Commensal rats in the genus Rattus are among the most ubiquitous and important pest species. Two species (Rattus norvegicus and Rattus rattus) have near-global distributions, now occurring in every continent except Antarctica. Rats damage infrastructure, consume agricultural yields, and contaminate food supplies, causing an estimated US$27 billion in damage each year in the United States alone (1). Rats also harbor and transmit more than 50 zoonotic pathogens and parasites to people, affecting public health around the world (2, 3). Associated diseases include leptospirosis, hantavirus pulmonary syndrome, murine typhus, and bubonic plague. Rats thrive in human-dominated landscapes by exploiting resources concentrated where human population density is high (4) and are often classified as urban exploiting species. As a result, rat population densities are expected to be higher in cities than in rural areas, with the potential to negatively affect more people (5). The very presence of rats also takes a measurable toll on the mental health of people living in contact with them (6).

Municipalities and property owners have been trying to reduce rat numbers for centuries. In recent decades, efforts at suppressing or eradicating rats have primarily been through the use of lethal rodenticide chemicals or traps rather than nonlethal options that would make the environment less suitable (e.g., securing food waste and removing harborage) (7). Globally, the control efforts associated with this “war on rats” cost an estimated US$500 million every year (8). At the municipal level, the strategies and intensity of these control efforts vary widely among cities. Rodent control is also inconsistent within cities over time, as priorities shift, budgets and staff fluctuate, and new control products or approaches are introduced.

One of the most intractable challenges associated with rodent control is tracking rat numbers in a consistent way over time, which is a necessary step to assess whether control efforts are effective (9). The common presumption cited in many media reports is that rat numbers are increasing around the world. Yet, rarely are formal scientific population surveys done for urban rats, as city budgets and agencies struggle to simply keep up with responding to rat complaints and infestations. Because of the absence of consistent long-term data, we are no closer to understanding the effectiveness of rat population control efforts (7). Tracking rat numbers over time is also needed for any basic understanding of the demography and population ecology of these urban rat populations, for which there has been little research on over the past 70 years (10).

While untested, the assertion that rat populations are growing is in line with potential biological responses to changing urban environments. As small mammals, rats must maintain internal body homeostasis and are limited by cold temperatures during winter (11, 12). Warming temperatures resulting from climate change or urban heat islands may extend the seasonal window for aboveground foraging and active breeding period for rats, supporting population growth (13). Increasing human population size and urbanization are also likely to provide more food waste as a resource and structural habitats that support rat populations. As rapid urbanization continues, it is critically important to track rat numbers and assess both the changes in their population ecology and our progress in controlling them. The human population living in cities is projected to increase 25% by 2050 (from 56% in 2020; World Bank), and total urban land cover across the world is also projected to increase 185% between the years 2000 and 2030 (14), providing even more suitable habitat and food waste for urban rats. This suggests that cities will become increasingly conducive environments for rats, but data are needed to confirm these effects and quantify their impact on rat numbers.

In this study, we use between 7 and 17 years (average of 12.2 years) of public rat sighting and inspection data from 16 cities around the world to quantify changes in rat numbers for each city and to evaluate trends across cities. Rat sightings coming from the public correlate well with relative abundance measures from trapping (15–17) and are an important proxy for rat numbers. We accessed rat reporting data for large US cities where such data are collected and available, or that we could request from cities that do not publicly report such data. To expand the geographic scope, we also requested data from other cities and rat researchers outside of the US, where public data collection is limited or not made available to the public. We limited our analyses to cities where the data collection methods and systems remained largely consistent during the study period to avoid possible trends stemming from changing data intake methods. We also assessed whether several relevant variables were linked with trends in rat numbers: human population density, ambient temperature changes over time, annual minimum temperatures, levels of urbanization, and socioeconomic variation among cities (see full Materials and Methods below). For reasons related to the biology of Rattus species, we hypothesized that most cities are experiencing an increasing trend in rat numbers and that rats are increasing fastest in larger cities with (i) more dense human populations, (ii) warmer winter temperatures, (iii) steeper temperature increases over time, (iv) less vegetation and greater urbanization, and (v) lower gross domestic product (GDP), as a proxy of socioeconomic resources available to implement rat control efforts.

For 11 of the 16 cities (69%) in our dataset, rat numbers significantly increased during the study period (Fig. 1Opens in image viewer). Washington, D.C., San Francisco, Toronto, New York City, and Amsterdam exhibited the five strongest positive trends, followed by Oakland, Buffalo, Chicago, Boston, Kansas City, and Cincinnati. The magnitude of trends among cities varied widely; for example, the trend in rat numbers in Washington, D.C. was three times greater than in Boston and 1.5 times greater than New York City. In contrast, Tokyo, Louisville, and New Orleans each had declining trends in rat numbers, with New Orleans experiencing the greatest decrease over the study period (Fig. 1Opens in image viewer). Dallas and Saint Louis did not show significant trends over time.

In a relative weights analysis, 40.7% of the variation in trend strength was linked to the mean temperature increase a city had experienced relative to long-term temperature averages (Fig. 2Opens in image viewer). Cities that had a greater rise in temperature over time had larger increases in rat sightings (Fig. 3Opens in image viewer). The percentage of a city’s land area that was vegetated (a proxy of urbanization) had a relative weight value of 34.3%, where cities with less vegetation experienced greater increases in rats. Human population density had a weight of 19.4%, followed by GDP (3.4%), and mean minimum temperature experienced by a city (2.3%). Overall, 66% of all variation in the trend data was explained by these five explanatory variables.

Separate regression analysis also found that the long-term temperature trend was strongly associated with rat numbers, with cities that have experienced the greatest warming (over long-term baseline temperatures) also experiencing faster increases in rat numbers (r2 = 0.478, P = 0.003; Fig. 3Opens in image viewer). Cities that have a higher percentage of green land cover experienced the inverse pattern, with slower growing or even decreasing rat numbers (r2 = 0.346, P = 0.017; Fig. 4Opens in image viewer). The human population density was positively correlated to rat trends (r2 = 0.29; P = 0.031). In addition, the change in vegetation cover between the years 1992 and 2020, which is a proxy for urbanization rates, was also associated with rat trends (r2 = 0.268; P = 0.04), meaning that cities that lost more vegetated areas (and became more urbanized) during the period saw greater increases in rat numbers. The trends in rat numbers were not linked to GDP (r2 = 0.057; P = 0.369) or mean minimum temperature in each city (r2 = 0. 027; P = 0.543).

We acquired rat sightings data primarily via public municipal databases, conducting searches for pest inspection reports and public complaint data. We started this study focusing on cities in the United States, which generally have public complaint reporting systems (e.g., “311” platforms). We looked for publicly available databases in the 200 largest cities (by population size) in the US. Cities typically file each complaint or inspection report into general categories. We sorted each year’s complaints by category, selecting complaints filed under terms such as “Rat Sighting” and “Rat Treatment.” The relevant search terms used for each city’s datasets are listed in table S1. The sorting and filtering process also involved eliminating cities that had noted shifts in reporting methods or data inconsistencies, such as long gaps in data availability. We also eliminated cities that used a generic “pest” category and did not distinguish between rodents and insects, for example. The last data collection phase involved reaching out to cities where the data were unclear or unavailable to request complete information. We excluded any cities that did not have the equivalent of at least 7 years of records. It is less common for cities outside of the US to make their public complaint data available through a public portal. To expand the insights of our study beyond the US, we reached out to researchers and municipal officials in numerous cities around the world. We were able to acquire data that met our criteria for three cities outside of the US.

Public complaint data are not the same as data coming from a systematic survey for rats. However, traditional surveying techniques like trapping or surveys of “active rat signs” (41) are rarely conducted by cities, requiring the use of other data sources to assess rat populations to overcome these data limitations (67). There are risks of bias in public reporting data if all residents are not equally likely to report a rat sighting based on differences in familiarity with rats, renter status, or socioeconomic or comfort levels interacting with the city government. Yet, studies in Chicago, Barcelona, Amsterdam, Rotterdam, and Eindhoven found that there was a close association between public rat complaints and systematic trapping estimates of relative abundance (15–17). In addition, in the current study, we were taking total public complaints across the entire city, and across multiple years, to look at the change in numbers, not the absolute numbers, that bypasses any potential risk of within-city heterogeneity or temporal heterogeneity in reporting bias.

After collecting and filtering the necessary data, we tabulated the number of monthly rat reports in each city for as many months as were available. We also acquired data on the human population density and population change via census reporting. Temperature change over historic baseline mean temperatures and mean temperatures were obtained from national weather agency data (e.g., the National Oceanic and Atmospheric Administration in the US and the Royal Netherlands Meteorological Institute in the Netherlands; table S1) and published reports. City-level GDP was used as a measure of municipal socioeconomics, obtained from publicly available reports. More information on these measurements can be found in table S1. We excluded several variables that were correlated with the other predictor variables; for example, mean annual temperature was excluded because it was correlated with mean minimum temperature, which was the more physiologically relevant variable. Percent urban land cover was inversely correlated with percent vegetation land cover, so we only used percent vegetation based on our hypotheses related to the urban heat island impacts on the thermal environment for rats. We also did not include mean maximum temperature, which was associated with the long-term temperature warming based on well-established latitudinal trends. We did retain two pairwise variables that are correlated because of long-established correlative links. First, human population density is correlated with GDP, which is a well-known association in economics due to urban scaling and Zipf’s Law (scaling based on disparity in city size). For example, Ribeiro et al. (68) found that across >5000 cities in 96 countries, GDP consistently associates with human population due to that scaling. Second, human population density is correlated with long-term temperature warming, and this pattern is known to be coincident on the basis of urbanization geography and anthropogenic heat emission (69) rather than at the subregional scale that most of our temperature data were collected at.

We also calculated the urban and vegetated land cover in each city to look for an association with land cover and rat trends. We obtained these global data as raster images from the European Space Agency (ESA) through their Climate Change Initiative (CCI) and analyzed them in QGIS mapping software. These data were 300-m resolution and included a time series from 1992 to 2020. We added the municipal boundaries for all 16 cities to this QGIS project, added a buffer of 1 km (to conjoin all noncontiguous city parcels), and then used the Zonal Statistics tool to calculate the number of cells of each land cover type. For the ESA-CCI data, we collated the land cover classes that had tree and shrub cover, herbaceous cover, and shrubland (i.e., cover class codes 12, 40, 60-62, 70-72, 80-82, 90, 100, 110, 120-122, 160, 170, and 180) in a “vegetation” category, and the “urban areas” class (code 190) into an “urban” category. We then calculated the proportion of each city that had vegetated and urban land cover. Low percentages of vegetation within cities could be explained by the low spatial resolution of the raster leading to the vegetation being outweighed by urban land covers in the raster’s classification at the 300 m–by–300 m resolution.

The trend in the number of sightings for each city was analyzed using a Mann-Kendall test, which analyzes data for the magnitude of trends in time series data. The z statistics were calculated for each city to analyze the magnitude, direction, and significance of the rat trends. The Mann-Kendall test is a nonparametric analysis that does not assume a priori distributions or homoscedasticity and is not sensitive to outliers or nonnormally distributed data (70, 71). We performed a relative importance (or weights) analysis to partition variance among the five environmental variables and quantify their association with the trend z statistic. Relative weights analysis is based on the standard regression framework but accounts for multicollinearity among the explanatory variables (72). We ran this in the “relaimpo” package within R. Lastly, we performed linear regression between the z statistic and each of the five variables to estimate the correlation coefficient for each relationship.