fr/jahia/webdav/site/hec/ shared/sites/tenenhaus/acces_anonyme/home/articles/PL_PM_5.pdf. Terêncio, D.P.S., Sanches Fernandes, L.F., Cortes, R.M.V., Pacheco, F.A.L., 2017. Improved framework model to allocate optimal rainwater harvesting sites in small watersheds for agro-forestry uses. J. Hydrol. 550, 318–330. Terêncio, D.P.S., Sanches Fernandes, L.F., Cortes, R.M.V., Moura, J.P., Pacheco, F.A.L., 2018. Rainwater harvesting in catchments for agro-forestry uses: a study focused on the
balance between sustainability values and storage capacity. Sci. Total Environ. 613–614, 1079–1092. Tintó, A., Real, J., Mañosa, S., 2010. Predicting and correcting electrocution of birds in Mediterranean areas. J. Wildl. Manag. 74, 1852–1862. https://doi.org/10.2193/2009-521. Wendell, M.D., Sleeman, J.M., Kratz, G., 2002. Retrospective study of morbidity and mortality of raptors admitted to Colorado State University Veterinary Teaching Hospital during 1995 to 1998. J. Wildl. Dis. 38, 101–106. https://doi.org/10.7589/0090-355838.1.101. Zambrano, T., Martins, M., 2004. Identificação e análise dos impactos ambientais das pequenas indústrias de São Carlos – SP. XXIV Encontro de Engenharia de Produção. Florianópolis, pp. 5326–5333.
Contents lists available at ScienceDirect
Science of the Total Environment journal homepage: www.elsevier.com/locate/scitotenv
Preservation of wild bird species in northern Portugal - Effects of anthropogenic pressures in wild bird populations (2008–2017) Andreia Garcês a, Isabel Pires b,c, Fernando A.L. Pacheco d, Luís F. Sanches Fernandes a, Vanessa Soeiro e, Sara Lóio e, Justina Prada b,c, Rui Cortes a, Felisbina L. Queiroga a,b,⁎ a
Centre for the Research and Technology of Agro-Environment and Biological Sciences, University of Trás-os-Montes and Alto Douro, Vila Real, Portugal Department of Veterinary Science of the University of Trás-os-Montes and Alto Douro, Vila Real, Portugal CECAV - University of Trás-os-Montes and Alto Douro, Vila Real, Portugal d Chemistry Research Centre, University of Trás-os-Montes and Alto Douro, Vila Real, Portugal e Wildlife Rehabilitation Centre in Parque Biológico de Gaia, R. Cunha, Avintes, Portugal b c
H I G H L I G H T S
G R A P H I C A L
A B S T R A C T
• Wild birds in northern Portugal are affected by several anthropogenic factors. • Increased number of medium companies contributed to increased mortality of wild birds. • Habitat destruction and fragmentation is linked to wild bird mortality.
a r t i c l e
i n f o
Article history: Received 6 August 2018 Received in revised form 2 October 2018 Accepted 2 October 2018 Available online 03 October 2018 Editor: Damia Barcelo Keywords: Anthropogenic pressures Mortality North of Portugal Partial least squares-path modelling Wild birds
a b s t r a c t In this study we aim to obtain a holistic view over the consequences of human-induced threats to the wild bird populations in the study area, based on data collected from a Wildlife Rehabilitation Centre (WRC) through the method of Partial Least Squares-Path Modelling (PLS-PM). The study area comprised 76 rural and urban municipalities located in northern Portugal. Within PLS-PM the threats (“anthropogenic pressures”) are termed exogenous latent variables, while the final environmental consequence (wild bird mortality) is termed endogenous latent variable. Latent variables are concepts assessed by numerical parameters. The PLS-PM results identified as most significant pressures, the number of small and medium companies in the municipalities, both in traumatic and nontraumatic deaths. Although the pattern of weights is similar regardless of the general cause of death, traumatic causes seem to play a more prominent role given the larger weights in the relevant specific causes (number of companies). The high coefficients of determination (R2 N 0,8) reveal that the variance of wild bird mortality is largely explained by the variance of the pressures, which indicates a cause-effect relationship between the independent (pressures) and dependent (mortality) variables. The ample coverage of northern Portugal with a huge dataset suggests that this cause-effect relationship is typical from this region. The use of a sophisticated statistical method PLSPM and its incorporation into a Geographic Information System (GIS) revealed to be an important tool for analysing wildlife impacts of environmental and human factors. The results obtained with this model indicate that there is a substantial negative impact of human activity on wild bird mortality. © 2018 Published by Elsevier B.V.
⁎ Corresponding author at: CITAB, University of Trás-os-Montes and Alto Douro, Vila Real, Portugal. E-mail address: [email protected] (F.L. Queiroga).
https://doi.org/10.1016/j.scitotenv.2018.10.022 0048-9697/© 2018 Published by Elsevier B.V.
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
1. Introduction In the last decades, the Earth has undergone profound and irreversible changes in the landscape. The intensification of crop and livestock production, the spatial increase in density of infrastructures such as highways, dams, buildings or power lines are some of these alterations (Santos et al., 2016). These human induced changes on the natural ecosystems have a detrimental effect on the wild populations. A major consequence of these changes is the loss and fragmentation of habitats, which in a short time span may trigger population decline and species extinction due to the decrease of resources and overpopulation in small areas (Ewers and Didham, 2007). There are numerous hazards for birds in urban centres. Collision with infrastructures (Avery, 1979; Bayne et al., 2012), collision with vehicles (Hernandez, 1988), electrocution (Bevanger, 1998; Tintó et al., 2010), traps (Rodríguez et al., 2010), hunting (Desmarchelier et al., 2010), climate change, pollution, pesticides or introduction of exotic species are some threats ultimately resulting from human activity (Acevedo-Whitehouse and Duffus, 2009; Erickson et al., 2005; Gross, 2015; Santos et al., 2018). The diversity and intensity of anthropogenic pressures on the ecosystems increases the stress level in wild animals (Orshay et al., 2008), forcing them to modify their behaviour to avoid or to mitigate stressors. Along these lines, alterations have already been observed in the patterns of movement, activity, diet or reproduction (Birnie-Gauvin et al., 2017; Ditchkoff et al., 2006; Sol et al., 2013) in several species. Although there is a tendency for the loss of biodiversity in urban centres, occasionally new ecological niches arise leading some species to prosper (Bradley and Altizer, 2007). One success story is that of peregrine falcon (Falco peregrinus), which successfully reproduced and increased its population in recent decades in several cities in North America and Europe (Ditchkoff et al., 2006; Luniak, 2004). The structural equation model (SEM) is an algorithm that facilitates estimation of causal relationships between two or more complex latent variables. Latent variables are concepts (e.g., “anthropogenic pressure”) measured by a number of observable indicators (e.g., “population density”, “road density”). The method of Partial Least Squares-Path Modelling (PLS-PM) is the Partial Least Squares approach to Structural Equation Modelling. It involves the design of a network composed of connections among latent variables, as well as among latent variables and measurable indicators, which are assumed to represent effect-cause relationship. Having defined the network, the method finds a linear regression model by projecting the predicted variables and the measurable variables onto a new common space and quantifies the connections in the sequel (Garson, 2016; Kumar Gorai et al., 2015; Sanches Fernandes et al., 2018). The PLS-PM model is composed of two sets of linear equations: 1) the structural or inner model, which specifies the relationship between latent variables; 2) the measurable or outer model, which defines the relationship between latent and associated measured variables. This model is very attractive because of its minimum demand on the sample size, and because it can achieve high levels of statistical significance as it is based on a series of ordinary least squares (OLS) regressions (Reinatz et al., 2009). The PLS-PM is widely used by researches from different disciplines in information research, strategic management, ecology and others (Hair et al., 2011; Kumar Gorai et al., 2015; Sanches Fernandes et al., 2018). However, to our knowledge the use of PLS-PM in environmental assessments involving wild birds has not been attempted. The purpose of this study is to take that step forward and use the Partial Least Squares-Path Modelling (PLS-PM) method to investigate the impact of multiple and diverse anthropogenic pressures in the mortality of wild bird during a period of 10 years (2008–2017) in northern Portugal. The notable flexibility of PLS-PM to efficiently handle large and complex sets of databases (Garson, 2016; Kumar Gorai et al., 2015; Sanches Fernandes et al., 2018), coupled with the dimension, spatial coverage and quality (detail on causes of death and ample coverage of anthropogenic pressures), is expected to produce a unique view on the impacts on human activity on wild bird populations found in
2997
urban and rural centres as well as their spatial incidence in northern Portugal. 2. Materials and methods A total of 6058 free-living birds were admitted into the Wildlife Rehabilitation Centre (WRC) in Parque Biológico de Gaia, located in Avintes, Portugal (41° 05′ 48.50″N–8° 33′ 21.34″W), from January 2008 to December 2017, and their admission records were analysed. The data collected comprises individuals that died during treatment and those that were euthanized, in compliance with its unfavourable prognosis for release into the wild. The data was separated according to the bird's taxonomic order and age classes (adult N 1 years, juvenile ˂ 1 year), based on external characteristics. Gender information was not included because it was absent in the majority of the medical records. The birds were collected from municipalities exclusively located in northern Portugal. 2.1. Definition of latent and measured variables for the PLS-PM model The PLS-PM exercise was based on two latent variables: “mortality” and “anthropogenic pressures”. The measured variables used to form “mortality” were first divided for each animal into two groups: nontraumatic and traumatic, in agreement with previous studies (MolinaLópez et al., 2011; Morishita et al., 1998; Rodríguez et al., 2010; Wendell et al., 2002). In brief, non-traumatic causes included infectious diseases, parasitic infections, captivity (animals kept in captivity illegally that presented dehydration, stress, muscular atrophy, feather problems, injuries, parasites or diet related metabolic problems), nutritional, enterotoxaemia (a paretic syndrome associated with enteritis, frequently observed in seagulls), toxics, and unknown origin. Traumatic causes included fish hooks and nets, collision with vehicles, collision with buildings, electrocution, predation, shooting and trauma of unknown origin. 2.2. Dataset preparation and model implementation The XLStat software (Addinsoft, 2014) was used to run PLS-PM with one exogenous (independent) and one endogenous (dependent) latent variable (LV), termed “anthropogenic pressures” and “mortality”, respectively. The model was run twice to investigate non-traumatic and traumatic “mortality”. The LVs were connected with a path and described by a number of measured variables (MV), namely the aforementioned death causes (LV “mortality”) and a list of anthropogenic pressures: density of burned areas, annual precipitation, number of water reservoirs, number of domestic landfills, wind farms (number and total power), road density, human population density, residential buildings density, non-residential buildings density, small companies (with ˂10 employees, 91.1% of all companies), medium companies (with 10–250 employees, 8.6%), large companies (with ˃250 employees, 0.3%), number of farms (Pordata, 2018). According to PORDATA portal (Pordata, 2018), the small companies are distributed in a similar way by the various economic activities, while the medium companies are mostly related to extraction, treatment and distribution of water (27.2%), quarries (21.1%), manufacturing industry (17.4%), construction (7.7%) and transport (7.4%). The large companies relate to distribution of water (53.1%) and electricity (18.8%). The degree of multicolinearity between the measured variables was tested using Variance Inflation Factors (VIF). In general, VIFs N10 are regarded by many practitioners as a sign of severe or serious multi-collinearity (Hair Jr. et al., 1995). However, Values of the VIF of 10, 20, 40, or even higher do not, by themselves, discount the results of regression analyses (O'Brien, 2007). The block comprising the LV and the associated MVs is termed construct. Constructs may be unidimensional, in case all measured variables are on the same scale or multidimensional otherwise. The construct linking mortality to death causes is unidimensional because the quantification of each cause is based on countings of animals dying from a non-traumatic or traumatic event. The construct “anthropogenic
2998
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
pressures” is multidimensional because the associated parameters are measured in a diversity of different scales. Robustness tests are available for unidimensional constructs, namely the Cronbach's alpha and the Dillon-Goldstein's Rho. According to Tenenhaus et al. (2005), unidimensional constructs are internally consistent when the alpha and Rho statistics are both N0,7. The links between the two LV were quantified through path coefficients while the links between LV and MV were quantified through weights, using the complex combination of multivariate statistical analyses embedded in PLS-PM. The coefficient of determination (R2) measures the inner model goodness of fit. The path coefficients account for the influence of the pressures on mortality. The weights represent the contribution of each measured variable to its latent variable. The “goodness of fit” parameter (GoF; e.g., Tenenhaus et al., 2005), complemented with the bootstrap procedure (Hair et al., 2005), can be used to test the outer model robustness. In this study, the PLS model was designed as formative model, where the MV were viewed as causes of the LV (Sanches Fernandes et al., 2018). In the XLStat software the data used to source the PLS-PM algorithm were compiled in a worksheet comprising n rows and p columns and representing the sample size and characterization, respectively. The sample size indicates the number of municipalities from northern Portugal for which data were available on the causes of death of wild birds. The sample characterization encompassed the number of measured parameters, which includes the causes of mortality and the anthropogenic pressures. The data and results of PLS-PM were spatially represented in a Geographic Information System (GIS), namely the ArcMap software of ESRI (2010) widely used in environmental studies
(see supplementary data) (e.g., Bellu et al., 2015; Sanches Fernandes et al., 2015; Terêncio et al., 2017, 2018).
3. Results 3.1. Study population The 6058 animals examined were categorized in 22 orders, as follows: Order Accipitriformes (n = 496, 8.2%), Order Anserifomes (n = 90, 1.5%), Order Apodiformes (n = 451, 7.4%), Order Bucerotiformes (n = 17, 0.3%), Order Caprimulgiformes (n = 74, 1.3%), Order Charadriiformes (n = 2055, 33.9%), Order Ciconiformes (n = 47, 0.8%), Order Columbiformes (n = 302, 5%), Order Coraciiformes (n = 17, 0.3%), Order Cuculiformes (n = 10, 0.2%), Order Falconiformes (n = 79, 1.3%), Order Galliformes (n = 12, 0.2%), Order Gaviidae (n = 1, 0.01%), Order Gruiformes (n = 15, 0.2%), Order Otidiformes (n = 2; 0%), Order Passeriformes (n = 1739, 28.7%), Order Pelecaniformes (n = 38, 0.6%), Order Phoenicopteriformes (n = 2, 0%), Order Piciformes (n = 46, 0.8%), Order Podicipediformes (n = 2, 0%), Order Procellariiformes (n = 5, 0.1%), Order Strigiformes (n = 439, 7.2%) and Order Suliformes (n = 119, 2%). Regarding the age, 4844 (80%) were adult and 1214 (20%) juveniles. Gender information was unavailable because it was absent from the majority of the medical records as described in material and methods section. Concerning the causes of death, 2606 (43%) had a non-traumatic origin and in 3452 (57%) cases, death was related to trauma. Additional information is available as supplementary data.
Fig. 1. Map with the spatial distribution of wild birds admitted to the Wildlife Rehabilitation Centre of Parque Biológico de Gaia (from 2008 to 2017). The wild birds were collected from distinct municipalities of northern of Portugal (number of events per municipality is indicated in the colour scale). (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
2999
3.2. Spatial distribution of wild bird mortality and causes of death
3.3. Spatial distribution of anthropogenic pressures
Fig. 1 shows the spatial distribution of the 6058 animals collected from 76 municipalities of northern Portugal covered in this study. The areas shaded as red are the municipalities where a larger number of animals were collected (250–2000 animals). The municipalities that contributed the most to the wild bird mortality database were Vila Nova de Gaia (n = 1875), Porto (n = 1540) and Matosinhos (n = 659). In 43% of the cases, the death was due to a non-traumatic origin. Fig. 2 shows the concomitant spatial distribution in the studied region. Independently of the cause of death, the municipalities with the greatest number of cases are located near the Atlantic coast (west), being coloured in red and orange in the figure. When looking at the specific cause of death (Fig. 2a–g), the municipalities ranked in the top position are: a) captivity-Porto (n = 246); b) nutritional-Vila Nova de Gaia (n = 164); c) parasitism-Matosinhos (n = 4); d) poisoning-Porto (n = 3); e) enterotoxaemia- Porto (n = 147); f)Infectious disease-Vila Nova de Gaia (n = 90) and g)unknown non-traumatic-Vila Nova de Gaia (n = 435). In 57% of the cases, the cause of death was due to trauma. Fig. 3 shows the spatial distribution of the main causes per municipality. As in the previous model, it becomes clear that the municipalities with the greatest number of cases are located near the coast, (Fig. 3a–g): a) collision with vehicles-Vila Nova de Gaia (n = 26); b) predation-Vila Nova de Gaia (n = 35); c) Collision with building-Vila Nova of Gaia (n = 24); d) gunshot-Santo Tirso (n = 3); e) electrocution-Santo Tirso (n = 11); f) fish hooks and netsVila Nova de Gaia (n = 33) and g) unknown trauma-Vila Nova de Gaia (n = 868).
Fig. 4 represents the spatial distribution of anthropogenic factors analysed in the municipalities from where the wild bird data were collected. The factors of population density, residential and non-residential buildings, medium and large companies and annual precipitation increase from the inland to the coastal areas. Regarding precipitation, the result exposes the influence of a reduction in the annual rainfall, related to the location of Gerês and Barroso mountains in the northwest of Portugal. For the other parameters, the results reflect the long standing asymmetry between the rural inland and the urban coast, which is consequence of governmental policies that lacked the integrated development of territories. The inland-coast trend is reversed for the factors burnt areas, water reservoirs, small companies, and farms, which are predominant in the inland areas. This result is not surprising considering the prevalence of a rural environment in the inland areas. For example, the largest forested regions are concentrated in these zones justifying the larger burnt areas. The irrigation of a larger number of farms combined with a lower annual rainfall justifies the larger number of water reservoirs. The sparse population contributes to the dominance of small companies. The factors related to wind farms are larger at the central area, where various mountain ridges allow for effective wind speeds. Finally, the factors roads and domestic landfills do not show any marked trend, which probably means that these public services are evenly distributed through the entire region. 3.4. Relation between mortality and anthropogenic pressures Fig. 5 characterizes the spatial distribution of trauma and nontrauma mortality based on the measured number of cases (Fig. 5a and
Fig. 2. a) to g): Maps with the spatial distribution of wild birds' mortality of non-traumatic origin in wild birds admitted to the Wildlife Rehabilitation Centre of Parque Biológico de Gaia (from 2008 to 2017). The wild birds were collected from distinct municipalities of northern of Portugal (number of events per municipality is indicated in the colour scale). (For interpretation of the references to color in this figure, the reader is referred to the web version of this article.)
3000
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
Fig. 3. a) to g): Maps with the spatial distribution of wild birds mortality of traumatic mortality in wild birds admitted to the Wildlife Rehabilitation Centre of Parque Biológico de Gaia (from 2008 to 2017). The wild birds were collected from distinct municipalities of northern of Portugal (number of events per municipality is indicated in the colour scale).
b) as well as based on the results of PLS-PM (Fig. 5c, d). The correspondence between the measured and modelled distributions is noticeable, which favours the applicability of PLS-PM in this study. We must remark that modelled anthropogenic pressures (Fig. 5e and f) most resemble the spatial distribution of medium companies (Fig. 4l). The diagrammatic representation of PLS-PM results in Fig. 6 illustrates the relationship between anthropogenic pressures and nontraumatic/traumatic causes of mortality. The corresponding path coefficients are 0.927 (non-trauma case) and 0.908 (trauma-case). The VIF values of anthropogenic pressures were b10 in most cases. Therefore, the vast majority of these variables are not multi-collinear. The exceptions were the number of small and medium companies. The VIF of non-traumatic causes were N10 for captivity, enterotoxaemia, toxics, and unknown origin, and b10 otherwise. The traumatic causes of death were ascribed VIF N 10 for predation, collision with vehicles, drowning and unknown origin, and b10 otherwise. These results would indicate the need to remove collinearity using standardization, principal component analysis or other methods, before running the PLS-PM model. It should be noted, however, that the collinear anthropogenic pressures are indicator (dummy) variables that represent a categorical variable with three or more categories (i.e., percentages of small, medium and large companies, which always sum 100%). The same applies to the traumatic and non-traumatic causes of death because the overall mortality is always the sum of deaths per cause. In these cases, interpretation of regression results will not be affected by the high VIFs (Allison, 1999), the reason why no attempts were made to remove multi collinearity from the independent variables. The PLS-PM model (panel A) identified the mortality associated with captivity and nutritional disorders as the main causes of death of nontraumatic origin associated with the anthropogenic pressures under study. According to the results, the main anthropogenic pressures
responsible for the wild bird mortality were small (w = 1.54) and medium (w = 1.56) companies. In a second line of influence, human population density (w = −0.77) and non-residential buildings (0.54) played also a role. Some other pressures were apparently less important considering their smaller absolute weights: residential buildings (w = −0.12), roads (w = −0.20), burnt areas (w = 0.16) and precipitation (w = 0.08). The remaining pressures (water reservoirs, annual precipitation, farms, large companies and wind farms) had insignificant or no impact on the mortality. It is worth mentioning that weights are positive for most measured variables and negative for a few others while the negative weights are usually small (≤0.2) or very small (≤0.1). In general, negative signs in the context of PLS-PM are considered an occurrence of Simpson's paradox related to the existence of improbable causal relationships (Pearl, 2009), or with instability caused by collinearity (Kock and Lynn, 2012), among other problems. In this study, a reason for the occurrence of negative signs may be the “reverse coding” of regionalized variables (i.e., variables distributed in space). For example, the number of farms is a variable that increases towards the inland area (Fig. 4n) whereas population density (Fig. 4h) or the number of medium companies (Fig. 4l) increase in the opposite direction. When these variables are assembled in a construct of a PLS-PM model (Fig. 6a and b), the reflex on the latent variable (anthropogenic pressures) of this negative spatial correlation between farms and the other two variables should be that weight signs of population density and number of medium companies are the same (e.g. +0.67 and +2.86 in the trauma mortality model; Fig. 6b) and differ from the farms sign (−0.07), regardless of the potential impact of these three anthropogenic pressures on bird mortality. Besides reverse coding of regionalized variables, the results suggest other causes for the negative signs, because variables such as density of roads (−0.2 in both models) do not show any marked spatial trend (Fig. 4g). A deeper inspection of such
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
3001
Fig. 4. a) to n): Maps with the spatial distribution of the selected anthropogenic pressures in the municipalities from where animals were originated and further admitted to the Wildlife Rehabilitation Centre of Parque Biológico de Gaia (from 2008 to 2017).
3002
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
Fig. 5. a) to f): Comparative spatial distribution of measured and modelled total trauma and non-trauma mortality model in the animals admitted to the Wildlife Rehabilitation Centre of Parque Biológico de Gaia from municipalities of northern Portugal (from 2008 to 2017).
causes would be justified if negative weights were large. Because they are small or very small, no further explanation for negative signs was explored. The PLS-PM model identified the mortality due to predation and trauma of unknown origin as the main causes of death of traumatic origin associated with the anthropogenic pressures under study. The main anthropogenic pressures identified were also small companies (w = 3.23) and medium companies (w = 2.86). Combined with the small companies' pressure, the human population density (w = 0.67), residential buildings density (w = 0.74), large companies (w = 0.21), non-residential buildings (w = −0.14) and roads density (w = −0.20) were also considered impacting pressures. A coupled representation of weights associated to traumatic and non-traumatic death causes is depicted in the radial diagram of Fig. 7. The plots are generally similar, but it seems that traumatic mortality is
more sensitive to the impact of pressures given the larger weights in the most relevant pressures (small and medium companies). It is recognized that the magnitude of weighs depends on a number of circumstances, namely sample size, number of measured variables per latent variable, among others (e.g., Jakobowicz, 2006). However, our interpretation is supported by the fact that non-trauma and trauma models were based on the same sample and anthropogenic pressures. The blue and red lines describe the impacts of pressures on traumatic and non-traumatic mortality, respectively. As regards the small and medium companies the model sensitivity is higher for traumatic mortality. The unidimensional construct “mortality” + death cases is internally consistent, characterized by Cronbach's alpha and Dillon-Goldstein's Rho always N0.9. The inner models of PLS-PM (relation between the two latent variables) are reliable because their coefficients of
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
3003
Fig. 6. (A and B): Diagrammatic illustrates the relationship between anthropogenic pressures and traumatic/non-traumatic causes of mortality. Panel A refers to non-traumatic model and panel B to traumatic model. The coloured rectangles represent the measurable variables, the coloured circles the corresponding anthropogenic pressures. The arrows represent the link between latent variables. The blue coloured rectangles represent the death causes and anthropogenic pressures with a bigger impact on the wild bird mortality, followed by the green and orange coloured rectangles. The rectangles shaded as yellow have quite small impact. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
determination (R2) are high in the non-traumatic (R2 = 0.86) and traumatic (R2 = 0.82) cases. The outer models (relationship between latent and measured variables) are satisfactory because their GoF and corresponding bootstrap values are moderate (N0.7 for the non-trauma case and N0.8 for the trauma case). 4. Discussion Nearly three quarters of the land without ice is now dominated by anthropogenic biomes with altered ecosystems, leading to their destruction or fragmentation (Ellis and Ramankutty, 2008). These changes have enormous negative consequences on the wild populations, both directly and indirectly. (Fleming and Bateman, 2018). Our data show that areas with the highest density of dead animals are located near the coast, particularly in urban areas such as Vila Nova de Gaia and Porto. In the maps depicting the spatial distributions of anthropogenic pressures (Fig. 4), it is possible to observe that the
majority of the selected anthropogenic pressures are also located in this area. The dominant ones are: burnt areas, annual precipitation, landfills, roads, population density, residential and non-residential buildings. Nevertheless, medium-sized and large companies predominate. Consequently, as expected the greater number of injured animals was found in these areas. The high density of the human population in urban areas also increases the probability of finding injured animals when compared with more rural ones. Another reason for finding a greater concentration of dead animals in urban areas is the benefits provided by urban biomes. The higher concentration of prey, the increase in the success of catches provided by artificial light and infrastructures or the use of man-made nesting structures are some of the attractions for some species to adopt this biome as a habitat (Fleming and Bateman, 2018; Molina-López et al., 2011). Nevertheless, the negative consequences for wildlife populations living in close proximity to humans are substantial, as showed by the results of our work. The use of PLS-PM evidenced the existence of a high
3004
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
Fig. 7. Radial diagram coupled representation of PLS-PM weights associated to anthropogenic pressures. The red line is associated to the weight of pressure in non-traumatic mortality and the blue line to the weight of pressure in traumatic mortality. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
relationship between mortality and anthropogenic pressures, supported by the high coefficients of determination (R2) obtained (close to 1). For the first time, to the authors' best knowledge, a spatial correlation is made between the mortality of a great variety of wild bird species and potentially threatening anthropogenic pressures, during an extensive period of time (2008–2017). The PLS-PM method reveals that small and medium companies represent the main anthropogenic pressures responsible for the mortality of free-living birds, both from traumatic and non-traumatic origin. Two thirds of these companies are related to commerce (20%), agriculture and animal production (8%), restoration (7%) or construction (7%) These are located mostly nearby the coast where the human density population is higher, and the larger metropolitan areas are located (Pordata, 2018). The reason why small and medium companies are associated with mortality can be justified by the presence of specific infrastructures that favour the occurrence of accidents (e.g., large transparent glass windows, liquid reservoirs, chimneys), pollution (smoke, gas, contamination of water with chemicals, landfills) and a great movement of people and vehicles. Small and medium industries may also contribute to a large concentration of species in the surrounding areas, since their characteristic infrastructures can provide nesting and shelter areas. Moreover, the accumulation of trash and wastes favour a high concentration of prey such as rats and insects facilitating the hunting task. The residential and non-residential buildings also had an important impact in the mortality of wild birds and can be linked to collision with windows (Avery, 1979; Bolshakov et al., 2013; Dunn, 1990; Klem, 1990). Small and medium-sized companies, despite the reduced number of workers and less complex facilities, seem to have a greater impact on birds' mortality than large companies. Although our data does not allow us to identify with precision the cause for this association, we believe that this may be associated with the dispersal of infrastructure within a given area leading to a greater fragmentation and habitat destruction (Zambrano and Martins, 2004). Habitat fragmentation and destruction has been described as one of the main threats linked to the decline of wild populations overall (Destro et al., 2018; Grimm et al., 2000; Hunter, 2007). In light of our data, small and medium-
companies should be targeted for specific monitoring programs in order to clarify strategies to mitigate their apparent negative impact on wild birds' mortality in northern Portugal. Some studies from the last decades have shown that urban areas are rendering the animals more susceptible to be infected by an ample variety of agents due to the warmer conditions, the stress induced by human presence and activity, pollutants, specific interactions with human and pets, landfills and outbreaks of zoonotic diseases (Acevedo-Whitehouse and Duffus, 2009; Bradley and Altizer, 2007; Brearley et al., 2013; Daszak et al., 2001). All these factors can be related to the negative impact of anthropogenic pressures over the mortality of these wild animals. One particular case is the enterotoxaemia syndrome, the second most common cause of non-traumatic death in our study. This syndrome was mainly observed in aquatic birds, particularly in Larus michaellis and other seagulls. It is characterized by the paresis of the pelvic limbs and enteritis. Even though frequently observed, the cause of this syndrom is not well established. In previous works carried out by our team (Garcês et al., 2018) and by others in the southern part of the country (Soares, 2014), a connexion with environmental toxics or viral origin was hypothesised, but a definitive conclusion was not reached. The stress induced by the excess of animals in urban areas in search for food together with human activities, pollution and climate changes, may be among the main contributors for the animals' enhanced susceptibility to toxics and pathogens contributing to their mortality (Acevedo-Whitehouse and Duffus, 2009). The results obtained in relation to death by trauma confirm previous results of other teams (Burton and Doblar, 2004; Erickson et al., 2005; Garcês et al., 2018). Around the globe, every year, billions of birds die due to anthropogenic causes where collision with windows and vehicles, power lines, wind turbines, cat predation or commercial fishing can be included (Banks, 1979; Erickson et al., 2005; Fowler, 1988; Gottdenker et al., 2008; Kalpakis et al., 2009; Loss et al., 2013; MolinaLópez et al., 2011; Morishita et al., 1998). Based on the investigated database, it can be concluded that human activity has a significant impact on the mortality of wild birds in northern Portugal These results confirmed the outcomes of earlier studies related to the mortality of wild birds all around the globe where the
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
predominant cause of death was associated with the direct or indirect action of humans. In the future, it would be useful introduce new variables to better explain LVs such as livestock and pet density, electricity distribution density, indicators of territory fragmentation (e.g. fractal measurements, dispersion, shape and aggregation of patches, etc.), indicators of climate change in order to have a better cause-effect relationship. In addition, because the PLS-PM involve several multidimensional components it would be appropriate to encompass other spheres such as the atmosphere, hydrosphere and lithosphere in order to enlighten our current understanding about human and wildlife interactions.
Acknowledgments This research was funded by the INTERACT project – “Integrated Research in Environment, Agro-Chain and Technology”, no. NORTE-010145-FEDER-000017, in its line of research entitled BEST - Bioeconomy and Sustainability, co-financed by the European Regional Development Fund (ERDF) (NORTE-01-0145-FEDER-000017) through NORTE 2020 (North Regional Operational Program 2014/2020). Moreover, this study was also supported by the Research Project UID/AGR/04033/2013 financed by the Portuguese Foundation for Science and Technology (FCT). Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi. org/10.1016/j.scitotenv.2018.10.022.
References Acevedo-Whitehouse, K., Duffus, A.L.J., 2009. Effects of environmental change on wildlife health. Philos. Trans. R. Soc. Lond. Ser. B Biol. Sci. 364, 3429–3438. https://doi.org/ 10.1098/rstb.2009.0128. Allison, P.D., 1999. Multiple Regression: A Premier. Pine Forge Press, Thousand Oaks, California (202p). Avery, M., 1979. Review of avian mortality due to collisions with manmade structures. Bird Control Semin. Proc. 2–11. Banks, B.R.C., 1979. Human Related Mortality of Birds in the United States. U.S. Fish and Wildlife Service, Washington, D. C. Bayne, E.M., Scobie, C.A., Rawson-clark, M., 2012. Factors in fluencing the annual risk of bird – window collisions at residential structures in Alberta, Canada. J. Compil. 39, 583–592. https://doi.org/10.1071/WR11179. Bellu, A., Sanches Fernandes, L.F., Cortes, R.M.V., Pacheco, F.A.L., 2015. A framework model for the dimensioning and allocation of a detention basin system: the case of a floodprone mountainous watershed. J. Hydrol. 533, 567–580. Bevanger, K., 1998. Biological and conservation aspects of bird mortality caused by electricity power lines: a review. Biol. Conserv. 86, 67–76. https://doi.org/10.1016/ S0006-3207(97)00176-6. Birnie-Gauvin, K., Peiman, K.S., Raubenheimer, D., Cooke, S.J., 2017. Nutritional physiology and ecology of wildlife in a changing world. Conserv. Physiol. 5, 1–18. https://doi.org/ 10.1093/conphys/cox030. Bolshakov, C.V., Bulyuk, V.N., Sinelschikova, A., Vorotkov, M., 2013. Influence of the vertical light beam on numbers and flight trajectories of night-migrating songbirds. Avian Ecol. Behav. 24, 35–49. Bradley, C.A., Altizer, S., 2007. Urbanization and the ecology of wildlife diseases. Trends Ecol. Evol. 22, 95–102. https://doi.org/10.1016/j.tree.2006.11.001. Brearley, G., Rhodes, J., Bradley, A., Baxter, G., Seabrook, L., Lunney, D., Liu, Y., McAlpine, C., 2013. Wildlife disease prevalence in human-modified landscapes. Biol. Rev. 88, 427–442. https://doi.org/10.1111/brv.12009. Burton, D.L., Doblar, K a., 2004. Morbidity and mortality of urban wildlife in the midwestern United States. Proc 4th Int Urban Wildl Symp, pp. 171–181. Daszak, P., Cunningham, A.A., Hyatt, A.D., 2001. Anthropogenic environmental change and the emergence of infectious diseases in wildlife. Acta Trop. 78, 103–116. https://doi.org/10.1016/S0001-706X(00)00179-0. Desmarchelier, M., Santamaria-Bouvier, A., Fitzgérald, G., Lair, S., 2010. Mortality and morbidity associated with gunshot in raptorial birds from the province of Quebec: 1986 to 2007. Can. Vet. J. 51, 70–74. Destro, G.F.G., De Marco, P., Terribile, L.C., 2018. Threats for bird population restoration: a systematic review. Perspect. Ecol. Conserv. 16, 68–73. Ditchkoff, S.S., Saalfeld, S.T., Gibson, C.J., 2006. Animal behavior in urban ecosystems: modifications due to human-induced stress. Urban Ecosyst. 9, 5–12. https://doi.org/ 10.1007/s11252-006-3262-3. Dunn, E., 1990. Bird mortality from striking residential windows in winter. J. F. Ornithol. 64, 302–309. Ellis, E., Ramankutty, N., 2008. Putting people in the map: anthropogenic biomes of the world. Ecol. Environ. 6, 439–447. https://doi.org/10.1890/070062.
3005
Erickson, W.P., Johnson Jr., G.D., Y., D.P., 2005. A Summary and Comparison of Bird Mortality from Anthropogenic Causes with an Emphasis on Collisions 1 Fatality Rates 1029–1042. Ewers, R.M., Didham, R.K., 2007. The effect of fragment shape and species' sensitivity to habitat edges on animal population size: contributed papers. Conserv. Biol. 21, 926–936. https://doi.org/10.1111/j.1523-1739.2007.00720.x. Fleming, P.A., Bateman, P.W., 2018. Novel predation opportunities in anthropogenic landscapes. Anim. Behav. 138, 145–155. https://doi.org/10.1016/j.anbehav.2018.02.011. Fowler, P.A., 1988. Seasonal endocrine cycles in the European hedgehog, Erinaceus europaeus. J. Reprod. Fertil. 84, 259–272. Garcês, A., Soeiro, V., Lóio, S., Prada, J., Silva, F., Pires, I., 2018. Necropsy findings and causes of mortality in wild birds in a Center for Rehabilitation of Wild Animals in the North of Portugal (Hallazgos de Necropsia y causas de mortalidad en aves silvestres en un Centro de Rehabilitación de Animales Silvestres en el). Rev. Electrónica Veterinária 19, 1–20. Garson, G.D., 2016. Partial Least Squares: Regression and Structural Equation Models. Statistical Associates Publishers, Asheboro, NC. Gottdenker, N.L., Walsh, T., Jimenez-Uzcategui, G., Betancourt, F., Cruz, M., Soos, C., Miller, E.R., Parker, P.G., 2008. Causes of mortality of wild birds submitted to the Charles Darwin Research Station, Santa Cruz, Galápagos, Ecuador from 2002–2004. J. Wildl. Dis. 44, 1024–1031. https://doi.org/10.7589/0090-3558-44.4.1024. Grimm, N., Grove, J., Pickett, S., Redman, C., 2000. Integrated approaches to long-term studies of urban ecological systems. Bioscience 50, 571–584. Gross, M., 2015. Europe's bird populations in decline. Curr. Biol. Mag. 25, R483–R485. https://doi.org/10.1016/j.cub.2015.05.057. Hair Jr., J.F., Anderson, R.E., Tatham, R.L., Black, W.C., 1995. Multivariate Data Analysis. 3rd edn. Macmillan, New York. Hair, J., Anderson, R., Tatham, R., Black, W., 2005. Análise Multivariada De Dados. Bookman, Porto Alegre. Hair, J., Ringle, C., Sarstedt, M., 2011. PLS-SEM: indeed a silver bullet. J. Mark. Theory Pract. 18, 139–152. Hernandez, M., 1988. Road mortality of the little owl (Athene noctua) in Spain. J. Raptor Res. 22, 81–84. Hunter, P., 2007. About sustainability a comprehensive foundation. EMBO Rep. 8, 316–318. https://doi.org/10.1038/sj.embor.7400951. Jakobowicz, J., 2006. Understanding PLS path modeling parameters estimates: a study based on Monte Carlo simulation and customer satisfaction surveys. COMPSTAT'06, 17th Symposium on Computational Statistics, held in Rome, Italy, in September 2006. Kalpakis, S., Mazaris, A.D., Mamakis, Y., Poulopoulos, Y., 2009. A retrospective study of mortality and morbidity factors for common buzzards Buteo buteo and long-legged buzzards Buteo rufinus in Greece: 1996–2005. Bird Conserv. Int. 19, 15–21. https:// doi.org/10.1017/S095927090800806X. Klem, D., 1990. From collisions with windows. J. F. Ornithol. 61, 115–119. Kock, N., Lynn, G.S., 2012. Lateral collinearity and misleading results in variance-based SEM: an illustration and recommendations. J. Assoc. Inf. Syst. 13 (7), 546–580. Kumar Gorai, A., Tuluri, F., Tchounwou, P.B., 2015. Development of PLS–path model for understanding the role of precursors on ground level ozone concentration in Gulfport, Mississippi, USA. Atmos. Pollut. Res. 6, 389–397. https://doi.org/10.5094/ APR.2015.043. Loss, S.R., Will, T., Marra, P.P., 2013. The impact of free-ranging domestic cats on wildlife of the United States. Nat. Commun. 4, 1–7. https://doi.org/10.1038/ncomms2380. Luniak, M., 2004. Synurbization: adaptation of animal wildlife to urban development. Proc. 4th Int. Urban Wildl. Symp. 50–55. Molina-López, R.A., Casal, J., Darwich, L., 2011. Causes of morbidity in wild raptor populations admitted at a wildlife rehabilitation centre in Spain from 1995–2007: a long term retrospective study. PLoS One 6. https://doi.org/10.1371/journal.pone.0024603. Morishita, T.Y., Fullerton, A.T., Lowenstine, L.J., Gardner, I. a, Brooks, D.L., Dale, L., Brooks, D.L., 1998. Morbidity and mortality in free-living raptorial birds of northern California: a retrospective study, 1983–1994. Medicine (Baltimore) 12, 78–81. O'Brien, R.M., 2007. A caution regarding rules of thumb for variance inflation factors. Qual. Quant. 41, 673–690. Orshay, K.E.J.F., Ohnson, P.I.T.J.J., Tock, M.E.S., En, C.A.P., 2008. Multiple anthropogenic stressors and the structural properties of food webs. Ecology 89, 2692–2699. https://doi.org/10.1002/ecy.2064. Pearl, J., 2009. Causality: Models, Reasoning, and Inference. Cambridge University Press, Cambridge, England. Pordata, 2018. Pordata (Base de Dados de Portugal Contemporâneo). [WWW Document]. URL. https://www.pordata.pt/, Accessed date: 20 July 2018. Reinatz, W., Haenlein, M., Henseler, J., 2009. An empirical comparation of the efficacy of covariance-based and variance-based SEM. Int. J. Res. Mark. 26 (332–244). Rodríguez, B., Rodríguez, A., Siverio, F., Siverio, M., 2010. Causes of raptor admissions to a Wildlife Rehabilitation Center in Tenerife (Canary Islands). J. Raptor Res. 44, 30–39. https://doi.org/10.3356/JRR-09-40.1. Sanches Fernandes, L.F., Terêncio, D.P.S., Pacheco, F.A.L., 2015. Rainwater harvesting systems for low demanding applications. Sci. Total Environ. 529, 91–100. Sanches Fernandes, L.F., Fernandes, A.C.P., Ferreira, A.R.L., Cortes, R.M.V., Pacheco, F.A.L., 2018. A partial least squares – path modeling analysis for the understanding of biodiversity loss in rural and urban watersheds in Portugal. Sci. Total Environ. 626, 1069–1085. https://doi.org/10.1016/j.scitotenv.2018.01.127. Santos, M., Bessa, R., Cabral, J.A., Leal-Pacheco, F., Leitão, D., Moreira, F., Pinto, M., Lecoq, M., Silva, J.P., 2016. Impacts of land use and infrastructural changes on threatened Little Bustard Tetrax tetrax breeding populations: quantitative assessments using a recently developed spatially explicit dynamic modelling framework. Bird Conserv. Int., 1–18 https://doi.org/10.1017/S0959270915000258. Santos, R.M.B., Fernandes, L.F.S., Varandas, S.G.P., Pereira, M.G., Sousa, R., Teixeira, A., Cortes, R.M.V., Pacheco, F.A.L., 2018. Impacts of climate change and land-use scenarios on
3006
A. Garcês et al. / Science of the Total Environment 650 (2019) 2996–3006
Margaritifera margaritifera, an environmental indicator and endangered species. Sci. Total Environ. 511, 477–488. https://doi.org/10.1016/j.scitotenv.2014.12.090. Soares, S.P.V., 2014. Paretic Syndrome in Gulls (LARIDAE) in The South of Portugal. University of Lisbon. Sol, D., Lapiedra, O., González-Lagos, C., 2013. Behavioural adjustments for a life in the city. Anim. Behav., 1–12 https://doi.org/10.1016/j.anbehav.2013.01.023 Special is. Tenenhaus, M., Vinzi, V., Chatelin, Y., Lauro, C., 2005. PLS path modeling. Comput. Statist. Data Anal. 48, 159–205 Available at:. https://studies2.hec.fr/jahia/webdav/site/hec/ shared/sites/tenenhaus/acces_anonyme/home/articles/PL_PM_5.pdf. Terêncio, D.P.S., Sanches Fernandes, L.F., Cortes, R.M.V., Pacheco, F.A.L., 2017. Improved framework model to allocate optimal rainwater harvesting sites in small watersheds for agro-forestry uses. J. Hydrol. 550, 318–330. Terêncio, D.P.S., Sanches Fernandes, L.F., Cortes, R.M.V., Moura, J.P., Pacheco, F.A.L., 2018. Rainwater harvesting in catchments for agro-forestry uses: a study focused on the
balance between sustainability values and storage capacity. Sci. Total Environ. 613–614, 1079–1092. Tintó, A., Real, J., Mañosa, S., 2010. Predicting and correcting electrocution of birds in Mediterranean areas. J. Wildl. Manag. 74, 1852–1862. https://doi.org/10.2193/2009-521. Wendell, M.D., Sleeman, J.M., Kratz, G., 2002. Retrospective study of morbidity and mortality of raptors admitted to Colorado State University Veterinary Teaching Hospital during 1995 to 1998. J. Wildl. Dis. 38, 101–106. https://doi.org/10.7589/0090-355838.1.101. Zambrano, T., Martins, M., 2004. Identificação e análise dos impactos ambientais das pequenas indústrias de São Carlos – SP. XXIV Encontro de Engenharia de Produção. Florianópolis, pp. 5326–5333.