|Year : 2019 | Volume
| Issue : 2 | Page : 127-133
Ecological niche modeling for the prediction of cutaneous leishmaniasis epidemiology in current and projected future in Adana, Turkey
Cukurova University, Karaisali Vocational School, Karaisali, Adana, Turkey
|Date of Submission||22-Mar-2018|
|Date of Acceptance||15-Nov-2018|
|Date of Web Publication||31-Jul-2019|
Cukurova University, Karaisali Vocational School, Karaisali, Adana
Source of Support: None, Conflict of Interest: None
Background & objectives: Cutaneous leishmaniasis (CL) is widespread in the tropical and subtropical regions of the world including, Tukey. Environmental determinants for the CL endemic areas in Turkey are relatively poorly understood. The aim of the present study was to develop a model based on ecological niche modeling (ENM) to predict the distribution of CL in endemic areas of Adana Province in Turkey.
Methods: The environmental data from different sources were extracted and information on 1831 native CL cases, obtained from the Provincial Health Directorate of Adana were recorded. The location information obtained from the Ministry of Health database were used for modeling the current probability of CL occurrence and predicting its future distribution using ENM analyses. ArcGIS and MaxEnt models were used to explore the ecological conditions of the disease.
Results: According to the MaxEnt model, the area under the curve (AUC) values for the current and projected future of CL were 0.868 and 0.867, respectively. The environmental variables, Bio1 (Annual mean temperature), Bio4 (Temperature seasonality) and DEM (Digital elevation model) were found to be associated with the presence of human cases of Leishmania infantum for both the time periods in the study area.
Interpretation & conclusion: The AUC curves and risk map generated by the ENM model indicate that the future status of CL is likely to be stable in the northern part of Adana, but the southern part will be affected by climate changes (change of temperature) with a large number of patient-reporting. The results of the study could be used as a reference for CL and vector control studies. The ENM could be useful for researchers in vector control studies and better understanding of the epidemiology of vector-borne diseases in a specific area.
Keywords: Adana; cutaneous leishmaniasis; ecological niche modeling; MaxEnt; Turkey
|How to cite this article:|
Artun O. Ecological niche modeling for the prediction of cutaneous leishmaniasis epidemiology in current and projected future in Adana, Turkey. J Vector Borne Dis 2019;56:127-33
|How to cite this URL:|
Artun O. Ecological niche modeling for the prediction of cutaneous leishmaniasis epidemiology in current and projected future in Adana, Turkey. J Vector Borne Dis [serial online] 2019 [cited 2020 Jul 6];56:127-33. Available from: http://www.jvbd.org/text.asp?2019/56/2/127/263726
| Introduction|| |
Leishmaniasis is a neglected tropical disease which is caused by intracellular Leishmania protozoan transmitted by infected Phlebotomine female sandflies. According to the World Health Organization (WHO), almost 12 million people from 98 countries worldwide are currently infected with leishmaniasis, while 350 million people are at risk. Ofthe two million new cases diagnosed every year, three-fourth are cutaneous leishmaniasis (CL) cases. Cutaneous and visceral leishmaniasis (VL), the most common forms of leishmaniasis in Turkey, are frequently reported from rural areas with low status socioeconomic level. These are transmitted by proven vector sandflies such as Phlebotomus tobbi, P. sergenti and P. similis. The cutaneous form is commonly seen in the southeast part of Turkey. It usually causes ulcers on the face, arms and legs. Although the ulcers heal spontaneously, they may cause serious disability and leave severe and permanently deforming scars,. The main area of sandfly distribution currently lies in the Mediterranean region of Turkey. Cutaneous leishmaniasis is a major public health problem for metropolitan municipalities, especially in Adana, located in the eastern Mediterranean region,.
Various environmental data obtained from a range of sources, along with the human infection cases are used for predicting the prevalence and distribution of vector- borne diseases, on the basis of geographical information systems (GIS) tool and other modeling softwares. They have been used in mapping and forecasting the epidemiology of prevalent diseases in a study area with high probability. Environmental variables like normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), digital elevation model (DEM), and emis- sivity (EMIS) have been used by many studies in Turkey to show the relationship between sandfly distribution and environmental factors effect,,,,.
Similarly, studies based on ecological niche modeling (ENM), conducted with inexpensive and powerful computers could be used for predictive modeling of various species, environmental requirements and geographic distributions. A study carried out in Mymensingh and Gazipur districts in Bangladesh has successfully used ENM modeling for determining the distribution of VL and its vector P. argentipes. In Tunisia, 19 bioclimatic variables were used to estimate CL geographical distribution and Leishmania major potential distribution.
Thus, the goal of the present study was to design a new prediction model for CL distribution for all the districts of Adana using ENM in terms of current and projected future (2070) bioclimatic conditions; and to determine the role of environmental factors affecting its distribution using MaxEnt modeling techniques.
| Material & Methods|| |
Adana is the fourth major city of Turkey, located in the East Mediterranean region, covering an area of 14,032 km2 and having a human population of 1.7 million [Figure 1]. It has 15 districts, 828 villages and 745 rural areas. Mediterranean climate and a dry-hot summer subtropical climate are common in Adana. The mean temperature between 1950 and 2014 was 17.5 °C (max 23.5 °C; min 12.1 °C).
The study was conducted in 2017 summer. The environmental data included three variables (altitude, slope, and aspect) derived from remotely sensed data, and 19 bioclimatic variables as shown in [Table 1]. The present (2017) and projected future (2070) bioclim variables were downloaded from WorldClim website, version 1.4. All the bioclimatic variables had a nominal resolution of approximately 1 km2. The current bioclimatic data were developed from monthly average climate data between 1950 and 2000. For 2070 prediction, downscaled and calibrated global climate model (GCM) data, ‘Had- GEM2-ES’, representative concentration pathway 60 (rcp60) were used. The current and projected (2070) bio- climatic variables included the climatic conditions such as temperature, isothermality, and annual precipitation.
Human CL cases
Earlier studies have shown that CL is an endemic disease in Adana and 1831 cases have been reported between 2008 and 2016. Various data of CL cases which are necessary to developed ENM, were obtained from the Ministry of Health, Republic of Turkey through bilateral agreement. The CL cases were reported from all the districts of Adana, except Tufanbeyli (northeast) throughout the nine consecutive years. In addition, Kozan, located in the southern part of the study area, had higher CL cases in related years [Table 2]. The spatially unique human CL cases were used as presence data to develop an ENM.
Ecological niche modeling
For constructing the ENM, the distribution of cases was used/analysed, as if it was a vector sandfly species. The distribution of sandflies could be well-determined by the same environmental variables used to model the distribution of the CL. For generating this, maximum entropy (MaxEnt) model (a robust model based on a maximum entropy algorithm) was applied using the MaxEnt v3.3.3, a freely downloadable software (http://www.cs.princeton. edu/~schapire/maxent/). A file was prepared for the proven vector species P. tobbi with several environmental variables in ASCII format and was entered into MaxEnt software. Of the 1831 CL presence cases (location), 1373 (75%) were included (as training data) for constructing the model, and the remaining 25% data were used in testing the model (validation). The software was used with its default parameters with 10,000 as the maximum number of background absences, 0.00001 convergent thresholds, 15 replicates and 5000 as the maximum numbers of iterations and a logistic output presenting a continuous presence probability ranging from 0 to 1. A Jackknife analysis was performed to calculate the contribution of each variable in the modeling process and to determine the most effective factors in the distribution of CL cases in current and future. The area under the curve (AUC) and receiver operating characteristic curve (ROC) were determined for the model. Furthermore, a probability threshold representing the 10th percentile training presence points was selected as a cut-off probability, used to convert continuous probability maps into binary maps for the present and future scenario (2070). Jackknife test plays an important role in the calculation of a substantial number of contributing variables in the generated model. The three most effective contributing variables (with highest percentage) helpful in better understanding of environmental requirements of CL, were separated from the other variables. For the model prediction in current and projected future (2070), the AUCs were categorized as high predictive power (AUC >0.5), random chance (AUC = 0.5) and worse than random (AUC <0.5).
| Results|| |
A total of 1831 CL patients from 15 districts of Adana, reported between 2008 and 2016, were included in the current study. Running the models 25 times, current and projected future (2070) AUC values were calculated as training data: 0.868, test data: 0.882; and training data: 0.867; test data: 0.884, respectively [Figure 2]a and [Figure 2]b.
|Figure 2: AUC values of MaxEnt model for CL cases in Adana, Turkey: (a) current (2017); and (b) projected future (2070).|
Click here to view
The ENMs predicting the possible distribution of presence of CL cases for the current and future (2070) are shown in [Figure 3]a and [Figure 3]b, respectively. The maps predict that, CL cases distribution (shown in yellow and red) will expand to the central and northeastern parts of Adana. Cutaneous leishmaniasis presence probability foci will be wider in the future than today, especially for Karaisali, Saricam, Kozan and Imamoglu districts. The generated map shows that, the more the number of CL patients in southeastern part of the city the more likely light blue color will turn to a green color. Further, the prediction map shows that CL cases will not be reported from the northern part of the Adana in the future (2070) [Figure 2]b.
|Figure 3: The prediction map for the presence of CL cases: (a) current (2017); and (b) projected future (2070). Blue areas show low probability of occurrence, and red areas show high probability of occurrence.|
Click here to view
The results of the Jackknife analysis performed with 19 variables describing climatologies, and three geographic variables [Table 1] are shown in [Figure 4]. The cross-comparison of 22 data groups in MaxENT software, revealed that, only six variables [Bio1 (Annual mean temperature) current, Bio4 (Temperature seasonality) current, DEM current, Bio1–2070, Bio4–2070 and DEM–2070] were effective and have contributed most to the model development. Though, Bio1, Bio4 and DEM were significantly associated with the presence of human cases of L. infantum during both the time periods. The total accuracy (training gain) of both the models in presence of all the variables was approximately 0.92 [Figure 4]. [Figure 5] displays the six influential variables observed in the present study. The occurrence of CL cases was inversely related to Bio1, directly proportional to Bio4 and DEM for the present predication, whereas in the projected future (2070), these have an inverse relationship with the appearance of CL cases [Figure 5].
|Figure 4: Results of Jackknife test on MaxEnt model used to determine the effective variables on the distribution of CL cases: (a) current (2017); and (b) projected future (2070).|
Click here to view
|Figure 5: Response of CL cases to significantly associated variables in context of (A) current situation (2017); and (B) projected future (2070): (a) Annual mean temperature; (b) Temperature seasonality; and (c) Digital elevation model.|
Click here to view
| Discussion|| |
Computer software packages, such as genetic algorithm for rule-set production (GARP), ArcGIS and Max- Ent are commonly used for geographical estimation of the data obtained from field works or other sources for analyzing and predicting species distributions, mapping, disease prevalence etc. Earlier studies have reported that the distribution pattern of the vector arthropods is of great importance for better understanding of the epidemiology of vector-borne diseases.
The MaxEnt model has been employed for generating ENM in several studies, for the prediction of disease epidemiology and the possibility of future distribution of disease/vector. However, ArcGIS is essential in order to generate and convert data for usage in MaxEnt software to produce ENM. The ENM in the present study supported the findings of earlier predictions by accurately estimating the occurrence of CL in present situation and projected future (2070). Reliability of prediction depends on the AUC values obtained, i.e. (AUC >0.5) indicates higher predictive power; (AUC = 0.5) indicates random chance; and (AUC <0.5) indicates worse predication. Both the AUC values (present and future) of this study, were above 0.5.
Among the diseases considered to be important by WHO, CL has been the topic of various multidisciplinary international studies worldwide. An average of10 to 13% of CL cases are reported every year in Adana, which is a CL endemic area. It also has a higher CL transmission risk than central Anatolia cities,,. Thus, the determination of possible future emergence of CL, shed light into a better prediction of the distribution of vector sandfly species such as P. tobbi, which has been proved as a specific vector of L. infantum in 2009. Studies based on ENM have also been carried out in South America, Middle East, Middle Asia and north Africa.
A total of 1831 patients reported between 2008 and 2016 were included in this study while another similar study carried out in Bangladesh included 3671 patients. The disease predication and AUC values of current (0.868), and future (0.867) were similar (0.842) to that calculated in Bangladesh study. However, the influential variables were different which included, LST (Land use/ land cover category), NDVI, Precipitation seasonality (Bio15), Precipitation of the warmest quarter (Bio18), Drainage and general soil type (GST).
Similarly, a comprehensive study carried out in Tunisia, using 24 environmental layers focusing on both the CL cases and L. major (specific vector of P. papatasi) for the prediction of the geographic distribution of CL, observed that the compound topographic index and land cover were most influential variables which were different from the present study. However, similar to this study, elevation was significantly associated with the presence of human cases of CL.
Another study in Afghanistan, where 148,945 new cases of leishmaniasis were recorded from 20 provinces between 2003 and 2009, used MaxEnt model to construct the ENM of leishmaniasis. The mean AUC value of the training model was 0.929, and that of the testing model was 0.756, which was an indication of a good performance.
Similar study conducted in Hormozgan Province of Iran in 2017 used altitude and 19 bioclimatic factors in the ENM model. During 2005–2015, a total of 2531 CL cases were reported from the study area. The AUC for P. papa-tasi and P. sergenti was 0.870 and 0.886, respectively. The high AUC values (>0.5) for both dominant vector sandfly species indicate that the risk of CL disease is high in their study areas. In the present study, the AUC values of vector sandflies could not be calculated. Risk maps were created only by considering the distribution of CL patients in the study region,.
| Conclusion|| |
The risk map generated by the ENM model along with the AUC curves indicates that the future status of CL is likely to be stable (as evident by the blue color in Results) in the northern part of Adana, but the southern part will be affected by climate changes (change of temperature) with a large number of patient reporting. The information about the environmental variables would be sufficient in predicting distribution of the disease and the vector sandfly species in present time and future. The findings of the study could be used as a reference for CL and vector control studies/strategies of Adana Provincial Health Directorate in overcoming CL which particularly affects the people living in rural areas.
| Acknowledgements|| |
The author is grateful to Dr Hakan Kavur from Cuku-rova University, Adana, Turkey, for assistance in obtaining the CL cases locations and interpretation of the results.
| References|| |
WHO. Control of the Leishmaniases. Geneva: World Health Organization 2010; p. 186.
Ok Z, Balcioglu IC, Ozkan AT, Ozensoy S, Ozbel Y. Leishmaniasis in Turkey. Acta Tropica 2002; 84(1): 43–8.
Simsek FM, Alten B, Caglar SS, Ozbel Y, Aytekin AM, Kaynas S, et al
. Distribution and altitudinal structuring of Phlebotomine sand flies (Diptera:Psychodidae) in southern Anatolia, Turkey: Their relation to human cutaneous leishmaniasis. J Vector Ecol
Kavur H, Eroglu F, Evyapan G, Demirkazik M, Alptekin D, Koltas IS. Entomological survey for sand fly fauna in Imamoglu Province (Cutaneous leishmaniasis endemic region) of Adana, Turkey. J Med Entomol
2015; 52(5): 813-8.
Koch LK, Kochmann J, Klimpel S, Cunze S. Modeling the climatic suitability of leishmaniasis vector species in Europe. Nature
2017; 7(13325): 1-10.
Ostfeld RS, Glass GE, Keesing F. Spatial epidemiology: An emerging (or re-emerging) discipline. Trends Ecol Evol
Gonzalez C, Wang O, Strutz SE, Gonzalez-Salazar C, Cordero VS, Sarkar, S. Climate change and risk of leishmaniasis in north America: Predictions from ecological niche models of vector and reservoir species. PLoS Negl Trop Dis
2010; 4(1): e585.
Ozbel Y, Balcioglu IC, Olgen MK, Simsek FM, Ozensoy Toz S, Ertabaklar H, et al
. Spatial distribution of Phlebotomine sandflies in the Aydin mountains and surroundings: The main focus of cutaneous leishmaniasis in western Turkey. J Vector Ecol
2011; 36(Suppl 1): S99-105.
Olgen MK, Ozbel Y, Balcioglu IC, Demir S, Simsek F, Ozensoy Toz S, et al
. A new approach for determining the spatial risk levels for visceral and cutaneous leishmaniasis related with the distribution of vector species in western part of Turkey using geographical information systems and remote sensing. Kafkas Univ Vet Fak Derg
2012; 18(Suppl A): A77-84.
Kavur H, Artun O. Geographical information systems in determination of cutaneous leishmaniasis spatial risk level based on distribution of vector species in Imamoglu Province, Adana. J Med Entomol
2017; 54(5): 1175-82.
Artun O, Kavur H. Investigation of the spatial distribution of sandfly species and cutaneous leishmaniasis risk factors by using geographical information system technologies in Karaisali district of Adana province, Turkey. J Vector Borne Dis
2017; 54(3): 233-9.
Kavur H, Artun O, Evyapan G, Demirkazik M, Alptekin D, Koltas IS. Determination of sandfly fauna and environmental parameters in a cutaneous leishmaniasis Endemic Region in Karaisali, Adana, Turkey. Cukurova Med J
2018; 43(2): 1-8.
Phillips S J, Anderson RP, Schapired RE. Maximum entropy modeling of species geographic distributions. Ecological Modelling
Abdullah AY, Dewan A, Shogib RI, Rahman M, Hossain F. Environmental factors associated with the distribution of visceral leishmaniasis in endemic areas of Bangladesh: Modeling the ecological niche. Trop Med Int Health
2017; 45(13): 1-15.
Chalghaf B, Chlif S, Mayala B, Ghawar W, Bettaieb J, Harrabi M, et al
. Ecological Niche Modeling for the prediction of the geographic distribution of cutaneous leishmaniasis in Tunisia. Am J Trop Med Hyg
2016; 94(4): 844-51.
USGS-EarthExplorer Home. Reston, United States: The United States Geological Survey. Available from: https://earthexplorer. usgs.gov/
(Accessed on 10 August 2017).
Turkish Ministry of Health. Leishmaniasis data of Adana province, Turkey. Ankara, Turkey: Ministry of Health 2017; p. 20.
Sofizadeh A, Rassi Y, Vatandoost H, Hanafi-Bojd AA, Mollalo A, Rafizadeh S, et al
. Predicting the distribution of Phlebotomus papatasi
(Diptera: Psychodidae), the primary vector of zoonotic cutaneous leishmaniasis, in Golestan province of Iran using ecological niche modeling: Comparison of MaxEnt and GARP models. J Med Entomol
Svobodova M, Alten B, Zidkova L, Dvorak V, Hlavackova J, Myskova J. Cutaneous leishmaniasis caused by Leishmania infantum
transmitted by Phlebotomus tobbi. Int J Parasitol
Adegboye AO, Adegboye M. Spatially correlated time series and ecological niche analysis of cutaneous leishmaniasis in Afghanistan. Int J Environ Res Public Health
2017; 14(309): 1-14.
Bojd AAH, Khoobdel M, Ahmadi MS, Azizi K, Afshar AA, Jaberhashemi SA, et al
. Species composition of sandflies (Diptera: Psychodidae) and modeling the spatial distribution of main vectors of cutaneous leishmaniasis in Hor- mozgan Province, southern Iran. J Med Entomol
2018; 55(2): 292-9.
[Figure 1], [Figure 2], [Figure 3], [Figure 4], [Figure 5]
[Table 1], [Table 2]