Correction to: Nature https://doi.org/10.1038/nature07035 Published online 26 June 2008

The BrO spectrum shown in the right-hand panel of Supplementary Fig. 3 was a software training spectrum that was inadvertently included as a typical BrO spectrum from the Cape Verde campaign. The training spectrum was generated using the BrO spectrum in Fig. 1 from Saiz-Lopez et al.1, which was adapted in order to test the DOAS retrieval software at the low BrO concentrations expected at Cape Verde. For doing this, the residual in the spectrum from Fig. 1 in Saiz-Lopez et al.1 was modified by the addition of an artificial BrO absorption equivalent to the typical level expected at Cape Verde. That is, an artificial spectrum was reverse-engineered by removing the BrO absorption from the Fig. 1 spectrum and then adding the BrO level expected at Cape Verde. Hence, the residual output of the program was very similar for the BrO fits in Fig. 1 from Saiz-Lopez et al.1 and that shown in Supplementary Fig. 3.

A new version of Supplementary Fig. 3 is shown below. The right-hand panel is a BrO DOAS spectrum taken on 10th April 2007 at 09:45 local time at Cape Verde (solid line). The dashed line illustrates the fitted reference BrO spectrum (indicating a BrO mixing ratio of 2.2 ± 0.3 ppt). The IO spectrum fit (left-hand panel) is unchanged from the original Supplementary Fig. 3.

Typical fits of IO and BrO reference spectra (dashed lines) to the DOAS atmospheric spectra (solid lines) taken at San Vicente, Cape Verde. The IO fit is from the 24th of March 2007 at 09:00 local time, representing a mixing ratio of 1.2 ± 0.2 pptv. The BrO fit is from 10th April 2007 at 09:45 local time and indicates a BrO mixing ratio of 2.2 ± 0.3 pptv.