The first goal of this study was to assess free-roaming cat populations in New York City, and measure the effects of a one year TNR effort. We found considerable variation in cat density across sites and within neighbourhoods. This is consistent with previous studies of free-roaming cats in Brooklyn, NY (Calhoon and Haspel 1989). Our data indicate that cat populations can differ substantially across an urban matrix, even between locations less than 2 km apart, such as our treatment and control sites. Such observed variation could be due to differences in colony caretakers in between neighbourhoods, though we would not expect those differences to be considerable between sites within neighbhourhoods. Given that our sampling was limited to linear transects, there could be differences in habitat suitability between sites not measured in this study. For example, Calhoon and Haspel (1989) suggest that cat populations are limited by shelter availability, which was not assessed in this study. Our data also demonstrated substantial variation between sampling periods in kittens and juveniles observable within study sites. We observed a significant increase in sterilized individuals in our treatment locations, which we attribute to our TNR efforts implemented in those areas between sampling periods. Furthermore, we did find a small increase in sterilized individuals in our control sites (Fig. 3), indicating there may be other groups implementing TNR in these neighbourhoods, yet there were not significant changes in the number of sterilized individuals. Alternatively, the increase in sterilised individuals in control sites could also represent the movement of sterilised but “unmarked” individuals from treatment areas.

Examination of our models found that none of the variables representing site, age, ET (previously sterilised), nor site*age and site*ET interactions had any influence on the encounter probability of cats. These results suggest that observability was not affected by sterilisation. However, other differences between sites could have impacted the likelihood of encountering cats. Prior to the study, we predicted that site, ET, and site*ET would have a significant relationship with encounter probability, given that TNR can lead to decreases in aggressive behaviours, increases in sociability, and limitations on movements of free-roaming cats (Finkler and Terkel 2010; Gunther et al. 2011). Furthermore, as our TNR study was conducted in half of the study sites, and only occurred on adult individuals, we expected to observe a relationship between age and site. Given that none of these covariates significantly affected our likelihood of observing individuals, this suggests high homogeneity of study sites. Our limited sample size (two control and two treatment sites) may also have influenced our lack of statistical significance. Additionally, previous studies have highlighted how socioeconomic status in a neighbourhood can affect free-roaming cat populations as well as the welfare of the cats in those areas (Finkler et al. 2011a, b). Analysis of cat populations in Brooklyn, NY, found that food resources do not limit cat population growth, but instead populations are more affected by shelter availability (Calhoon and Haspel 1989), which may correlate with neighbourhood socioeconomic status or proportion of abandoned buildings. Given the landscape homogeneity across our study sites, it appears populations were similar across neighbourhoods.

We observed substantial turnover of individuals in all study sites over our study period. That is, only a small proportion of individuals identified in the first sampling period (2011) were subsequently observed in the second sampling period (2012). There are three possible hypotheses to explain this phenomenon. First, cat populations may be experiencing high immigration rates. Previous studies have compared owned and unowned cat home ranges using radio-telemetry (Schmidt et al. 2007a; Horn et al. 2011). While those studies were conducted in a lower density urban area, results showed that unowned cats exhibited considerably larger home ranges than owned cats (approximately 40-80 times larger), though this effect might vary by sex, as males often have larger home ranges than females (Say and Pontier 2004, though see Haspel and Calhoon 1989a). Secondly, larger home ranges of unowned cats potentially reduce their detection probability along transects. If high immigration, in conjunction with large home range of unowned cats, were occurring in a dense urban area such as New York City, likelihood of resighting between sampling periods would be substantially reduced. This may suggest that our estimates are biased toward individuals with small home ranges. Alternatively, our data could indicate that cats may remain in an area for a certain period of time (during the length of our study) before moving to a new location. Lastly, high mortality, particularly among juvenile cats, in conjunction with high dispersal rates could also account for the high turnover of individuals that we observed. Such effects have been observed in urban areas in France (Devillard et al. 2003). The lack of knowledge regarding free-roaming cat population dynamics in highly urban areas prevents us from understanding how populations are growing and changing within the urban matrix. More research is needed to understand how cats move throughout the urban landscape.

Using a sight-resight analysis, we estimated cat densities per kilometer in each of our study locations. Previous studies which have examined cat density in urban areas have applied transect data to estimates calculated per hectare or kilometer2 making comparisons between our studies difficult (Calhoon and Haspel 1989; Schmidt et al. 2007b; Finkler et al. 2011c). These studies of free-roaming cat density show highly variable results, ranging from 0.009 cats/km2 in a small urban area (Schmidt et al. 2007b) to more than 900 cats/km2 in a densely urban landscape in Tel Aviv (Finkler et al. 2011c). Internationally, cat density tends to increase with urbanization and the concurrent provision of large sources of food such as garbage or cat food (Liberg et al. 2000). In general, we believe that our population estimates represent an under-estimate of the true cat population. Given the nature of our study environment, where roads were almost exclusively bound by buildings on both sides, our transects encompassed the linear space along sidewalks. Furthermore, based on logistical constraints of working in a highly urban area, we were unable to assess cats that may have been located in more cryptic locations that are inaccessible to humans. In addition, our assessments took place during daylight hours, when cats are generally less active (Haspel and Calhoon 1993; Horn et al. 2011). Therefore, reporting cat densities along a two dimensional area provides a more accurate of free-roaming cats. Assuming an equal distribution of cats along streets and in block interiors is unrealistic, as free-roaming cat movement may be influenced by human activities, which are more abundant along streets and sidewalks. Accurate understanding of cat populations in urban areas requires a targeted approach in different landscape types, as cats may not perceive all areas as equally suitable habitat.

The second goal of our study was to examine the effects of one year of Trap-Neuter-Return effort on free-roaming cat populations. Previous examinations of this management strategy have been variable, and are likely heavily influenced by feasibility of effort, population size, and duration of project (Longcore et al. 2009). Given the increase in cat populations in highly urban areas around North America, it is crucial that we gain an understanding of the utility of different management strategies in these unique urban landscapes. Most published studies are based on a long-term effort, occasionally used in conjunction with a secondary approach (e.g. adoption of socialized individuals) (Levy et al. 2003; Foley et al. 2005; Natoli et al. 2006). Although our study represents only a single year of effort, we did observe a significant increase (50%) in sterilised individuals in our treatment sites. While 50% sterilisation is generally considered insufficient to observe any changes in population size (Andersen et al. 2004; Schmidt et al. 2009; McCarthy et al. 2013), the longer term effects of such an effort over several years remain unclear. Our results suggest that significant changes to population sterilisations rates are feasible in highly urban environments. However, more research is needed on movement patterns, mortality and immigration rates in cities to accurately tailor TNR efforts to urban landscapes. Additionally, empirical data are needed on the effect on TNR on population size, not simply sterilisation rates, in highly urban areas.

In gauging the results of a one-year TNR effort, we found several limitations to our trapping strategy constrained our ability to trap more cats. Our inability to access inner-block locations for trapping may have prevented us from trapping less socialized, more feral individuals. Furthermore, residential areas in New York City are still highly trafficked areas, which imposed logistical issues on our methods, such as our ability to observe cats along transects, set up a trap and lure immediately upon locating a cat, and access other locations behind homes or businesses. This was particularly the case for Harlem-Treatment which likely resulted in the lower number of cats seen and trapped compared to Bed-Stuy-Treatment. We also had difficulty in collaborating with several cat colony caretakers, an affiliation which can be extremely useful in the trapping component of TNR efforts (Levy et al. 2003; Lohr et al. 2013). Some caretakers also were not identified until we had been working in the community for months, or took weeks to provide cat location information. Colony caretakers can provide unique access to free-roaming, particularly feral, cat colonies, and in combination with public education and outreach, could maintain positive public perceptions of TNR efforts thus improving the efficacy of TNR programs (Haspel and Calhoon 1989b; Centonze and Levy 2002; Ash and Adams 2003). Our trapping and counting efforts were additionally restricted by time and safety concerns. These difficulties working within a highly urban environment would be typical in other cities of similar sizes and would need to be addressed prior to implementation of a successful TNR effort.

In conclusion, our results suggest that free-roaming cat populations are robust in a highly urban landscape such as the neighbourhoods of Harlem and Bedford-Stuyvesant. Implementation of a management strategy such as TNR to limit population growth in these areas would require a substantial effort and must include a community component to build trust and communication, whereby cat colony caretakers would be of assistance. Furthermore, while we did find a statistically significant increase in the proportion of sterilised individuals in our treatment locations after only one year of effort, this increase was not enough to expect any decline in population numbers. A longer duration in TNR effort will be necessary to observe any quantifiable difference in population structure. One of the greatest concerns in cat population dynamics is the immigration rate, which may have been reflected in the overall turnover of individuals across years. Further research is necessary to better understand cat population dynamics in urban areas. Detailed understanding of movement patterns, survival estimates and dispersal rates would help explain the high turnover of individuals that we observed in all of our sites. Currently, we have little understanding of what factors influence urban cat population dynamics across North America. More data are necessary on how to effectively and humanely manage this controversial species.