a–f, Relationships between alkalinity and dye for a representative experiment day (n = 20 independent experiments, 29 September 2016 shown here) and control day (n = 10 independent experiments, 30 October 2016 shown here). a, b, Dye concentrations; c, d, alkalinities; e, f, alkalinity anomalies versus dye concentrations. Linear regressions were fit to alkalinity–dye data using least-squares residuals. d, On control days, the observed (measured) alkalinities closely agree with predicted values for each station. Comparing the upstream and downstream alkalinity–dye ratios provides an estimate of the effect of CO2 enrichment on NCC, as described in the Methods. e, On experiment days (n = 20 independent experiments), if CO2 suppresses NCC, the drawdown in alkalinity is smaller in areas with more CO2 (and more dye) than in areas with less CO2 (and less dye). This effect yields a positive correlation between dye and alkalinity (that is, a positive alkalinity–dye slope) that increases as the water mass moves across the reef—in other words, the alkalinity–dye slope at the downstream transect is greater than that of the upstream transect. f, On control days (n = 10 independent experiments), when dye but no CO2 was added, alkalinity and dye were not correlated, and the mean alkalinity–dye slopes for the upstream and downstream transects did not differ from zero.