Unveiling the urban colonization of the Asian water monitor (Varanus salvator) across its distribution range using citizen science

Study species and area

The Asian water monitor (Varanus salvator, Laurenti 1768; order Squamata; family Varanidae) has six subspecies recognized: Varanus s. salvator, V. salvator andamanensis, V. salvator bivittatus, V. salvator celebensis, V. salvator macromaculatus and V. salvator ziegleri (Auliya, 2006; Quah et al., 2021; Auliya & Koch, 2020). It is one of the most widely distributed varanids, ranging from Sri Lanka in the west, to the Celebes Islands (Indonesia) to the east and South China to the north, occurring mostly in Southeast Asia countries (Gaulke & Horn, 2004; Bennett et al., 2010; Quah et al., 2021). The Asian water monitor is among the biggest lizards in the world, reaching almost 3 m from head to tail. Males mature at a smaller size but grow to reach larger body sizes (Shine & Harlow, 1998; Frýdlová et al., 2011) and have longer tails. Observations made in Bangkok (Thailand) show how they have bimodal diurnal activity, hunting/scavenging in the morning (06:00–08:00 h) and the afternoon (15:00–17:00 h), spending the rest of the day basking and floating (Trivalairat & Srikosamatara, 2023). This diurnal activity was also confirmed in the Sundarbans (Bangladesh) by Rahman, Rakhimov & Khan (2017). Considering data from Sumatra, this species extend egg-laying season from April to October, with the possibility to produce more than one clutch each per year, ranging from five to approximately 20 eggs (Shine, Harlow & Keogh, 1996). They are able to thrive in both terrestrial and aquatic environments (Zhao, Zhao & Zhou, 1999; Gaulke & Horn, 2004; Weijola, 2010; Weijola & Sweet, 2010; Bennett et al., 2010), including highly human-altered landscapes such as farmlands and cities (Cota, Chan-Ard & Makchai, 2009). They have a very flexible diet, that includes invertebrates, eggs, fish and even carrion (Bennett et al., 2010; Grismer, 2011; Rahman, Rakhimov & Khan, 2017; Yu et al., 2021). V. salvator has been intensively harvested for the skin industry (Shine, Harlow & Keogh, 1996; Traeholt, 1998; Khadiejah et al., 2019), the traditional medicine (Mardiastuti et al., 2021) and also for its consumption as food (Arida et al., 2021).

Literature review

To review the scientific literature available about the Asian water monitor we followed and adapted the guidelines proposed by Haddaway et al. (2015). The main steps used for our study are summarized in a flow diagram in the supplementary material (SP1). Briefly, first we studied peer-reviewed articles published in journals available on Scopus and Web of Science databases. The search was applied to the title, including any dates. We also included an additional non-systematic search using Google Scholar (Gehanno, Rollin & Darmoni, 2013; Piasecki, Waligora & Dranseika, 2018). Moreover, we used the “snowball” procedure, including those articles related to our topic found in the selected and analyzed references (Lozano et al., 2019). Initially we considered the inclusion of the so-called grey literature, attending to the additional contribution of technical reports and other sources beyond scientific articles (Haddaway & Bayliss, 2015). We only considered articles written in English, and we discarded those articles without an online version and also those where the link provided and parallel searches did not lead to the referred article. The review was conducted using a search with the term “Varanus salvator” AND “urban environment”, but previously we used other potential combinations that did not yield adequate results for the objective of the work (i.e.: articles not related to Varanus lizards, articles in urban ecosystems but focused on human well-being and other aspects related to urban life of humans, etcetera). Regarding the topic in the studies we reviewed, in our final list we discarded those studies based on reviews and theoretical/conceptual articles without their own data and analyses. Lastly, to avoid the heterogeneity of inclusion criteria inherent to different observers, only one of the authors carried out the search and the subsequent exploration of the articles.

We revised the content of the retained articles by a two-step process. First, we screened titles and abstracts of the first 200 articles obtained in our search, ordered by relevance and without limiting to any date. We do not consider as ‘urban’ those articles conducted in human modified landscapes such as agricultural areas but not considered properly cities. Similarly, articles focused on the use of V. salvator in laboratory conditions for medical research were discarded. Secondly, we read the main text of the articles to gather basic information to separate the studies conducted in urban environments, and to obtain information regarding the year of publication, the topic addressed and the country in which the study was conducted. We classified the articles retained according to defined categories: behavior, diet, physiology, distribution, habitat selection, parasitology, and conservation (see details in SP2). A given article can be included in more than one category, if it focuses on more than one aspect according to our classification.

GBIF data collection and analyses

We acquired occurrence data of Asian water monitors from the Global Biodiversity Information Facility (GBIF) (http://www.gbif.org), one of the largest sources of open data, that includes information from different citizen science online platforms, hence being of great interest to ecologists working on the spatial distribution of species (Telenius, 2011; Beck et al., 2014; Ivanova & Shashkov, 2021). As a first step, we downloaded 4,584 occurrences (Gbif.org, 2022). We then filtered by year (>2010) and verified veracity of data one by one by spatial confirmation using QGIS (v. 3.16) (i.e., avoiding those records that for any reason appeared on the sea instead of on terrestrial ecosystems). We also checked the linked image (we only considered data with pictures) from iNaturalist (GBIF uses observations from iNaturalist as well) associated with each occurrence, discarding, when necessary, any doubtful data (for example when the individual in the image is in captivity or when the record was not correctly identified and could refer to other Varanus species). We only kept data within urbanized areas from all the distribution range of the species.

As a second step, for subsequent analysis we selected the top 5 cities with more records. For those cities, we delimited urban areas as raster cells with a value higher than 20 for Human Footprint (HFP) index (Venter et al., 2018), as a clear boundary could be detected between less anthropized environments surrounding the city and the urban limits at this threshold. We chose this layer because of its common usage (Santini et al., 2021) and its great potential to indicate human influence, quantifying the impacts of human disturbance on a global scale. Since observations can have spatial biases with proven consequences on species distribution modelling and spatial analysis (Hortal, 2008), we sought to reduce them by calculating the ratio of observation, therefore, avoiding working with occurrence abundance. In order to gather enough information about sampling efforts, this ratio used all squamate occurrences available for the study area from 2010–2023 as a proxy (Gbif.org, 2023), including V. salvator, and the observations were divided by the number of different observers (Squamate observations/N° Observers) per pixel (≈1 km²) (Chauvier et al., 2021). By doing this, we obtained an estimation of how the sampling effort is distributed along our cities in order to support if the accumulation of water monitor observations could correspond to real abundance. We selected squamates because they are prevalent in our study areas and they are the taxonomic group of our study species. The observation patterns could differ between the different types of squamates and those of V. salvator. However, the V. salvator records are not abundant enough to calculate an observation ratio of its own, so the comprehensive records within the entire taxonomic group (squamates) should adequately represent the distribution of sampling effort across the cities for this particular group. Afterwards, we calculated the distance of those pixels with at least one V. salvator observation to the urban border defined by the HFP, which allowed us to explore possible patterns of distribution inside cities.

Additionally, we analyzed the approximated location where these observations occur within the urban matrix. For this aim, we preliminary checked if we could determine patterns that reveal greater sampling efforts, and therefore, higher V. salvator presence in urban green areas (parks and big gardens) in comparison to other urban environments such as neighborhoods, small gardens, natural and artificial streams, etc. For this, we determined as “Park” those areas with a Vegetation Continuous Field (VCF) value higher than 7, as in the previous case, it seemed to be a natural break to distinguish urban from green areas. This last raster layer is global representation of surface vegetation cover (DiMiceli et al., 2015) and was generated using Google Earth Engine derived from a 250 m resolution from MODIS. In both cases, HFP and VCF, there was a pixel size of 0.00833° (∼1 km² at the equator).

Literature review

Our search was conducted on 8th April 2021. We retained 148 scientific articles after the steps applied to discard or include them (see SP3 to see a detailed list of the articles reviewed). Most of the articles (96) were found in Scopus, Web of Science or in both sources. Moreover, we added 49 scientific articles found in Google Scholar and three articles detected by snowball. Although we initially considered grey literature, we did not find technical reports, conferences and similar publications, so we only included scientific articles. Most of the articles (117) were published after the 2000s, and only 31 articles before. There are 13 countries represented in those studies (Bangladesh, China, India, Indonesia, Laos, Malaysia, Myanmar, Philippines, Singapore, Sri Lanka, Thailand, Timor-Leste, Vietnam). However, in 46 cases the country was not described, or the study was conducted in laboratory conditions (i.e., the species was not studied in its habitat). Considering the total of studies reviewed, only 17 articles focus in urban habitats. We found that the studies of urban Asian water monitors started around the 2000s, and between 0–2 articles focused on urban populations of this species are published yearly to date. Moreover, we detect that all the studies analyzed in urban habitats were conducted in three countries: 12 in Thailand, three in Sri Lanka and two in Indonesia (Fig. 1A). The predominant topics of those studies were the behavior of these reptiles in cities, their diet, and to a lesser extent the distribution of the species, while studies focused on physiological aspects, parasitology, habitat selection and the conservation of the species are only represented with one study (Fig. 1B).

Urban occurrence using GBIF data

We show how V. salvator occurs in Asian urbanized areas from Sri Lanka in the west to East Indonesia, with most records concentrated in cities of Thailand, Sri Lanka, Singapore, peninsular Malaysia, Java and Sumatra (Fig. 2A). Using this source, we did not find data from Vietnam, Laos, Cambodia, Timor Leste, Philippines and China. Looking more closely to our data, we detect the five cities with the most records: Colombo (n = 36), Kuala Lumpur (n = 99), Jakarta (n = 42), Bangkok (n = 481), Singapore (n = 1,641) (Fig. 2B). For Colombo and Jakarta we observe that sampling effort (i.e., more observations per observer) tends to accumulate near the city center (Fig. 3). Singapore, Kuala Lumpur and to a lesser extent Bangkok, show a balanced distribution of sampling effort with a slight increase towards the centered areas mainly for the first two. Bangkok, however, shows a maximum observation ratio at a distance of 6.25 km from the border. These observation spots (1 km² pixels with at least one V. salvator observation) are not equally distributed either (Fig. 3). Thus, mainly for Colombo and Jakarta we see gaps on the distribution of black dots along the distance to border axis, which indicates uneven observation patterns or incomplete sampling within the city. Bangkok, Kuala Lumpur and Singapore show a more constant representation of observations along the border to city-center gradient.

We also studied the distribution of records within the urban matrix and how this sampling effort is distributed among habitat types. Colombo, Jakarta and Bangkok experience at least some degree of net positive increase in the intensity of sampling when increasing VCF (Fig. 4). In particular, Jakarta experiences a high point at around a value of 7.25, near the stablished threshold for being considered as park (VCF > 7). Kuala Lumpur reaches its peak around a value of 12.25 for VCF. After that, the effort decreases. Singapore, however, again shows a homogenous pattern of sampling effort when it comes to different types of habitats. Colombo, Jakarta and Kuala Lumpur presented some data gaps for the VCF variable (see Fig. 4), while in Singapore and Bangkok the data are better distributed. We observed that spotting areas of V. salvator occur mainly in green urbanized environments: in Colombo (87.5%), Jakarta (77.4%), Kuala Lumpur (53.1%) and Singapore (76%) a great percentage of the spotting areas occur in to these defined park areas. Only in Bangkok, (26.9%) we observed a greater number of observations out of parks (Fig. 4).

Literature review

Our results confirm the apparent lack of scientific literature on the colonization and ecology of Asian water monitors (V. salvator) in urban environments, with only 17 of the 148 articles reviewed conducted in urban environments, and few cities of three countries (Indonesia, Sri Lanka and Thailand) frequently represented in those studies. Most of the articles reviewed are short notes and reports of field observations, mostly about behavior, diet and the how the species thrive in this habitat. Those records confirm the generalist feeding habits of the species, which in cities consumes diverse birds, mammals, fish, amphibians and reptiles (Bundhitwongrut et al., 2008; Karunarathna, Amarasinghe & De Vos, 2008; Stanner, 2010; Cota & Sommerlad, 2013; Mahaprom & Kulabtong, 2018), but also exploit carrion and are attracted to human waste (Kulabtong & Mahaprom, 2015; Lawton et al., 2008). Regarding the behavior, the studies reviewed describe the mating of the species, the intraspecific relations in different seasons, the use of underwater burrows, its daily activity patterns, with both diurnal and nocturnal habits, and how the Asian water monitors preferentially use aquatic habitats, especially the hatchlings and juveniles (Rathnayake et al., 2003; Cota, 2011a; Cota, 2011b; Karunarathna et al., 2017). In this sense, the urban environment facilitates observations of basic ecological and behavioral aspects that could be more difficult to observe in other contexts, mostly due to the more elusive behavior of non-urban individuals of many species, compared in some studies in both urban and non-urban environments (Evans, Boudreau & Hyman, 2010; Isaksson, AD & Gil, 2018). Nevertheless, it is possible that our findings may be biased due to the small sample size obtained through the revision, together with the apparent absence of studies offering comparisons across a broader geographical scope or studies that articulate their hypotheses within the theoretical framework of urban ecology.

Urban occurrence using GBIF data

Attending to our results, V. salvator is colonizing cities in different countries within its distribution range, with differences regarding how they exploit the urban matrix among cities. This is a common pattern in studies focused on urban wildlife, with population and species relationships with urbanization varying due to underlying reasons such as the urban landscape, the presence of corridors, the dimensions and spatial arrangement of habitat patches (Ortega-Álvarez & MacGregor-Fors, 2009; Fontana et al., 2011) of tourism, and also cities with big parks used by both local people and tourists.

In our study, in Singapore we saw a slight increase in sampling effort when approaching the city center, but most of the occurrences are recorded at parks and gardens. Within a city, not all parks or green areas hold a continuum of values for VCF. However, in our data we find few gaps between dots, which reflects a good distribution of samplings for the VCF variable. In Colombo the urban matrix is imbricated with green avenues, parks and gardens. Therefore, the Asian water monitor appears in many parts of the city homogenously, explaining its ability to reach particularly central parts of the city. A similar situation is observed in Jakarta, with higher distribution of observation ratio relying on the city center, and many of these observations occurring in gardens and water courses throughout the city. Bangkok demonstrates a similar pattern, with a smooth increase in the observation ratio as one approaches the city center and a tip around 6.25 km away from the border, and most of the records concentrated in small urban parks. In this case, V. salvator maybe uses the water canal system and even the sewage system to move within the city, disperse and connect parks and the outside of the city. Lastly, in Kuala Lumpur, where the city core is densely occupied by human-made buildings, the observation ratio tends to be similar even along the border-center gradient with a little increase in these core areas. Also, in this case, the ratio of observation in green areas is similar to that in the urban matrix. Similar studies in other reptiles, especially lizards, also contribute to explain how these animals exploit cities, with different selection according to the anthropogenic intensity and habitat availability observed within city boundaries. Thus, Winchell et al. (2018) show, in a study conducted in Puerto Rico, how Anolis stratulus tends to occur in more “natural” urban patches while Anolis cristatellus seems more attracted to human infrastructures. Moreover, Dékány, Kövér & Babocsay (2015) in their study of Podarcis muralis in Budapest (Hungary) also show higher densities in more diverse and with semi-natural elements urban patches. Regarding the data used, the lack of occurrences in certain areas might reflect different factors, such as cultural, low tourism, population density, lack of devices like smartphones to take pictures or even the legislation of the country itself that prevents their citizens from sharing data with the rest of the world (Capdevila et al., 2020; Callaghan et al., 2021; Walker, Smigaj & Tani, 2021). It should be also highlighted that citizen science is frequently subjected to representation biases that can affect spatial analysis or species modelling (Kadmon, Farber & Danin, 2004; Franklin, 2010; Boria et al., 2014), in such a way that observations accumulate near accessible paths, roads or in urban areas. In that sense, we realized that most of our observations are concentrated towards the city center in populated cities with high levels.