Abstract
Henipavirus infection (NiV), a fatal disease transmitted by bats has caused human death and disease in India and Bangladesh. With an average case fatality rate of more than 70%, NiV has the potential to cause widespread outbreaks that can have a devastating impact on human health. Spillover of NiV from bats to humans is influenced by several factors mainly human behaviour and human interaction with bats. Climatic conditions and habitat destruction influence the shedding of the virus and the transmission of the disease to humans. Limited research has been conducted on the factors driving these spillover events. As the world has recently witnessed the devastating impact of the COVID-19 global pandemic, a forewarning of spillover events will enable the adoption of adequate measures to prevent and control future outbreaks. Our study maps the likely changing distribution of NiV reservoir Pteropus medius bat species using future climate and land-use change predictions. We use existing data to predict which districts within India and Bangladesh have increased risk of spillover of NiV in the future. We use Species Distribution Modelling to predict the likely change in the distribution of P. medius under different future scenarios and the concurrent increase in the risk of NiV spillover events. We focus on the influence of anthropogenic changes in different socioeconomic scenarios to predict future spillover events. We consider climate and social goals set by the International Panel on Climate Change (IPPC) to create accurate predictions of regions at risk of NiV spillovers. We find that the risk of NiV spillover events in India and Bangladesh will likely increase. More districts are predicted to be at risk of NiV spillovers under the high population growth and persistent environmental degradation scenario than under moderate population growth and medium challenges to achieve climate goals. This highlights the significance of population growth and climate change when considering disease outbreaks and public health. Our findings will enable the authorities in the predicted spillover regions to take public health measures to prevent and control NiV outbreaks.
Introduction
Bats have become infamous as reservoirs of viral diseases [1]. Amidst the COVID-19 pandemic, concerns of another fatal disease, the Nipah virus (NiV) infection has taken a backseat. The causation of NiV is the Henipavirus of the family Paramyxoviridae [2]. NiV, with a case fatality rate of >70%, can cause respiratory infection and encephalitis with persistent neurological defects in humans [3]. Since no definitive preventative nor treatment protocols have been set for this disease, NiV has been listed as a Priority Disease for Research and Development for diagnosis, prevention, and treatment by the World Health Organisation (WHO) [4]. The Coalition for Epidemic Preparedness Innovations (CEPI) has also prioritized NiV for vaccine development [5]. However, our understanding of the dynamics of NiV transmission and its pathology remains vague, as the conditions leading to its transmission have not been consistent across spillover events in different regions.
The first human case of NiV was reported in 1998 in Malaysia. The disease was transmitted zoonotically from pigs to farmers who were in direct contact with the affected animals. Thus, pigs acted as an intermediate host in the Malaysian outbreak. The disease spread to Singapore through the import of these infected pigs [6]. In 2001, a second NiV outbreak was reported in Bangladesh. Since 2001, outbreaks have occurred annually in Bangladesh and in West Bengal, India [7]. The route of transmission in Bangladesh was found to be through the consumption of date palm sap contaminated with bat urine or saliva. Bats consume date palm sap from collection pots and shed the virus as they lick the sap, or by urinating in the pots [7]. People involved in the practice of date palm sap consumption have been found to be more likely to become infected with NiV [7]. In 2018, a NiV outbreak occurred in Kerala, in the South of India, with a severe case fatality rate of 91% [8]. Very recently in September 2021, NiV caused the death of a 12-year-old in Kerala [9].
The main difference between the Malaysian and Bangladesh/Indian outbreaks, was the route of transmission. While the Malaysian outbreak involved an intermediate host, NiV outbreaks in Bangladesh and India was mainly through consumption of date palm sap contaminated with bat urine or saliva. Studies have shown that the fruit bat Pteropus medius is the main reservoir of the virus in India and Bangladesh [2]. P. medius has been confirmed to carry NiV since the 1950s, posing a risk of spillover to humans [10]. Either spillovers prior to 2001 have passed undetected, or a previously absent combination of factors could have driven the first spillover event in 2001 and subsequest spillovers.
NiV is an RNA virus that can mutate rapidly to adapt to changing conditions. This enables the virus to overcome barriers to its transmission and infectivity [11]. NiV can mutate to infect new host species, thereby increasing the chance of spillover through different domesticated species [12]. Antibodies against NiV have been found in cattle, goats, and pigs in Bangladesh [13]. Even though no zoonotic transmission of the disease through livestock has yet been reported in Bangladesh and India, the detected antibodies suggest these domestic species are susceptible to NiV. Mutations can then enable the transmission of the virus to humans. Mutations can also enhance the transmissibility of the virus between humans [11]. Until now, sustained human-to-human transmission of NiV has not been recorded as the virus has not been passed on for more than five generations in the transmission chain [14]. Transmissibility, indicated by the Basic Reproductive number (R0), takes into account the number of susceptible people the infection is likely to be passed on to by a single infected person. Diseases with R0>1 have the ability to cause pandemics as they can cause sustained chains of transmission. NiV with a R0 of 0.48 has not yet caused large scale outbreaks [15]. However, with the possibility of mutating to a form with R0>1, the threat of a Nipah virus pandemic originating from the densely populated and globally connected regions of Asia seems very plausible [16]. As the life span of the virus is threatened by unfavourable environmental conditions such as high temperature, mutations that enhance the survival of the virus in unfavourable conditions also increase the chance of spillovers. The provision of an energy-rich medium such as date palm sap may enhance the survival of the virus as it provides cooler temperatures and moisture [17].
Viral shedding by bats has been found to be variable, as shedding occurs more extensively during periods of decreased immunity often instigated by stressful conditions such as unfavourable climatic conditions or suboptimal nutrition [18]. Spillover events in Bangladesh have been mainly recorded in winter. This could be due to the combination of increased foraging and thermoregulatory stress in winter and the practice of mainly consuming date palm sap during cold months [19,20]. Climate change and its impact on different host species distribution and diseases has been explored previously, with a consensus that a changing climate will lead to shifting distribution ranges and in turn introduce diseases carried by these species to new areas [21]. In addition, deforestation, habitat destruction, and urbanisation impact host species distribution and disease occurrence [22]. Destruction of wild habitat has increased human-wildlife contact, creating more opportunities for spillovers to occur [23]. As wild habitats and forests are being destroyed, some species of bats seek human-dominated areas for roosting and foraging. They roost near human dwellings and exploit alternate resources provided by people [2]. Date palm sap is one such alternate resource. Fermented date palm sap is also consumed by humans in certain areas [3]. Its alcohol content of around 4%, is insufficient to kill the virus which requires 60-70% alcohol concentration to lose its viability [3,24]. Previous studies have established human activities and behaviour as the main drivers of spillover events [19,25]. Mere presence of NiV circulating within the bat population cannot lead to spillovers. Anthropogenic changes that create opportunities for human-wildlife interaction, seem to be the prime drivers of spillover, possibly compounded by factors that influence viral shedding in host wild animals [19].
Another example of human behaviour driving spillover events is the persecution of bats and the destruction of their roosts following NiV outbreaks [26]. Disturbed bats carry the virus to new areas as they disperse to find new roosts causing the further spread of the disease [27].
Bat roosts are being increasingly found in densely human-populated areas in Thailand, Fiji, Pakistan, and Malaysia [28–31]. Bats’ ability to adapt to changing conditions caused by humans may create potential for new spillover points that are yet to be identified. Studies have been conducted exploring the impact of changing climate and predicting spillover events of NiV and Hendra virus, a related Paramyxovirus [32–34]. However, the impact of anthropogenic changes has not been adequately considered in predicting future outbreaks. Our study aims to predict areas susceptible to spillovers in the future, by considering likely land-use changes and changing climatic conditions under four Shared Socioeconomic Pathways (SSPs). These SSPs consider greenhouse gas emissions and radiative forcing under different socioeconomic scenarios [35] that are influenced by population growth, urbanisation, economic growth, and the development of green technology. SSP1, the lowest emission scenario, follows the path of sustainability, with few challenges to achieving climate and socioeconomic goals. SSP2 is a ‘middle of the road’ scenario with moderate population growth and medium challenges to achieve climate goals. SSP3 predicts high population growth in developing countries and persistent environmental degradation with strong challenges to achieving climate goals. SSP4, the ‘worst-case scenario’ is characterised by high social disparity, poor technological development, and failure to achieve climate and socioeconomic goals. We use Species Distribution Modelling to predict the distribution of P. medius under different socioeconomic scenarios with changing climate and human land-use changes in the current and future scenarios. As we assume the SSP2 and SSP3 scenarios are more likely to occur due to moderate challenges to policy adoption in the case of SSP2 and minimal efforts to adhere to climate policies in the case of SSP3 [35], the results of only these two scenarios are presented and discussed in the main text of this study. The predictions for the SSP1 and SSP4 scenarios have been included in the supporting information. We hypothesize an expansion in the range of P. medius and a northward shift in accordance with the changing climate. The distribution of P. medius is used as a proxy to identify new areas at risk of NiV spillover events.
Methods
Data collection
We collected spatial layers from open access sources and databases. Current bat occurrence data across India and Bangladesh was collected from the Global Biodiversity Information Facility (GBIF) and the India Biodiversity Portal [36–39]. Data was also extracted from literature with records of P. medius observations. The terms “Pteropus medius” and “Indian Flying Fox” were used as search terms in Google Scholar and Web of Science to extract papers with occurrence records of the species. We filtered the results to include only articles focusing on bat populations in India or Bangladesh [40–46]. Only bat occurrences recorded since 2015 were included in this study to generate a current distribution model of the species.
Spatial data for 130 layers was initially collected from several databases for current and future scenarios. The future data for four Shared Socioeconomic Pathways were included in the study. Layers include climatic variables from WorldClim and Envirem, 2010 livestock distribution data from the Food and Agriculture Organisation (FAO), vegetation data from National Centre for Neoenvironmental Information, land-use data from Oak Ridge National Laboratory (ORNL) Distributed Active Archive Center (DAAC), data on human modification of terrestrial systems from NASA Socioeconomic Data and Applications Center (SEDAC), and human population density data from Landscan. Land-use changes influence the level of contact between bats and humans and contribute to spillovers. Data on human population density and land-use changes were collected as they were expected to contribute the most in determining future distribution of P. medius. Climate data and livestock density were included to determine their influence on the P. medius distribution.
Data processing
We examined the occurrence data for any duplicate records across different databases and extracted the corresponding coordinates in decimal form. Spatial layers were processed in Quantum GIS (QGIS version 3.18) [47]. Layers were clipped to the combined geographic extent of India and Bangladesh. A uniform pixel size of 2.5 arc minutes was set for all the layers. The aligned layers were saved as ASCII files.
Statistical analysis of variables
A covariance report was generated for all the current and future variables separately using the covar function in QGIS. The main variables were extracted using a Python script[48] Variables with covariance coefficient of >0.7 [49] were considered to be highly correlated to at least one other variable and excluded from the models. A set of spatial variables each for the current and future scenarios was determined (Tables 1 & 2).
Species Distribution Modelling
We used MaxEnt [50] to create Species Distribution Models (SDMs) using presence-only data. MaxEnt uses maximum entropy machine learning to create SDMs with information provided by spatial layers. We used 25% of our data set as training data. Regularization multiplier was set to 3. We performed Jackknife tests to measure the information contributed by each variable. Based on the individual contributions of each variable, their response curves, and jackknife results, the number of layers was further reduced to form the final set of variables (Table 1 & 2). Species Distribution Models for current and future scenarios were generated. Future projections were made for four SSPs for the years 2040, 2060, 2080 and 2100, and for two General Circulation Models (GCMs)-BCC-CSM2-MR and MIROC-ES2L as these show high performance for precipitation and temperature predictions, respectively [51]. We made predictions for two models for comparison to check the accuracy of predictions across different models.
Quantifying model performance
Statistical analysis of the distribution maps was performed in MaxEnt [50]. The Area under Curve (AUC) of the Receiver Operating Characteristic Curve (ROC) was used to measure model performance. AUC values between 0.7 - 0.9 indicate good to very good performance [64]. The p value of the 10th Percentile training presence data was used to determine the statistical significance of each model.
Generating categorical maps
We created categorical distribution maps using district boundaries within India and Bangladesh. We inferred the risk of spillover events from the probability of presence of P. medius within a district. For example, a district with a very high probability of P. medius presence according to the future SDM was categorised as ‘very high risk’ of NiV spillover. Categories include ‘very high risk (0.75-1)’, ‘high risk (0.5-0.75)’, ‘moderate risk (0.25-0.5)’, and ‘low risk (0-0.25)’. Each individual raster map generated with MaxEnt was added to a district level vector polygon map of India and Bangladesh, where each polygon represents a district. A graduated symbology was applied to the vector files to infer and categorise the risk of spillover in each district based on the probability of occurrence of P. medius.
Results
Current distribution
The model shows the current distribution of P. medius has an AUC of 0.829 with a standard deviation of 0.025 and p<0.01. Population density contributed the most information (53.8%) followed by continentality (12.7%) and annual PET (12.2%). Other variables have a contribution of less than 10%. Jackknife tests showed human modification of land has importance (Fig. S2-a) in addition to population density, annual PET and continentality.
The model predicts a wide range of distribution for P. medius across the South, the East and West Coast, and the Northern states of Haryana, Uttar Pradesh, and Bihar (Fig. 1). Central and North-Eastern regions are also predicted to be suitable. The model also predicts the species to occur across the Eastern and Central regions of Bangladesh.
The scale ranges from low suitability areas (blue) to high suitability areas (red). The likely current range of the P. medius extends along East and West coast, and the Central and Northern region of India and Eastern and Central regions of Bangladesh.
Human modification of land (Fig. S2-a) seems to increase the suitability for the species whereas the effect of population density (Fig S2-b) stabilizes after a rapid rise. The distribution also seems to increase initially with an increase in annual PET (Fig S2-c), peaks and then decreases. Increasing continentality (Fig S2-d) decreases the distribution of the species.
Future distribution
The results of the future projections have an AUC of 0.808 indicating the soundness of the model. The p value for the 10th percentile training presence is <0.01. The highest contributor is the urban land layer with a contribution of 33.7%, followed by temperature seasonality (13.8%), and pastureland (11.5%). Other variables have a contribution of less than 10%. Jackknife analysis shows that urban land cover is the most important variable with model performance decreasing considerably when the layer is removed (Fig S1-b). Models for the two GCMs show mostly similar predictions. Across all four scenarios, the future projections predict a Northward shift from the Southern states where the species is predicted to be more prevalent according to the current model.
Increase in urban land cover (Fig. 4-a) increases the suitability of the land for the species whereas increase in pastureland (Fig. 4-b) decreases suitability of the land. Increasing annual mean temperature (Fig. 4-c) up to 25°C seems favourable for the bat species, with a further increase in temperature limiting its distribution.
Similar to continentality (Fig. 4-d), an inverse relation seems to exist with temperature seasonality.
Shared Socioeconomic Pathway 2
The predictions for the Shared Socioeconomic Pathway 2 scenario show high suitability of the Western Coast and the Northern states in 2040 and 2060 (Fig. 2). The 2080 and 2100 predictions show an expansion of the range with the Central, Western, and North-Eastern states becoming more suitable for the species. It also shows a shifting of range towards the Eastern boundary of the Northmost state of Jammu and Kashmir.
(Middle of the road scenario with moderate population growth and medium challenges to achieve climate goals) [35]; Future projections for Pteropus medius presence for the years a) 2040; b) 2060; c) 2080; and d) 2100 for the two Global Circulation Models BCC-CSM2-MR and MIROC-ES2L are presented, along with the map depicting the present distribution of P. medius. Risk of spillover events was inferred through the probability of presence of P. medius within a district. Categories include ‘very high risk (0.75-1)’, ‘high risk (0.5-0.75)’, ‘moderate risk (0.25-0.5)’, and ‘low risk (0-0.25)’. According to the predictions, an expansion and Northward shift of range can be observed as we proceed to the end of the century.
Shared Socioeconomic Pathway 3
A considerable increase in the suitability of the North-Eastern states of India and the eastern regions of Bangladesh can be predicted under the SSP3 scenario across 2040, 2060, and 2080 (Fig. 3). A shift in the range towards the Northern states of Himachal Pradesh and Uttarakhand in India can also be observed. However, the projections for 2100 show a significant decrease in the range of the species with several areas in the Central, Western, and North-Eastern parts becoming unsuitable.
(High population growth in developing countries and persistent environmental degradation with strong challenges to achieving climate goals)[35]; Future projections for Pteropus medius presence for the years a) 2040; b) 2060; c) 2080; and d) 2100 for the two Global Circulation Models BCC-CSM2-MR and MIROC-ES2L are presented along with the map depicting the present distribution of P. medius. Risk of spillover events was inferred through the probability of presence of P. medius within a district. Categories include ‘very high risk (0.75-1)’, ‘high risk (0.5-0.75)’, ‘moderate risk (0.25-0.5)’, and ‘low risk (0-0.25)’. According to the predictions, range expansion is noticed till 2080 after which there is a decrease in the range of P. medius.
a. Urban land response curve (g_22); b. Pastureland response curve (g_17); c. Temperature seasonality response curve (bioc_04); d. Annual mean temperature response curve (bioc_01). Suitability of habitat increases with increasing urbanisation and decreasing pastureland, annual mean temperature, and seasonality of temperature.
Discussion
Current distribution
The current model (Fig. 1) shows an extensive range for P. medius across the East and West coast of India, and in the Northern and Central states. In the predicted current distribution, population density, continentality, and annual potential evapotranspiration contribute the most with other potentially relevant variables contributing little.
Species distribution and land suitability increases with increasing human population density indicating the suitability of human-occupied areas for P. medius as they gain access to alternate resources [2]. This also highlights an increased chance of spillover as human-bat contact increases [23]. The human modification of land resulting in the conversion of wild habitats to human-dominated lands, similarly, creates favourable conditions for the bat species (2). This is in agreement with the previous research that P. medius were more likely to roost in fragmented forests near areas with high human population density [65].
Increasing continentality, a measure of the difference between the extremes in land temperatures influenced by moderating effects of the ocean, causes a decrease in habitat suitability. Areas that are farther away from the ocean experience extreme climate as the moderating effect of the ocean is reduced. Extreme temperatures in inland areas may create unfavourable conditions for the bat species and their preferred vegetation. Increasing annual potential evapotranspiration, if coupled with decreased precipitation, may affect vegetation as the soil becomes unsuitable [66]. This decreases the suitability of the area as these bats prefer specific trees for roosting and feeding [65].
Future distribution
Our findings generally predict an expansion in the range of P. medius and a shift towards the North (Fig. 2 & 3). The expansion or shift in the range of P. medius could cause NiV spillover in these new areas. Over time, species distribution is predicted to decrease in the high emission scenarios (Fig. 3). This highlights the adverse effects of anthropogenic and climate change on the host species distribution.
The response curves for the future projections show an increase in the suitability of the area for P. medius with increasing urbanisation (Fig. 4). This implies that the species benefits from urban landscapes as they are provided with alternate resources. This will increase the chances of spillover events as human-bat contact increases [23].
An increase in pastureland (Fig. 4-b) likely results in cutting down of trees that serve as roosting sites. Habitat degradation will lead to further spillover events as human-bat contact increases [23].
The inverse relationship of temperature seasonality (Fig. 4-c) may be an indication of the pecies’ preference for stable temperatures. The suitability for annual mean temperature (Fig. 4-d) declines after peaking at 24-25°C indicating that as the temperature rises with climate change, habitat suitability will decrease. This stress could culminate in increased viral shedding and possible spillover events facilitated by the increased contact between bats and humans as a result of habitat degradation and provision of alternate resources [67].
SSP2
With sustainable development, moderate population growth, and global warming of not more than 3°C by the end of the century, the models predict an expansion of P. medius range (Fig. 2). As the climate warms slowly and sustainable practices decrease human modification of habitat, the species adapts and expands its range northward. Even as the distribution expands, if habitat-degrading practises are minimised and spillover routes such as through date palm sap and zoonotic transmission are controlled [7], the SSP2 scenario may prove to be the most beneficial in terms of conservation concerns and reducing the chances of spillover.
SSP3
Under the medium-high emission scenario, where environmental goals are not prioritised and population increases in developing nations such as India and Bangladesh, up until 2080, there is predicted to be an expansion of the range of the species with a shift towards the North (Fig. 3). This expansion coupled with stressing factors such as habitat degradation and increased human population could increase the chances of spillover in these regions as human contact with the bat species increases. This warrants the application of control measures in these areas to prevent spillovers and to limit human-to-human transmission of NiV. As we move towards 2100 under this scenario a reduction in the suitable areas can be noticed as the high temperatures become unsuitable for the bats. The northward shift is still prominent as the relatively colder climate warms up and becomes suitable to P. medius populations. The decrease in the range towards the South and North-East might be of conservation concern and the bats may undergo a population decline. The few areas where their population is still dense may witness spillover events as viral shedding may increase due to unsuitable temperatures and human activities [67].
The regions at risk of NiV spillover should be targeted to prevent and control outbreaks. Public education regarding NiV, and its transmission routes should be of prime concern as human behaviour seems to be a major driving factor of the spillovers [19]. Increasing surveillance in high-risk areas through combined efforts of the public health, veterinary and wildlife sectors will help in the early detection of outbreaks to implement adequate control measures.
Limitations
The accuracy of our predictions depends on the accuracy of the data used. As spillovers are dynamic events with several contributing factors, several unknown factors may influence spillover events that have not been accounted for in our study. Data for some known factors such as date palm consumption records are unavailable. A future study including such records can be conducted to provide better predictions of spillovers.
Conclusion
The effects of a pandemic can be devastating as observed with the current the COVID-19 outbreak. NiV, a fatal disease passed on by bats has the potential to cause such a pandemic by mutating to a form that is easily transmitted from bats to humans and within the human community. Our study predicts an increase in the risk of NiV spillovers throughout India and Bangladesh in the future under all socioeconomic scenarios. To minimize the risk of NiV spillover, following the SSP2 ‘middle of the road’ scenario would be practical. Efforts to control human population growth, minimize emissions, and ensure sustainable development to retard global warming will be elemental in preventing NiV spillover events. Stressing factors that increase viral shedding by bats such as anthropogenic modification of land and changing climate, and human behaviours that increase contact with wildlife enhance the chances of disease outbreaks in a future scenario where climate and social goals are not prioritized. Our study emphasizes the importance of adhering to the climate policies set by the IPCC by coordinated efforts by governments and inter-governmental organizations. Our findings can be practically applied by the adoption of public health measures in the areas predicted to be at risk of NiV spillover from P. medius to humans. This could minimise the probability of spillovers and maximise the control of future outbreaks. Governments and health facilities that have successfully dealt with the outbreaks should compile and pass on their collected knowledge to areas that will be at risk of NiV spillovers in the future. The World Health Organisation should gather information to provide an action plan that can be readily deployed to control outbreaks of NiV. Measures should be taken to inform the public about the disease and the various routes through which it spreads. Public education programmes discouraging consumption of raw and fermented date palm sap, and fruits discarded by bats should be undertaken. Further study needs to be conducted on the transmission of the virus and to identify potential spillover routes that can emerge in the future.
Data Availability
All data produced are available online at https://www.kaggle.com/varshabalachandra/risk-mapping-nipah-virus-spillover
https://www.kaggle.com/varshabalachandra/risk-mapping-nipah-virus-spillover
Acknowledgements
We are extremely grateful to Vishnu Hariharan Anand for his valuable support whilst conducting the statistical analysis.
References
