Potential Impact of Flood on Schistosomiasis in Poyang Lake Regions Based on Multi-Source Remote Sensing Images

Flooding may be the most important factors contributing to the rebound of Oncomelania hupensis in endemic foci. This study aimed to assess the risk of schistosomiasis japonica transmission impacted by ooding around the Poyang Lake region using multi-source remote sensing images. Normalized Difference Vegetation Index (NDVI) data collected by the Landsat 8 satellite was used as an ecological and geographical suitability indicator of O. hupensis snail habitats in the Poyang Lake region. The ood-affected water body expansion was estimated using dual polarized threshold calculations based on the dual polarized synthetic aperture radar (SAR). The image data were captured from Sentinel-1B satellite in May 2020 before the ood and in July 2020 during the ood. The spatial database of snail habitats distribution was created by using the 2016 snail survey in Jiangxi Province. The potential spread of O. hupensis snails after the ood was predicted by an overlay analysis of the NDVI maps of ood-affected water body areas. In addition, the risk of schistosomiasis transmission was classied based on O. hupensis snail density data and the related NDVI. heavy rainfall along the middle and lower reaches of the Yangtze River has resulted in severe ood disasters along the Yangtze and around the Poyang Lake area in The highest recorded water level exceeded that of previous record highs following ooding in 1998, and has led to dyke collapse and water logging in many urban areas in Control of schistosomiasis in ood-affected areas has been impacted by increased contact with infested water during the ght against oods, and increased spreading of snails attributed to ooding. The aim of this study was to predict the expansion of O. hupensis snail habitats impacted by ooding and to assess the associated potential risk of schistosomiasis transmission in the Poyang Lake area using multi-source remote sensing image data and Sentinel-1B satellite radar images. polarized threshold calculations for the ooding time period [18]. SAR images were segmented by estimation of small backscattering coecient thresholds of water on SAR images. Segmentation results were saved as classication result les and classication results were post-processed. Incorrect extraction images were removed through a human-computer interaction system to yield the nal water body data. Sentinel-1B satellite data were processed using the SAR Scape module in ENVI software version 5.3, which mainly included radiometric calibration, ltering processing, terrain correction and geocoding. Radiometric calibration allows the transformation of image intensity into a backscattering coecient which was calculated using the following formula:


Snail distribution
Data on snail distribution was obtained from a snail survey carried out in Jiangxi Province in 2016 [10]. The geographical and environmental characteristics of snail distribution were extracted for the 13 counties (cites, districts) in which schistosomiasis is endemic, and the spatial database of snail distribution in Poyang Lake areas was created accordingly. A 75% subset of the snail distribution data was randomly selected as a model training dataset, and the remaining 25% were assigned as the model validation dataset.
Multi-source remote sensing images Remote sensing image data collected by the Landsat 8 satellite, with a spatial resolution of 30 m following geometric correction, convolution interpolation and resampling, were extracted from the NASA EarthData database (https://earthdata.nasa.gov/). Operational Land Imager (OLI) multi-wave remote sensing images were obtained for May 2016, the time period during which the snail survey was conducted. Sentinel-1B synthetic aperture radar (SAR) images were obtained from the European Space Agency (ESA) Earth Online database (https://earth.esa.int/), and remote sensing image data were obtained from the Sentinel-1B dual-polarized (HV + HH) SAR data for May 15, 2020 corresponding to the time period before the ood and for July 14, 2020, corresponding to the peak ood time period. An interferometric wide swath mode was assigned as the image mode.
Remote sensing image data inversion NDVI data were used as a measure of vegetation coverage in snail habitats, calculated using the following formula: where NIR indicates the re ectance in near-infrared wavelengths, and R indicates the re ectance in visible red wavelengths. Maximum and minimum NDVI values for snail habitats around the Poyang Lake area were calculated based on a dataset of 75% of snail distribution sites.
Flood-cased expansion of water body areas were identi ed using dual polarized threshold calculations for the ooding time period [18]. SAR images were segmented by estimation of small backscattering coe cient thresholds of water on SAR images. Segmentation results were saved as classi cation result les and classi cation results were post-processed. Incorrect extraction images were removed through a human-computer interaction system to yield the nal water body data. Sentinel-1B satellite data were processed using the SAR Scape module in ENVI software version 5.3, which mainly included radiometric calibration, ltering processing, terrain correction and geocoding. Radiometric calibration allows the transformation of image intensity into a backscattering coe cient which was calculated using the following formula: where is the backscattering coe cient of each pixel, A is the digital number of original images, K is an absolute calibration factor, and is an incident angle.
Because the extraction of the water body data may be interfered with by the intrinsic speckle noise in SAR images, a Frost ltering algorithm (5 × 5) was employed to control the output of a wave lter based on the local statistical characteristics of images. The speci c side-view mode of SAR images may lead to the occurrence of foreshortening, layover and shadow in mountains with terrain undulations, which affects the correct analysis of imaging data. Due to the occurrence of terrain undulations in the study area, terrain corrections of SAR images were performed to reduce the error of the water body data caused by geometric characteristics. Geocoding and radiometric calibration of ltered intensity data were also carried out using a digital elevation model (DEM) to generate backscattering coe cient images with dual-polarized (HV + HH) geographic coordinate systems. The index used for the water body data extraction was calculated based on the backscattering coe cients for HV and HH polarizations using the following formula:

Risk prediction of schistosomiasis transmission
Areas neighboring snail habitats which overlapped with ooded areas were extracted from SAR data to determine possible snail distribution zones following ooding. Snail density was estimated from snail habitat-neighboring areas, and the NDVI values corresponding to snail distribution were calculated to predict potential snail spread and associated schistosomiasis transmission risk. A 25% subset of observed snail breeding sites were selected as a validation dataset, and the distribution of snail breeding sites predicted using model results in ArcGIS software version 10.1. Model predictions corresponding to the validation dataset were extracted and the predictive accuracy of the model assessed by identifying snail breeding sites correctly identi ed by the model.

Results
Distribution of O. hupensis snails around Poyang Lake areas Results of a snail survey carried out in in Jiangxi Province in 2016 found snail habitats mainly distributed in 13 marshland and lake counties (cities, districts), including Nanchang, Xinjian, Jinxian, High-tech Zone of Nanchang City, Yongxiu, Gongqingcheng, Lushan, Lianxi, Hukou, Duchang, Poyang, Yugan and Wannian. A total of 1,257 habitat settings were identi ed, with marshlands consisting of 74.94% of all identi ed habitat settings. Among 763 snail habitats identi ed, 99.48% were marshland areas. In a marshland area covering 126,756.75 hm 2 , snail habitats accounted for78,900.89 hm 2 . Snails were detected in 519 settings with a density of living snails ranging from 0.000,1 to 6.497,2 snails/0.1 m 2 and 0.01% to 84.44% occurrence within surveyed frames. Among these, 12 settings were identi ed with a density of living snails at 1 snail/0.1 m 2 and greater located in northern coast of the Poyang Lake (seven settings in Lushan, four in Hukou and one in Yongxiu), and eight settings with 50% and higher occurrence of frames with living snailslocated in northern coast of the Poyang Lake ( ve settings in Lushan, two in Yongxiu and one in Hukou) ( Table 1).
Environmental identi cation for suitable snail habitats NDVI values were estimated with a range of -1 to 0.61 in the Poyang Lake area based on Landsat 8 remote sensing image data. A total of 75% of snail distribution sites were randomly sampled from snail habitats identi ed by eld surveys as a training dataset to extract the optimum thresholds of the maximum and minimum value of NDVI, with 95% con dential intervals from 0.08 to 0.59. Figure 2 shows the distribution of suitable snail habitats.

Extraction of ood-cased water body expansion
Radar echo intensity was determined by brightness in SAR data. Because of low echo intensity in water bodies and high echo intensity in corresponding to land areas, water body areas in SAR images appeared as dark or black and the land areas as grayish white or dark grey. Pre-processed SAR images from Sentinel-1B satellite images for the time period before and after ooding are shown in Figure 3. The speckle noise was effectively inhibited, and a more obvious water-land boundary was seen on original radar images in which the water and land were well differentiated and the water pro le was more distinct. Figure 4 shows a histogram of pre-processed SAR image scattering values before and after ooding. There are two apparent peaks in the histograms in images presented in Figure 4, with segmented SAR image thresholds of between -25.5 and -24 dB before observed by the visual interpretation before and after ooding.
Segmentation results were saved as classi cation results and transformed to a vector le. Following post-classi cation data processing, incorrect extraction images were removed through a human-computer interaction to generate a nal water body dataset for the Poyang Lake area for the May 15 and July 14, 2020 time period.
Changes in water body areas before and after the ood disaster The distribution of water bodies in the Poyang Lake area was overlapped before and after ooding. The blue areas in Figure 5 indicate the distribution of water bodies before the ood, and red areas describe the expansion of water bodies during ooding. Examination of the main body of the Poyang Lake and neighboring water areas from Sentinel-1B SAR images showed that the water area was approximately 2,207 km 2 on May 15 and 4,403 km 2 on July 14, an increase of 2,196 km 2 compared to May and an increase of 25.4% on the historical mean level during the same period (3,510 km 2 ). The water body expanded by approximately 99.5% after ooding relative to the water body areas before ooding, mainly identi ed in Xinjian, Duchang, Poyang, Yongxiu and Yugan ( Figure 5).
Risk Predictions of potential snail diffusion and associated schistosomiasis transmission after the ood Data indicating ood-affected water areas were transformed into a binary image of potential snail habitat distribution. Areas of predicted snail diffusion exhibited a patchy clustered distribution. After submersion of snail habitats following ooding, snail habitats were likely to be in neighboring settings. Snail distribution was predicted to cover an area of approximately 759 km 2 , mainly occurring in the east of Yongxiu, south of Lushan, southwestern Poyang, southwestern Duchang, northwestern Xinjian and northwestern Yugan. This suggested that areas of the possible snail diffusion were predominately concentrated in marshlands around Poyang Lake ( Figure 6).
NDVI values of suitable snail habitats were calculated based on snail density data, with values ranging from 0.15 to 0.35 in high density snail habitats, 0.35 to 0.42 in medium density snail habitats, and from 0.08 to 0.15 greater than 0.42 in low density snail habitats. NDVI values of ood-affected areas were estimated and the risk of potential snail spread classi ed accordingly. We found areas at high risk of snail distribution predominantly located in northwestern Yongxiu, southwestern Duchang, south of Lushan and southwestern Poyang (Figure 7). These high risk areas are also indicative of neighboring areas suitable for snail breeding where snail habitats are likely to emerge following ooding, and with potential for schistosomiasis transmission. Validation of predicted snail habitats was carried out using observational data on snail breeding habitats in order to assess the predictive performance of models NDVI value of snail habitats ranged from 0.1 to 0.52 using the 25% validation dataset, with 87% prediction accuracy indicating that the NDVI values are a good predictor of snail habitats.
Multiple dykes were observed to collapse following ooding in Jiangxi Province. Three sites in Yongxiu and Poyang counties where dykes had collapsed were found to overlap with areas classi ed as snail distribution risk areas. The sites in Yongxiu and Poyang County where dykes were observed to have collapsed was predicted to be medium risk areas of snail distribution, (Figure 7), suggesting that snail habitats were likely to emerge in both sites.

Discussion
Schistosomiasis is a neglected tropical parasitic disease strongly associated with ecological and geographic factors. Changes in the natural environments which affect the breeding, reproduction and distribution of intermediate host snails, are likely to impact transmission of schistosomiasis [19,20]. Previous research has found snail distribution to be closely correlated with vegetation, humidity and temperature, as well as with human and livestock activities with O. hupensis snails favoring marshlands, ponds and ditches. As the geographical location of schistosomiasis is determined by snail distribution, surveillance of snail habitats is an important component in the national schistosomiasis control program in China.
Flooding frequently occurs along the south and middle and lower reaches of the Yangtze River, areas which are endemic for schistosomiasis. Flooding is considered to be one of the most important natural factors impacting the rebound of schistosomiasis in endemic foci. Periods of ooding impact the geographical expanse, reproduction and growth of O. hupensis snails, particularly juvenile snails and snail eggs, which spread via water ow [21][22][23]. Additionally, ood discharge or dyke collapse may facilitate expansion of snail populations into embankments resulting in the re-emergence of snails in areas where snails had not previously been present [24].
Currently, the endemicity of schistosomiasis in China is declining and the country is progressing towards elimination [11]. Many challenges to schistosomiasis elimination remain, however, due to natural, biological and social factors [6]. Prevalence of schistosomiasis is high in 11 counties of the 13 counties (districts) in the Poyang Lake area in Jiangxi Province, with the exception of the high-tech zone of Nanchang City and Wannian County. In 2018, three egg positive individuals were identi ed in Yugan County (eight egg positive cases in China) and two egg positive bovines identi ed in Duchang County (two egg positive bovines in China) in 2019, suggesting that risk of schistosomiasis transmission of remains in the Poyang Lake area. Interestingly, these areas were identi ed as high risk transmission areas by this study. In areas characterized by vast marshlands, multiple grasslands, dense vegetation, di culty in management of water levels, and extensive snail distribution where schistosomiasis was once hyper-endemic and transmission had been controlled or interrupted [9], S. japonicum transmission potential remains due to the distribution of O. hupensis snails following ooding. Relaxing of control interventions would therefore likely result in reemergence of schistosomiasis infection.
Remote sensing image data has been widely employed for the surveillance schistosomiasis and intermediate host snails habitats [26][27][28][29]. Satellite-based remote sensing collects ecological and geographical data, such as land coverage, vegetation, soil type, surface moisture and rainfall, which may be used to monitor environmental changes, thereby assessing the suitability of snail breeding habitats of association potential of schistosomiasis transmission [30][31][32]. Some limitations are inherent in data of this type however such as a short wavelength and potential effects of cloud amount and severe weather. Most notably, during ooding continuous cloudy and rainy conditions may lead to failure in the accurate acquisition of image data in ood-affected areas. Radar images, characterized by full-time and full-weather, high-coverage, high-resolution observations and high revisit rate are not affected by meteorological conditions or light levels, and have been widely employed in the elds of disaster monitoring, agriculture and oceanography [33,34].
In the current study, the distribution snail habitats was estimated using Landsat 8 satellite remote sensing image data, and areas where water body overlapped before and after ooding identi ed using Sentinel-1B dual-polarized image data. The water body area was estimated to have expanded by approximately 99.5% in the Poyang Lake region after ooding relative to the water body area before ooding. Snail habitats are likely to emerge in oodaffected areas after ooding has receded. All of these regions were found to be adjacent to, or connected with, original snail habitats, predominantly distributed in southwest and northwest of Poyang Lake. Snail distribution was predicted to mainly occur mainly in Yongxiu County, Lushan County, Poyang County, Duchang County, Xinjian District and Yugan County. It is predicted that multiple regions at high risk of schistosomiasis transmission are located in Yongxiu County, Duchang County, Lushan County and Poyang County. The classi cation of potential snail habitats is consistent with the spatial distribution of schistosomiasis transmission risk in the Poyang Lake regions based on schistosomiasis case reports [29,35].

Conclusion
Remote sensing techniques were an effective quantitative tool for the rapid assessment of snail distribution, providing insights into e schistosomiasis risk for informing control programs. It is recommended that monitoring of O. hupensis and epidemiological surveys of S. japonicum infections should be conducted by schistosomiasis control institutions in the coming 2 to 3 years, so as to prevent the spreading of snail populations and to reduce the risk of transmission of schistosomiasis.  Figure 1 Geographical location of the study area. Poyang Lake is located in north of Jiangxi Province and northern coast of the middle and lower reaches of the Yangtze River and the natural geographical features of the Poyang Lake are very suitable for snail breeding. A total of 13 counties (districts) around the Poyang Lake areas, where schistosomiasis is prevalent, were included in this study. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 2
Distribution of snail habitats in Poyang Lake areas. Snail data were captured from the 2016 snail survey in Jiangxi Province. Normalized Difference Vegetation Index (NDVI) was employed to measure the vegetation coverage in snail habitats, which was collected from Landsat 8 satellite remote sensing images from the NASA EarthData database Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.   Changes of water bodies before and after the ood disaster in Poyang Lake areas. The blue regions indicate the water bodies on May 15, 2020 before the ood, and the red regions indicate the expansion of water bodies on July 15 during the ood period. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 6
Prediction of snail habitats based on ood-affected settings. Snail spread might cover an area of approximately 759 km2, which mainly occurred in east of Yongxiu, south of Lushan, southwestern Poyang, southwestern Duchang, northwestern Xinjian and northwestern Yugan. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.

Figure 7
Risk classi cation of spread of Oncomelania hupensis. The NDVI values of suitable snail habitats were calculated based on snail density data, and the risk was classi ed as high density, medium density and low density snail habitats with reference to the NDIV values. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors.