Attribution of Sahel Rainfall Variability: What Can Flawed Models Teach Us?

The drought experienced in the Sahel in the 1970s and 1980s was unique in the 20^th century record: it was exceptionally severe, widespread, and long-lasting. Much literature has thus suggested that anthropogenic forcings were at least partially to blame for such an extreme. Specifically, it was proposed that, as sulfate aerosols associated with the burning of oil and coal in North America and Europe cooled the north Atlantic, tropical rainfall moved to the south, drying out the Sahel. The cause of the recovery of Sahel rainfall since the mid 80s is even more debated: it could be the response to a decline in aerosols, to a concomitant increase in greenhouse gases (GHG), or to natural variability.

We systematically interrogate the CMIP5 single-forcing simulations to (i) estimate the relative importance of each external forcing and of natural variability in producing the simulated 20^th century variations in Sahel rainfall, (ii) evaluate whether CMIP5-class models simulate forced response and natural variations that are consistent with observations, (iii) build optimal fingerprints for rainfall variations that highlight distinctions in the response to different forcings and (iv) estimate the time of emergence of the GHG signal.

The results highlight the importance of anthropogenic and volcanic aerosols over GHG in generating 20^th century forced Sahel rainfall variability in models. The correlation and RMSE between the forced signal and observations are both significantly separable from noise (p<.05). But while the correlation (0.35) is substantial at about half of what can be achieved in SST-forced simulations, the RMSE—at 96% of the observed variance—is only marginally better than that of a constant, climatology prediction, reflecting the fact that the forced variance is disproportionately small compared to the observed variance

Fingerprints of the Sahel rainfall response to a decrease in aerosol emissions and an increase in greenhouse gas emissions are built from the different patterns in spatial coherence, in seasonality, and in the relation between frequency and intensity of rain events. Using as much information as possible in constructing these fingerprints helps us separate the influence of different factors on current Sahel precipitation. We conclude that the rainfall recovery of the past 30 years was not due to increase GHG, but the similarity of the aerosol-forced variability to natural variability hampers our ability to confidently attribute it to either one or the other.