Precision mapping of snail habitat in lake and marshland areas: integrating environmental and textural indicators using Random Forest modeling

doi:10.21203/rs.3.rs-3991100/v1

Background

Schistosomiasis japonica poses a significant health issue in China, largely due to the spatial distribution of Oncomelania hupensis, the only intermediate host of Schistosoma, which directly affects schistosomiasis incidence. This study therefore aimed to address the limitations in existing remote sensing studies, particularly the oversight of spatial scale and seasonal variations in snail habitats by introducing a multi-source data-driven Random Forest approach.

Methods

This method effectively integrates bottomland and ground-surface texture data with traditional environmental variables for a more comprehensive and accurate snail habitat analysis. Four distinct models focusing on lakes and marshlands in Guichi, China, were developed: the baseline model, including ground-surface texture, bottomland variables, and environmental variables; Model 1, including only environmental variables; Model 2, including ground-surface texture and environmental variables; and Model 3, including bottomland and environmental variables.

Results

The baseline model outperformed the others, achieving a true skill statistic of 0.93, accuracy of 0.97, kappa statistic of 0.94, and area under the curve of 0.98. The findings identified key high-risk snail habitats, particularly along major rivers and lakes in a belt-like distribution, particularly near the Yangtze River, Qiu Pu River, and surrounding areas of Shengjin Lake, Jiuhua River, and Qingtong River.

Conclusions

This study providing vital data for effective snail monitoring, control strategies, and schistosomiasis prevention. This approach may also be applicable in locating other epidemic hosts with similar survival and ecological characteristics.

Multi-source data

Schistosomiasis

Potential risk area

Machine learning

Recognition

Schistosomiasis is a severe parasitic disease with a significant effect on human health. It is highly prevalent in subtropical and tropical regions of Asia, Africa, South America, and the Middle East [1]. The primary parasitic species responsible for human infection are Schistosoma japonicum, Schistosoma mansoni, and Schistosoma aegypti. Among these, schistosomiasis japonica is predominantly found in Southeast Asia and South China [2]. In China, schistosomiasis japonica remains a major public health challenge because of its complex etiology and multiple transmission pathways [3]. The transmission of schistosomiasis japonica is closely related to the distribution of its only intermediate host, Oncomelania hupensis (snail), which governs the spread of schistosomiasis [4]. Consequently, accurate identification and subsequent monitoring of snail habitats has become essential for controlling and mitigating schistosomiasis.

The traditional approach to snail inspection is both labor-intensive and time-consuming, often yielding suboptimal results [5]. Snail reproduction is intricately linked to climatic conditions, environmental attributes, and socioeconomic factors. Therefore, numerous studies have employed remote sensing technology to identify snail habitats. For instance, Walz et al. extracted seven environmental factors from remote sensing images, using a habitat suitability index to identify snail habitats [6]. Rao et al. explored the relationship between snail habitats and environmental factors through correlation analysis to predict snail breeding sites [7]. Similarly, Xue et al. extracted three important environmental factors and established a multiple regression analysis model to predict snail distribution [8]. Further advancing this field, Xia et al. determined additional environmental factors from medium-resolution images and constructed ecological niche models to identify high-risk areas for snail habitats, contributing to the progress in this research area [5]. Collectively, these studies converge on a prevalent methodology for snail detection, involving the extraction of environmental variables from remote sensing images and the application of models to assimilate and interpret the environmental characteristics associated with snail habitats. This approach facilitates the identification and mitigation of snail habitat risks.

Previous studies primarily focused on natural and socio-environmental factors, often overlooking the importance of ground-surface texture information in snail habitat analysis. Remote sensing images capture various textural attributes reflecting the spatial distribution structure of ground objects [9]. These attributes are invaluable for discerning the surface characteristics of distinct features [10]. Snail habitats often exhibit unique land surface features identifiable through these textural attributes. Additionally, past studies neglected a key variable closely related to snail habitats – the distribution of bottomland areas characterized by "winter land-summer water." This refers to water levels rising from May to October (wet season), submerging the whole substratum where snails reside, rendering the area unsuitable for snail activity. This bottomland area has shown a close correlation with the distribution of snails; however, information on bottomland distribution has been used only to test whether an area is suitable for snail habitation and has not been introduced as a continuous variable in the model [11, 12]. Therefore, this study hypothesized that harnessing texture information and quantifying the distance from snail habitats to these bottomlands can significantly improve precision in identifying snail habitats.

This study aimed to identify snail habitats distributed in Guichi, China, a typical schistosomiasis-endemic region along the Yangtze River. To achieve this objective, two new variables were incorporated into our analysis: (i) texture information of the ground surface and (ii) data on bottomland distribution, alongside traditional environmental variables. Using this enriched dataset, a model specifically designed for snail habitat identification was constructed using the Random Forest (RF) algorithm. This model primarily facilitates the probabilistic delineation of snail habitats and helps generate a habitat probability distribution map.

Study area

Located in the lower-middle section of the Yangtze River Basin in Guichi is a focal area for schistosomiasis japonica research in China. Covering an area of 2,516 km², its geographic features are notable for numerous rivers such as Qiupu, Jiuhua, Qingtong, and Baiyang in the north. By contrast, the central and southern areas of the region are predominantly hilly. The combination of a favorable climate, plentiful water resources, and lush vegetation fosters an environment conducive to snail reproduction, a key factor in schistosomiasis transmission [13]. This study primarily focuses on the lake areas in the Guichi district, as shown in Figure 1.

Fig. 1 Study area location for snail habitat identification in the lake area of Guichi, Chizhou City, Anhui Province, China. The study area is located in the northern part of Guichi, near the south bank of the middle and lower reaches of the Yangtze River.

Data source

Snail data

Snail distribution data were obtained from field surveys conducted throughout Guichi between March and May 2021. This dataset includes geographic information (latitude and longitude coordinates) on snail presence, snail population types, and environmental characteristics of their habitats. The dataset also features 108 survey sites covering snail populations in lakes and marshlands. For the control group, an equal number of snail-absent points were selected as the control group based on criteria such as areas without snails, including rooftops, roads, construction sites, and permanent water bodies.

Covariate data

Environmental Data

Covariates previously identified as significant in determining snail habitat distribution were chosen [5]. The environmental data includes precipitation, distance to bottomland, distance to water bodies, land surface temperature, normalized difference vegetation index, wetness, land use, and nighttime light, abbreviated in the snail habitat identification model as PRE, DB, DW, LST, NDVI, WET, LU, and NL, respectively. Table 1 details the sources and resolution of all environmental data.

Natural environmental factors in the study area were derived from Landsat 8 images. NDVI was used to assess vegetation cover in the study area [14], the Mono-window algorithm to estimate land surface temperature [15], and the Kirchhoff Transform (K-T) to gauge wetness [16]. The moisture content derived from the K-T provides insights into soil and vegetation moisture. We computed the DW variables using spatial analysis methods in ArcGIS 10.2 (ESRI Inc., Redlands, CA, USA).

Previous studies explored the relationship between snail habitats and waterbody distribution [17]. However, they often overlooked the effect of seasonal changes on water distribution, which is critical for the reproduction of snails. The present study addressed this gap by focusing on bottomland distribution, which reflects these seasonal variations more accurately. For identifying the bottomland area, sentinel-1 radar images, sensitive to waterbodies, were assessed, and the Sentinel-1 dual-polarized water index (SDWI) algorithm proposed by Jia et al. was used to segment the water and land portions of the study area [18]. The waterbody in the study area during the abundant water period (May to October) and the dry water period (November to April) were segmented out, and spatial analysis was conducted using ArcGIS10.2, identifying bottomland areas characterized by "winter land-summer water." The nearest distance from the sampling point to the bottomland was calculated using the Coverage tool in ArcGIS10.2, deriving the variable "DB."

Table 1. Summary of the environment variables used in this study.

Covariate	Spatial Resolution	Source	Reference
PRE	30m	China Meteorological Data Service Centre	http://data.cma.cn[41]
DB	/	SDWI	Jia et al., 2019 [18]
DW	/	Open Street Map	http://www.openstreetmap.org[42]
NDVI	30m	Band math	Rouse et al., 1974 [14]
LST	30m	Mono-window algorithm	Qin et al., 2003 [15]
WET	30m	Optical image inversion	Huang et al., 2002 [16]
LU	10m	EULUC-China	Gong et al., 2020 [33]
NL	500m	NPP-VIIRS	Chen et al., 2021 [40]

Note: PRE, precipitation; DB, distance to bottomland; DW, distance to waterbody; NDVI, normalized difference vegetation index; LST, land surface temperature; WET, wetness; LU, land use; NL, nighttime light.

Texture Data

Texture information reflects the spatial information of ground objects, and it is essential for identifying features of ground objects. Therefore, we introduced the texture information to characterize the spatial distribution of snail habitats. Eight texture feature indicators were used: mean (B1), variance (B2), uniformity (B3), contrast (B4), dissimilarity (B5), entropy (B6), second-order moment (B7), and correlation (B8); these were extracted from Landsat 8 images using the Gray-Level Co-occurrence Matrix (GLCM), a commonly used texture analysis method particularly suitable for optical image texture features [19]. The eight indicators were computed in ENVI 5.3 software (Exelis inc., Boulder, CO, USA). Table 2 shows the texture information categories and their labels.

Table 2. Types of texture indicators used in the study.

Acronyms	Textural indicator	Meaning of the indicator
B1	Mean	Degree of regularity of texture
B2	Variance	Deviation of the image element value from the mean value
B3	Standard Deviation	Deviation of the image element value from the mean value
B4	Contrast	Local gray level uniformity of an image
B5	Dissimilarity	Local gray level uniformity of an image
B6	Entropy	Amount of information an image has
B7	Angular Second Moment	Uniformity of the gray level distribution of an image
B8	Correlation	Degree of similarity between elements

Methodology

To generate a probabilistic risk map of snail habitats quickly, the RF method was used for identifying snail habitats based on multi-source data. The specific process is shown in Figure 2.

Fig. 2 Technical workflow of snail habitat identification study. D: input data, I: model indicators, B: model building, V: validation indicators.

In the specific processing process, the following research hypothesis was tested: incorporating texture information from Landsat 8 images and quantifying the proximity of snail habitats to bottomland improves snail habitat identification accuracy. The baseline model, Model 0, integrated environmental variables, ground-surface texture information, and the "DB" variables. For comparative analysis, three additional models were constructed: Model 1, a traditional model that incorporated only environmental variables (PRE, DW, LST, NDVI, WET, LU, and NL); Model 2, combining environmental variables with ground-surface texture information; and Model 3, merging environmental variables with the "DB" variable. To ensure consistency in the number of covariates, randomly permuted texture information was added as a control in models lacking texture information (Model 1 and Model 3) [20]. RF models were constructed by Python’s Scikit-learn machine learning library [21], with the number of trees set at n=200.

The dataset for model validation was partitioned into two segments: 75% allocated for the training set and 25% for the validation set. The training set was utilized to fit the model, allowing it to learn the relationship between the input features and the target variable. To prevent statistical randomness, the method of 10-fold cross-validation was employed to ascertain optimal parameters and evaluate models using test set data. Model performance was assessed using four metrics: true skill statistic (TSS) [22], accuracy (ACC) [23], kappa [24], and area under the curve (AUC) [25]. These metrics facilitated a comparative analysis of model performances [17,26]. Subsequently, the best-performing model was applied to predict snail habitat probabilities across the study area.

Model comparison

Table 3 presents a comparative analysis of model performances. The baseline model emerged as the most effective, registering the highest scores across all metrics (TSS=0.93, ACC=0.97, kappa=0.94, AUC=0.98). It was closely followed by Model 2 (TSS=0.87, ACC=0.93, kappa=0.85, AUC=0.96) and Model 3 (TSS=0.77, ACC=0.91, Kappa=0.81, AUC=0.93). By contrast, Model 1 exhibited comparatively weaker performance (TSS=0.68, ACC=0.85, kappa=0.53, AUC=0.91).

Table 3. Type and comparison of models

Model	Model 0	Model 1	Model 2	Model 3
Whether Environmental Variables is Included	Yes	Yes	Yes	Yes
Whether Texture Information is Included	Yes	No	Yes	No
Whether "DB" variable is Included	Yes	No	No	Yes
TSS	0.933	0.676	0.871	0.771
ACC	0.969	0.846	0.926	0.908
Kappa	0.938	0.528	0.852	0.813
AUC	0.980	0.910	0.960	0.930

Note: TSS, true skill statistic; ACC, accuracy; AUC, area under the curve.

Importance of variables

Figure 3 illustrates the variable importance rankings in the baseline model. The vertical axis represents variable categories, while the horizontal axis signifies the importance value (IV) of the variables. Out of 15 covariates, four—comprising two environmental and two texture variables—had an importance value exceeding 0.06. The most significant variable was DB (IV=0.69), markedly outperforming others in importance. This was followed by WET (IV=0.09), B6 (IV=0.07), and B7 (IV=0.06). The variables NL, LST, B1, NDVI, B8, B3, PRE, LU, B2, B4, and B5 had importance scores below 0.06.

Fig. 3 Variable importance ranking of the baseline model (Model 0). The vertical coordinate indicates the category of the variable and the horizontal coordinate indicates the importance of variables. B5, dissimilarity; B4, contrast; B2, variance; LU, land use; PRE, precipitation; B3, standard deviation; B8, correlation; NDVI, normalized difference vegetation index; B1, mean; LST, land surface temperature; NL, nighttime light; B7, angular second moment; B6, entropy; WET, wetness; DB, distance to bottomland.

Risk probability map of snail habitats

Figure 4 presents the risk probability maps used in this study. It details the baseline model’s identification of high-risk snail habitats (probability of snails having more than 0.7), predominantly along the Yangtze River and Qiupu River in the northern part and central part of the study area, near Shengjin Lake in the south, and around the Jiuhua River and Qingtong River in the eastern part of the study area. Conversely, low-risk areas (probability below 0.3) include the northeastern residential zones, the western bank of the Qiupu River, and remote agricultural lands. Notably, three alternative models identified more extensive high-risk areas compared with the baseline model. Specifically, Model 1 and Model 3 suggest greater risk of snail habitat is higher in the western part of the study area than in the eastern part of the study area, whereas the high-risk areas identified in Model 2 were mainly concentrated in the waterbody regions.

Fig. 4 Probability map showing the distribution of snail habitats. a: Model 0, characterized by texture information of ground-surface, DB, and environmental variables; b: Model 1, characterized by environmental variables; c: Model 2, characterized by texture information and environmental variables; d: Model 3, characterized by DB and environmental variables.

This study aimed to identify high-risk zones of snail habitats using remote sensing imagery. Therefore, a model for identifying snail habitats was developed, which integrates traditional environmental variables with ground-surface texture information and the DB variable. The comprehensive approach effectively pinpointed high-risk sites for snail habitats in the study area. The generated probability map of snail habitat serves as a tool for assisting relevant personnel in precisely focusing on snail control and schistosomiasis prevention education, thereby improving the effectiveness of snail detection and eradication efforts.

Table 3 indicates that, among the four models, the baseline model, incorporating ground-surface texture information and the variable "DB," outperforms others in predictive accuracy. Conversely, Model 1, reliant solely on environmental factors, exhibits the lowest accuracy. This discrepancy underscores the substantial contributions of the DB variable and ground-surface texture information in enhancing model performance. The analysis of variable importance in the optimal model (i.e., the baseline model) reveals that DB is the most critical predictor, with a relative importance score significantly surpassing other variables. This finding aligns with the results of Li et al. [27], confirming the significant correlation between snail habitat and bottomland distribution. A comparison between Model 2 and Model 3 highlights significant performance enhancements resulting from incorporating texture information. Texture indicators, particularly B6 and B7, have proven to be highly influential, surpassing the importance of previously recognized variables such as NL, LST, NDVI, and so on. These findings are supported by those reported in recent studies [28–30].

The superior performance of the baseline model can be attributed to its comprehensive evaluation of spatial factors associated with snail habitats and its consideration of seasonal aspects of reproduction. Among these factors, texture information has been shown in past studies to help models distinguish between different vegetation types [10, 31]. The dominant snail habitat community, Cyperaceous, exhibits specific patterns or structures at the textural scale, and incorporating textural information helps capture these differences. Additionally, texture information aids in understanding the spatial structure and distribution of features, such as the banded distribution of vegetation on bottomland along the Yangtze River [32], and the vegetation cover, thus improving the model's accuracy in identifying the distribution of vegetation and snail habitats. Previous studies highlight the importance of seasonal alternation between land and water in providing an ideal environment for snail growth and reproduction, emphasizing that areas with long periods of inundation are not conducive to snail habitation [5, 17]. The actual snail distribution in the present study area is consistent with this characteristic and with our identification results. Therefore, the model's accuracy improves by considering bottomland distribution characteristics.

Figure 4 depicts risk probability maps generated by the baseline model and the other three models. These maps identify key high-risk areas for snail habitats along major rivers and lakes, contrasting with low-risk areas in residential and agricultural regions. This detailed mapping enables public health authorities, including the local Centers for Disease Control and Prevention, to focus on snail detection and eradication efforts, customizing strategies for physical extermination and drug delivery in areas with higher snail probabilities. Additionally, the maps support targeted educational initiatives in high-risk regions, raising awareness about schistosomiasis prevention and control. The methodology applied in this study is promising for broader application in similar contexts. Guichi was chosen as the study area due to its typical snail habitats in lake and marshland environments, and its conditions favor snail survival, making it a critical area for schistosomiasis japonica in China [13, 34–36]. The developed model has potential for identifying snail habitats in other endemic regions. Additionally, this study's approach can assist in locating breeding sites for intermediate hosts of other diseases, such as Zika, dengue fever, and malaria, which are heavily influenced by environmental factors [37–39]. The approach used in this study offers a comprehensive understanding of the relationship between these hosts and their environments.

This study acknowledges certain constraints requiring additional discussion. First, the focus of this study was limited to lakes and marshlands, potentially diminishing the applicability and efficacy of the model in hilly and mountainous terrains with different influencing factors governing snail habitats. Second, the temporal scope was restricted to image data from 2021. For a more holistic understanding of snail habitats and their evolution, future research should engage in a spatial-temporal change analysis, utilizing image datasets spanning multiple years.

This study reveals that high-risk areas for snail distribution in Guichi, China, are predominantly found along rivers, particularly near the Yangtze River, Qiu Pu River, and surrounding areas of Shengjin Lake, Jiuhua River, and Qingtong River. Our model, upon validation, outperformed traditional models in identifying high-risk areas. These findings are critical for directing focused interventions, such as targeted snail control measures, preventive education on schistosomiasis, and other related activities. Furthermore, the methodology of this study has potential for broader application in identifying breeding sites for other disease hosts influenced by environmental factors, enhancing the efficiency and precision of disease prevention and control efforts.

Acknowledgements

Not applicable.

Author contributions

Xuedong Zhang: Writing - original draft, Validation, Funding acquisition. Zelan Lv: Writing - review & editing, Writing - original draft, Software, Methodology, Formal analysis, Data curation. Jianjun Dai: Investigation, Resources, Project administration. Yongwen Ke: Validation, Project administration, Investigation. Yi Hu: Writing - review & editing, Validation, Supervision, Conceptualization.

Funding

This work is primarily being funded by Beijing Key Laboratory of Urban Spatial Information Engineering (NO. 20230101) and National Natural Science Foundation of China (NO. 81773487).

Availability of data and materials

All the environmental data is available in a public repository at https://figshare.com/articles/dataset/environmental_data_of_snail_habitat_identification/25092638, and data on snail habitat can be available on request from the corresponding author.

Ethical Approval and Consent to participate

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors have no competing interests to declare.

McManus DP, Dunne DW, Sacko M, Utzinger J, Vennervald BJ, Zhou XN, Schistosomiasis. Nat Rev Dis Primers. 2018;4(1):13. https://doi.org/10.1038/s41572-018-0013-8.
Qiu C, Lu DB, Deng Y, Zou HY, Liang YS, Webster JP. Population genetics of Oncomelania hupensis snails, intermediate hosts of Schistosoma japonium, from emerging, re-emerging or established habitats within China. Acta Trop. 2019;197:105048. https://doi.org/10.1016/j.actatropica.2019.105048.
Xu J, Lv S, Cao CL, Li SZ, Zhou XN. [Progress and challenges of schistosomiasis elimination in China]. Zhongguo Xue Xi Chong Bing Fang Zhi Za Zhi. 2018;30(6):605–9. https://doi.org/10.16250/j.32.1374.2018249. Chinese.
Leonardo L, Varona G, Fornillos RJ, Manalo D, Tabios IK, Moendeg K, et al. Oncomelania hupensis quadrasi: Snail intermediate host of Schistosoma japonicum in the Philippines. Acta Trop. 2020;210:105547. https://doi.org/10.1016/j.actatropica.2020.105547.
Xia C, Hu Y, Ward MP, Lynn H, Li S, Zhang J, et al. Identification of high-risk habitats of Oncomelania hupensis, the intermediate host of schistosoma japonium in the Poyang Lake region, China: A spatial and ecological analysis. PLoS Negl Trop Dis. 2019;13(6):e0007386. https://doi.org/10.1371/journal.pntd.0007386.
Walz Y, Wegmann M, Dech S, Vounatsou P, Poda JN, N'Goran EK, et al. Modeling and Validation of Environmental Suitability for Schistosomiasis Transmission Using Remote Sensing. PLoS Negl Trop Dis. 2015;9(11):e0004217. https://doi.org/10.1371/journal.pntd.0004217.
Rao DD, Hu F, Lu CF, Lv SB, Lin DD, Liu Y. [Study on Relationship between Oncomelania Hupensis and Environmental Remote Sensing Index Based on Landsat8 Image]. Jiangxi Sci. 2018;036(4):599–604610. https://doi.org/10.13990/j.issn1001-3679.2018.04.014. Chinese.
Xue JB, Xia S, Zhang LJ, Abe EM, Zhou J, Li YY, et al. High-resolution remote sensing-based spatial modeling for the prediction of potential risk areas of schistosomiasis in the Dongting Lake area, China. Acta Trop. 2019;199:105102. https://doi.org/10.1016/j.actatropica.2019.105102.
Anwer RM, Khan FS, Joost VDW, Molinier M, Laaksonen J. Binary patterns encoded convolutional neural networks for texture recognition and remote sensing scene classification. ISPRS J Photogramm Remote Sens 138 APR. 2017;74–85. https://doi.org/10.1016/j.isprsjprs.2018.01.023.
He S, Peng P, Chen Y, Wang X. Multi-Crop Classification Using Feature Selection-Coupled Machine Learning Classifiers Based on Spectral, Textural and Environmental Features. Remote Sens. 2022;14(13):3153. https://doi.org/10.3390/rs14133153.
Xue JB, Wang XY, Zhang LJ, Hao YW, Chen Z, Lin DD, et al. Potential impact of flooding on schistosomiasis in Poyang Lake regions based on multi-source remote sensing images. Parasit Vectors. 2021;14(1):116. https://doi.org/10.1186/s13071-021-04576-x.
Qiu J, Han D, Li R, Xiao Y, Zhu H, Xia J, et al. Satellite Imagery-Based Identification of High-Risk Areas of Schistosome Intermediate Snail Hosts Spread after Flood. Remote Sens. 2022;14(15):3707. https://doi.org/10.3390/rs14153707.
Su Q, Bergquist R, Ke Y, Dai J, He Z, Gao F, et al. A comparison of modelling the spatio-temporal pattern of disease: a case study of schistosomiasis japonica in Anhui Province, China. Trans R Soc Trop Med Hyg. 2022;116(6):555–63. https://doi.org/10.1093/trstmh/trab174.
Rouse JW, Haas RH, Schell JA, Deering DW. Monitoring vegetation systems in the Great Plains with ERTS (Earth Resources Technology Satellite). In: Third ERTS (Earth Resources Technology Satellite).Symposium, Greenbelt. 1973; Vol. 1, pp. 309–317. https://api.semanticscholar.org/CorpusID:133358670.
Qin ZH, Li WJ, Zhang MH, Arnon K, Pedro B. Estimating of the essential atmospheric parameters of mono-window algorithm for land surface temperature retrieval from Landsat TM 6. Remote Sensing for Natural Resources. 2003, (2): 37–43. Chinese. https://doi.org/10.3969/j.issn.1001-070X.2003.02.010.
Huang C, Wylie B, Yang L. Derivation of a tasseled cap transformation based on Landsat 7 at-Satellite reflectance. Int J Remote Sens. 2002;23(8):1741–8. https://doi.org/10.1080/01431160110106113.
Zhang J, Yue M, Hu Y, Bergquist R, Su C, Gao F, et al. Risk prediction of two types of potential snail habitats in Anhui Province of China: Model-based approaches. PLoS Negl Trop Dis. 2020;14(4):e0008178. https://doi.org/10.1371/journal.pntd.0008178.
Jia SC, Xue DJ, Li CR, Zheng J, Li WQ. Study on new method for water area information extraction based on Sentinel – 1 data. Yangtze River. 2019;50(2):213–7. https://doi.org/10.16232/j.cnki.1001-4179.2019.02.038. Chinese.
Guo SY, Li L, Zhang LJ, Li YL, Li SZ, Xu J. From the One Health Perspective: Schistosomiasis Japonica and Flooding. Pathogens. 2021;10(12):1538. https://doi.org/10.3390/pathogens10121538.
Jagadesh S, Zhao C, Mulchandani R, Van Boeckel TP. Mapping Global Bushmeat Activities to Improve Zoonotic Spillover Surveillance by Using Geospatial Modeling. Emerg Infect Dis. 2023;29(4):742–50. https://doi.org/10.3201/eid2904.221022.
Pedregosa F, Varoquaux G, Gramfort A, Michel V, Thirion B, Grisel O, et al. Scikit-learn: Machine Learning in Python. J Mach Learn Res. 2011;12(85):2825–30. https://dl.acm.org/doi/10.5555/1953048.2078195.
Allouche O, Tsoar A, Kadmon R. Assessing the accuracy of species distribution models: prevalence, kappa and the true skill statistic (TSS). J Appl Ecol. 2006;43(6):1223–32. https://doi.org/10.1111/j.1365-2664.2006.01214.x.
Atanda O, West J, Stables T, Johnson C, Merrifield R, Kinross J. Flow rate accuracy of infusion devices within healthcare settings: a systematic review. Ther Adv Drug Saf. 2023;14:20420986231188602. https://doi.org/10.1177/20420986231188602.
Atanda O, West J, Stables T, Johnson C, Merrifield R, Kinross J. Flow rate accuracy of infusion devices within healthcare settings: a systematic review. Ther Adv Drug Saf. 2023;14:20420986231188602. https://doi.org/10.1177/20420986231188602.
Kumar R, Indrayan A. Receiver operating characteristic (ROC) curve for medical researchers. Indian Pediatr. 2011;48(4):277–. https://doi.org/10.1007/s13312-011-0055-4. 87.
Gong YF, Hu XK, Hao YW, Luo ZW, Feng JX, Xue JB, et al. Projecting the proliferation risk of Oncomelania hupensis in China driven by four shared socioeconomic pathways (SSPs): a multi-scenario comparison and integrated modeling study. Adv Clim Chang Res. 2022;13(2):258–65. https://doi.org/10.1016/j.accre.2022.02.004.
Li Y, Guo S, Dang H, Zhang L, Xu J, Li S. Oncomelania hupensis Distribution and Schistosomiasis Transmission Risk in Different Environments under Field Conditions. Trop Med Infect Dis. 2023;8(5):242. https://doi.org/10.3390/tropicalmed8050242.
Zheng JX, Xia S, Lv S, Zhang Y, Bergquist R, Zhou XN. Infestation risk of the intermediate snail host of Schistosoma japonicum in the Yangtze River Basin: improved results by spatial reassessment and a random forest approach. Infect Dis Poverty. 2021;10(1):74. https://doi.org/10.1186/s40249-021-00852-1.
Gong YF, Luo ZW, Feng JX, Xue JB, Guo ZY, Jin YJ, et al. Prediction of trends for fine-scale spread of Oncomelania hupensis in Shanghai Municipality based on supervised machine learning models. Zhongguo Xue Xi Chong Bing Fang Zhi Za Zhi. 2022;34(3):241–51. https://doi.org/10.16250/j.32.1374.2021247. Chinese.
Xue J, Hu X, Hao Y, Gong Y, Wang X, Huang L, et al. Transmission Risk Predicting for Schistosomiasis in Mainland China by Exploring Ensemble Ecological Niche Modeling. Trop Med Infect Dis. 2022;8(1):24. https://doi.org/10.3390/tropicalmed8010024.
Cheng G, Ding H, Yang J, Cheng Y. Crop type classification with combined spectral, texture, and radar features of time-series Sentinel-1 and Sentinel-2 data. Int J Remote Sens. 2023;44(4):1215–37. https://doi.org/10.1080/01431161.2023.2176723.
Li ZJ, Chen HG, Gong P, Zeng XJ, Liu YM. Study on Relationship between Vegetation and Spatial Distribution of Oncomelania Snails in Poyang Lake region. Chin J Schisto Control. 2010;22:132–5. https://api.semanticscholar.org/CorpusID:89000107. Chinese.
Gong P, Chen B, Li X, Liu H, Wang J, Bai Y, et al. Mapping essential urban land use categories in China (EULUC-China): preliminary results for 2018. Sci Bull (Beijing). 2020;65(3):182–7. https://doi.org/10.1016/j.scib.2019.12.007.
Hu Y, Li S, Xia C, Chen Y, Lynn H, Zhang T, et al. Assessment of the national schistosomiasis control program in a typical region along the Yangtze River, China. Int J Parasitol. 2017;47(1):21–9. https://doi.org/10.1016/j.ijpara.2016.09.003.
Gong YF, Zhu LQ, Li YL, Zhang LJ, Xue JB, Xia S, et al. Identification of the high-risk area for schistosomiasis transmission in China based on information value and machine learning: a newly data-driven modeling attempt. Infect Dis Poverty. 2021;10(1):88. https://doi.org/10.1186/s40249-021-00874-9.
Zhao GM, Zhao Q, Jiang QW, Chen XY, Wang LY, Yuan HC. Surveillance for schistosomiasis japonica in China from 2000 to 2003. Acta Trop. 2005 Nov-Dec;96(2–3):288–95. https://doi.org/10.1016/j.actatropica.2005.07.023.
Wang T, Fan ZW, Ji Y, Chen JJ, Zhao GP, Zhang WH, et al. Mapping the Distributions of Mosquitoes and Mosquito-Borne Arboviruses in China. Viruses. 2022;14(4):691. https://doi.org/10.3390/v14040691.
Ding F, Fu J, Jiang D, Hao M, Lin G. Mapping the spatial distribution of Aedes aegypti and Aedes albopictus. Acta Trop. 2018;178:155–62. https://doi.org/10.1016/j.actatropica.2017.11.020.
Alexander J, Wilke ABB, Mantero A, Vasquez C, Petrie W, Kumar N, et al. Using machine learning to understand microgeographic determinants of the Zika vector, Aedes aegypti. PLoS ONE. 2022;17(12):e0265472. https://doi.org/10.1371/journal.pone.0265472.
Chen ZQ, Yu BL, Yang CS, Zhou YY, Yao SJ, Qian XJ, et al. Earth Syst Sci Data. 2021;13(3):889–906. https://doi.org/10.5194/essd-13-889-2021. An Extended Time Series (2000–2018) of Global NPp-VIIRS Like Nighttime Light Data from a Cross-Sensor Calibration.
China Meteorological Data Service Centre. : http://data.cma.cn. Accessed 19 July 2022.
Open Street Map. : http://www.openstreetmap.org. Accessed 25 September 2022.

Precision mapping of snail habitat in lake and marshland areas: integrating environmental and textural indicators using Random Forest modeling

Status:

Version 1

Abstract

Background

Methods

Results

Conclusions

Figures

Background

Methods

Results

Discussion

Conclusion

Declarations

References

Status:

Version 1