In more recent publications, the range of Central Asian Desert Monitor Varanus griseus caspius (Eichwald, 1831) for entire Central Asia, as well as for individual republics, is shown either quite conditionally and in general terms, or distribution maps are presented as points of findings. We summarized and analyzed all available sources of information regarding the distribution of the Central Asian Desert Monitor, plotted the existing points of findings on the map, conducted modeling of the potential distribution of the monitor, and, comparing the result with the map of the anthropogenic modified landscape, delineated the modern range of the Desert Monitor in Central Asia. Modeling the potential range using various environmental factor variables allowed us to supplement factual data and construct a more detailed map of the Desert Monitor’s range. The modeling was performed using the maximum entropy method in the Maxent software. The modeling results showed that the most significant factors for model construction were the mean temperature of the driest quarter, as well as the normalized difference vegetation index (March), precipitation during the warmest quarter, precipitation (March), and solar radiation (April). Consolidating all available information, we were able to update the distribution map of the Desert Monitor in Central Asia, and the classification of findings by periods along with the anthropogenic landscape map allowed us to assess the degree of its range change since the mid-last century. It was found that the monitor’s range has significantly decreased in many regions due to the “expansion” of an anthropogenically altered landscape. The most considerable range reduction, relative to the past, occurred in Uzbekistan, Tajikistan, and Kyrgyzstan. To a lesser extent, the range decreased in Turkmenistan, and very slightly in Kazakhstan.

anthropogenic impact, Central Asia, Desert Monitor, ecological niche models, environmental factors, GIS, MaxEnt, range dynamics, species distribution modelling

The range of the Desert Monitor Varanus griseus occupies most of the Sahara-Gobi Desert region. From west to east, it is distributed from North Africa, the Middle East, Central and South Asia to Pakistan and northwestern India. The southern limits of the Desert Monitor are reached in Yemen, and to the north, it extends to the Aral Sea and the city of Kyzylorda in the southern part of Kazakhstan. Within the entire range of the Desert Monitor, three subspecies are distinguished: the nominative subspecies V. g. griseus (Daudin, 1803), inhabiting the western part of the range – from northwestern Africa to the Zagros in Iran; the Central Asian subspecies V. g. caspius (Eichwald, 1831), occurring in the сentral part of the species’ range – from the Zagros to the mountain systems of Pakistan; and the Indian subspecies V. g. koniecznyi Mertens, 1954, native to the eastern part of the range – in the plains of Pakistan and northwestern India (Stanner 2004).

The Central Asian Desert Monitor V. g. caspius was described by E.I. Eichwald in 1831 from the eastern coast of the Caspian Sea, the Dardzha peninsula in Turkmenistan (Eichwald 1831). More than half a century later, in the 1860–1880s, the first findings of the Desert Monitor in other parts of Central Asia had been published: from the vicinity of the city of Khujand (Khojent) in Tajikistan, the foothills of the Kopetdag in Turkmenistan, and from the eastern part of the Kyzylkum desert in Uzbekistan (Zaroudnoï 1885; Nikolsky 1899; Chernov 1959). For the territory of Kazakhstan, the first record of the Desert Monitor on the western shore of the Aral Sea was published by M.A. Levanevsky (1894), and N.A. Zarudny (1915) reported on the occurrence of the Desert Monitor in 1890 on the island of Kaska-Kulan on the eastern coast of the Aral Sea.

At the beginning of the 20^th century, data began to appear from other regions, mostly from Turkmenistan, but they were still scarce and did not provide a complete understanding of the distribution of the Desert Monitor in Central Asia. Targeted studies on the Central Asian Desert Monitor were only initiated in the second half of the 20^th century. During this period, numerous works were published that quite fully reflected the distribution of the Central Asian Desert Monitor in the region (Paraskiv 1956; Chernov 1959; Bogdanov 1960; Yakovleva 1961; Bogdanov 1962; Salikhbaev et al. 1967; Rumin 1968; Yadgarov 1968; Vashetko and Kamalova 1974; Kamalova 1978; Said-Aliev 1979; Shammakov 1981; Zakharidze 1981; Khodzhaev 1982; Makeev 1982; Tsellarius 1982; Ataev 1985; Kovshar 1986; Bondarenko 1989a, 1989b; Brushko et al. 1990; Tsellarius et al. 1991; Brushko 1995; Tsellarius et al. 1995; Tsellarius and Tsellarius 1996; Tsellarius et al. 1997).

In the 21^st century, field studies were conducted to clarify the distribution of the population in Kazakhstan (Chirikova et al. 2019) and Uzbekistan (Khodzhaev et al. 2019; Abduraupov et al. 2021). For the territories of Turkmenistan, Tajikistan, and Kyrgyzstan, only fragmentary data on the modern distribution of the Desert Monitor exist (Milko and Panfilov 2007; Shestopal 2008; Solovyova et al. 2013; Davletbakov et al. 2015; Shestopal and Rustamov 2018; Khidirov et al. 2021).

Despite the significant amount of information on the distribution of the Desert Monitor available today, information on the species’ regional status remains uncertain. The expansion of anthropogenic impacted landscapes leads to the “displacement” of the monitor from its previous habitats. In this regard, it is necessary to conduct monitoring of the populations of this species, especially in those territories for which there are no current data.

The aim of this study is to define the boundaries of the modern distribution of the Desert Monitor for the territory of Central Asia. In more recent publications, the range for entire Central Asia, as well as for individual republics, is depicted either in an approximate and very generalized manner (Stanner 2004; Ananjeva et al. 2006), or distribution maps are presented as individual points of occurrence (Milko and Panfilov 2007; Chirikova et al. 2019; Abduraupov et al. 2021), or in the form of an atlas with a square grid (Sindaco and Jeremčenko 2008). The only exception is Uzbekistan, for which the most detailed range of the Desert Monitor has been published (Nuridjanov et al. 2016).

In this work, we summarized and analyzed all available sources of information regarding the distribution of the Central Asian Desert Monitor, plotted the existing points of findings on the map, conducted modeling of the potential distribution of the monitor, and, comparing the result with the map of the anthropogenic modified landscape, delineated the modern range of the Desert Monitor in Central Asia.

Following the comprehensive analysis of all available data sources, a total of 820 occurrences of the Desert Monitor were collected, with 354 in Kazakhstan, 9 in Kyrgyzstan, 180 in Uzbekistan, 45 in Tajikistan, 140 in Turkmenistan, 86 in Iran, and 9 in Afghanistan.

In the first stage of modeling using the full set of variables, Maxent analysis showed that at rarefaction distances of 15, 25, and 50 km, 49, 38, and 46 variables, respectively, contributed zero to the model; 44, 54, and 40 variables contributed less than one percent each; and only 13, 14, and 20 variables contributed more than one percent (Suppl. material 2: table S2).

After filtering the list of variables – excluding those that were highly correlated or contributed nothing to the first model (Suppl. material 2: table S3) – 24, 26, and 26 variables were identified as the most significant at rarefaction distances of 15, 25, and 50 km, respectively. The highest average value between average test AUC and average TSS was obtained for the rarefaction of the initial point set at 15 km, with Maxent settings where feature classes were specified as LQH and the regularization multiplier was set to 0.5. The AUC and TSS values for all combinations of Maxent settings are given in Suppl. material 2: table S4. The list of environmental variables used in the best iteration of the final model, as well as their contributions, is provided in Table 3.

The Area Under the Curve (AUC) of the final model was 0.93, indicating that its predictive value is “excellent” (AUC > 0.90) (Araujo et al. 2005). The True Skill Statistic (TSS) value of the same model was 0.78, which is considered “useful” (0.6 < TSS < 0.8) (Coetzee et al. 2009).

The cutoff threshold according to the “Equal training sensitivity and specificity (Cloglog)” rule, as determined by Maxent, was 0.34 for the best model. The resulting model, with values below the threshold excluded and with overlaid zones of human settlements and irrigated croplands, is shown in Fig. 1.

As a result of summarizing all the gathered literature, collection, and survey data on the actual findings of the Desert Monitor, as well as our own materials, and comparing them with the obtained model and map of anthropogenic impact, the range of the Desert Monitor in Central Asia was delineated (Fig. 2).

Figure 1. 

Best final SDM of the Desert Monitor. Yellow dots show the points used for modeling, dark gray areas represent human settlements and irrigated croplands.

Figure 2. 

Range of the Desert Monitor in Central Asia.

As shown in Table 3, five variables contributed 75% to the construction of the final model. These are: the mean temperature of the driest quarter (BIO9), the Normalized Difference Vegetation Index for March (NDVI3), precipitation during the warmest quarter (BIO18), precipitation for March (PREC3), and solar radiation in April (SRAD3). The mean temperature of the driest quarter alone accounted for more than 25% of the model’s contribution. Climatic factors have complex interdependencies with each other and the biology of the species, making the limiting effect of any given variable multifaceted. It can be hypothesized that a decrease in the mean temperature of the driest quarter—typically the summer months when egg incubation occurs—may result in embryos not developing sufficiently before the onset of autumn cooling. The significant contribution of “solar radiation for April” and “NDVI for March” to the model is likely crucial, especially in the northern parts of the range. By this time, the soil must warm up enough to allow the monitors to emerge from their hibernation burrows by early April. Furthermore, vegetation presence is linked to the availability of food resources at the time of their emergence. Precipitation of warmest quarter and precipitation for March humidity level, which, on one hand, also affects the availability and abundance of food objects during spring activity, on the other hand – on the moisture balance in the soil at the depth of the nesting chamber, where a sufficient soil moisture reserve must be formed to preserve the monitor’s clutches during the summer months. However, the limiting effects of these factors can vary greatly across different parts of the species’ range. For instance, while the length of the warm season is critical in the north, where embryos must develop before the autumn cold snap, this is not as relevant in the southern areas, where frost is absent.

The distribution model is calculated based on a limited set of variables and cannot account for all environmental factors. Therefore, the result is hypothetical and does not always match the actual situation – that is, despite the high suitability of certain territories according to the model, in reality the animal may be absent from these areas for various reasons, including natural and human-induced ones. In our case, such areas include a significant part of the right bank of the Syr Darya River (up to the Karatau range) in Kazakhstan, for which the model predicts very high habitat suitability for the Desert Monitor. However, despite the dense human population of this area and its frequent visits by scientists, the Desert monitor has only been recorded south of the 42^nd latitude. The most straightforward explanation for this could be the anthropogenic factor – the entire right bank of the Syr Darya River, north of the known records of the Desert monitor, is home to one of Kazakhstan’s largest cities, Shymkent, as well as many villages and roads. Moreover, a large part of this territory is currently or was recently used in agriculture. The negative attitude of the local human population towards this animal cannot be ignored. An encountered Desert Monitor often causes fear among residents. As a result, despite its protected status (in Kazakhstan, the Desert Monitor is listed in the Red Book), the animal is usually killed. Our surveys and observations have confirmed this in most cases (Brushko and Chirikova 2010; Nuridjanov et al. 2016; Chirikova et al. 2019).

In the vicinity of the city of Kyzylorda, despite the presence of several Desert Monitor observations, including a relatively recent one (Dzhikibaev 2010), it seems the Desert Monitor does not live there. This is indicated by both the modeling results and the scant number of findings from this area. Surveys of residents and inspectors confirmed that the Desert Monitor is not found in these regions. Such a distant encounter from the main habitats of the Desert Monitor can be explained by intentional or accidental transportation of the animal. The same point of view was previously expressed by Z.K. Brushko (1995). The authors are aware of two such cases: in one instance, a Desert Monitor climbed into the engine compartment of a car in the Eastern Kyzylkum Desert and revealed itself only after stopping more than a hundred kilometers away (B.M. Gubin, pers. comm.), and in another, an individual was transported to the city of Shymkent in the back of a truck along with sand. In our view, the record at the western foothills of the Karatau range, near the settlement of Zharykbas (Leontyevka) (Chirikova et al. 2019), has a similar nature.

Several records of the Desert Monitor were known near the Aral Sea in the past. It is difficult to assess the reliability of this data and the nature of the stay of the noted individuals, however, it can be confidently stated that today the Desert Monitor is not found in these territories. This is evidenced by both the data from Uzbek colleagues (Nuridjanov et al. 2016; Abduraupov et al. 2021; R.A. Nazarov, pers. comm.), and its absence in the fauna lists of the “Barsakelmes State Nature Reserve” on the Kazakh territory of the Aral Sea.

Classifying the points by historical periods well demonstrates the pace of data accumulation on the distribution of the Desert Monitor. However, the local nature of research at different times does not fully allow assessing the changes at the boundaries of the entire range. Nevertheless, comparing the distribution model, the time and places of locality records, as well as the map of settlements and irrigated arable lands, it can be concluded that in the territory of Uzbekistan, large areas of the Desert Monitor’s range disappeared in the second half of the 20^th century as a result of anthropogenic landscape transformation – this includes the entire Tashkent and Syrdarya regions, the eastern part of the Jizzakh region, significant territories along the river floodplains in the Samarkand, Kashkadarya, and Bukhara regions, as well as much of the Fergana Valley, where the Desert Monitor, apparently, has only survived in narrow sections along the southern, northeast, and northwest foothills, unsuitable for agriculture, and on a small segment in the central part of the valley (R.A. Nazarov, pers. comm.).

For Tajikistan, current observations/records are known from the foothill areas. The plains and historical areas of the Desert Monitor’s range along the river valleys have been completely transformed by agricultural activity.

In Turkmenistan, the Desert Monitor inhabits a larger territory of the country. However, despite its widespread distribution and abundance, which even led to its removal from the latest edition of the republic’s Red Book (Shammakov and Geokbatyrova 2016), we managed to find about two dozen references of records in recent publications. According to the modeling results and the absence of encounters, the Desert Monitor does not live in the northern part of Turkmenistan around the Kara-Bogaz-Gol Bay and the Sarykamysh Lake. Additionally, areas previously suitable for the Desert Monitor have been anthropogenically transformed – these are the northern foothills of the Kopetdag mountain system and the territories around the cities of Tejen and Mary.

Regarding modeling, relatively recently, the range of the Desert Monitor in Central Asia was modeled (Malakhov and Chirikova 2018), using a methodology different from ours. That study focuses on identifying the biological characteristics of the Desert Monitor, however, the correspondence of the obtained model to the actual distribution of the species is not discussed, nor the threshold for the suitability of the environmental conditions, which the authors consider to be the species distribution boundary. To compare the results, we attempted to classify the raster image of the model from this article and “draw” a boundary on the model for two values: 50% – as a median value on the scale, and 40% – a value at which the majority of the initial points are “captured” by the model. At 50%, many observational records of the Desert Monitor in Kazakhstan, Uzbekistan, and Tajikistan were outside the model, while at 40%, the entire territory of Kazakhstan south of the latitude of Kyzylorda is displayed as “suitable” habitat for the Desert Monitor, including the Mangyshlak Peninsula and the Moynkum Desert east of the Karatau range. That is, under different assumptions of suitability from the model, either a large number of Desert Monitor locality records “fall out,” or the model outlines suitable territories too “broadly,” which poorly corresponds to actual data, especially regarding the northern boundary of the species’ range.

The territory of Iran in our work is captured incidentally and adjunctly, nonetheless, we have the opportunity to compare the calculation results with the recent work of Shadloo et al. (2021). This work uses a different set of species presence points and a different modeling methodology. The model in the mentioned article agrees well with ours in the western part of Iran but differs in the eastern part – in our case, the model demonstrates “suitable” conditions on more plain areas. In the work of Malakhov and Chirikova (2018), observational records from the territory of Iran were not used, therefore the model in this area is the result of extrapolation and shows the opposite result – plain territories have weak suitability or are unsuitable at all, while mountainous areas are indicated as highly suitable.

Thus, it can be stated that different modeling methodologies can provide varied, sometimes opposite, results. The primary criterion for evaluating the modeling result in such cases should be the model’s correspondence to credible and contemporary actual data, in cases, of course, when such data are available.

Consolidating all available information, we were able to update the distribution map of the Desert Monitor in Central Asia, and the classification of findings by periods along with the anthropogenic landscape map allowed us to assess the degree of its range change since the mid-last century. It was found that the monitor’s range has significantly decreased in many regions due to the “expansion” of an anthropogenically altered landscape. The most considerable range reduction, relative to the past, occurred in Uzbekistan, Tajikistan, and Kyrgyzstan. To a lesser extent, the range decreased in Turkmenistan, and very slightly in Kazakhstan.

Based on the available data, the contemporary range of the Desert Monitor in Central Asian countries appears as follows. In Kazakhstan, the Desert Monitor is widespread in the extreme south of the country in the Turkistan and Kyzylorda regions along the right and left banks of the Syr Darya River. To the west of the Syr Darya, the Desert Monitor inhabits the Kyzylkum sands up to the republic’s border on the south and west of this section, reaching its northern distribution limits at the latitude of the settlement Baygekum village (44.3°). To the east of the Syr Darya River, the Desert Monitor’s range in the south is limited by the Shardara reservoir, in the southeast by the Keles River, and in the northeast by an approximate line from the Koksaray reservoir to the city of Saryagash. In Uzbekistan, the Desert monitor inhabits a significant part of the republic to the east and south of the city of Nukus. The northern boundary of its distribution here runs approximately from the southern part of the Tabakum sands to the east of the city of Nukus, south of the Tasqudyk sands, through the Central Kyzylkum desert, along the northern foothills of the Bukantau range, and further northeast towards the Tasqudyk, well to the southeast of the Taspan village. The Desert Monitor is absent from cultivated river valleys around major cities: Tashkent, Gulistan, Jizzakh, Khujand, Fergana, Andijan, Urgench, Bahara, Samarkand, Karshi, and Termez. In the Fergana Valley, the Desert Monitor is found on narrow strips of foothills. In Kyrgyzstan, the Desert Monitor has only survived in small areas in the south of the Jalal-Abad and Batken regions along the lower belt of mountains along the northeastern and southern edges of the Fergana Valley, respectively. In Turkmenistan, the Desert monitor inhabits a large part of the country. The northern boundary of its distribution here runs in a broken curve from the city of Turkmenbashi, south of the Kara-Bogaz-Gol Bay, south of the city of Gyzylgaya, crosses the Uchtagan sands, south of the Gaplaňgyr Nature Reserve, along the northern edge of the Karakum sands towards city of Dashoguz. The Desert Monitor is absent from cultivated areas along the Karakum Canal and the Amu Darya River floodplain, as well as from the river floodplains around the cities of Ashgabat, Tejen, Mary, Dashoguz, and Turkmenabat. In Tajikistan, the Desert Monitor inhabits the plain and low mountain parts of the western half of the republic, including a small section of the foothill zone in the south of the Fergana Valley, while absent from the agriculturally transformed river valleys of the Syr Darya, Kafirnigan, Vakhsh, Kyzylsu, and Panj rivers.

Modeling the potential range using various environmental factor variables allowed us to supplement factual data and construct a more detailed map of the Desert Monitor’s range. The modeling results showed that the most significant factors for model construction were the mean temperature of the driest quarter, as well as the normalized difference vegetation index for March, precipitation during the warmest quarter and precipitation for March, and solar radiation in April. It can be assumed that these factors are limiting for the distribution of the Desert Monitor.

To verify the predictions of the obtained Desert Monitor distribution model, future field surveys are necessary in territories where its occurrence has not been previously confirmed. In Kazakhstan, these are the areas of the Kyzylkum desert south of the city of Kyzylorda; in Uzbekistan – the northeastern part of the Kyzylkum desert south of Taspan village and the sands from the city of Nukus along the right bank of the Amu Darya River towards the cities of Zarafshan and Bukhara; and in Turkmenistan – the northern part of the Balkan and Lebap Region, and Dashoguz Region.

To conserve the Desert Monitor in all Central Asian republics, it is necessary to implement social national, and international programs aimed at raising awareness among local populations about the species as a whole, its threats, vulnerabilities, and the importance of its conservation.

We express our gratitude to colleagues who provided information on Desert Monitor encounter sites, including M.A. Kulemin, G.S. Nazarbek, B.M. Gubin, D.A. Milko, K. Beishebaev, as well as to V.F. Orlova and N.B. Ananjeva for the opportunity to work with the collections of the Zoological Museum of Moscow University and the Zoological Museum of the Zoological Institute of the Russian Academy of Sciences. Special thanks to D.A. Nuridjanov and M.A. Gritsyna for allowing the use and publication of a substantial volume of data on Uzbekistan. We also sincerely appreciate everyone who participated in joint expeditions or assisted us in gathering factual information during this multi-year project. We are grateful to M. Auliya and R.A. Nazarov for valuable remarks, corrections and useful advice on writing this article, and to the anonymous reviewers, whose comments and suggestions have contributed to the improvement of this work.

YuZ and VF co-led the study design and wrote the manuscript. YuZ conducted field research, collected field materials and survey data, and processed literature. VF conducted modeling of the potential species distribution and created the maps. YuZ and VF analyzed the data and interpreted the results.

Previously unpublished species occurrence points used in this study are provided in the supplemental materials. All references to published data are provided in the “Material and methods” section.