Historical surveys reveal a long‐term decline in muskrat populations

Abstract The muskrat (Ondatra zibethicus) is an iconic species in Canada, valued for both its fur and its integral role in wetland ecosystems, and widely regarded for its perseverance. However, the resilience of this semiaquatic mammal seems to be in question now as increasing evidence points to widespread population declines. Recent analyses of harvest data across North America suggest a reduction in their numbers, but this has not been widely corroborated by population surveys. In this study we replicated historic muskrat house count surveys at two large Great Lakes coastal wetlands and present confirmation that declines in muskrat harvest correspond to actual declines in muskrat abundance. At the Point Pelee National Park marsh and the Matchedash Bay‐Gray Marsh wetland we found that mean muskrat house counts declined by 93% and 91% respectively between historic surveys 40–50 yrs ago and contemporary surveys over the past 7 yrs. The factors responsible for these dramatic declines remain unclear but there may be a relationship with changes in the habitat quality of these wetlands that have occurred over the same time frame. Not only is the loss of muskrats an issue for the resulting loss of the wetland ecosystem services they provide, but it may be an indication of broader marsh ecosystem degradation. As such, a scarcity of muskrats should be considered a red flag for the state of biodiversity in our wetlands. Continued surveys and ongoing research are needed to shed more light on the current status of muskrat populations and their marsh habitats across their native range.

harvest data across North America suggest a reduction in their numbers, but this has not been widely corroborated by population surveys. In this study we replicated historic muskrat house count surveys at two large Great Lakes coastal wetlands and present confirmation that declines in muskrat harvest correspond to actual declines in muskrat abundance. At the Point Pelee National Park marsh and the Matchedash Bay-Gray Marsh wetland we found that mean muskrat house counts declined by 93% and 91% respectively between historic surveys 40-50 yrs ago and contemporary surveys over the past 7 yrs. The factors responsible for these dramatic declines remain unclear but there may be a relationship with changes in the habitat quality of these wetlands that have occurred over the same time frame. Not only is the loss of muskrats an issue for the resulting loss of the wetland ecosystem services they provide, but it may be an indication of broader marsh ecosystem degradation. As such, a scarcity of muskrats should be considered a red flag for the state of biodiversity in our wetlands. Continued surveys and ongoing research are needed to shed more light on the current status of muskrat populations and their marsh habitats across their native range.

K E Y W O R D S
fur harvest, muskrat, Ondatra, population decline, Typha, wetlands lower today, muskrats remain a major source of income for fur trappers and are still among the most prevalent species trapped for fur (Fur Institute of Canada, 2019).
While the muskrat played a major role in the early fur trade and colonization of North America by Europeans (White et al., 2015), the species has been of cultural significance as a traditional clothing and food item and as a spiritual symbol among Indigenous people long before European explorers arrived on the continent. For example, in an Anishinaabe story of creation, the muskrat (Wa-zhushk) comes to the rescue to help rebuild the Earth after a great flood and decimation of life, and is said to embody humility, courage, and determination (MacGregor, 2013).
As a wetland obligate and a significant consumer of marsh vegetation, the muskrat plays several important roles in wetland ecosystems. Their foraging, travel, and house-building activities create numerous small openings in marshes, thereby increasing the interspersion of open water and emergent vegetation, which often results in increased structural diversity and plant species richness in wetlands (Connors et al., 2000;Keddy, 2010;Nyman et al., 1993).
Different biotic communities are known to respond positively to such enhancements in habitat diversity (Wilcox & Meeker, 1992); in particular, a greater density and diversity of marsh birds and waterfowl have been found in wetlands with an equal ratio of open water to emergent vegetation (Kaminski & Prince, 1981;McDonnell, 1983;Weller & Fredrickson, 1973). Muskrat houses (both new and old) can also create important loafing and nesting sites for marsh birds. For example, the black tern (Chlidonias niger) will use muskrat houses and feeding platforms as nesting substrate (Hickey & Malecki, 1997) and it is quite common to find Canada geese (Branta canadensis) and sometimes trumpeter swans (Cygnus buccinator) loafing or nesting on old muskrat structures. Less recognized are the benefits that muskrats provide to snakes and turtles, many of which are species at risk. In an Illinois, USA, study, a large number of spotted turtle (Clemmys guttata) captures occurred in deep open water pools associated with muskrat lodges where muskrat grazing decreased the vegetative cover, and it was believed that these pools served as refugia for the turtles during periods of high temperature and/ or drought (Litzgus & Brooks, 2000). Numerous vertebrate species, including more than ten herptile species, have been observed using muskrat houses, burrows, cleared pathways, and other features for thermoregulation, nesting, cover, and ease of travel (S. Gillingwater, unpublished data; Kiviat, 1978). Furthermore, muskrat activity can influence mussel abundance (Diggins & Stewart, 2000), invertebrate communities (De Szalay & Cassidy, 2001;Nummi et al., 2006), microbial activity (Wainscott et al., 1990), and nutrient cycling (Connors et al., 2000). Muskrats are also an important food item for many predators, such as red foxes (Vulpes vulpes), coyotes (Canis latrans), raccoons (Procyon lotor), raptors, and especially mink (Neovison vison) (McDonnell, 1983).
Early research on muskrats illustrated the density-dependent nature of muskrat populations and a variety of abundance cycles (Clark & Kroeker, 1993;Erb et al., 2000;Errington, 1951). In their analysis of almost one hundred individual time series of muskrat harvest data from the Hudson's Bay Company in Canada, Erb et al. (2000) found that the mean period length of muskrat population cycles differed between ecozones, ranging from 3.7 to 8.6 yrs, with the shorter periods tending to occur at higher latitudes and in eastern regions. In some cases, however, the time series did not exhibit any periodicity. Evidence exists for both population-intrinsic factors (e.g., social factors) and extrinsic factors (e.g., disease, environmental variability, and trophic interactions) explaining the observed patterns of these cycles (Errington, 1963 andErrington et al., 1963;Bulmer, 1974;Weller & Fredrickson, 1973). Despite some commonalities, these patterns lack consistency across geographic areas, and in many cases, muskrat population dynamics and their mechanisms of regulation remain unclear.
The muskrat is a prolific species, typically having 2-3 litters per year and an average litter size of 6.5 kits (Boutin & Birkenholz, 1987).
As well, most muskrats are able to breed the same year they are born and have high dispersal capabilities. These demographic characteristics make muskrats relatively resilient to harvest and other population pressures (e.g., disease, predation) as a small number of individuals can quickly multiply and enable population recovery.
They are also reasonably flexible in their habitat requirements, and as noted by Errington (1951), muskrats often demonstrate a remarkable ability to survive up to the very edges of what may be considered habitable range.
Today, however, this marsh denizen may not be thriving like it once was. There is a growing body of literature suggesting declines in muskrat populations from myriad locations across North America over the past 10-20 yrs. Most recently, Gregory et al. (2019) reported that muskrat harvest on Prince Edward Island declined by more than half when comparing the average harvest from 1977-1988 to 1988-2016. More broadly, Ahlers and Heske (2017) analyzed harvest data from 1970 to 2012 across the United States and, after controlling for pelt prices, found strong evidence that muskrat populations declined during this time period. Prior to that study, analyses by Roberts and Crimmins (2010) revealed a 75% decline in muskrat harvest across the northeastern United States and eastern Canada from 1986Canada from to 2006. This decline was thought to be indicative of regional declines in muskrat abundance as the study's authors found that a previously strong correlation between harvest and pelt prices had weakened in the latter years of their analysis. In other words, they considered it was likely that recent changes in muskrat harvests were reflecting underlying population change (and not simply changes in harvest effort) because they did not find a strong relationship between harvest levels and pelt prices, which had previously defined the harvest dynamics of muskrats and other furbearer populations (Bailey, 1981;Scognamillo & Chamberlain, 2006). As well, in the same study the authors reported a lack of periodic fluctuations in the modern muskrat harvest data (as compared to the historic data, which exhibited mild periodicity), providing further support for widespread population decline (Roberts & Crimmins, 2010).
In Ontario, similar to trends across Canada, provincial fur harvest records show a clear decline in muskrat harvest over the past 100 yrs, most notably over the past 30 yrs (Figure 1). In fact, the average annual number of muskrats harvested in the past 30 yrs has declined by more than 90% from the mean in the previous 20-yr period (late 1960s to late 1980s). While some of this decline can be explained by changes in fur harvest reporting procedures that occurred in Ontario in the late 1980s, as well as economic factors and cultural shifts in trapping, the magnitude and time span of the decline seem too great to be explained by these factors alone. For example, the spatial pattern of change in muskrat harvests from 1972 to 2004 suggests an actual decline in muskrat populations may have occurred because the pattern did not conform to the expectation arising from spatially homogeneous decline in fur price and therefore trapper effort (Gorman, 2007). Furthermore, the time span of the recent period of low muskrat harvest numbers (30 yrs) far exceeds that of the maximum population cycle length reported for muskrats (Erb et al., 2000). As well, the low annual numbers harvested in recent years are well below the lower limit of published historic muskrat population fluctuations (Statistics Canada, 2011).
Since many furbearers (including muskrats) are difficult and costly to census (Erb & Perry, 2003), harvest data are often the only type of information related to species' population trends that are available for a long time series or across a large geographic area (White et al., 2015). However, there are problems with relying on harvest data to infer wildlife population trends, and these have been noted by many of the same authors that have analyzed harvest trends of furbearers (e.g., Gregory et al., 2019;Roberts & Crimmins, 2010). For example, harvest data can encompass trapping seasons of different lengths, can be missing data for some years, or can be deficient due to lack of full reporting. More importantly, harvest data can be biased by trapping effort, which can be affected by the number of trappers or by changing economic factors (e.g., pelt price) that might influence trapper behavior (Landholt & Genoways, 2000).
Although long-term census data on muskrat populations are generally rare, a few researchers have published evidence of declines in local muskrat populations from direct count surveys. For example, Benoit and Askins (1999) reported that counts of muskrat houses from marshes of the Quinnipiac and Connecticut Rivers decreased dramatically (78% and 100%, respectively) between 1965 and 1990.
As well, Ward and Gorelick (2018)  Island for muskrats and reported a mean density of just 0.07 houses/ ha. All of these results are much lower than the typical densities of 2.1-3.6 muskrat houses/ha reported previously in the literature for cattail-dominated marshes in Canada (Messier & Virgil, 1992;Proulx & Gilbert, 1984).

Year
Indigenous people across Canada (Brietzke, 2015;Straka et al., 2018; S. Mallany, personal communication, December 2016) and have also been received from a variety of sources by us. The underlying theme among these reports is that trappers and other long-term land users are simply not finding muskrats in the numbers they used to, despite efforts to do so.
Collectively, these recent studies and reports point to potential widespread declines in muskrat populations. However, robust empirical data on long-term trends in muskrat populations based on field observations are generally scarce in the literature. Thus, more information is needed to confirm that declines in muskrat harvest correspond to real declines in muskrat abundance (Ahlers & Heske, 2017).
Fortunately, we learned of annual muskrat field surveys having been conducted at multiple locations in Ontario, Canada, between 1950 and 1990 and sought to exploit this untapped source of historic muskrat population data.
Our specific objectives in this study were to locate and revisit sites where historic muskrat survey data exist for Ontario, to replicate the historic survey methods as closely as possible, and to then compare contemporary survey results with the historic data to determine whether there have been empirical changes in the muskrat populations in these areas. We hypothesized that declines observed in muskrat harvest are due at least in part to real declines in muskrat populations. Therefore, we expected to observe evidence of fewer muskrats in contemporary versus historic muskrat surveys.

| Study area
We found two sites in Ontario with collections of multi-year historic muskrat survey data from coastal wetlands: Point Pelee National Park and Matchedash Bay-Gray Marsh ( Figure 2). In both cases, the historic data were more than 30 yrs old. The methods and data from these surveys are published in internal government reports, and we were able to obtain copies from the respective offices responsible for the resource management of these sites. We replicated the historic muskrat surveys in the marshes of these two locations between  Both wetlands consist of approximately 700 ha of robust emergent marsh vegetation (MNRF, 2015), which is considered to be suitable muskrat habitat (Bellrose & Brown, 1941;Clark, 1994;Proulx & Gilbert, 1983

| Survey methods
We aimed to replicate the historic survey methods as closely as possible to obtain comparable contemporary survey results. We undertook this task by following the detailed descriptions and survey maps provided in the old reports to relocate the same survey areas for each site. At Point Pelee National Park, muskrat surveys were initiated in the 1950s by Parks Canada staff and standardized in 1963 when the marsh was divided into 14 survey zones, with a 15th zone added in 1971 (Reive, 1978). The zones remained consistent in subsequent years; however, a few boundary adjustments were made in 1979 to accommodate changes in the vegetation structure of the marsh (Reive, 1979).
The surveys were conducted annually until 1980, after which time no further muskrat surveys were undertaken (Bremner & Reive, 1980). At were concerned that muskrat houses might be buried and thus not detected. We verified this suspicion by returning to the site early the following spring (2 months later) and paddled the perimeter of the entire marsh by canoe, counting several muskrat houses that we missed due to snow depth in the winter. We know that these were not houses that were newly built in the interval between our winter and spring visits because we conducted the survey in May only a few weeks after ice-out, which is not a time when muskrats conduct house building in this region. Muskrats typically build houses in the fall before freeze-up, and these remain in place over the winter until they start to deteriorate (if no longer in use) the following spring-summer (Dozier, 1948 As was done for the historic surveys, we used our annual counts of muskrat houses as indices of abundance for each site, a common method first described by Dozier (1948) and since undertaken by many researchers as a means to infer abundance and track annual change in muskrat populations (e.g., Greenhorn et al., 2017;Kroll & Meeks, 1985;Proulx & Gilbert, 1984;Toner et al., 2010).

| Point Pelee Marsh
We found very low numbers of muskrat houses at Pelee over our four contemporary survey years between 2014 and 2019 as compared to the historic survey period from 1968 to 1980 (Figure 4).
The mean (SE) number of muskrat houses found at Pelee between 1968 and 1980, excluding incomplete survey zones and years, was 719 (202.9). The 95% confidence interval for house counts during these years was 267-1171. In contrast, we derived a mean (SE) house count of 47 (7.4) from our 2014-2019 surveys, excluding the same zones as we did with the historic data set. The 95% confidence interval for the house counts during these latter years was 24-71.
Our contemporary surveys of Pelee demonstrated a 93% decline in the mean number of muskrat houses from historic levels 40-50 yrs prior ( Figure 5). Retaining the count results for the excluded survey zones still resulted in a dramatic difference in the mean number of houses found between each survey period (791 historically vs. 80 recently) and a 90% decline in house numbers. Excluding the imagerybased estimates that we added for the areas we could not access, our mean (SE) house count from the ground count data alone was 50 (6.9) for all survey zones and 39 (6.5) for the standardized zones, representing declines of 94% and 95%, respectively, from historic times.

| Matchedash Bay-Gray Marsh
Similar to Pelee, we found relatively low numbers of muskrat houses at Matchedash in each of our five recent survey years from 2014 to 2018 when compared to the historic survey period from 1979 to

(Figure 4). The mean (SE) number of muskrat houses found at
Matchedash between 1979 and 1986 was 650 (81.9). The 95% confidence interval for house counts averaged across these years was 450-850. In contrast, we observed a mean (SE) house count of 57 (9.1) during 2014-2018, using our estimated maximum house counts derived from a combination of ground and imagery-based surveys.
The 95% confidence interval for the house counts averaged across these latter years was 32-82.
Our contemporary surveys of Matchedash demonstrated a 91% decline in the mean number of muskrat houses from historic levels 30-40 yrs prior ( Figure 5). Excluding the imagery-based estimates that we added for the areas we could not access, our strictly groundbased survey results showed an even greater decline of almost 95%, as our mean (SE) house count from those data alone was 35 (5.8).

| D ISCUSS I ON
Our surveys demonstrate substantial declines in muskrat popu- It has been suggested that most survey techniques for muskrats are only feasible across small geographic extents and may not be useful for examining large-scale population trends (Roberts & Crimmins, 2010). While this may be true, we have two independent data sets from widely separated wetlands illustrating declines of similar magnitude over a similar time frame, which are also coincident with declines in our provincial harvest data over the same approximate time period. Although our findings are limited to two individual wetlands and therefore extrapolating the trends to a broader region would be premature, when considered alongside the provincial harvest records our data are consistent with a broader trend of muskrat decline across the Great Lakes basin in Ontario. It would be worthwhile to discover and collect additional survey data from other areas to shed more light on the scale of the observed trends.
The greater than 90% decline in muskrat house abundance at each of our study sites is startling, yet we are confident that the survey data indicate a true decline of this magnitude. In many wildlife surveys, there is potential for overestimating or underestimating abundance as a result of observer error. In the case of muskrat house surveys, some houses may be missed, and some structures may be falsely labeled as active houses when they are in fact only feeding structures or inactive houses, or vice versa. However, we assume that in our study, on average, detection probabilities were similar from year to year (except for the survey years that were deliberately eliminated from the analysis because the surveyors reported poor conditions that hindered their ability to detect houses) and that in most cases muskrat structures were correctly identified. We attempted to replicate the historic survey methods as closely as possible so that the results we obtained could be largely attributed to actual population change and not simply methodology differences.
Furthermore, considering that our final house counts for the recent survey years at each site were adjusted based on imagery data that were used to supplement our ground counts, and it is more difficult to distinguish new houses from old ones on aerial photographs, it is more likely that we have overestimated contemporary muskrat abundance by including houses in our tally that are no longer in use.
Therefore, our estimates of the magnitude of muskrat population change at each site are likely conservative.
Muskrat numbers have been shown to fluctuate periodically; thus, it is possible that low house counts may be reflecting a low phase in a normal population cycle. Erb et al. (2000) examined ninety-one historic time series of muskrat harvest data from boreal, taiga, and southern Arctic regions of Canada and found overall that mean period length ranged from 3.7 to 8.6 yrs. If the same pattern holds true in southern Ontario and the start of our surveys corresponded to a population low phase, we would expect to have seen a greater increase in our recent house count numbers over the 5-6 yrs we conducted the surveys. Based on a visual examination of the data, we observed no such increasing trend, and similarly, there is no increasing trend evident in muskrat harvest data collected over the same time period. Instead, it appears that muskrat populations at Point Pelee and Matchedash Bay-Gray Marsh are currently persisting at very low levels and are much reduced from the levels seen 30-50 yrs ago. Furthermore, harvest data suggest that these declines are widespread.

| Reasons for decline
It is unclear whether the apparent declines in muskrat abundance that we have demonstrated are a result of a broad-scale underlying cause or site-specific factors.
In recent studies of muskrat occurrences in Lake Ontario-St.
Lawrence River coastal wetlands, it was found that water-level regulation of that system had a negative influence on muskrat abundance (Greenhorn et al., 2017;Toner et al., 2010). While the two Great Lakes that our study sites occur on are not subject to such tightly controlled water levels as Lake Ontario, both Lake Erie and Lake Huron have experienced record high and extended low water levels over the duration that our investigation spans (NOAA, 2020), primarily brought on by weather extremes and climate variability over the past few decades (Gronewold & Rood, 2019). The presence of water, either through excessively high water levels or periods of drought, has been considered the greatest selection pressure affecting muskrats (Ahlers et al., 2015;Bellrose & Brown, 1941;Errington, 1963;Proulx & Gilbert, 1983;Virgl & Messier, 1996;Ward & Gorelick, 2018). Essentially, the water level in a marsh is suitable for muskrats when it is deep enough to maintain travel routes and allow access to houses and feeding areas underwater (even in cold winters when ice cover can be very thick), yet shallow enough to permit the growth of emergent aquatic vegetation. It is also important for muskrats that any changes in water level occur gradually, so that houses are neither flooded out nor left high and dry. We have not conducted a detailed analysis of water-level patterns at Pelee and Matchedash in relation to muskrat occupancy; therefore, this is an avenue that warrants closer examination.
Closely related to water levels is the hypothesis that habitat change has led to dramatic reductions in muskrat numbers. For example, Ward and Gorelick (2018) suggested that the loss of critical wetland habitat due to drying is the primary driver responsible for the decline of muskrats in the Peace-Athabasca Delta over the past half-century and potentially across their entire native range. While the amount of wetland habitat on the landscape available to muskrats is certainly an important factor influencing their population levels, the sizes of the wetlands in our study area have not changed appreciably over the time span of our investigation. Province-wide, Ontario has experienced significant loss of wetlands over the past prior to the recent period of muskrat decline that we have demonstrated. Therefore, it seems unlikely that a loss of habitat area is a major factor driving the recent declines we have seen in muskrat populations in Ontario over the past 30-50 yrs.
A more plausible explanation for muskrat decline in Ontario is a change in the overall quality of muskrat habitat, particularly in the structure and composition of Great Lakes coastal wetlands that were previously occupied in large numbers by muskrats as shown in this study. Proulx and Gilbert (1983)  The well-documented changes in both the structure and composition of coastal wetlands over the past 50 yrs may have contributed to the dramatic decline in muskrats at our study sites, yet this remains unclear. In the absence of an experimental study or control sites where such habitat changes have not occurred and muskrat numbers have not declined, it would be premature to suggest that this type of habitat change is the underlying mechanism that has driven muskrat declines across a broader regional scale. Therefore, this hypothesis warrants further investigation in future studies looking at changes in muskrat abundance.
Several other explanations for the decline in muskrat populations are possible but seem less likely for a variety of reasons. Overharvest by trappers is often postulated as the reason behind furbearer population declines; however, the demographic characteristics of muskrats make them resilient to overharvest (Boutin & Birkenholz, 1987 Predation and disease are important natural aspects of muskrat ecology that can affect their abundance over the short term, but which typically balance out over the long term. Errington (1963) suggested that significant mortality of muskrats due to predation usually only occurs when muskrats are already vulnerable as a result of disease or habitat changes such as loss of cover, drought/flooding, or freeze-outs. As well, muskrat populations typically respond to predation mortality by a variety of compensatory processes, such as increased reproduction and survival among the remaining individuals (Errington, 1951). Mink (Neovison vison) are considered the most important predator of muskrats (McDonnell, 1983), yet there is evidence from harvest data of long-term mink population decline in Ontario (Gorman, 2007) and no evidence of local mink population booms at our study sites (personal observations). Disease outbreaks among muskrats are typically localized and usually occur when muskrat populations are under some form of stress, such as overcrowding, food shortage, or drought (Errington, 1963). Also, as with predation, the effects of disease on muskrat population size are often short-term because of compensatory responses in other sources of mortality or in reproduction. We are unaware of any documented outbreaks of disease in muskrats at either of our study sites or elsewhere in Ontario over the time period that our study covers (Ganoe et al., 2020).

| Implications of muskrat decline
A marked decline in muskrat populations is concerning not only for the value of muskrats in their own right, but also for the broader economic, ecological, and cultural benefits that muskrats and wetlands provide. Muskrats are highly valued by trappers and are an important part of our natural heritage. In many Indigenous cultures, the muskrat is revered for the life-giving role it is said to have played in the creation of the Earth as we know it. As such, a loss of muskrats comes with significant costs, many of which we might already be experiencing (Papworth et al., 2008).
Muskrats manipulate marshes (Higgins & Mitsch, 2001), promoting greater diversity and providing ecological services to humans and other wetland species. Many of the species that benefit from muskrats are of conservation concern themselves. Thus, the loss of muskrats from our marshes may compound existing problems within these wetlands and across broader landscapes containing wetlands.
Understanding the current state of the muskrat might help us to better understand the current state of our wetland ecosystems.
There is a clear need for more research on the relationship between muskrat populations and the various factors that influence their abundance, especially if we are to understand what has led to their recent declines and if we wish to prevent further loss of this iconic species.

ACK N OWLED G M ENTS
We gratefully acknowledge the assistance and marsh access pro- Financial support for this study was provided by the Ontario Ministry of Natural Resources and Forestry; in-kind support was provided by Parks Canada (Point Pelee National Park).

CO N FLI C T O F I NTE R E S T
The authors declare there are no competing interests or conflicts of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data are archived in the Dryad Data Repository at https://doi. org/10.5061/dryad.b2rbn zsf2.