Review Wildlife health and supplemental feeding: A review and management recommendations Maureen H. Murray a, ⁎ , Daniel J. Becker b,c , Richard J. Hall b,c,d , Sonia M. Hernandez a,e a Warnell School of Forestry and Natural Resources, University of Georgia, 180 E Green St, Athens, GA 30602, United States b Odum School of Ecology, University of Georgia, 140 E Green St, Athens, GA 30602, United States c Center for the Ecology of Infectious Diseases, University of Georgia, 140 E Green St, Athens, GA 30602, United States d Department of Infectious Diseases, College of Veterinary Medicine, University of Georgia, Athens, GA 30602, United States e Southeastern Cooperative Wildlife Disease Study, University of Georgia, 589 D. W. Brooks Drive, Athens, GA 30602, United States a b s t r a c t a r t i c l e i n f o Article history: Received 19 July 2016 Received in revised form 13 September 2016 Accepted 24 October 2016 Available online 12 November 2016 Humans provide supplemental food to wildlife under many contexts, ranging from professional feeding areas for game species to backyard bird feeders. Such resources bene fi t wildlife by providing reliable resources during pe- riods of food shortages, but may also alter the risk of pathogen transmission and development of disease. While several reviews have summarized the detrimental effects of supplemental food on infection risk, we conducted a comprehensive review to quantify support for mechanisms by which intentional wildlife feeding in fl uences host condition (i.e. malnutrition and stress) and pathogen transmission on a global scale and provide a framework to mitigate these risks. We also examined whether the purpose of feeding, whether for game management, conser- vation, tourism, or in residential areas, in fl uenced health outcomes. We found 115 studies that evaluated the health of wildlife with supplementary feeding, representing 68 species in 35 countries, although nearly half (46% of studies) were from North America. Supplemental feeding tended to increase the risk of pathogen trans- mission by increasing contact rates between hosts (95%) and promoting pathogen accumulation at feeders or the surrounding environment (77%). Provisioned food was also often a source of immunosuppressive contaminants (80%). Feeding associated with tourism frequently increased wildlife stress, rates of injury, pathogen prevalence, or malnutrition (85%), while feeding for conservation purposes had mostly positive effects on wildlife health (63%). We recommend adopting feeding practices that validate the nutritional appropriateness of wildlife feed for the target species, make food available at lower densities for short periods at unpredictable times and places to prevent aggregation, and avoid feeding during times of migration, pulses of new recruits, and epidemics. These strategies will help retain the recreational and management bene fi ts of wildlife provisioning while mitigating negative effects for many species around the world. © 2016 Elsevier Ltd. All rights reserved. Keywords: Supplemental feeding Provisioning Wildlife disease Pathogen transmission Fitness Nutrition Contents 1. Introduction . 164 2. Literature review . 164 3. Processes by which provisioning affects wildlife health . 166 3.1. Changes in host condition and immune function 166 3.1.1. Nutritional consequences of food quality and quantity . 166 3.1.2. Contaminants and toxins in wildlife feed . 167 3.1.3. Crowding, stress, and immunosuppression . 167 3.2. Behavioral and demographic changes that alter direct pathogen transmission 168 3.2.1. Host aggregation at feeding stations . 168 3.2.2. Host demographic rates and pathogen transmission . 169 3.3. Mechanisms that alter indirect pathogen transmission . 169 3.3.1. Environmental buildup of pathogen infectious stages . 169 Biological Conservation 204 (2016) 163 – 174 ⁎ Corresponding author at: 180 E Green St, Athens, GA 30602, United States. E-mail address: mhmurray@uga.edu (M.H. Murray). http://dx.doi.org/10.1016/j.biocon.2016.10.034 0006-3207/© 2016 Elsevier Ltd. All rights reserved. Contents lists available at ScienceDirect Biological Conservation j o u r n a l h o m e p a g e : w w w . e l s e v i e r . c o m / l o c a t e / b i o c 3.3.2. Changes in host community composition that affect pathogen transmission 169 4. Managing supplemental feeding to mitigate health impacts . 169 4.1. Improving food quality and providing nutritionally complete diets 169 4.2. Design and maintenance of feeders to lower exposure to pathogens . 169 4.3. Timing of feeding . 170 4.4. Location and spacing of feeding stations 170 4.5. Understanding the biology of target wildlife species . 170 4.6. Purpose of feeding: management vs. human enjoyment 171 4.7. Using feeding as an intervention strategy . 171 5. Synthesis and future directions . 171 5.1. Understanding how multiple mechanisms interact and different scales of organization 171 5.2. Reciprocal effects between host condition and provisioning. 172 6. Conclusions. 172 Acknowledgements 172 Appendix A. Supplementary data . 172 References. 172 1. Introduction Around the globe, humans intentionally provision wildlife with food for recreational and management purposes. Managers provide food to declining wildlife populations that are food limited (Ewen et al., 2015), to attract wildlife away from speci fi c areas or food types (i.e. di- versionary feeding; Conover, 2002), to attract animals for research or to deliver medicine (e.g. Miller et al., 2000). Managers and hunters also maintain feeding stations to support high densities of game species for sport hunting. As one example, up to 2.8 trillion tons of shelled corn is offered to ungulates every year in the U.S. (cited in Oro et al., 2013). Feeding wildlife is also a popular activity for public recreation. Over 80 million households provide up to 450 million kg of seed for wild birds in residential feeders each year in the U.S. and U.K. (Jones, 2011; Robb et al., 2008). Outside of residential areas, feeding is also commonly used to promote sightings and encounters with wildlife for tourists (Orams, 2002). Intentionally feeding wildlife can bene fi t both wildlife and humans. For some species, consistent access to a reliable food source can de- crease the likelihood of starvation or nutritional stress (Page and Underwood, 2006; Wilcoxen et al., 2015). For example, vultures ( Gyps sp.) in Europe and Asia were provisioned with poison-free carcasses at “ vulture restaurants ” to divert them away from diclofenac-laced car- casses that were causing mass mortalities (Gilbert et al., 2007). Feeding wildlife can also provide psychological and health bene fi ts to people by allowing them to interact with wildlife, a form of connecting with na- ture (Rowan and Beck, 1994; St Leger, 2003). Despite these known ben- e fi ts, feeding wildlife is often controversial (Brown and Cooper, 2006; Milner et al., 2014) because it can also promote negative ecological ef- fects (reviewed in Boutin, 1990; Oro et al., 2013) including foraging shifts in the behavior of individual animals (e.g. begging for food from humans; Samuels and Bejder, 2004). Feeding can also have communi- ty-level implications beyond target species. For example, 98% of visits to Northern bobwhite ( Colinus virginianus ) feeders were made by non-target species (Guthery et al., 2004), and some wildlife dispropor- tionately bene fi t from subsidies to alter species interactions such as predator-prey dynamics. Similarly, anthropogenic food provided unin- tentionally in campgrounds supports higher densities of corvids, which increases predation on nests of threatened species such as mar- bled murrelets ( Brachyramphus marmoratus ; Marzluff and Neatherlin, 2006). One negative consequence of particular conservation concern is the increased risk of pathogen transmission around feeding stations, including cross-species transmission (Becker et al., 2015; Becker and Hall, 2014; Tollington et al., 2015). This may happen in several ways. If anthropogenic food is of low quality, it may compromise wildlife condition and immune function. Because wildlife might aggre- gate around food sources, high local densities associated with supplemental feeding can increase the transmission of pathogens among animals and lead to the accumulation of infectious stages in the environment. Animals with access to reliable food sources often move less and form more sedentary populations, with further conse- quences for pathogen transmission (e.g. Flack et al., 2016; Satter fi eld et al., 2015). While several studies have investigated the links be- tween supplemental feeding and the spread of pathogens of eco- nomic importance in ungulates in North America (e.g. Sorensen et al., 2014), provisioning occurs in many other contexts around the world and may have important consequences for the health of spe- cies of conservation concern. Because the effects of anthropogenic food on wildlife infection risk are complex and varied, Becker et al. (2015) recently performed a meta-analysis assessing the positive and negative effects of anthro- pogenic food on infection outcomes in wildlife. They concluded that intentionally managed and recreational food resources generally in- creased infection risk for provisioned wildlife, whereas accidental provisioning (e.g. feeding on garbage or agricultural fi elds) fre- quently had small or negative effects on infection measures (Becker et al., 2015). Here, we build on this result by conducting a comprehensive literature review to quantify support for the mecha- nisms by which intentional provisioning can alter the condition of wildlife and factors in fl uencing exposure to pathogens for a diverse range of taxa and across different reasons for feeding. We also out- line a framework by which managers can anticipate and prevent health de fi cits in species provisioned for the purposes of species management or recreation. We focus on intentional provisioning, de fi ned as food deliberately supplied to free-living or semi-domesti- cated wildlife, because wildlife managers might be able to mitigate or reverse negative health effects more easily than with unintention- al provisioning. In reviewing the existing literature, we focus on three mechanisms by which feeding can affect wildlife infection risk, morbidity, condition, or fi tness: through changes in condition and susceptibility to infection, altered contact rates with infected in- dividuals (direct transmission), and altered contact rates with envi- ronmental pathogen stages that accumulate around feeding sites (indirect transmission). Our goals are to provide a greater under- standing of the mechanisms affecting the health of provisioned wild- life and to suggest novel intervention tools to prevent adverse health effects. 2. Literature review We searched for studies examining the health of provisioned wild- life using the Google Scholar and Web of Science databases. We used the following search terms and Boolean operators: (( “ anthropogenic food ” OR “ supplemental food ” OR feeding OR provisioning OR “ supple- mental feed* ” OR “ supplemental resource* ” OR “ resource subsidies ” OR 164 M.H. Murray et al. / Biological Conservation 204 (2016) 163 – 174 supplementation) AND (disease* OR infect* OR pathogen* OR transm* OR condition OR health) AND wildlife). We collected papers from 1961 to December 15, 2015. We retained journal publications or gradu- ate theses that examined effects of intentional provisioning, by either professionals or the public, on some aspect of wildlife health including pathogen or parasite prevalence, load, or diversity; immune function; stress levels; body condition; and growth, reproduction, or survival of free-living or semi-domestic wildlife. For retained studies, we recorded whether the study was observational or experimental and the type of feeding involved: by managers to support game populations or to sup- port small or declining populations for conservation; or by the public for recreation in residential areas or for tourism. If provisioning was done by researchers to test the effects of existing feeding programs or recreational activities on wildlife health, we classi fi ed that study under the category which the feeding emulated. For example, a research study that manipulated the spacing of existing feeding stations for elk (e.g. Creech et al., 2012) would be classi fi ed as game management. We also recorded the target host and pathogen species, other health ef- fects measured by the study (Table S1), and the study latitude and longitude. Our database searches yielded 1594 records. We removed 51 dupli- cates and retained 115 studies that met our criteria of including provi- sioned wildlife and at least one measure of infection risk, immune function, or nutritional de fi ciency (see Supplementary material for more detail and PRISMA diagram). The majority of records (n = 1326) were removed because they consisted of captive feeding or sup- plement trials on domestic animals or aquaculture, human studies, re- view papers, or theoretical studies. We found studies examining 68 different wildlife species representing 5 orders, mostly mammals (52%) followed by birds (38%) and reptiles, rays, and fi sh (3% each). Feeding game populations was most common (37%), followed equally by conservation, tourism, and residential (all 21%). These studies spanned 35 countries representing all continents on which humans re- side, although nearly half (46%) were in North America (Fig. 1). Most studies (63%) reported observational data from existing feeding pro- grams, but several manipulated existing feeding programs (18%) or held captive experiments (19%). Of the 115 studies, 42% found positive effects on one or more measures of health, 20% found no effect, and 38% found negative effects and we noted large differences between health effects for different feeding types and health metrics. Because they are important metrics for the success of wildlife feeding programs, we also recorded any changes in wildlife survival and reproduction. Wildlife survival and reproduction increased with supplemental feeding in 75% of studies, and this was mainly driven by populations fed for con- servation purposes (11 of 13 of studies). However, this increase may belie important health effects, as the prevalence of infections by patho- gens or parasites increased with supplemental feeding in 95% of studies (Fig. 2 and Table 1) across all feeding types. Most studies of tourism feeding found negative consequences for measures of animal health (85%), and conservation was the only feeding type with mostly positive health effects (63%; Fig. 3). We found evidence for several mechanisms Fig. 1. Map showing locations of the 115 studies that examined changes in wildlife health including survival, nutrition, and pathogen prevalence (Table S1) with supplementary feeding programs in 35 countries. Fig. 2. The proportion of 115 studies in our review that report negative, positive, or no health effects associated with the four most common health outcomes measured with wildlife provisioning. “ Health effect ” indicates the overall effect of that factor on host health, with a decrease in infection measures considered a positive health effect (summarized in Table S1). The number of studies we found in each category is displayed inside the wedges. 165 M.H. Murray et al. / Biological Conservation 204 (2016) 163 – 174 by which supplemental feeding could negatively affect wildlife health (Table 1), discussed below. 3. Processes by which provisioning affects wildlife health The articles in our literature review identi fi ed three broad mecha- nisms by which provisioning alters the health of wildlife: changes in host condition and immunity from nutritional status or stress, changes in host behavior that affect directly transmitted pathogens, and changes in exposure to pathogens in the environment. Here we discuss the sup- port in the literature for each of these processes in detail. 3.1. Changes in host condition and immune function 3.1.1. Nutritional consequences of food quality and quantity Wildlife nutrition is notoriously dif fi cult to study (Barboza et al., 2009) as it requires repeated fi eld observations and often captive feeding trials for robust results. Moreover, the best physiological assess- ments of nutrition require invasive sampling (capture, immobilization, and removal of blood or tissue) to examine levels of vitamins, fatty acids, plasma protein, and several other biochemical analyses (Page and Underwood, 2006). In other cases, noninvasive measures might in- clude fecal or urine analyses, but these have yet to be validated for most wildlife species. Nutritional studies also often require various experi- mental trials that are logistically dif fi cult to conduct on wildlife. Owing to these challenges, our knowledge of what food to provide wildlife lags behind nutritional guidelines for domestic species. Thus, managers and wildlife enthusiasts often use feed intended for other species (e.g. commercially prepared pelleted diets) or based on convenience or af- fordability (e.g. millet in birdseed). Our literature review produced mixed results on health effects linked to nutrition provided in supplemental food (Table 1). While nearly half of the studies found negative effects of provisioning on mea- sures such as protein or micronutrient de fi ciencies (42%), other studies showed bene fi ts to the health of provisioned wildlife (36%). In some cases, provisioning was found to increase host body condition (Kaneko and Maruyama, 2005) and reduce physiological measures of nutritional stress. For example, provisioned white-tailed deer ( Odocoileus virginianus ) had higher levels of fecal nitrogen than unprovisioned white-tailed deer over the winter when ungulates are typically nutritionally stressed (Page and Underwood, 2006). In terms of consequences for infection, food quality and quantity can alter immune function and susceptibility to infection. For example, be- cause protein is crucial for the proper functioning of T cells and the adaptive immune system (Cooper et al., 1974), protein-de fi cient diets can lower immune function (Taylor et al., 2013). To illustrate, free-living ungulates that consumed diets low in protein had higher gastrointesti- nal parasite loads (Ezenwa, 2004). This relationship between dietary protein and immunity can be dif fi cult to tease apart in natural popula- tions, in part, because parasites can directly affect protein levels. For Table 1 The frequency of the fi ve most common mechanisms for changes in wildlife health (Table S1) and three mechanisms of pathogen transmission with supplemental feeding found in 115 studies. Health effect Mechanism Negative No effect Positive Contact rates 10 – – Immune function 3 2 2 Stress hormones 4 – 1 Contaminants 4 1 – Nutrition 5 2 4 Transmission mode Intermediate host 3 1 3 Non-close contact 17 6 1 Close contact 24 5 – Fig. 3. The proportion of 115 studies in our review that report negative, positive, or no health effects associated with existing intentional supplementary feeding programs by: (i) managers to game populations such as red deer ( Cervus elaphus ; Vicente et al., 2007) and (ii) populations of conservation concern such as Spanish Imperial Eagles ( Aquila adalberti ; Blanco et al., 2011); and (iii) the public in residential areas, for example Japanese badgers ( Meles anakuma ; Kaneko and Maruyama, 2005) or (iv) as tourists during excursions such as those to feed southern stingrays ( Dasyatis americana ; Semeniuk et al., 2009). The number of studies we found in each category is displayed inside the wedges. Fitness and health effects examined by papers are summarized in Table S1. Images from WikiMedia commons. 166 M.H. Murray et al. / Biological Conservation 204 (2016) 163 – 174 example, gastrointestinal nematodes promote the loss of protein into the gastrointestinal tract, reducing absorption (Holmes, 1993). Provid- ing adequate dietary protein is especially of concern for herbivores, whose protein requirements are often underestimated across different life stages (e.g. lactation; Brown and Cooper, 2006) and likely unmet with grain and corn versus seasonally available vegetation. For example, provisioned elk ( Cervus elaphus ) can miss seasonal and protein-rich young forbs by abandoning traditional migration in favor of grains and hay provided on feeding grounds (Jones et al., 2014). Beyond host immunity, nutritionally incomplete food can cause other problems for wildlife health. One of the most notorious exam- ples is “ Angel Wing ” , a deformity mostly of waterfowl, thought to be caused by diets high in calories but de fi cient in vitamins and min- erals fed to wild birds (Flinchum, 1997). Thus, birds offered anthro- pogenic food such as bread and chips in parks are at a higher risk of angel wing. Similarly, a tourist-fed population of endangered North Bahamian Rock Iguanas ( Cyclura cychlura ) were de fi cient in potassi- um relative to an unfed population, likely because grapes (the fruit- of-choice provided by tourists) contain up to 10 times less potassium than naturally available vegetation (Knapp et al., 2013). This same population was routinely offered ground beef and had levels of tri- glycerides (constituents of fat) nearly four times higher than iguanas not provisioned, a fi nding similar to an overweight population of tourist-fed Barbary macaques ( Macaca sylvanus ; O'Leary, 1996). 3.1.2. Contaminants and toxins in wildlife feed When provisioned food contains toxic or pathogenic contaminants, severe negative health consequences can occur, in some cases promot- ing signi fi cant die-offs of wildlife. We found only fi ve studies examining effects of contaminants in provisioned feed on wildlife health, of which four detected negative effects (Table 1). One of the most common causes of wildlife feed contamination is fungal growth, which can produce toxic metabolic byproducts known as mycotoxins. Feed can become contaminated with mycotoxins when it is grown, stored prior to distribution, and held in feeding stations. Two common types of mycotoxin in feed are a fl atoxin and ochratoxin A, which can cause immunosuppression, anorexia, organ failure, and cancer at high concentrations or with chronic exposure (Hussein and Brasel, 2001; Moore et al., 2013). Because of these health risks, govern- ment agencies have established legal limits for mycotoxin concentra- tions in domestic animal feed (e.g. EFSA, 2006). Mycotoxins are typically detected in feed containing grain or corn, and a fl atoxin has been detected in supplemental corn provided for wild northern bob- whites ( Colinus virginianus ; Oberheu and Dabbert, 2001), turkeys ( Meleagris gallopavo ; Schweitzer et al., 2001), and white-tailed deer ( Odocoileus virginianus ; Brown and Cooper, 2006). Illegal concentra- tions of a fl atoxin have been detected in up to 40% of bags of “ deer corn ” in Texas (Milner et al., 2014), which could cause acute illness in birds. Several mortality events of captive endangered whooping cranes ( Grus americana ) and sandhill cranes ( Grus canadensis ) were associated with feed contaminated with the fungus Fusarium spp. and two species of mycotoxins, T2 toxin and deoxynivalenol (Olsen et al., 1995). Al- though not necessarily left intentionally as forage for wildlife, moldy corn, rice, and peanuts containing extremely high concentrations of af- latoxin were implicated in the deaths of 7500 waterfowl in Texas (Robinson et al., 1982) and over 10,000 waterfowl in Louisiana in the winter of 1988 – 99 (cited in Lawson et al., 2006). Mycotoxins have also been detected in commercial birdseed (Henke et al., 2001) and chronic doses could impair bird immunity even at low concentrations. While grain-based feed provided to wildlife can promote the growth of harmful contaminants, carcasses provided to carnivores can also har- bor toxins or pathogens. For example, Spanish Imperial Eagles ( Aquila adalberti ) fed rabbits medicated with antibiotics and antiparasitics ex- hibited depressed immune systems (Blanco et al., 2011). Thus, it is of crucial importance to verify that provisioned feed is nutritionally complete, appropriate for the target species, and free of deliberately added or accidental contaminants to maintain wildlife immune function. 3.1.3. Crowding, stress, and immunosuppression Feeding stations can attract large numbers of animals, leading to crowding and competition for resources and space. We found 12 studies examining effects of feeding on wildlife crowding and stress with likely or measured consequences for host immunity (Table 1). Four of fi ve studies that measured stress directly found higher stress levels in provisioned wildlife; for example, elk on feedgrounds had higher fecal glucocorticoid levels driven by higher elk densities (Forristal et al., 2012). However, only 3 of 7 studies found negative consequences for immune function. Wildlife often conform to the resource-matching rule, in which for- agers distribute in an ideal free manner according to food resource den- sity (Pulliam and Caraco, 1984). Because population density usually scales with food abundance, the resource-matching hypothesis predicts that individual fi tness measures (fecundity, survival, condition) should be comparable between low- and high-resource habitats (Rodewald and Shustack, 2008). However, supplemental food in human-modi fi ed landscapes can promote overmatching or overexploitation of habitat, leading to increased stress and starvation risk (Faeth et al., 2005; Shochat et al., 2004). Habitats containing supplemental food can thus act as ecological traps (Dwernychuk and Boag, 1972) by appearing to be of high quality due to resource abundance but are functionally poor quality due to disproportionately high levels of crowding and competi- tion. For example, adult birds frequently outcompete young birds at supplemental feeding stations, resulting in poorer juvenile condition and higher mortality (Duriez et al., 2012; Wunderle, 1991). Similarly, winter feeding can reduce avian reproductive success by attracting birds to areas with insuf fi cient natural resource availability in spring (Plummer et al., 2013). However, other species such as raccoons ( Procy- on lotor ) can tolerate very high densities of conspeci fi cs if enough food is provided (Rosatte, 2000), and so these effects may be species- and con- text-speci fi c. Greater food competition and starvation risk could function as a chronic stressor with immunosuppressive effects. Acute stressors ac- tivate the hypothalamic – pituitary – adrenocortical (HPA) axis, which suppresses non-essential functions while enhancing energy alloca- tion to those important for immediate survival (Reeder and Kramer, 2005). Increased glucocorticoid production during HPA acti- vation is temporarily immunosuppressive until the stressor sub- sides; however, chronic stressors can cause long-term immune impairment (Padgett and Glaser, 2003). Tourist-provisioned vervet monkeys ( Chlorocebus pygerythrus ; Fourie et al., 2015) exhibit ele- vated indices of chronic stress in provisioned populations, suggested at least in part from greater competition. Such patterns could be par- ticularly pronounced in wildlife that display dominance hierarchies, owing to unequal resource partitioning between social classes and to direct in fl uences of hierarchy on cellular and humoral components of immune function (Hawley et al., 2007b; Wunderle, 1991). In support of this idea, tourist-fed southern stingrays ( Dasyatis americana ) displayed poorer condition and immune function from greater com- petition around concentrated food, and markers of immune suppres- sion were highest in individuals that displayed more frequent signs of aggressive interactions (Semeniuk and Rothley, 2008; Semeniuk et al., 2009). Even if provisioned wildlife are not directly food limited, competi- tion could arise from crowding in limited roosting or breeding territory (Davis and Maerz, 2009). Because crowding is a well-supported stressor linked with impaired resistance to pathogens (Goulson and Cory, 1995), regardless of the source of competition with supplemental feeding, heightened competition and stress could enhance individual suscepti- bility to infection. 167 M.H. Murray et al. / Biological Conservation 204 (2016) 163 – 174 3.2. Behavioral and demographic changes that alter direct pathogen transmission 3.2.1. Host aggregation at feeding stations Anthropogenic food is often unevenly distributed on the landscape and promotes aggregations of wild animals at higher abundances and densities than natural foraging activities. Abundant and concentrated resources can increase home range overlap, decrease home range size, increase group size, and reduce territoriality (Robb et al., 2008). These aggregations promote higher rates of intra- and interspeci fi c contact be- tween foragers (e.g. (Dhondt et al., 2007; Wright and Gompper, 2005), providing ideal conditions for the transmission of many pathogens. We found 29 studies that examined a pathogen or parasite transmitted through close proximity or direct contact, and nearly all (95%) found ev- idence of increased transmission with provisioning (Table 1). Further, all of the studies (n = 10) that measured contact rates in response to provisioning found positive relationships (Table 1). Studies where infection status was assumed to result from increased pathogen transmission due to increased aggregation at supplemental food resources are on the rise (e.g. Trichomonas gallinae infections of fi nches at feeders; Lawson et al. 2012). In some cases, the duration of supplemental feeding or behavior while feeding was directly correlated with infection status (e.g. Brucella abortus and gastrointestinal parasite infections in elk increased with time spent at supplemental feeders; Cross et al., 2007; Hines et al., 2007) and the risk of Mycoplasma gallisepticum infection of house fi nches ( Haemorhous mexicanus ) in- creased with time an individual spent at a feeder (Adelman et al., 2015; see Box 1). Supplemental food can also promote inter-speci fi c contact and cre- ate opportunities for novel interactions between species. Contact be- tween wild and domestic species can be higher when crop residues are left as feed for wildlife, promoting the spread of shared pathogens (e.g. bovine tuberculosis in cattle and white-tailed deer Odocoileus virginianus ; Brook et al., 2013). In urban areas, mixed fl ocks of birds at bird feeders or in parks can be composed of species that would naturally never come into contact, because feeding typically favors avian assem- blages largely composed of exotic invasives (Galbraith et al., 2015). Sim- ilarly, mesocarnivores that would typically avoid each other, like raccoons and skunks ( Mephitis mephitis ), engaged in more frequent and aggressive interactions between each other and with feral cats when cat food was experimentally provided to urban yards they were accustomed to visit (Theimer et al., 2015). Aggregations of prey species at feeding sites can attract predators and increase the likelihood of tro- phic transmission of pathogens. For instance, birds of the family Columbidae are hosts for Trichomonas gallinae and Cooper's hawks ( Ac- cipiter cooperii ) in urban areas where rock doves (pigeons; Columba livia ) aggregate, have both a higher proportion of pigeon in their diet and a higher prevalence of trichomoniasis (Boal et al., 1998). Cross-species transmission at supplemental feeding sites could im- pact human, livestock, and wildlife health. For example, the aggregation of wild boar ( Sus scrofa ) at arti fi cial watering and feeding sites in Spain was signi fi cantly associated with the presence of tuberculosis lesions (likely caused by Mycobacterium bovis ), which could increase exposure of domestic pigs and cattle to this pathogen (Vicente et al., 2007). The development of clinical respiratory diseases shared between humans and chimpanzees ( Pan troglodytes schweinfurthii ) was associated with periods where chimpanzees were supplemented with bananas at Gombe National Park, illustrating that disease transmission from humans to wildlife at feeding sites can have dire consequences for en- dangered species (Lonsdorf et al., 2011). Experimental studies, although rare, can provide crucial tests of un- derlying mechanisms linking host aggregation with pathogen transmis- sion. Forbes et al. (2015) experimentally manipulated vole ( Microtus agrestis ) populations to understand how supplemental feeding might affect infection with a common respiratory pathogen, Bordetella bronchiseptica . Supplemented voles experienced greater transmission Box 1 How supplemental feeding influences pathogen transmission and im- pacts: a case study of an emerging pathogen in a wild songbird. Background A B Mycoplasma gallisepticum (henceforth Mg ) is a bacterial pathogen of domestic poultry. In January 1994, infection was detected in house finches ( Haemorhous mexicanus , henceforth HOFI; Ley et al., 1996), a songbird introduced to the eastern US where it is now abundant. Symptoms include conjunctivitis (A; Ley et al., 1997) and lethargic behavior (Kollias et al., 2004); most birds recover from infection but severe cases can result in death (Kollias et al., 2004). Within 3 years Mg had spread from the mid-Atlantic states throughout the eastern US, where it was estimated to have killed 228 million HOFI (Dhondt et al., 1998); it reached native HOFI populations in the western US in 2002 (Duckworth et al., 2003). C D Feeding facilitates Mg transmission • Feeder density predicts regional HOFI density (Fischer and Miller, 2015); and Mg infection prevalence increases with local and regional HOFI density (Altizer et al., 2004b; Hochachka and Dhondt, 2000). • Seasonal peaks in prevalence coincide with winter aggregations of HOFI at feeders when natural food is scarce (Altizer et al., 2004a, 2004b; Hosseini et al., 2004). • Mg is transmitted via fomites deposited on bird feeders (B; Dhondt et al., 2007). • Infected birds stay longer at feeders (Hotchkiss et al., 2005), and those that stay longer are more likely to initiate epidemics (Adelman et al., 2015). • Time spent at feeders by uninfected HOFI predicts their infection risk (Adelman et al., 2015). • Infection risk is higher at tube feeders where fomites are concentrated than at platform feeders (Hartup et al., 1998). • Conjunctivitis associated with Mg infection has been observed in related feeder birds, Purple Finch ( Haemorhous purpureus ; C) and American Goldfinch ( Spinus tristis , D; (Hartup et al., 2000) Does feeding mitigate pathogen impacts? • HOFI infected via fomites develop mild disease and recover more quickly, suggest- ing that feeders might immunize against severe infections (Dhondt et al., 2007). • Consistent with a positive effect of feeding on Mg impacts, the steepest HOFI declines following Mg spread occurred where the number of people feeding birds fell dramatically (Fischer and Miller, 2015). • Symptomatic birds associate with smaller flock sizes and more likely to feed alone, potentially indicating behavioral avoidance (Hawley et al., 2007a). • The popularity of feeding allowed scientists to track spread of an emerging wild- life disease, and disseminate information on good feeder hygiene practice, via a pioneering citizen science project (Dhondt et al., 1998). All images taken by R. Hall. 168 M.H. Murray et al. / Biological Conservation 204 (2016) 163 – 174 and infection rates due to aggregation at feeding stations (Forbes et al., 2015). Similarly, raccoons foraging at experimental feeding sites had higher contact rates and endoparasite prevalence and diversity (Wright and Gompper, 2005). 3.2.2. Host demographic rates and pathogen transmission Most (75%) of the papers we found that examined the effects of sup- plemental feeding on rates of survival and reproduction reported posi- tive effects (Fig. 2). While this may indicate that supplemental feeding bene fi ts wildlife, it is important to consider the implications of arti fi cial- ly high survival and reproduction on pathogen transmission. Higher re- productive rates can produce more naïve susceptible hosts that lack acquired immunity and hence can enhance pathogen transmission (Lloyd-Smith et al., 2005). Higher rates of survival and reproduction also lead to higher population densities, which promotes density-de- pendent pathogen transmission (see Section 3.2.1). Providing a reliable source of easily-accessible food could also prolong the survival of infect- ed individuals, extending the infectious period over which transmission occurs. Apart from enhancing pathogen transmission, providing supple- mental resources could also promote health problems if it reduces the likelihood of dispersal, thereby increasing risk of inbreeding (Bowler and Benton, 2005). Although each of these mechanisms is well support- ed by theory, more work needs to be done in fee-living provisioned pop- ulations to examine any empirical support. 3.3. Mechanisms that alter indirect pathogen transmission 3.3.1. Environmental buildup of pathogen infectious stages Many pathogens are transmitted when susceptible hosts encounter infectious stages that are shed into the environment by infected animals via feces, respiratory secretions, saliva, and other routes. These infec- tious stages can persist outside of hosts for periods ranging from over an hour (e.g. Trichomonas gallinae in bird baths; Purple et al., 2015) to weeks (e.g. Salmonella enterica in soil; Jensen et al., 2006), months (e.g. avian in fl uenza in water; Stallknecht et al., 1990), or years (e.g. Chronic Wasting Disease prions in soil; Miller et al., 2000). High densi- ties of wildlife around feeding locations and greater sedentary behavior can increase the likelihood of accumulation of infectious stages in the surrounding environment and thus pathogen transmission (e.g. an- thrax; Hugh-Jones and de Vos, 2002). In our review, the majority (77%) of the 13 studies that investigated risk of environmentally trans- mitted pathogens found support for pathogen accumulation at provi- sioned sites as a mechanism for increased transmission. Infective stages are more likely to accumulate in the environment when populations spend more time on or near feeding stations and re- peatedly resample the same habitats. For example, monarch butter fl ies that breed year-round in non-native milkweed have higher parasite prevalence than migratory monarchs (Satter fi eld et al., 2015). Patho- gens can also be deposited onto feeders and feed itself, promoting path- ogen transmission via fomite actions. For example, Mycoplasma gallisepticum in ocular secretions can be deposited by house fi nches and remain on feeders for hours or longer (Dhondt et al., 2007; Box 1). 3.3.2. Changes in host community composition that affect pathogen transmission By introducing supplemental resources, humans can change the rel- ative abundance of wildlife species that will either use this novel re- source or capitalize on subsidized prey. Some of these species, including the target species of feeding, can be competent hosts for path- ogens, in that they promote pathogen replication. Other species can be dilution or incompetent hosts, those that are functionally “ dead ends ” for pathogens (Keesing et al., 2006). If feeding supports higher abun- dances of dilution hosts it may decrease the transmission of pathogens and/or lower the likelihood that parasites with complex life cycles will fi nd intermediate or fi nal hosts (Johnson and Thieltges, 2010). In other cases, feeding might increase the abundance of ampli fi cation hosts, species in which pathogens can multiply to high concentrations (LoGiudice et al., 2003). Further, provisioning has been sh