Environmental Determinants for Snail Density in Dongting Lake Region: An Ecological Study Incorporating Spatial Regression

ABSTRACT.

This study explored the environmental determinants of different months on snail density measured in April at different types of snail habitats (marshlands, inner embankments, and hills) by considering spatial effects. Data were gathered from surveys on snails that were conducted in Hunan Province in April 2016, and information was collected on environmental variables. To investigate the environmental factors influencing snail density in various types of snail habitats, the ordinary least square model, spatial lag model, and spatial error model were all used. The environmental determinants for snail density showed different effects in the three types of snail habitats. In marshlands, snail density measured in April was associated positively with the normalized difference vegetation index (NDVI) and was associated negatively with flooding duration and annual hours of sunshine. Extreme temperatures correlated strongly to snail density measured in April (P < 0.05). In areas inside embankments, snail density measured in April increased with a decreased distance between snail habitat and the nearest river (P < 0.05). In hills, extreme heat, annual hours of sunshine, NDVI in September, and annual average land surface temperature (LST) were associated negatively with snail density measured in April, whereas index of moisture (IM) was associated positively with snail density measured in April (P < 0.05). The effects of LST and hours of sunshine on snail density measured in April varied with months of the year in the three different types of snail habitats (P < 0.05). Our study might provide a theoretical foundation for preventing snail transmission and subsequent spread of schistosomiasis.

INTRODUCTION

Schistosomiasis is a chronic and debilitating disease that exacerbates poverty in endemic areas.¹ According to the WHO, 290.8 million people required preventive treatment worldwide, and schistosomiasis contributed to more than 200,000 deaths in 2018.² In China, schistosomiasis is caused by Schistosoma japonicum, with Oncomelania hupensis as the unique intermediate host.^3–6 Although various control measures have controlled schistosomiasis successfully in most parts of China,⁷ the elimination of schistosomiasis remains a challenge.⁸

The distribution of O. hupensis snails is largely determined by the ecology and topography of an endemic area.⁹ In China, schistosomiasis-endemic areas are classified into three types according to ecogeographic characteristics: 1) marshland and lake regions, 2) mountainous and hilly regions, and 3) plain regions with waterway networks,¹⁰ and the type of marshlands and lake regions can be further divided into two subtypes of marshlands outside embankments and areas inside embankments according to the variability of hydrology.¹¹ Marshlands outside embankments are frequently flooded in summer; areas inside embankments are rarely flooded. In addition, the different geographic environments between subtypes and the effect of human activities inside embankments might contribute to the varied impact on snail density. Hunan Province retains the first two types.¹¹ Snail density is closely associated with environmental factors such as water distribution, land surface temperature (LST), and vegetation coverage.¹² Researchers have explored the effects of environmental factors on snail density, but their differences with regard to the types or subtypes of snail habitats and seasonal variations remain unclear.¹³

The dependent data have spatial autocorrelation and spatial dependence, which was ignored in previous studies using correlation analysis, normal regression analysis, generalized additive models, or Bayesian models and might bias study results.^14–17 According to previous studies, spatial autocorrelation will increase precision falsely in the estimation of coefficients if spatial autocorrelation is not taken into account.¹⁸ Hence, if there is clear evidence that the data are spatially dependent, the spatial regression models should be accounted for. The spatial lag model (SLM) and the spatial error model (SEM) are the most commonly used spatial regression models, preventing some statistic problems (e.g., unstable parameters and unreliable significance tests),¹⁴ and scholars in the social sciences have done a lot of research with them.¹⁹ This study focused on determining the impacts of environmental factors on snail density measured in April, and their variations with months of the year and types of snail habitats by considering spatial effects, which might provide a theoretical basis for controlling snail and subsequent transmission of schistosomiasis.

MATERIALS AND METHODS

Study area and data source of snail density.

This study was conducted in Hunan Province, China, situated between 24°37′ to 30°08′ N latitude and 108°47′ to 114°16′ E longitude. This area is classified as the East Asian monsoon climate zone and the warm climate, abundant rainfall, and vegetation provide a suitable environment for the survival and breeding of O. hupensis.²⁰ As a result, S. japonicum has been endemic there for a long time.²¹ Previous studies have shown there are three subtypes of snail habitats in Hunan Province.^10,11 In our study, an inner embankment is defined as an area close to the lake but surrounded by protective embankments, where snails can migrate effortlessly from marshlands to inner embankments by sluice gates. The marshland and inner embankment have rather constant topography. The habitat area of snails is large and the groundwater level is relatively stable.²² Compared with the marshland and inner embankment, snails are more isolated and dispersed in the mountainous and hilly regions, and their primary habitats are wasteland, river flats, and ditches.²³ These plots are tiny and are scattered sporadically, with uneven terrain and frequent fluctuation in groundwater level. The survival of snails depends mainly on seasonal surface water. Snails also find it difficult to migrate because of the difficult terrain in hilly areas, and their population distribution is often steady. However, during the flood season, snails might migrate along waterways and ditches, making the elimination of snails in the mountainous and hilly regions more difficult.²⁴

The snail surveys were carried out in April 2016. Systematic quadrat sampling combined with random quadrant sampling was adopted for snail collection. For the current habitats, systematic quadrat sampling was used. If no snails were found, random quadrant sampling was investigated using 0.11-m² quadrats.²⁵ The frames were set 20 × 20 m apart when the area outside the embankments was < 200,000 km² or 50 × 30 m apart when the area was > 200,000 km². For the inner embankments and hilly regions, the frames were set 5 × 10 m apart. After the exclusion of the sites missing snail density data (one site in marshlands and two sites in hills), there were 1,564 sites of current snail habitats, including 829 in marshlands, 200 inside embankments, and 535 in hills (Figure 1). All captured snails were examined by microscope after crushability to identify living ones.¹¹ The density of snails was calculated as the average number of living snails per frame (no. of snails/0.11 m²). Logarithmic transformation was performed to normalize the data. The latitude and longitude of snail-infested sites were obtained by a handheld global positioning system.

Figure 1.

Geographic distribution of snails in Hunan Province.

Data sources of environmental variables.

Water levels.

Water level data were obtained from the Hunan Flood Prevention Information System, including daily water levels at 8 am at 14 hydrological stations from January 1 to December 31, 2015, and were imported into ArcGIS version 10.0 (ESRI Inc., Redlands, CA). Ordinary Kriging interpolation was performed, and the spherical model was applied as the semi-variogram. Data for elevation were collected through Google Earth (Google Ltd., http://www.google.com/earth/), with a resolution of 90 m. Current snail habitats in marshlands were considered to be flooded if the water level was higher than the elevation. Flooding days were defined as days from the onset of flooding to the offset in a year. The flooding duration was calculated as the flooding days divided by 30.

Distance to the nearest river.

The raster data of major rivers, which are typically in the third order or higher, were downloaded (http://bbs.3s001.com/thread-97987-1-1.html [accessed May 17, 2012]). The point-to-straight distance tool in ArcGIS was used to calculate the distance from a snail habitat of an inner embankment (an area close to the lake but surrounded by protective embankments) to the nearest river. Similarly, the distance to marshlands refers to the distance from a snail habitat of an inner embankment to marshlands.

Slope.

The slopes of snail habitats in hills were calculated using the function Slope in the Spatial Analyst toolbox in ArcGIS. The raster data of elevation was used as the base map with a resolution of 90 m.

The definitions and sources of other data (e.g., normalized difference vegetation index (NDVI], LST, index of moisture [IM], and hours of sunshine) are presented in Table 1. All environmental considered in this study were presented in Supplemental Table 1.

Table 1.

Definition and data sources

Variable	Definition	Label	Resolution	Data source
Flooding duration (month)	Duration of flooding of snail habitats in marshlands	Flooding duration	–	Hunan Flood Prevention Information System
Distance to the nearest river (km)	Distance to the nearest river	Distance to the nearest river	–	http://bbs.3s001.com/thread-97987-1-1.html (accessed May 17, 2012)
IM	Index of moisture, from the station’s construction to the mid-1990s	IM	500 m	http://www.resdc.cn/ (accessed June 30, 2020)
NDVI	Monthly and annual average NDVI, 2015	NDVI	250 m	http://www.resdc.cn/ (accessed June 30, 2020)
Sunshine hours	Monthly and annual hours of sunshine, 2015	Sunshine hours	1 km	http://www.geodata.cn/data/datadetails.html?dataguid=278608814236557&docId=7492 (accessed December 31, 2015)
Extreme cold (°C)	Minimum temperature in January, 2016	Extreme cold	1 km	https://worldclim.org/data/index.html# (accessed January 1, 2020)
Extreme heat (°C)	Maximum temperature in August 2015	Extreme heat	500 m	https://worldclim.org/data/index.html# (accessed January 1, 2020)
LST (°C)	Monthly and annual average LST, 2015	LST	5,600 m	https://data.tpdc.ac.cn/zh-hans/data/055dfa65-e097-4000-9bda-992def32969f/ (accessed January 1, 2020)
Ln (snail density)	Natural log-transformed snail density, April 2016	Ln (snail density)	–	Hunan Institute for Schistosomiasis Control

IM = index of moisture; LST = land surface temperature; NDVI = normalized difference vegetation index.

Statistical analysis.

First, we calculated descriptive statistics for snail density and environmental variables. Standard ordinary least squares (OLS) linear regressions were fitted to explore the association of the natural log (snail density) and the environmental determinants in different snail habitats. Both univariate and multivariate models were fitted, and only the significant variables (P < 0.1) from univariate models were included in multivariable models. The environmental determinants included flooding duration, distance to marshlands, distance to the nearest river, land use, slope, accumulated temperature > 0°C, annual average temperature, IM, annual precipitation, hours of sunshine, NDVI, extreme temperatures, and LST. All variables were checked for multicollinearity, and the variables with multicollinearity were excluded in multivariate analysis.

Second, two spatial regression models, the SLM and the SEM, were fitted. To control and prevent multicollinearity, only variables found to be statistically significant in the OLS regression model were used in these two spatial models. The SLM incorporates spatial autocorrelation directly into the model by including a spatial lag term. The SEM works similarly to the SLM, except the spatial autocorrelation term applies to the error terms rather than their dependent variable values. GeoDA version 1.18 (University of Chicago, Chicago, IL) was used to construct the spatial weight matrix based on distance. The geometric center was the longitude and latitude coordinates of the snail habitats, and the method was the distance band (specified bandwidth).

The SLM can be expressed as

$$Y=\rho WY+X\beta +\epsilon ,\epsilon \sim N(0,{\sigma }^2),$$ (1)

where Y is an $n\times 1$-dimensional dependent-variable vector, representing snail density; X is the independent variable; $\rho$ is the spatial regression coefficient; W is the space weight matrix; $\beta$ is the regression coefficient; and $\epsilon$ is the random error term, which is generally considered to be normally distributed.

The SEM can be expressed as

$$Y=X\beta +\lambda Wu+\epsilon ,\epsilon \sim N(0,{\sigma }^2),$$ (2)

where u is the random error vector, $\lambda$ is the spatial error coefficient of the $n\times 1$-order dependent-variable vector, and W is the spatial weight matrix of $n\times n$ order.

Third, Moran’s I test was adopted to examine the spatial autocorrelation of snail density. If a spatial correlation existed, the Lagrange multiplier (LM) test was adopted to reveal spatial dependence in the form of a spatial lag-dependent variable and spatial error dependence. When the lag dependence and error dependence were both found to be statistically significant using the LM test, the robust LM test was used to reveal additional spatial dependence. If any of the test outcomes was statistically significant, the spatial model was chosen; models with a greater statistical value are a better choice.²⁶

RESULTS

In our study, the median snail density measured in April in marshlands, inner embankments, and hills was 0.15 (range, 2.54–3.46 × 10^–4)/0.11 m², 0.27 (range, 5.33–1.39 × 10^–4)/0.11 m², and 0.41 (range, 2.01–2.16 × 10^–3)/0.11 m², respectively. The average flooding duration was about 3 months in marshlands. Snails inside embankments were mainly distributed in areas with an average distance of 3.26 km to the nearest river. The mean slope in hills was 1.23°. The mean annual hours of sunshine was 1,367.71 hours in marshlands, 1,412.72 hours inside embankments, and 1,336.02 hours in hills, respectively. Extreme cold and extreme heat in marshlands were ∼2.45°C and 31.35°C, respectively. Extreme heat in hills was ∼31.41°C. The annual average LST in hills was 16.78°C (Table 2). The descriptive statistics for environmental determinants with significance in different months of the year were presented in Supplemental Table 2.

Table 2.

Description of environmental variables in different types of snail habitats

Variables	Marshlands (n = 829)		Inner embankments (n = 200)		Hills (n = 535)
	Range	Mean ± SD	Range	Mean ± SD	Range	Mean ± SD
Flooding duration (month)	0.00–9.17	3.03 ± 3.5	–	–	–	–
Distance to the nearest river (km)	–	–	0.2–14.65	3.26 ± 3.01	–	–
Slope (°)	–	–	–	–	0.00–12.03	1.23 ± 1.45
Aat0 (°C/day)	5,924.90–6,240.50	6,114.02 ± 68.69	5,861.00–6,240.50	6,056.37 ± 64.16	5,781.20–6,305.50	6,077.75 ± 112.76
Annual average temperature (°C)	16.30–17.10	16.76 ± 0.19	17.87–18.79	18.35 ± 0.15	17.42–18.98	18.49 ± 0.20
Moisture index	39.04–59.66	46.71 ± 3.64	39.44–64.04	45.25 ± 4.89	44.54–63.54	50.67 ± 3.87
Annual precipitation (mm)	1,368.17–1,674.31	1,545.96 ± 82.01	1,366.53–1,697.90	1,493.81 ± 92.14	1,312.88–1,710.12	1,450.40 ± 112.37
Annual hours of sunshine	1,266.00–1,515.00	1,367.71 ± 57.24	1,292.00–1,493.00	1,412.72 ± 45.09	1,209.00–1,468.00	1,336.02 ± 58.84
Annual NDVI	0.02–0.90	0.63 ± 0.18	0.04–0.90	0.73 ± 0.14	0.15–0.90	0.72 ± 0.11
Extreme cold (°C)	1.69–3.16	2.45 ± 0.33	1.89–3.05	2.22 ± 0.31	1.70–3.67	2.35 ± 0.41
Extreme heat (°C)	30.88–31.91	31.35 ± 0.17	31.07–32.02	31.41 ± 0.14	30.80–32.69	31.41 ± 0.36
Annual average LST (°C)	14.60–17.24	16.11 ± 0.53	15.26–17.21	16.59 ± 0.48	15.22–18.22	16.78 ± 0.49

AAt0 = accumulated temperature > 0°C; LST = land surface temperature; NDVI = normalized difference vegetation index.

A separate analysis was conducted for the environmental determinants of snail density in marshlands, inside embankments, and on hills. For the environmental factors related to snail density in marshlands, the White test indicated the existence of heteroskedasticity (P < 0.001), and the Moran’s I analysis indicated a significant spatial autocorrelation (Moran’s I = 0.993, P < 0.01) among snail densities. The LM test showed that both the lag dependence and error dependence were statistically significant (P < 0.001). The robust LM test showed a greater statistical value for the SLM (robust LM = 1,673.961) than the SEM (robust LM = 6.306), indicating that the SLM was a better choice compared with the SEM, and the SLM had a lower Akaike information criterion (AIC) value (AIC = 701.405) than the SEM (AIC = 1,480.45). Similar results were found for the environmental determinants of snail density inside embankments and on hills. Therefore, the SLM was adopted to determine the effects of the environmental factors on snail density.

Table 3 shows the results of the multivariable SLM for environmental determinants in marshlands. Eleven variables were related significantly to snail density measured in April (P < 0.05). Flooding duration, annual hours of sunshine, and LST in October and November were associated negatively with snail density measured in April, whereas NDVI in April and LST in February, March, May, and June were associated positively with snail density measured in April. Extreme cold and extreme heat related strongly to snail density measured in April, with regression coefficients of 0.9315 and 0.594, respectively.

Table 3.

Multivariable spatial lag model for environmental factors associated with snail density measured in April in marshlands

Variables	Coefficient	z Value	P value
Constant	−0.0004	−0.0995	0.9208
Flooding duration	−0.0481	−15.2170	< 0.001
Annual average temperature	−0.1669	−1.5384	0.1239
Annual hours of sunshine	−0.0056	−23.5410	< 0.001
NDVI in April	0.0555	2.3202	0.0203
Extreme cold	0.9305	18.7203	< 0.001
Extreme heat	0.5943	14.3604	< 0.001
LST in February	0.0932	6.0150	< 0.001
LST in March	0.0839	6.7252	< 0.001
LST in May	0.0193	3.4726	< 0.001
LST in June	0.2892	6.5513	< 0.001
LST in July	0.0335	0.5361	0.5919
LST in October	−0.4992	−8.9954	< 0.001
LST in November	−0.2782	−11.2760	< 0.001

LST = land surface temperature; NDVI = normalized difference vegetation index.

Table 4 shows the results of the multivariable SLM for environmental determinants inside embankments. Fifteen variables were related closely to snail density measured in April (P < 0.05). The distance to the nearest river had a negative impact on snail density measured in April. LST in June, November, and December; hours of sunshine in January, August, and November; and NDVI in January, April, and August were associated positively with snail density measured in April. LST in April; hours of sunshine in July, October, and December; and NDVI in February were associated negatively with snail density measured in April.

Table 4.

Multivariable spatial lag model for environmental factors associated with snail density measured in April inside embankments

Variables	Coefficient	z Value	P value
Constant	3.9611	0.6292	0.5292
Moisture index	0.0463	1.9528	0.0508
Distance to the nearest river	−0.2832	−15.0320	< 0.001
Distance to marshlands	−0.0426	−1.9115	0.0559
Sunshine hours in January	0.2838	2.2364	0.0253
Sunshine hours in July	−0.2157	−4.3911	< 0.001
Sunshine hours in August	0.5371	5.9259	< 0.001
Sunshine hours in October	−0.4817	−5.3816	< 0.001
Sunshine hours in November	0.3522	2.0500	0.0403
Sunshine hours in December	−0.6778	−5.1612	< 0.001
LST in April	−0.4318	−5.2392	< 0.001
LST in June	0.6248	5.1599	< 0.001
LST in November	0.1832	2.8367	0.0045
LST in December	0.3813	3.3169	< 0.001
NDVI in January	5.0492	2.8355	0.0045
NDVI in February	−7.8196	−3.6446	< 0.001
NDVI in April	1.4669	2.0050	0.0449
NDVI in August	1.6350	3.1263	0.0017
NDVI in December	−2.0861	−1.7046	0.0882

LST = land surface temperature; NDVI = normalized difference vegetation index.

Table 5 shows the results of the multivariable SLM for the environmental determinants on hills. Eleven variables were related significantly to snail density measured in April (P < 0.05). Extreme heat, annual hours of sunshine, NDVI in September, LST in January and February, and annual average LST were associated negatively with snail density measured in April. Hours of sunshine in February and May, IM, and LST in May and July were associated positively with snail density measured in April.

Table 5.

Multivariable spatial lag model for environmental factors associated with snail density measured in April in hills

Variables	Coefficient	z Value	P value
Constant	25.3899	1.5603	0.1186
Moisture index	0.1108	5.8849	< 0.001
Extreme heat	−0.8469	−2.6217	0.0087
Sunshine hours in February	0.2486	3.9635	< 0.001
Sunshine hours in March	0.1291	1.3041	0.1922
Sunshine hours in May	0.2024	4.6198	< 0.001
Sunshine hours in June	0.1248	1.6918	0.0906
Sunshine hours in August	0.1092	1.3175	0.1876
Sunshine hours in October	0.0873	0.9995	0.3175
Annual hours of sunshine	−0.0749	−3.7905	< 0.001
NDVI in September	−0.7142	−2.1359	0.0326
LST in January	−0.7628	−4.7687	< 0.001
LST in February	−0.2424	−2.6882	0.0071
LST in May	0.2244	2.4093	0.0159
LST in July	0.3737	2.6168	0.0088
Annual average LST	−0.9133	−3.0876	0.0020

LST = land surface temperature; NDVI = normalized difference vegetation index.

DISCUSSION

This study investigated environmental determinants of different months for snail density measured in April at different types of snail habitats using the SLM. The SLM avoids unstable parameters and unreliable significance tests, and provides information on spatial relationships among the parameters. Our results showed that the effects of NDVI, flooding duration, extreme temperatures, LST, distance between snail habitats and rivers, and IM on snail density measured in April varied among different types of snail habitats (marshlands, inner embankments, hills).

In marshlands, snail density measured in April was associated negatively with flooding duration. Snail eggs must hatch in water or damp soil, and then develop in the water for the first few weeks. However, long-term flooding may lead to developmental disorders of juvenile snails and the death of adult snails.²⁷

In our study, the NDVI in April and August had a stimulatory effect on snails measured in April inside embankments. However, the NDVI in September had a negative effect on snail density measured in April on hills. This might be related to the fact that sufficient sunlight was needed to maintain the temperature of snail habitats when the temperature was low in autumn and winter. Similarly, the NDVI inside embankments had different effects in January and February. This might be because vegetation in January is helpful to maintain a suitable temperature for the hibernation of snails in winter.²⁸ However, snails begin to wake up in February, and less vegetation ensured more sunlight and a temperature increase in the spring. In marshlands, the difference of the NDVI among snail habitats was only observed in April. This is possibly a result of the sparse vegetation before April and the high vegetation density after April in marshlands, resulting in a similarity of vegetation density among snail habitats.

The distance to the nearest river was found to be an important environmental determinant for snail density measured in April inside embankments. The distance to the nearest river can change the humidity condition of snail habitats, which is essential to snail growth.¹³ A previous study found that snails were distributed mainly in the 1-km buffer zone of the main channel of the Beijing-Hangzhou Grand Canal.²⁹ In our study, snails were distributed at an average distance of 3.26 km to rivers. This is probably because the environment inside embankments with appropriate living conditions was suitable for snails.

The IM had a positive impact on snail density measured in April on hills. The effect of the IM on snail density is realized mainly by soil moisture, which is affected by rainfall, vegetation, soil, topography, and other factors.³⁰ Hills have the lowest soil moisture of the three types of snail habitats because of the low rainfall and the red soil’s poor ability to retain water,¹⁵ making the IM a more important factor affecting snail density than the temperature in relatively arid hilly areas.¹⁷

The relationship between changes in temperature and snail density measured in April is complex.³¹ Extreme cold showed a strong positive correlation with snail density measured in April in marshlands, and extreme heat showed a negative correlation with snail density measured in April on hills. A possible reason for the results is that the most optimal temperature for survival is 27°C, indicating that either too high or too low a temperature is unfavorable to snails.¹⁹ We found that LST in marshlands correlated negatively with snail density in relatively cold weather. This might be because snails begin to hibernate around November, leading to the inhibition of temperature to snails.²⁸ A previous study indicated that annual average LST had both positive and negative effects on snails at different times and regions when the annual average LST was between 16.74°C and 19°C.¹³ In our study, the mean annual average LST was 16.78°C on hills. As a result, the annual average LST showed a negative effect on snail density measured in April in hilly areas. Inside embankments, the LST in April showed a negative effect on snail density, whereas the LST in November and December showed a positive effect on snail density. This might be a result of complex human activities inside embankments. The LST on hills measured in April showed a negative effect on snail density measured in January and February, and a positive effect in July, which was inconsistent with results from a previous study.⁹ Thus, further studies are needed. Moreover, extreme heat related positively to snail density measured in April in marshlands. This might be related to the flooding in marshlands during the summer, resulting in snails being submerged in the cool water.

Results inside embankments showed that hours of sunshine had a positive effect on snail density measured in April in the colder months (January and November), but a negative effect in July. Moreover, the hours of sunshine on hills measured in April was related positively to snail density in February and May. This might be because optimal hours of sunshine can provide suitable living conditions for snails, but above-optimal hours of sunshine can inhibit the survival of snails.³² In our study, the factor of hours of sunshine was not a major determinant in marshlands, whereas snail density measured in April was affected significantly by the hours of sunshine on hills, which was consistent with the results from a previous study.³² There are some points for further consideration. First, the hours of sunshine inside embankments measured in April was related negatively to snail density in December. This might be a result of the disturbance of high temperatures associated with sunlight during the hibernation of snails. Also, the results of hours of sunshine on embankments were inconsistent with previous studies in August and October.¹⁵ Further studies are needed.

Environmental modification is the core strategy for the control and elimination of snail population.²¹ We suggest the following environmental modification strategies in light of our findings. First, snails need a moderate amount of moisture to survive, and it is difficult for snails to survive in dry, hot soil. So, to allow for direct sunlight and to reduce soil moisture, the mechanical soil burial method can be used to dig deep underground.⁷ This will help control snail populations. Second, although snails prefer to live in moist soil, they cannot develop normally when submerged in water for an extended period. Our results also showed that flooding duration was associated negatively with snail density. Therefore, new fishponds can be excavated to submerge snails completely with water.³³ Third, given that snails require moist soil, appropriate temperature, vegetation, and other factors for reproducing, the ditch or channel with snail distribution can be hardened with cement or concrete to change the environment in which snails breed and thus deprive them of the requirements for breeding.³⁴ Fourth, according to our results, we can modify the microecological environment of snails, such as the IM, NVDI, and other environments factors of snail habitats by clearing lush surface vegetation and planting trees or interplanting crops, hence controlling the snail population and eliminating the snails.³⁵ In addition, by using environmental data from previous years, it is possible to estimate the snail density for the future year and intensify the control and elimination efforts of snails in potential high-density areas.

There are some limitations to our study. First, we did not explore the nonlinear relationships between environmental determinants and snail density. Second, we did not consider temporal autocorrelation in our spatial regression models, which might lead to biased estimates of standard errors of parameters. Third, overfitting might have occurred because of the limited sample size. We included monthly data for variables instead of annual averages to maximize the utility of information from data. Fourth, some environmental variables were extracted and estimated from remote sensing images. The resolution of these data might affect the accuracy of our results. Last, because the snail density was obtained using a cross-sectional survey, all results should be interpreted with caution.

CONCLUSION

The snail density measured in April was influenced by environmental determinants, which showed substantial variations in different snail habitats and months of the year. We found that the NDVI, flooding duration, extreme temperatures, distance between snail habitat and the nearest river, annual hours of sunshine, annual average LST, and the IM were important determinants of April snail density. The impacts of the NDVI, LST, and hours of sunshine on April snail density varied in different months of the year. According to our findings, various environmental modifications could be used to control and monitor snail populations.

Note: Supplemental tables appear at www.ajtmh.org.