Upstream watershed condition predicts rural children's health across 35 developing countries

Abstract

Diarrheal disease (DD) due to contaminated water is a major cause of child mortality globally. Forests and wetlands can provide ecosystem services that help maintain water quality. To understand the connections between land cover and childhood DD, we compiled a database of 293,362 children in 35 countries with information on health, socioeconomic factors, climate, and watershed condition. Using hierarchical models, here we find that higher upstream tree cover is associated with lower probability of DD downstream. This effect is significant for rural households but not for urban households, suggesting differing dependence on watershed conditions. In rural areas, the effect of a 30% increase in upstream tree cover is similar to the effect of improved sanitation, but smaller than the effect of improved water source, wealth or education. We conclude that maintaining natural capital within watersheds can be an important public health investment, especially for populations with low levels of built capital.Globally diarrheal disease through contaminated water sources is a major cause of child mortality. Here, the authors compile a database of 293,362 children in 35 countries and find that upstream tree cover is linked to a lower probability of diarrheal disease and that increasing tree cover may lower mortality.

Fig. 1

Location of household clusters in the sample. Our health and environmental database covers 293,362 individuals in 35 countries (highlighted in red). The database links 2001–2012 geo-referenced Demographic and Health Survey’s (DHS) individual and household information with data on temperature, precipitation, and upstream watershed conditions

Fig. 2

Factors associated with the probability childhood diarrheal disease. Variables reducing the probability of diarrheal disease (DD) have odds ratios lower than 1 to the left of the red vertical line. Odds ratios and 95% confidence intervals (horizontal lines) show statistically significant associations between socioeconomic variables and the probability of DD in urban and rural households. The bottom two variables indicate that the associations between upstream watershed variables (human activity and tree cover) and the probability of DD are significant (confidence intervals do not cross the red vertical line) for rural, but not urban households. See Supplementary Table 1 for model coefficients

Fig. 3

Marginal effects of policy options for reducing childhood diarrheal disease. Reductions in the probability of diarrhea and 95% confidence intervals for wealth, education, improved sanitation, improved water, and increases in tree cover of 10, 20, and 30%. Marginal effects of tree cover are only statistically significant for the rural subset. The effect of a 30% increase in tree cover is approximately the same as the effect of an improved sanitation facility, but lower than the effect of improved water, education, and wealth

Fig. 4

Factors associated with childhood diarrheal disease differ between rural households with improved and unimproved water. Variables reducing the probability of diarrheal disease (DD) have odds ratios lower than 1 to the left of the red vertical line. Odds ratios and 95% confidence intervals (horizontal lines) show statistically significant associations (confidence intervals do not cross the red vertical line) between socioeconomic variables and the probability of DD in both groups. However, the associations between watershed variables (human activity and tree cover) and the probability of DD are significant for rural households with unimproved water, but not for rural households with improved water. See Supplementary Table 1 for model coefficients

Fig. 5

Factors associated with diarrheal disease in rural areas with high and low upstream human activity, including age and upstream tree cover dummy variables. Variables reducing the probability of DD have odds ratios lower than 1 to the left of the red vertical line. Odds ratios and 95% confidence intervals (horizontal lines) show a statistically significant (confidence intervals do not cross the vertical line) negative association between upstream tree cover and the probability of DD for both areas of high and low upstream human activity, suggesting multiple mechanisms for the effect of tree cover. See Supplementary Table 3 for model coefficients