Projecting spatiotemporal dynamics of severe fever with thrombocytopenia syndrome in the mainland of China

Severe fever with thrombocytopenia syndrome (SFTS) is an emerging infectious disease with increasing incidence and geographic extent. The extent to which global climate change affects the incidence of SFTS disease remains obscure. We use an integrated multi‐model, multi‐scenario framework to assess the impact of global climate change on SFTS disease in China. The spatial distribution of habitat suitability for the tick Haemaphysalis longicornis was predicted by applying a boosted regression tree model under four alternative climate change scenarios (RCP2.6, RCP4.5, RCP6.0, and RCP8.5) for the periods 2030–2039, 2050–2059, and 2080–2089. We incorporate the SFTS cases in the mainland of China from 2010 to 2019 with environmental variables and the projected distribution of H. longicornis into a generalized additive model to explore the current and future spatiotemporal dynamics of SFTS. Our results demonstrate an expanded geographic distribution of H. longicornis toward Northern and Northwestern China, showing a more pronounced change under the RCP8.5 scenario. In contrast, the environmental suitability of H. longicornis is predicted to be reduced in Central and Eastern China. The SFTS incidence in three time periods (2030–2039, 2050–2059, and 2080–2089) is predicted to be increased as compared to the 2010s in the context of various RCPs. A heterogeneous trend across provinces, however, was observed, when an increased incidence in Liaoning and Shandong provinces, while decreased incidence in Henan province is predicted. Notably, we predict possible outbreaks in Xinjiang and Yunnan in the future, where only sporadic cases have been reported previously. These findings highlight the need for tick control and population awareness of SFTS in endemic regions, and enhanced monitoring in potential risk areas.


| INTRODUC TI ON
Tick-borne diseases constitute a major threat to human health and are rapidly becoming a global issue.Severe fever with thrombocytopenia syndrome (SFTS) is a tick-borne disease caused by a novel bunyavirus designated as SFTS virus (SFTSV; Yu et al., 2011), now renamed as Dabie bandavirus in the genus of Bandavirus, family Phenuiviridae.
Since its emergence nearly a decade ago, SFTS has affected a rapidly growing number of people, with consistently expanding geographic distribution, mostly in China, Korea, Japan, Vietnam, Myanmar, and other Asian countries (Rattanakomol et al., 2022;Takahashi et al., 2014;Tran et al., 2019;Win et al., 2020;Yu et al., 2011;Yun et al., 2014).With severe concern about high case fatality and wide geographic distribution, SFTS was listed among the top 10 priority infectious diseases by the World Health Organization in 2017 (Mehand et al., 2018).Despite these public health implications, approved vaccines or specific therapeutics against SFTS remain unavailable.
Although human-to-human transmission has been reported, tick bite is recognized as the primary infection route, with Haemaphysalis longicornis identified as the predominant tick vector via its competence of transstadial and transovarial transmission of SFTSV (Hu et al., 2020;Luo et al., 2015).Originally native to East Asia, H. longicornis had spread to Australia, New Zealand, and several Pacific Islands since 1983 and was recently found in the eastern United States (Wormser et al., 2020).Given the wide distribution and dispersal of the predominant tick vector, it is suggested that the toll and potential threat of SFTS might be underestimated (Luo et al., 2015).
The spatial-temporal distribution and spread of SFTSV depend on complex interactions between multiple ecological, environmental, and societal variables.As environmental variables may affect ticks, their hosts and habitat, ecological data can potentially be used as proxies for the tick habitat in predicting SFTS incidence.Miao et al. (2020) adopted a two-stage GBRT model to simulate the distribution of SFTS, indicating that altitude, the coverage of closed-canopy woodland, and precipitation in the driest season were important drivers for the presence of SFTS cases in China.Wu et al. (2020) performed a random forest analysis on meteorological factors, demonstrating a significant association between the incidence of SFTS and monthly mean pressure, mean temperature, mean relative humidity, mean 2-min wind speed, duration of sunshine, and precipitation in a complicated and nonlinear pattern.Du et al. (2014) revealed that the temperature, precipitation, land cover area, normalized difference vegetation index, and duration of sunshine were significantly related to the spatial distribution of SFTS.Another study in Jiangsu revealed that altitude, yearly average temperature, yearly accumulated precipitation, and yearly average relative humidity accounted for 94.1% of the total effect in the model (Sun et al., 2021).In one study that used the distribution of H. longicornis ticks as a predictive variable, the boosted regression tree (BRT) model based on the national surveillance data for SFTS had disclosed a significant association between human case incidence and temperature, rainfall, relative humidity, sunshine hours, elevation, cattle density, and coverage of forest, with most of the factors showing nonlinear relationships with the risk of SFTS (Liu et al., 2015).
Heterogeneity of significant risk factors was often observed across studies, however, in most studies, the contribution of the yearly average temperature to the spatial distribution of SFTS is independently significant, which was often higher than most other factors.The increasing suitability for SFTSV transmission owing to global warming merits attention from the health community.Understanding the mechanisms driving disease dynamics and estimating the effect of climate change can help to compose public health and disease control strategies.However, this research was highly hindered by the lack of surveillance data on H. longicornis population in China.
In this study, we adopt a four-step strategy to predict the future incidence of SFTS.Using the field surveillance data of H. longicornis ticks, we simulate the BRT models to investigate their environmental suitability.Then, the future distribution of H. longicornis is predicted based on the fitted ensemble BRT models.At the third step, we explore the association between the number of SFTS cases and nine explanatory variables (mainly involving climate, land cover, terrain, and environmental suitability for H.

| Occurrence records of H. longicornis
In our previous study, we have compiled a dataset of 7000 unique spatial records of 124 tick species reported between 1950 and 2018 (Zhao et al., 2021).In the current study, we refreshed this database by supplementing the recent 4-year data from 2019 to 2022 by using the same study strategy.Briefly, the peer-reviewed Chinese and English literature published in 2019-2022 were reviewed from major databases (PubMed, China National Knowledge Infrastructure, and China WanFang database) using the search terms 'Haemaphysalis longicornis' or 'H.longicornis' and 'China'.All records of tick occurrence were extracted, and for each record, a geo-positioning of the location was compiled.Finally, the dataset was re-evaluated to include historical changes in tick taxonomy and the validity of novel observations.

| Climate data
Historical climate data were obtained from the China Meteorological Data Service Center (http://data.cma.cn).On the one hand, based on monthly mean temperature, maximum temperature, minimum temperature, and mean precipitation, 19 ecoclimatic variables (Bio1-19) were calculated.Considering that excessive bioclimatic variables may have negative impacts on model analysis, we used a clustering method to select the highly representative ecoclimatic covariates for further analysis.The cluster analysis was performed by R package NbClust version 3.0.1 (Charrad et al., 2014).Briefly, the correlation coefficients among the 19 variables were calculated (Figure S1).The distance between two variables was assigned "0" if the absolute value of correlation coefficient <0.8 and "1" otherwise, forming a binary distance matrix.Using Krzanowski and Lai indices to select the best clusters, 19 ecoclimatic predictors were classified into eight clusters, and only one from each cluster was chosen to model the environmental suitability of H. longicornis.Finally, bioclimatic data including bio1, bio2, bio3, bio4, bio5, bio12, bio14, and bio15 of multi-year average  were used in H. longicornis modeling.On the other hand, precipitation, mean temperature, and relative humidity spanned from 2010 to 2019 on a month level were used in SFTS modeling.The detailed information of historical climate data can be found in Tables S1 and S2.

| Human population data
Gridded datasets for the human population under Shared Socioeconomic Pathways (SSPs) were obtained from Science Data Bank (https://www.scidb.cn/en),with a spatial resolution of 0.5° and a period of 2010-2100.The human population estimates under SSP2 for the 2010s, 2030s, 2050s, and 2080s are applied in this study.

| Land cover and terrain data
The current and projected land cover data from 2010 to 2080 with a spatial resolution of 1 × 1 km 2 under different RCPs were obtained online (http://data.ess.tsinghua.edu.cn/data/Simulation/).
The gridded terrain data were downloaded from the Data Center for Resources and Environmental Sciences, Chinese Academy of Sciences (RESDC; http://www.resdc.cn),with a spatial resolution of 1 × 1 km 2 .The proportion of cropland, shrubland, grassland, and forest in each county were used as predictive variables in the current model.It should be noted that the above spatial covariates from differing grid geometries were aggregated to county level using ArcGIS 10.6 software (ESRI Inc.).

| Modeling of H. longicornis distribution in the mainland of China
A case-control study design was used to BRT models to predict the distribution of H. longicornis at the county level.Briefly, counties with at least one record of H. longicornis occurrence were considered as "case", while those counties where tick surveillance was conducted but without reporting H. longicornis occurrence were considered as "control".Counties without tick survey were excluded from the BRT modeling but kept for risk mapping.To counterbalance the potential sampling bias of the survey counties, we constructed a logistic regression model for the selection of tick survey counties with total 14 variables (eight ecoclimate, five land cover, and one terrain variable) used as potential predictors.The outcome response of this model is valued "1" for all tick survey counties and "0" for non-survey counties.The predictors were selected using a backward procedure with a significance level of .05.The reciprocal of the predicted sampling probability for all surveyed counties was first rescaled to have a mean of one and then used as a weight in the BRT model for the H. longicornis distribution (Albery et al., 2020;Little & Rubin, 2014;Pandit et al., 2018).
BRT combines boosting and classification and regression tree by combining many decision trees to optimize the prediction performance of the model, and it allows nonlinear relationships between outcomes and covariates.The BRT model can be written in terms of M-trees added together: where i is the county, Y i , the outcome that whether county i was reported with at least one record of H. longicornis occurrence, and x i the vector of predictors for county i. M is the number of trees, h(x i ,γ m ) returns the terminal node of a tree defined by parameters γ m (representing both selected predictors and splitting values) for an input x i , and β m is the expansion coefficient.The process of solving γ m is the learning process of a single decision tree.
The R packages gbm version 2.1.8.1 (Greenwell et al., 2022) and dismo version 1.3-14 (Hijmans et al., 2023) were adopted to construct BRT models based on the R version 4.2.2 statistical programming environment (R Core Team, 2022).A tree complexity of five, a learning rate of .005,and a bagging fraction of 75% were used for the primary analysis as previously studied (Ma et al., 2022).A training set consists of 75% of the data points that are randomly sampled without replacement, and the remaining 25% is used as the testing set.A two-stage bootstrapping procedure is employed to provide a more robust and parsimonious estimate of the model parameters, which can be found elsewhere (Zhang et al., 2023).The output of a BRT model consists of both predicted probabilities of occurrence and relative contribution (RC) of each predictor (Elith et al., 2008).The RC of each predictor is expressed as a percentage, with all RCs summed to 100%.In the first stage, the split-and-fitting step is repeated 10 times to screen for important predictors.Predictors that had a RC ≥2% remained in the second stage, at which the split-and-fitting step is repeated for 100 times to increase the robustness of the modeling performance.In this study, area under the receiver operator characteristic curve (ROC-AUC; Ge et al., 2022) was adopted to assess the predictive power of the models.

| Modeling risk for SFTS incidence
The GAM that has been successfully used to simulate other vectorborne disease (i.e., scrub typhus; Ding et al., 2022) was performed to explore the associations between monthly the number of SFTS cases and nine explanatory variables.The explanatory variables included climate factors (mean temperature, relative humidity, and precipitation), land cover data (proportion of cropland, shrubland, grassland, forest), terrain data, and the environmental suitability of H. longicornis.The GAM model can be written as follows.
where i is the county, t the time (month), and f means a smoothing function.Y i , t , P i , t , T i , t , and RH i , t represent the number of SFTS cases, precipitation, mean temperature, and relative humidity at county i in month t, respectively.ES i represents the environmental suitability of H. longicornis at county i. Dem i denotes the average elevation of county i; Crop i , Forest i , Grassland i , and Shrub i are the proportion of cropland, forest, grassland, and shrub land cover type, respectively, at county i; a term of fixed effect (Area i ) was included to account for the effects of unknown or unavailable variables in the model, which is set at the city level.The log transform of human population size is used as an offset value; Month/ Year represents the time stratification effect; means the error term.
The following steps of data manipulation and analysis were employed to fit the model.

| The distribution and current environmental suitability of H. longicornis in China
We compiled a database encompassing the geo-referenced occurrence of H. longicornis ticks reported across the mainland of China between 1950 and 2022.Based on the database, H. longicornis was distributed in 490 of all 1249 counties surveyed for tick existence.We mapped the distribution of H. longicornis based on the year when it was first detected in each county (Figure 1a),   S2).Annual mean temperature and isothermality contributed with the greatest effect to the model (RC of 20.58% and 18.07%, respectively), followed by temperature seasonality (14.53%), total precipitation (11.27%), the proportion of forest (6.46%), and proportion of green space (6.3%; Figure S3).We thus predicted a wider geographic range suitable for the H. longicornis than already recorded (1496 predicted vs. 490 recorded counties; Figure 1b).

| Projected environmental suitability of H. longicornis under future climate and land cover changes
The environmental suitability of H. longicornis was mapped under four RCP scenarios from the 2030s to the 2080s based on ensembled BRT models, indicating a more suitable habitat for H. longicornis ticks in the context of global warming.By the 2080s, under the RCP2.6 scenario, the total suitable habitat for H. longicornis remained largely unchanged compared with that in the 2030s (Figure S4).In contrast, the suitable areas under RCP4.5, 6.0, and 8.5 were predicted to increase by 9.5%, 17.2%, and 29.9%, respectively (Figures S5 and S6, Figure 2, Table S3).In particular, the geographic range of H. longicornis was predicted to expand northward in both eastern and central China, and to higher-altitude regions in northern China, where SFTS has been established in low-altitude land.
In contrast, the environmental suitability of H. longicornis is predicted to be reduced in the central and eastern China, for example, Shandong, Henan, Jiangsu, and Anhui provinces, but is increased in the northern and northeastern China, including Jilin, Heilongjiang province, Inner Mongolia, and northern Xinjiang autonomous regions.The change is particularly significant at an RCP 8.5 scenario.

| The spatiotemporal dynamics of SFTS
A total of 8902 laboratory-confirmed human SFTS cases were reported from 2010 to 2019 in 361 counties and 104 cities in 22 provinces across the mainland of China.Shandong province has the highest average annual incidence (0.311/10 5 ), followed by Henan (0.254/10 5 ), Hubei (0.172/10 5 ), Anhui (0.153/10 5 ), and Liaoning (0.107/10 5 ; Table 1).Based on five-level risk grading (low risk, annual incidence, 0), medium-low risk (annual incidence range 0-0.01/10 5 ), medium risk (>0.01/10 5 and ≤0.1/10 5 ), medium-high risk (0.1/10 5 and ≤1/10 5 ), and high risk (>1/10 5 ), 14 counties in Shandong, eight counties in Henan, four counties in Anhui, four counties in Hubei, three counties in Jiangsu, three counties in Zhejiang, and two counties in Liaoning were classified as high risk (Figure 3).The seasonality differed between northern and southern provinces, that is, Hubei, Anhui, and Henan similarly had peaking case incidence in May, while a later peaking season was shown for Shandong (in June) and Liaoning provinces (in July), where endemic regions are of a higher altitude (Figure 3).The cases number reported per year increased steadily, from 53 in 2010 to 1355 in 2016.Afterwards the cases in 2017-2019 showed no more increase, keeping over 1000 cases annually (Table S4).Approximately 91.04% of the patients occurred between April and September, and peaking in May (22.80% of total cases).This temporal pattern differed across study years, with more cases reported beyond the traditional seasons and reduced level observed during the peaking month.For example, the proportion

| Ecological drivers for the dynamics of SFTS
A good simulation of the spatiotemporal dynamics of SFTS was attained by the GAM model, with R, GCV, and deviance estimated as 0.68%, 0.061%, and 75.3%, respectively (Table S5).The number of SFTS cases was significantly associated with the habitat suitability of H. longicornis (OR:1.26,95%: 1.20-1.32,p < .001),revealing 1.26 times increase in human cases for every 10% increase in the environmental suitability of H. longicornis.Other significant predictors included higher elevations, higher proportions of cropland and shrub, and lower proportions of grassland.
Significant nonlinear associations are demonstrated for climate factors (Figure S7), for example, as precipitation increased, the case incidences show a trimodal upward and downward, revealing a bidirectional effect on the disease (F = 7.28, p < .001).In a similar way, a U-shape effect is observed from relative humidity when the level is kept below 50% (F = 13.88,p < .001).However, a nonlinear effect is shown when humidity is increased above 50%.For the mean temperature, we have observed a nonlinear but consistently positive association with the case number of SFTS (F = 33.85,p < .001).There is a rapid increase in the case number as the mean temperature is elevated when below 20°C, while the trend slows when the temperature is above 20°C.

| Predicting SFTS dynamics under future climate and land cover changes
We predict the annual incidence of SFTS in the 2030s, 2050s, and 2080s with or without human population reduction under four RCPs (Figure 4).In the context of fixed human population size, SFTS case incidence might be significantly increased as compared to the 2010s, which is likely to peak in the 2050s under RCP2.6, with an increased rate of 28.26% from the baseline incidence in the 2010s.
In the context of reduced human population size, the SFTS incidence is likely to peak in the 2080s under RCP2.6,representing a 23.05% increase from the baseline incidence in the 2010s.
The predicted risk probabilities show spatial variation in the future climate scenarios (Figures S8-S14, Figure 5).Predicted hot spots of SFTS by the 2030s, 2050s, and 2080s are similar to that observed in the 2010s, with number of high-risk counties remained constant in Shandong, Henan, Anhui, Hubei, Liaoning, and Jiangsu province.
Under the upper climate change scenario (RCP8.5),we predict an increased case incidence in the 2030s compared to the 2010s in most of the endemic regions, however, we predict a decreased incidence in Zhejiang by 2080s compared to 2050s.The incidence in Anhui, Henan, and Jiangsu is predicted to peak in the 2030s, while incidence TA B L E 1 Current and predicted future annual incidence of SFTS in endemic areas of mainland China (10 −5 ). in Shandong and Liaoning is predicted to be on the rising (Figure 5).

Future population
Henan province, which currently has the second highest incidence rate, is predicted to develop a lower incidence rate than the current level, significantly lower than that in Hubei province (i.e., 0.158/10 5 vs. 21.5/10 5 under RCP8.5 by the 2080s in the context of reduced human population size; Table 1).Three provinces with sporadic cases report were predicted to bear risk of SFTS, that is, central and western Fujian, northwestern Xinjiang, and southwestern Yunnan.
The seasonal profiles of case incidence in the five highly endemic provinces were predicted (Figure 6).Predicted seasonal variations were heterogeneous across provinces and were depended on specific combinations of climate conditions in future climate scenarios.
Compared with the 2010s, the SFTS epidemic season in Liaoning will shift forward, with the start of the epidemic will be at least 1 month earlier, while the end and peak of the epidemic will be 1 month earlier or unchanged.For Shandong and Anhui provinces, the duration of epidemic was predicted to reduce under different future climate scenarios compared to that of the 2010s.Henan and Hubei provinces were predicted to a have similar seasonality, that is, the predicted duration of SFTS epidemic was 1 month longer compared with that of the 2010s, and the month with the highest case incidence will keep steady or delayed under most scenarios.

| DISCUSS ION
The complex enzootic cycles of SFTSV involve a combination of interactions between multiple factors, including the tick species F I G U R E 3 Spatiotemporal dynamics of reported SFTS cases in the mainland of China from 2010 to 2019.(a) Geographic distribution of SFTS case incidence rate.Incidence is expressed as the number of cases per 100,000 per year.Five grades were defined based on the annual incidence: 0, >0 and ≤0.01/10 5 , >0.01/10 5 and ≤0.1/10 5 , 0.1/10 5 and ≤1/10 5 , >1/10 5 .(b) Number of SFTS cases per month by province.(c) The seasonality of SFTS from 2010 to 2019.The squares represent peak month for SFTS, and error bars represent high prevalence month for SFTS.The prevalence month was defined as follows: the cumulative case numbers captured in the interval between the start and end accounted for >90% of the total cases throughout the epidemic.Map lines delineate study areas and do not necessarily depict accepted national boundaries.Haemaphysalis longicornis ticks are susceptible to environmental factors, with temperature and relative humidity strongly influencing their development and host-seeking behavior (Heath, 1979(Heath, , 1981)).Previous experimental studies have shown that tick survival and host-seeking activity are substantially inhibited at low humidity exposure.Precipitation, on the other hand, influences the host-seeking activity of H. longicorns in a nonlinear way (Ginsberg et al., 2017).Our own study had predicted a wide distribution of H. longicornis in 1176 counties in China, with temperature seasonality and the coverage rate of shrub grasslands as two of the most important environmental drivers for its spread (Zhao et al., 2021).
Here by using a more comprehensive dataset of field investigation of H. longicornis, we predict its habitat suitability, demonstrating the annual mean temperature as the most important contributor to the model, with an inverted U-shape effect (Figure S3).This was consistent with the previous finding that H. longicornis tick was restricted to temperate regions (Jiang et al., 2018) and required high humidity of 75% and air temperatures >18°C to complete their life cycle (Heath, 2016).Although H. longicornis ticks can withstand a wide range of temperature with 40°C as its lethal limit, the preferred temperature range for incubation at favorable humidities was found to be between 18 and 32°C.On the other hand, its tolerance of dehydration is less wide, especially in the larva and adult stages (Heath, 2016).In a consistent manner, total precipitation is demonstrated as a significant contributor to the suitability of H. longicornis ticks.Unfed H. longicornis larvae are unable to experience a long-term dryness, therefore a higher rainfall would enhance their chance of survival, which moreover lead to more breeding sites in shrub or forest areas, conceivably increasing the population size of the ticks (Estrada-Peña & de la Fuente, 2014).
Here we have shown a bidirectional effect on SFTS from climate warming, that is, to drive increases in some regions and decreases in others, depending on current and future local climates relative to the optimum and thermal limits for H. longicornis and disease transmission.In regions where annual mean temperatures are regularly between 2 and 6°C, including in Jilin, Heilongjiang, Inner Mongolia, and northern Xinjiang, a warming climate will become more suitable for H. longicornis.We predict that SFTS suitability will increase in northeastern regions, especially in the Liaoning province.This is consistent with the geographic shift toward higher altitudes in provinces, such as in Yunnan, Fujian, and Xinjiang, in contrast to the decreased incidence predicted in Henan province.This is largely attributed to the environmental suitability of H. longicornis, which was predicted to be reduced in F I G U R E 4 The multi-GCM ensemble mean of the predicted incidence of national SFTS cases under different climate change scenarios (RCP2.6,RCP4.5, RCP6.0, and RCP8.5) in the 2030s, 2050s, and 2080s.Rhombic points and circular points indicate the mean estimates of SFTS incidence and error bars are defined as the range.Rhombic points indicate the case incidence calculated with a fixed future human population.Circular points indicate the case incidence calculated with a changed future human population.The spatiotemporal pattern of the SFTS also reflects a sensitive response to climate factors.According to the national reporting data on human cases, SFTS epidemic is highly seasonal with the peaking season spanning from May to August, also showing an inter-regional difference.Global warming will also negate the need for ticks to hibernate in winter, triggering their earlier activity in the spring or even during warmer winter days, which might be responsible for the extended circulation period of SFTS that was observed in the current study, particularly in the regions with higher latitudes, that is, Liaoning province.
There are several key limitations of this study.Five grades were defined based on the annual incidence: 0, >0 and ≤0.01/10 5 , >0.01/10 5 and ≤0.1/10 5 , 0.1/10 5 and ≤1/10 5 , >1/10 5 .The change in risk of SFTS was estimated by subtracting the incidence grade of the former time period from the incidence grade of the latter time period, with a minimum value of −4 and a maximum value of 4. Map lines delineate study areas and do not necessarily depict accepted national boundaries.

F I G U R E 6
The current and projected seasonality of SFTS.(a) The seasonality of SFTS from 2010 to 2019 in the main endemic provinces of the mainland of China.(b-f) the projected seasonality of SFTS from the 2030s to the 2080s under different climate change scenarios (RCP2.6,RCP4.5, RCP6.0, and RCP8.5) of Liaoning, Shandong, Anhui, Henan, and Hubei province, respectively.The squares indicate peaking month for SFTS incidence; error bars indicate peak seasons for SFTS incidence.The vertical dashed lines in panels B-F represent the beginning, peak, and end months of the SFTS epidemic, respectively.The peak seasons were defined as the months during which the reported cumulative case numbers accounted for >90% of the total cases throughout the year.
longicornis) by using a generalized additive model (GAM).Based on this model, we predict the spatiotemporal dynamics of SFTS under future climate and land cover changes, by substituting the future environmental suitability for H. longicornis obtained in Step 2 and environmental variables.| 6649 DING et al.China Information System for Disease Control and Prevention (CISDCP).Based on the guidelines released by the National Health Commission of China, the laboratory diagnosis of SFTS was made by meeting one of the following criteria: (1) positive SFTSV culture, (2) real-time RT-PCR positive for SFTSV RNA, and (3) seroconversion or at least fourfold increase antibody titer between two serum samples collected more than 2 weeks apart (Ministry of Health of People's Republic of China, 2011).Only laboratory-confirmed SFTS cases were included in this study.
The ensemble of general circulation models (GCMs) from the fifth phase of the Coupled Model Intercomparison Project (CMIP5) was used in this study.To project the future environmental suitability for H. longicornis, and to project the future spatial-temporal patterns of SFTS cases, future climate data simulated by three GCMs, the Hadley Global Environment Model 2-Earth System (HadGEM2-ES), Institute Pierre Simon Laplace Coupled Model Version Five A-Low Resolution (IPSL-CM5A-LR), and an atmospheric chemistry version of the coupled Model for Interdisciplinary Research on Climate/Earth System Model (MIROC-ESM-CHEM), under four Representative Concentration Pathways (RCPs: RCP2.6,RCP4.5, RCP6.0, and RCP8.5) were included in this study, which were obtained from the website of World Climate Research Programme (https://esgf-node.llnl.gov/searc h/cmip5/).In addition, those future climate data were extracted as three 10-year time slices: the 2030s (2030-2039), 2050s (2050-2059), and 2080s (2080-2089).
(a) Monthly SFTS cases were calculated from 2010 to 2019 at the county level; (b) climate, land cover, terrain data, and the environmental suitability of H. longicornis at the county level were compiled from 2010 to 2019 as spatial covariates; (c) the GAM model was fitted based on assembled data to explore the explanatory variables that contribute to the number of SFTS cases; (d) projected explanatory variables were used under four RCPs from three GCM models to predict SFTS cases incidence in the 2030s, 2050s, and 2080s.
indicating an expanding in the geographic distribution of it.The correlation analysis of all the 19 ecoclimatic variables yielded eight variables with good representation, including annual mean temperature, mean diurnal range, isothermality, temperature seasonality, max temperature of warmest month, mean temperature of coldest quarter, precipitation seasonality, and precipitation of driest quarter, which together with the landcover and terrain data, were entered into the BRT model to predict the environmental suitability of H. longicornis.The model has yielded a highly Projected environmental suitability of Haemaphysalis longicornis based on BRT model simulations under the RCP8.5 scenario from 2030 to 2080.(a) The projected suitability of H. longicornis in the 2030s; (b) The changes of projected suitability of H. longicornis from the 2010s to the 2030s; (c) the projected suitability of H. longicornis in the 2050s; (d) the changes of projected suitability of H. longicornis from the 2030s to the 2050s; (e) the projected suitability of H. longicornis in the 2080s; (f) the changes of projected suitability of H. longicornis from the 2050s to the 2080s.Map lines delineate study areas and do not necessarily depict accepted national boundaries. of H.longicornis presence Probability of H.longicornis presence Probability of H.longicornis presence Probability of H.longicornis presence Probability of H.longicornis presence of H.longicornis presence Probability of H.longicornis presence Probability of H.longicornis presence Probability of H.longicornis presence Probability of H.longicornis presence 0 during the off-season (between October and the next March) had increased from 4.56% in the year 2011 to 10.6% in 2019.While the peaking incidence in May had reduced from 34.1% in 2012 year to 17.4% in 2019.
vectors, reservoir hosts for the development and survival of the ticks, landscape and vegetation types, meteorological factors, and socioeconomic factors among others.All the assessments of published studies have only touched on the basic ecoepidemiology aspects of the disease, mostly on habitat, landscape, and climate while lacking in-depth investigation on the combination of interactions.In the current study, we present a multi-model, multi-scenario framework to assess the potential impacts of climate warming on the future risk of SFTS in the mainland of China.To our knowledge, this study is the first to develop an integrated approach based on SFTS case incidence, the predicted distribution of H. longicornis, historical and future data on the environmental factors, landcover data, and human population data in the mainland of China.
eastern provinces, for example, Henan, Jiangsu, Anhui, etc., but increased in northern and northeastern provinces, including Jilin, Heilongjiang, Inner Mongolia, and northern Xinjiang.Although altitude does not directly affect H. longicornis' development, it influences climate and vegetation type.As the global climate warms, high latitudes will become suitable for H. longicornis breeding and expansion.
Our estimates are subject to data limitations, particularly the lacking of data on H. longicornis ticks.The current survey locations of the ticks were F I G U R E 5 The projected SFTS incidence based on GAM model simulations under the RCP8.5 scenario and future human population count under the SSP2 scenario from the 2030s to the 2080s.The projected SFTS incidence in the 2030s (a) and the change from the 2010s to the 2030s (b); in the 2050s (c) and change from the 2030s to the 2050s (d); in the 2080s (e) and from the 2050s to the 2080s (f).
log 10 Pop i,t