Predicting the potential distribution of striped hyena Hyaena hyaena in Iran

Predictive potential distribution modelling is crucial in outlining habitat usage and establishing conservation management priorities. Association among species occurrence and environmental and spatial characteristics has been calculated with species distribution models. Herein, we used maximum entropy distribution modelling (MaxEnt) for predicting the potential distribution of striped hyena Hyaena hyaena in the entire country of Iran, using a number of occurrence records (i.e., 118) and environmental variables derived from remote sensing. The MaxEnt model showed a high rate of success according to AUC test scores (0.97). Our results are roughly congruent with previous studies suggesting that mountainous re gions in northern and western Iran, and the plains in central and eastern Iran are a suitable habitat for H. hyaena.


Introduction
The striped hyaena, Hyaena hyaena (Linnaeus, 1758) is distributed in most parts of Africa (except for the southern parts), the Middle East including the Arabian Peninsula, the Levant, Turkey, Iraq, Iran, and the Caucasus, and extends further into Central Asia and the Indian subcontinent. The species occupies varied types of environments and ecosystems (akay et al. 2011; reichmann 2005), extending from arid deserts (singh et al. 2010;Tichon et al. 2016;Wagner 2006) to dense Mediterranean shrubland (Abi-saiD & abi-saiD 2007;rosenberg et al. 2016). It was reported that hyaenas prefer open semi-arid habitats (alam et al. 2015;Wagner 2006), and occur in low densities in true deserts and dense vegetation (abisaiD & Dloniak 2015). Striped Hyaenas rely on large areas to support their resource requirements (alam et al. 2015;Wagner 2006). The distribution of H. hyaena in the Middle East and the Caucasus in particular, is now patchy as the result of a decline which has occurred over the past decades (kasparek et al. 2004). Thus, this species has the largest range of distribution among hyaenids (kruuk 1976). Hyaena hyaena has a wide range of distribution in Iran including mountainous, steppe and desert habitats throughout the country except north and northwestern provinces (ziaie 2008; karami et al. 2016).
According to the IUCN criteria, the world status of striped hyenas has been called as "lower risk: near threatened" in the conservation action plan prepared by the IUCN/SSC Hyaena Specialist Group (mills & hofer 1998). In Iran, it is mentioned in "Data Deficient" based on the latter reference and needs to be further investigated in the country.
According to eTemaD (1985), the Iranian hyaenas belong to the subspecies H. h. hyaena (Linnaeus, 1758). The striped hyena occupies an extensive range of habitats and may settle wherever it gets sufficient food supply, adequate cover, and access to water (kruuk 1976). It generally favors open or thorn bush areas in arid and semi-arid environments (rosevear, 1974; kruuk 1976; rieger 1978; leakey et al. 1999) to pine forests (akay et al. 2011) where water is available within 10 km (rieger 1979).
ziaie (2008) believed that the Iranian population of hyenas has severely declined. They live in steppes, semi-deserts, rocky fields and valleys with sparse trees (mills & hoffer 1998; Fig. 1); they avoid deserts, high altitude areas, dense bushes and forests. Striped hyenas prefer living inside caves (mills & hofer 1998; yilDirim 2010). These animals participate in the ecosystem by eating dead and decaying animal carcasses (mills & hofer 1998;bunaian et al. 2001;singh 2008;sTein et al. 2013). The dietary preferences of this species include vertebrate (bon et al. 2012) and invertebrate animals, various fruits, vegetables and human-sourced organic wastes (Wagner 2006). Being a dietary opportunist, hyenas are omnivorous carcass eaters. When starved, they can feed on melons, watermelons, grapes and some other vegetables (mills & hoffer 1998;yilDirim 2010). Recently, factors have generally led to major habitat loss (Fig. 2). Therefore, it is very important to protect the hyenas' habitats. As such, habitat management based on evaluation is strongly recommended as a practical solution. The prediction of distribution patterns of organisms using species distribution models (SDMs) could represent more information about the effects of environmental factors on potential geographical distribution of organisms (eliTh & leaThWick 2009;collevaTTi et al. 2013;eskilDsen et al. 2013). SDMs are informative tools to map and monitor distribution of organisms, and have become increasingly important in the context of awareness of environmental change and its ecological consequences (miller 2010).
Prediction models can forecast changes in distribution and abundance of species. These changes could be the result of future climatic conditions coupled with increases in habitat fragmentation that may result in species extinction (broWn et al. 1997;WalTher et al. 2002;rooT et al. 2003;Wake 2007). According to the results of predictive models, if climate change continues unchecked, 37% of global species might be extinct by the end of 2050 ( Thomas et al. 2004). The effects of climate change such as increasing temperatures have been felt from species to community levels (pounDs et al. 1999;WalTher et al. 2002;Thomas et al. 2004). This is a very serious issue for all species, and some regions may bear severe effects.
We used the maximum entropy (MaxEnt) modeling approach which is based on species presence data and environmental factors to predict potential distribution patterns. It is con sidered one of the most efficient techniques for predicting species distribution models based on presence only data (eliTh et al. 2006;phillips et al. 2006;eliTh et al. 2011). The aim of this study was to provide a comprehensive distribution map for striped hyena to recognize the environmental variables associated with the predicted dis tribution of this species using the MaxEnt distri bution modeling approach.

Material and methods
A total of 123 locality records for H. hyaena was gathered and after removing the duplicated records finally 118 points were used in the maximum entropy distribution modeling approach (Fig. 3). All records of H. hyaena were based on our own field work as well as those obtained from the Department of Environment of Iran. Twenty environmental variables including tempera ture, precipitation, seasonality, and altitude, all with 30-arc-sec onds resolution were obtained from the Worldclim data set (http://www.worldclim.org /;hiJmans et al. 2005) and built in ArcMap version 10. First, correlations between all 21 environmental variables were measured using Pearson's correlation coefficient in SPSS 16. The variables with a correlation coefficient > 0.7 were excluded from the analysis (rissler et al. 2006). Then, 10 out of 21 environmental variables were selected and used in this study ( Table 1).
The most common approach for testing the predictive performance is to divide data into 'training' and 'test' datasets. This approach creates relatively independent data for model testing (fielDing & bell 1997;guisan et al. 2006). Then, MaxEnt was carried out with default settings when separating records into training and test samples (75 and 25%, respectively) (phillips & DuDík 2008). Convergence threshold and maximum number of iterations were remained as default (0.00001 and 500, re spectively) and auto feature. The cross-validation has been used to evaluate the pre dictive performance of the model. In addition, Jackknife testing was produced to estimate the average contribution and response of each variable to the model. Our model was tested with 'area' under the receiver-operating characteristic curve (AUC) used extensively in assaying species' distribution models, and measures the ability of a model to differentiate between sites where a species is 'present' versus 'absent ' (phillips et al. 2006;eliTh et al. 2006). Models with AUC = 0.5 consider a performance equivalent to random; AUC > 0.7 as useful performance, AUC > 0.8 as good per formance, and finally AUC ≥ 0.9 as excellent performance (manel et al. 2001).

DADASHI-JOURDEHI A. et al., Predicting the potential distribution of striped hyena in Iran
Modeling of the potential distribution of H. hyaena revealed that the most suitable habitats are mountainous re gions in northern and western Iran, and plains regions in central and eastern Iran, as well (Fig. 5).

Discussion
According to the results of this study, the central parts of Iran which correspond to Dasht-e-Lut and Dasht-e-Kavir might not be suitable places to find the species. Moreover, north-western and central parts of Iran are not suitable habitat based on modeling. Generally, predicted distribution is approximately congruent with the known distribution pattern. According to singh et al. to abaDe et al. (2014), precipitation is one of the most important environmental factors that affects the distribution of spotted hyenas. The results of our study are congruent with the findings of these authors.
Various factors influencing the distributional pattern of hyenas, such as agricultural activities, livestock, mining, wild dogs, incursion and building of dams such as Mamlo dam located in the territory of the Khojir National Park. As a consequence, the most important cause of mortality of hyena in the country has been reported from Parchin road constructed in the middle of the Park. The model presented by rezaei et al. (2017) concluded that Hyaena hyaena is successful in finding suitable nesting areas in the Haftad-Gholleh protected area in Iran.
Along with the threat of road strike, illegal hunting is also increasing. Striped hyenas are usually widely distributed among local communities (moqanaki et al. 2015;farhaDinia et al. 2017), but they are rarely engaged in livestock's killing or damaging farm animals (farhaDinia et al. 2017). Traditionally, the organs of hyenas are used as a medicine and in superstitious beliefs, which have originated from the myth that different parts of the hyena's body are an effective means to ward off evil and to ensure love and fertility (karami et al. 2010;Tourani et al. 2012;rezaei et al. 2017). Habitat destruction and land modification stands as an important threat for the species and in some parts, farmers convert natural habitat into agricultural lands and start to kill hyenas to protect their crops. Land conversion reduces food access for hyenas, making predator-prey competition more disturbing. Unfortunately, there is no accurate data on the situation of Hyaena hyaena in neighboring countries, most data come from Iran.
Hence, identifying regions with high a suitability for hyenas can help to manage their conservation activities.
The International Union for the Conservation of Nature (IUCN) considers the striped hyenas to be Near-Threatened (arumugam et al. 2008). Although a high proportion of the species range is covered within the current network of protected areas in Iran (farashi et al. 2017), the species needs specific conservation programs, given that the majority of its suitable habitats lies near human settlements.
This study provides more information about suitable habitat and relevant environmental factors of the striped hyena. Certainly, more information about behavior and population size of the species could be obtained from ongoing research.