Answers are grounded in this paper and cite their sources.
Background Surveillance of Oncomelania hupensis relies on labor-intensive surveys with limited spatial coverage. Presence–absence models do not characterize variation in snail density. We modeled live-snail density spatially using corrected meteorological–hydrological predictors to guide surveillance. Methods We analyzed 2023 snail-surveillance data from Dongting Lake, Poyang Lake, and the Anhui section of the Yangtze River. Integrated nested Laplace approximation–stochastic partial differential equation (INLA–SPDE) models incorporated environmental covariates and spatial effects, with deviance information criterion–guided mesh and covariate selection. We compared corrected and uncorrected models with ordinary kriging, random forest, and random forest–kriging using spatial block cross-validation and an independent survey of 900 quadrats at nine marshland sites. Results The corrected INLA–SPDE model achieved an external root mean square error of 0.064 and mean absolute error (MAE) of 0.052 individuals/0.11 m 2 , reductions of 23.81% and 23.53%, respectively, versus the uncorrected model. Errors were lowest in Dongting Lake, followed by Poyang Lake and Anhui, but predictions exceeded observations at all nine sites. Predicted potential high-density areas (> 0.300 individuals/0.11 m 2 ) totaled 445.58 km 2 in Dongting Lake, 170.02 km 2 in Poyang Lake, and 82.55 km 2 in Anhui. Temperature, inundation duration, and vegetation effects were similar, but soil phosphorus and wilderness-distance effects varied regionally. Conclusions Corrected inputs improved performance in the reported comparisons. Spatial estimates may guide targeted surveillance and field verification, but the systematic overprediction and unvalidated high-density category limit direct operational interpretation.
No takes yet. Share an insight, caveat, or question.
Gong et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: