Species distribution modeling for disease ecology: a multi-scale case study for schistosomiasis host snails in Brazil

Abstract

Species distribution models (SDMs) are increasingly popular tools for profiling disease risk in ecology, particularly for infectious diseases of public health importance that include an obligate non-human host in their transmission cycle. SDMs can create high-resolution maps of host distribution across geographical scales, reflecting baseline risk of disease. However, as SDM computational methods have rapidly expanded, there are many outstanding methodological questions. Here we address key questions about SDM application, using schistosomiasis risk in Brazil as a case study. Schistosomiasis—a debilitating parasitic disease of poverty affecting over 200 million people across Africa, Asia, and South America—is transmitted to humans through contact with the free-living infectious stage ofSchistosomaspp. parasites released from freshwater snails, the parasite’s obligate intermediate hosts. In this study, we compared snail SDM performance across machine learning (ML) approaches (MaxEnt, Random Forest, and Boosted Regression Trees), geographic extents (national, regional, and state), types of presence data (expert-collected and publicly-available), and snail species (Biomphalaria glabrata,B. tenagophilaandB. straminea). We used high-resolution (1km) climate, hydrology, land-use/land-cover (LULC), and soil property data to describe the snails’ ecological niche and evaluated models on multiple criteria. Although all ML approaches produced comparable spatially cross-validated performance metrics, their suitability maps showed major qualitative differences that required validation based on local expert knowledge. Additionally, our findings revealed varying importance of LULC and bioclimatic variables for different snail species at different spatial scales. Finally, we found that models using publicly-available data predicted snail distribution with comparable AUC values to models using expert-collected data. This work serves as an instructional guide to SDM methods that can be applied to a range of vector-borne and zoonotic diseases. In addition, it advances our understanding of the relevant environment and bioclimatic determinants of schistosomiasis risk in Brazil.

Climate and urbanization drive changes in the habitat suitability ofSchistosoma mansonicompetent snails in Brazil

“…We used machine-learning based (XGBoost) Species Distribution Models (SDMs), 30 years of expert-collected occurrence data, and remote sensing data to map the habitat suitability for each of the three Schistosoma -competent Biomphalaria species. SDMs use species occurrence coordinates and corresponding environmental variables to interpolate habitat suitability across large geographic extents, machine learning SDMs are particularly flexible models that can handle non-linear relationships, collinear features (covariates), and complex interactions among features 19,20 . We trained the models on data collected from 2000-2020.…”

“…Next, we included high density urban areas in both analyses: > 1500 people per grid cell with contiguous grid cells totalling more than 150,000 people 47 . As population-based definitions of urban are inconsistent, as a separate analysis, we repeated the analysis with alternative definitions of urbanization as described in supplementp 2, and the results yielded similar results (Supplementpp 7, [17][18][19][20][21][22]. We used WorldPop to estimate population size and cumulative cost mapping -where every pixel is assigned the total cost of the lowest cost path (distance) to an urban pixel -to estimate the distance to the urban pixel for points collected between 2000 -2020 15 .…”

“…We used extreme gradient boosted regression tree models, using package Xgboost 50 , for each SDM as boosted regression models were determined to yield the best performance and interpretability for modeling snail habitat in Brazil 19 . Extreme gradient boosted regression is a machine learning algorithm that creates an ensemble of weak decision trees to form a stronger prediction model by iteratively learning from weak classifiers and combining them into a strong classifier (i.e., boosting) 17 .…”

“…SDMs use species occurrence coordinates and corresponding environmental variables to interpolate habitat suitability across large geographic extents. Machine learning SDMs are particularly flexible models that can handle non-linear relationships, collinear features (covariates), and complex interactions among features 29,30 . We trained the models on data collected from 2000-2020.…”

“…Agricultural crop cover and temporary crop cover were among the top six predictors for B. straminea and B. tenagopila, respectively, with habitat suitability increasing with crop cover (Figure 2-3). In the models using the UN definition of urbanization, agricultural mosaic was among the top five variables for B. tenagophila and B. straminea , with agricultural mosaic the second highest for B. straminea (Supplementpp [19][20][21][22].…”