REVENGE OF THE TREES: ENVIRONMENTAL DETERMINANTS AND POPULATION EFFECTS OF INFECTIOUS DISEASE OUTBREAKS ON A BREEDING COLONY OF DOUBLE-CRESTED CORMORANTS (PHALACROCORAX AURITUS) OVER A PERIOD OF 21 YEARS

One outcome of the past three decades of research and responses to new and newly important (emerging) infectious diseases has been a renewed recognition of the preeminence of environmental factors in determining the location, timing, scale, and outcome of epidemics in people and animals (Eisenberg et al. 2007). Short-term studies of such complex relationships can easily lead to false conclusions because both biological and environmental factors of major importance may fluctuate over time frames longer than the study periods (Pelton and van Manen 1996). In this article, we report the results of a continuous, multicomponent study of infectious disease epidemics on a large breeding colony of Double-crested Cormorants (Phalacrocorax auritus; hereafter, cormorants) in the southern boreal forest of Canada carried out from 1994 to 2014. We offer a hypothesis derived from these long-term data regarding environmental determinants of the epidemics observed, and explore the potential impacts of these epidemics on the colony population.

The Double-crested Cormorant is abundant and breeds in colonies distributed across the interior of Canada and the US. Over most of its range, this cormorant migrates north to breed in summer and south in winter. The continental population is between 1 and 2 million birds, including about 350,000 breeding pairs (Dorr et al. 2014). It is likely that the current population is similar in size to the population before European settlement of North America, but the species now is widely perceived to be overabundant (Wires 2014).

In 1990, an epidemic of Newcastle disease (ND), defined as a disease in any bird species caused by infection with strains of avian paramyxovirus-1 (APMV-1), which are highly pathogenic in chickens, was discovered in cormorants in western Canada, and at additional locations in Canada and the US in subsequent years (Wobeser et al. 1993; Kuiken 1999). The disease was not, and is not, present in commercial poultry in Canada or the US, and its discovery in cormorants posed an apparent new risk to the continental agricultural economy (Cattoli et al. 2011).

In 1994, research on ND and other causes of mortality in cormorants was initiated on Island A, a provincial wildlife sanctuary on Doré Lake (Saskatchewan, Canada), which is the site of a large nesting colony. Intensive study for 3 yr (Kuiken et al. 1998b) was followed by 18 additional years of monitoring the occurrence and causes of diseases producing large-scale mortality (epidemics) in cormorants during the breeding season; during that period, epidemic avian cholera (AC; infection with the bacterium Pasteurella multocida) was also discovered.

This article reports the results of these 21 yr of observation. In addition to the occurrence of epidemic diseases, data were collected on the pathogens causing epidemics, the range of species and age classes affected, the presence in freshly laid eggs of antibodies against APMV-1, and the total number of breeding pairs of cormorants. The potential impacts of the observed epidemics on the breeding population of cormorants on Island A were estimated. A model of APMV-1 transmission dynamics on Island A was constructed and was used to identify the parameters potentially most important in determining the occurrence and scale of epidemics and to examine whether changes over time in the proportion of nests in trees versus on the ground might influence disease occurrence, as was observed by Cleary (1977) and proposed by Kuiken et al. (1998b).

Island A is a small island (approximately 100×300 m) located in Doré Lake, Saskatchewan, Canada (54°46′N, 107°17′W). Colony-nesting waterbirds, particularly cormorants and American White Pelicans (Pelecanus erythrorhynchos), began nesting there in 1983 when the island was forested (Hanbidge 1989). Large-scale mortality of cormorants was reported for the first time on Island A by provincial conservation officers in 1992, at which time all trees on the island were dead. During the 21 yr of this study, in addition to cormorants, approximately 1,500 nesting pairs of pelicans, fewer than 10 pairs of Herring Gulls (Larus argentatus), one pair of Bald Eagles (Haliaeetus leucocephalus), but no other bird species, were found nesting on Island A in some years.

During 1994–96, the colony was monitored intensively (Kuiken 2016; Kuiken et al. 1998b). From 1997 to 2014, the colony was visited by two to four people one to three times each summer in late July and early August, the period during which hatchling cormorants have left their nests and are most vulnerable to bird-to-bird transmission of contagious pathogens (Kuiken et al. 1998b). At each visit, Island A was searched systematically and completely on foot for sick or dead birds. Dead birds were identified to species based on field characteristics; were categorized as nestling (down-covered and present in a nest, approximately 0–4 wk old), juvenile (postnestling subadult that has left the nest but is fed by parents off the nest or has just achieved flight and independence [fledged], approximately 4–8 wk old), or adult, based on location, size, and feathering; and were counted. When dead birds were discovered, necropsies to determine all potential causes of death were conducted on up to 10 freshly dead cormorants either on Island A or at the Western College of Veterinary Medicine in Saskatoon. When found, specimens of other species were also necropsied. Basic sanitary procedures were used to prevent unintended transport of pathogens from Island A to other locations.

In all cases, grossly evident and microscopic lesions were assessed by qualified veterinary pathologists. Samples of liver, spleen, kidney, heart, and brain were frozen and subsequently used for detection of APMV-1 by PCR or virus isolation (Wobeser et al. 1993; Kuiken et al. 1998b, Wise et al. 2004). Immunohistochemistry was also used to identify APMV-1 in formalin-fixed, histologic sections (Kuiken et al. 1999b). If a specimen was found to be infected with APMV-1, samples were sent to the National Centre for Foreign Animal Diseases, Canadian Food Inspection Agency, to identify the virus and to assess its pathogenicity to chickens by gene sequencing or by experimental infections (Clavijo et al. 1998).

Samples of liver and spleen were cultured for pathogenic bacteria at the veterinary diagnostic laboratory of the Western College of Veterinary Medicine. Samples were inoculated on blood and MacConkey agar plates (Oxoid Canada, Nepean, ON, Canada), and isolates were identified by standard laboratory procedures (Quinn et al. 2011).

The concentration of antibodies in the yolk of freshly laid eggs was used to assess the level of humoral immunity to APMV-1 in breeding female cormorants returning to the colony site in spring. From 1994 to 2003, 100 freshly laid eggs were collected randomly (one freshly laid egg from the nest nearest a randomly selected intersection point on a 1-m grid superimposed on a map of the island) across the colony in late May each year. Antibodies to AMPV-1 were measured in 1:10 or 1:5 dilutions of the yolk in phosphate-buffered saline by hemagglutination inhibition (Kuiken et al. 1998b).

In 4 different yr, freshly voided excrement from cormorants was collected on Island A in late May or early June to determine whether adult cormorants returning to the colony site were shedding AMPV-1, and in August 2008, samples of freshly voided excrement from cormorants of unknown age and one sample from the cloaca of a moribund juvenile cormorant were collected during an epidemic of ND (Table 1). Samples were held on ice for 1–3 d and then frozen at –80 C until analyzed. The PCR for AMPV-1 was performed at the Department of Viroscience, Erasmus University Medical Centre (Rotterdam, the Netherlands) by published methods (Wise et al. 2004).

The total number of active cormorant nests on Island A was counted each year from 1996 to 2014 using photographs taken directly above the colony by low-flying aircraft (Fig. 1). Photographs were taken in late May or early June to coincide with the peak of incubation activity. Each active nest thus counted was assumed to represent a breeding pair of cormorants (Somers et al. 2010).

Stage-specific mortality rates were used to estimate the capacity of the Island A breeding colony to sustain its breeding population of cormorants under various magnitudes of additive mortality from epidemic diseases, based on published annual mortality data (Dorr et al. 2014) and on data of reproductive success measured on Island A from 1994 to 1996 (Kuiken et al. 1999a). These calculations assumed that immigration and emigration of breeding cormorants on Island A did not occur and that mortality of juveniles from epidemics of ND and AC occurred in addition to published annual mortality rates for cormorants in their first year. Parameter values and calculations for these estimates are in Tables 2 and 3. The average number of cormorants fledged per nest (breeding pair), 1.4, is the rounded mean estimate for this colony in 1994–96 (1.38; Kuiken et al. 1999a). The average number of breeding pairs on Island A is from this study. Annual survival rates for different age classes of cormorants are from Dorr et al. (2014).

Outbreak dynamics for ND in juvenile cormorants were modeled using a compartmental epidemiologic model with four classes: susceptible individuals (S), which have not yet been exposed to APMV-1; exposed individuals (E), which have been infected and are incubating the virus but are not yet infectious; infectious individuals (I), which are actively shedding the virus; and individuals that have recovered (R) from infection and are no longer infectious. We assumed that recovered individuals acquire complete immunity and cannot become reinfected during the nesting season. We used the following first-order differential equations to represent the infection dynamics of ND in juvenile cormorants:

Model parameters: Model parameter values and initial conditions used for simulation are listed in Table 4. The rate at which exposed juvenile cormorants become infectious (k) is the inverse of the virus incubation period, which ranged from 2 to 4 d (mean=3.27 d) in experimental infection trials (Kuiken et al. 1998a). Recovery rate (α) is the inverse of the number of days cormorants remain infectious (shedding period), which was calculated as the mean shedding period for experimentally infected juvenile cormorants (Kuiken et al. 1998a). Disease-induced mortality rate (m) for infected birds is the inverse of survival time, which is short in juvenile cormorants infected with APMV-1 but for which no precise data are available. Postinfection survival time in infected young chickens is often 3 d or less (Saif and Fadly 2008); 3 d was used here as the initial parameter value. For simplicity, we assumed that mortality from sources other than disease during the juvenile period between leaving the nest and fledging was constant and negligible, as reported by Kuiken et al. (1999a). Finally, in the absence of empirical estimates of transmission rate of APMV-1 between infected and susceptible juvenile cormorants, we estimated the transmission rate (β) by fixing all other parameters to their literature-derived values and then calculating the value of β required to produce the estimated 50% average juvenile mortality by the end of the nesting season observed in this study (see Results).

Initial conditions: The total number of juvenile cormorants (N) was calculated as the average total number of nests counted from 1996 to 2014 (6,849) multiplied by the mean number of young produced per nest observed on Island A from 1994 to 1996 (1.38), yielding 9,452 juvenile birds. We assumed that all nestling cormorants were susceptible to infection at hatching but that only some individuals acquired infection from their parents before leaving the nest. The initial number of infected juveniles was estimated based on our detection of virus in only one of 1,650 fresh fecal samples collected during the period of egg incubation, an apparent prevalence of infection in incubating breeding adults of 0.00061. The total number of infected returning adults was estimated as that apparent prevalence multiplied by the total number of adult breeding birds (13,698, twice the average nest count), yielding 8.4 infected returning adults. Assuming that each infected adult is in a different nest and that all offspring in that nest are infected when they leave the nest, the number of infected juveniles initially in the colony is 11.6 (1.38 per nest×8.4 infected adults). The initial number of juvenile cormorants in the exposed and recovered classes was set at zero.

Sensitivity analysis: In this model, we were most interested in one outcome: the total mortality of juvenile cormorants during the 30-d prefledging period. The sensitivity of the model to variation in model parameters was estimated by carrying out a global sensitivity analysis using the Latin hyper cube (LHC) method with partial rank-correlation coefficient index (Blower and Dowlattabadi 1994; Marino et al. 2008). This method tests multiple combinations of model parameters drawn from a distribution to evaluate the relative influence of each parameter on model outcomes of interest. The LHC consisted of 1,000 simulations of different parameter combinations, with all model parameters varied by ±20% and parameter values sampled from a uniform distribution within this range. We also varied the number of infected juveniles at the start of the simulation by ±20% within the LHC framework to investigate the sensitivity of model outcomes to that initial condition.

Nesting behaviour scenarios: To determine the effect of the timing of juveniles leaving the nest on ND outbreak dynamics, we used the model to compare two nesting scenarios: (1) nestling cormorants leaving the nest at 6 wk old (corresponding to tree-nesting behavior), and (2) nestlings leaving the nest at 4 wk old (corresponding to ground-nesting behavior; Lewis 1929).

All analyses were carried out with packages deSolve and lhs in R, version 3.0.2 (R Core Team 2017).

Outbreaks of two diseases, ND and AC, resulted in large-scale mortality on Island A in 16 of the 21 nesting seasons monitored in this study (Table 5). No other causes of epidemic disease were detected during this study by the diagnostic methods employed: necropsy, virus culture, PCR for APMV-1, histopathology, and screening cultures for bacterial pathogens. In four of the five seasons without outbreaks, neither pathogen was detected in the few birds found sick or dead on Island A; in 2001, infection with APMV-1 was detected in two moribund nestlings found in one nest, but large-scale mortality from that or other causes did not occur during the 2001 nesting season.

Nearly all cormorants dying in these epidemics were 6–8 wk old, fully feathered, and approaching adult size and independence. Adult cormorants were never found among the sick or dead in outbreaks of ND, and they were rare but present among dead birds observed during AC epidemics. Although ND affected only the cormorants, AC also affected a few of the other bird species (Table 6).

In the intensively studied epidemic of ND in 1995, the apparent mortality rate among juvenile cormorants was 32% (Kuiken et al. 1998b). This was considered a minimum estimate because detection of dead birds on the colony site was less than 100% and because cormorants dead from ND were also found on a large adjacent island without nests. In assessments of subsequent epidemics, mortality data were limited to counts of dead cormorants made during single, brief visits to the colony site during the epidemics; thus, all are minimum mortality counts (Table 5). The largest number of dead birds counted during an epidemic was 3,000, counted at what was likely the peak of mortality in 2006. That is 27% of the 11,218 juveniles estimated to be in the Island A population for that year and is close to the minimum mortality estimate of 32% of juvenile birds on the colony site made in 1995. As reasoned by Kuiken et al. (1998b), because some affected birds die off the colony site, carcasses can quickly be eliminated from the site by scavenging and by wave action at the shoreline, and counts of dead birds on the colony site are imperfect, we estimated that ND and AC killed approximately 50% of juvenile birds in each of the 16 yr when epidemics occurred.

From 1996 to 2014, the average number of cormorant nests counted from aerial photographs was 6,849 (range, 2,912–11,094; Table 5). Thus, on average, 13,698 breeding cormorants nested on the colony each year. There was no evident relationship between nest counts and occurrence of epidemic diseases (Table 5).

From 1994 to 2003, the percentage of antibody prevalence to APMV-1 in the yolk of freshly laid cormorant eggs varied from 43% to 100% (Table 5). Yolks with antibody titers of 1:20 or greater were considered positive (Kuiken et al. 1998b). There was no evident relationship among antibody prevalence, breeding population size, or occurrence of ND. Epidemics of ND occurred in cormorants hatched in years with the lowest and with the highest recorded antibody prevalence in the eggs sampled (Table 5).

In the 1,650 fresh samples of excrement collected from adult cormorants in spring, RNA attributable to APMV-1 was found by PCR in only one sample, with a cycle threshold value of 37 (Table 1). We detected RNA attributable to APMV-1 in three of 98 freshly voided samples collected from full-grown cormorants of unknown age in August during an epidemic of ND and in a sample collected from the cloaca of one juvenile cormorant, moribund from ND, during the same epidemic.

The annual adult mortality rate of 0.15 (Dorr et al. 2014) implies that 15% of the breeding bird population of Island A must be replaced each year if the breeding population is to be maintained over time. Cormorants breed for the first time at 2 yr old (Dorr et al. 2014). In years with no additional mortality of juveniles from disease, the number of juveniles fledged per nest, combined with the expected annual mortality of 52% and 26% of these birds in their first and second years, respectively (Dorr et al. 2014), yields a surplus of available new birds able to begin breeding in year three: 126% of the number required for replacement. In years with additional mortality among juveniles of 40, 50, or 60% (the range considered in this study), only 76, 63, and 50%, respectively, of the number of new breeders required to maintain the average breeding bird population size would be available 2 yr later.

As shown in Table 5, during the 21 yr of this study, 5-yr periods with the fewest epidemic years consisted of three epidemic years and two disease-free years. The most-affected 5-yr periods consisted of 5 consecutive yr of epidemic disease. As shown in Tables 2 and 3, the overall estimated effect on the breeding bird population of Island A from 50% additional mortality of juveniles caused by disease occurring in three or in five breeding seasons during a 5-yr period is production of 88% or of 63%, respectively, of the required replacement breeders.

Infection dynamics: The transmission model developed for APMV-1 suggested that virus transmission and resulting mortality are limited during the first 10 d after juvenile cormorants leave the nest but increase rapidly thereafter until juveniles fledge, progressively gain independence, and begin to disperse from the colony site (Fig. 2). Infection dynamics were most sensitive to changes in transmission rate (β), disease-induced mortality rate (m), and the rate at which exposed juvenile cormorants become infectious (k; Fig. 3). Increases in transmission rate (β) and in the rate at which exposed birds become infectious (k) caused an increase in total mortality of juveniles at the end of the nesting season, whereas more rapid mortality of infected birds (m) slowed down transmission, resulting in lower total mortality. Infection dynamics were moderately sensitive to the rate at which infected individuals recovered from infection (α), with more rapid recovery slowing down transmission and reducing total mortality. An increase in the number of infected juveniles at the time of their departure from the nest (I₀) accelerated transmission, resulting in greater total mortality; however, infection dynamics were least sensitive to variation in that parameter.

Nesting behavior scenarios: Simulations of ground-nesting vs. tree-nesting scenarios suggested that the earlier departure from the nest observed in ground-nesting cormorants has a significant effect on the occurrence of major disease outbreaks (Fig. 4). Juveniles from nests on the ground that left the nest at 4 wk old suffered 14 times greater disease-induced mortality by the end of the nesting season (50.32% mortality) than did nestlings in tree-nests that left the nest 2 wk later (3.56% mortality).

In 16 of 21 consecutive breeding seasons, the cormorants using Island A as a breeding site experienced approximately 50% mortality of hatch-year offspring because of epidemic infectious diseases during their juvenile period. From these data, it was possible to estimate the substantial potential impact of disease-associated mortality on the colony population (Tables 2, 3) and to document that such negative effects could not be detected in the actual population data. The rapid growth of the cormorant population from the 1970s to the present (Dorr et al. 2014; Wires 2014) implies that mortality factors, including diseases, have had no important impact on the population of the species on a continental scale. Immigration of breeding birds probably compensated completely for the loss of replacement breeders because of epidemics on Island A.

A plausible mathematical model of contagious pathogen transmission dynamics in this colony indicated that juveniles from nests on the ground were 14 times more likely to die of infectious disease than were juveniles from nests in trees, the critical difference being the 4-wk (ground nests) versus 2-wk (tree nests) period during which pathogen transmission could occur (Kuiken et al. 1998b).

Island A was first colonized by cormorants in 1983 when it was a forested island. Initially, cormorants nested primarily in trees but also on the ground (Hanbidge 1989). Cormorant excrement is toxic to vascular plants (Ishida 1997), and all trees were dead by 1992. As trees died and fell over, an ever-larger proportion of the cormorants had to nest on the ground; by 1994, nearly all nests were on the ground. In his seminal study of the Double-crested Cormorant, Lewis (1929) noted that nestlings from nests on the ground left the nest and began to live in groups of postnestling juvenile birds at about 4 wk old and 2 wk earlier and younger than nestlings from nests built in trees. Kuiken et al. (1998b) noted that nestlings left ground nests at 4 wk old and that they began dispersing from Island A starting at about 8 wk old. Kuiken et al. (1998b) also noted that contact among nestlings from different nests did not occur until nestlings left the nest, but, thereafter, physical interactions among postnestling juveniles conducive to transmission of contagious pathogens was intensive. Juveniles from nests on the ground were, thus, at risk of pathogen transmission among postnestling juveniles for 2 wk longer than were juveniles from tree nests.

As trees died and an increasing proportion of cormorants on Island A came to nest on the ground, the average duration of the period of potential pathogen transmission among juvenile cormorants would have increased from approximately 2 wk toward 4 wk, simultaneously increasing the probability that a contagious pathogen present on the colony would cause a large-scale epidemic.

This revenge-of-the-trees hypothesis, in which deforestation by cormorants results in epidemics, may apply principally along the northern tier of the breeding range of the Double-crested Cormorant, where trees for nesting are often abundant at new colony sites, and where breeding phenology is highly synchronous and tightly compressed into the short boreal summer. Our model is highly supportive of this hypothesis in this setting. The duration of the contact period among juveniles had an enormous effect on total mortality (Fig. 4).

Our model of transmission dynamics was developed for AMPV-1 and not for Pasteurella multocida. Like the virulent strains of AMPV-1 affecting the Island A colony, virulent Pasteurella multocida is highly contagious and rapidly fatal (Samuel et al. 2005) and affected juvenile cormorants almost exclusively on Island A. Although several parameters required for our model were not available for P. multocida, it is probable that the transmission dynamics of this bacterium in cormorants on Island A was similar to that of APMV-1.

It is not known how the pathogenic strains of APMV-1 and Pasteurella multocida arrived on the Island A colony in epidemic years. Both pathogens are known to infect adult birds asymptomatically in other species and APMV-1 has produced asymptomatic infections in cormorants (Kuiken et al. 1998a; Samuel et al. 2005). Shedding of these pathogens by asymptomatic carrier birds seems the most plausible source among cormorants on Island A. In this study, APMV-1 was detected in only one of 1,650 samples of fresh excrement collected from adult birds between late May and early June. These data imply that shedding of this virus is not common, but can occur, among adult cormorants during the period of egg incubation. According to our model, as few as eight to nine infectious adults in a breeding population of 13,700 birds is sufficient to initiate an epidemic.

This research was carried out under annual scientific permits for wildlife research issued by the Government of Saskatchewan. Funding was provided by the Canadian Wildlife Health Cooperative, with additional contributions from Saskatchewan Environment and Resource Management; Canadian Food Inspection Agency; Canadian Wildlife Service; Wildlife Health Fund of the Western College of Veterinary Medicine; Animal Health Trust of Canada; Alberta Sports, Recreation, Parks and Wildlife Foundation; and the University of Saskatchewan. Parks Canada Agency provided aerial photographs. Technical and logistical support were provided by Ian Shirley, Marnie Zimmer, Gary Wobeser, Alex Reid, Ferdie Edquist, Ron Laliberté, Rhys Beaulieu, Charlene Berkvens, Asha Perera, Kamal Gabadage, Seth Cherry, Dan Frandsen, Dustin Guedo, Derek Harder, Fiona Moreland, E. Jane Parmley, Chris Somers, Theresa Stene, and many volunteers.