Solute concentrations and annual export from Amazon headwater streams following conversion from forest to intensive soybean agriculture

Background/Question/Methods

Across the globe, water quality and stream solute loading have been altered by conversion of wildland to cropland and pasture. Currently, the rate of conversion is highest in the tropics, where the effect of landuse change on stream function is least well understood. Within the tropics, the Brazilian Amazon is a globally significant center for agricultural expansion and intensification. We sampled seven headwater streams draining forest (N=3) or soy (N=4) to examine changes to stream solute export across multiple watersheds in Mato Grosso, Brazil, the main region of soybean agriculture expansion. We measured baseflow concentrations on a biweekly to monthly basis and stormflow in a subset of watersheds (2 forest, 2 soy) during a subset of storms (2-3 in each watershed). Additionally, we sampled groundwater, rain, and throughfall to characterize other flowpaths that may influence stream chemistry. We analyzed these samples for anions (NO₃^-, PO₄^3-, SO₄^2-, Cl^-), cations (NH₄⁺, Ca²⁺, Mg²⁺, Na⁺, K⁺, Al³⁺, Fe^3+/2+), silica, dissolved organic carbon (DOC), and d¹⁸O.

Results/Conclusions

Annual export from soy streams was higher (mean increase of 4.15 times) than export from forest streams for all solutes. This difference was significant for NH₄⁺, Ca²⁺, Mg²⁺, Na⁺, SO₄^2-, silica and DOC (p<0.03). Despite high fertilizer inputs and intensive management, there was no difference in concentration between landuses for any solute, with the exception of iron (p=0.001). Annual stream discharge from soy watersheds was 4.12 times higher than from forest watersheds (p=0.01), suggesting export increases were attributable to changes in discharge rather than solute concentrations. Despite this increase, export remained low (e.g. <1 kg NO₃^- N·ha^-1·yr^-1, <1.5 kg Ca²⁺·ha^-1·yr^-1).

 During storm events solute concentrations were sometimes, but not always, correlated with discharge. However, these relationships were not consistent across storms, streams, or within landuses. This suggests that extrapolating stream solute behavior during high flows using data from a small number of storms may be problematic.

Solute concentrations varied seasonally for all solutes (p<0.003), with the exception of K⁺. Peak concentrations generally occurred approximately one month after the first rains, when rainwater solute concentrations also peaked. These data suggest the increase in stream solute concentrations may not be solely due to flushing from accumulated litter and dry-season deposition but also due to regional changes in atmospheric particulates. In Mato Grosso, land clearing via burning begins after the first rains, and we hypothesize that this management strategy contributes to seasonally higher solute concentrations.