The potential to narrow uncertainty in projections of regional precipitation change

Abstract

We separate and quantify the sources of uncertainty in projections of regional (∼2,500 km) precipitation changes for the twenty-first century using the CMIP3 multi-model ensemble, allowing a direct comparison with a similar analysis for regional temperature changes. For decadal means of seasonal mean precipitation, internal variability is the dominant uncertainty for predictions of the first decade everywhere, and for many regions until the third decade ahead. Model uncertainty is generally the dominant source of uncertainty for longer lead times. Scenario uncertainty is found to be small or negligible for all regions and lead times, apart from close to the poles at the end of the century. For the global mean, model uncertainty dominates at all lead times. The signal-to-noise ratio (S/N) of the precipitation projections is highest at the poles but less than 1 almost everywhere else, and is far lower than for temperature projections. In particular, the tropics have the highest S/N for temperature, but the lowest for precipitation. We also estimate a ‘potential S/N’ by assuming that model uncertainty could be reduced to zero, and show that, for regional precipitation, the gains in S/N are fairly modest, especially for predictions of the next few decades. This finding suggests that adaptation decisions will need to be made in the context of high uncertainty concerning regional changes in precipitation. The potential to narrow uncertainty in regional temperature projections is far greater. These conclusions on S/N are for the current generation of models; the real signal may be larger or smaller than the CMIP3 multi-model mean. Also note that the S/N for extreme precipitation, which is more relevant for many climate impacts, may be larger than for the seasonal mean precipitation considered here.

Appendices

Appendix 1

1.1 Testing the internal variability estimates

We test our estimates of the internal variability component of regional precipitation using the constant forcing pre-industrial control integrations of the same GCMs, in which there are no complicating transient forcing effects to consider.⁴ This estimate can then be compared to the estimate derived from the transient integrations.

Figure 8 shows the standard deviation of running decadal means of regional precipitation, expressed as a percentage of the mean precipitation, for both the pre-industrial control integrations (panel a) and the twenty-first century transient integrations (panel b). The patterns are very similar, giving us confidence in our estimates of the decadal variability component of the uncertainty.

Fig. 8

Estimates of the decadal internal variability of regional precipitation. a Mean standard deviation from the pre-industrial control integrations of the GCMs used. b Mean standard deviation for the twenty-first century integrations of the GCMs used. c From the CRU TS3.0 land observations for 1901–2006 (Mitchell and Jones 2005), averaged into 10° boxes, using a 1971–2000 climatology. Grey denotes no data. Note the different colour scale for the observations. Units are % of the mean precipitation

Boer (2009) also analysed the CMIP3 pre-industrial control integrations and found that the magnitude of the zonal mean decadal variability, relative to the mean precipitation, peaks in the tropics, and increases slightly with increased radiative forcings. The maps in Fig. 8 suggest that it is Africa which dominates the estimates of zonal mean decadal variability in the tropics, and the Sahel region is well known to have large decadal variability in precipitation (e.g., Rowell et al. 1995), possibly caused by changes in the Atlantic Multi-decadal Oscillation (AMO) (e.g., Folland et al. 1986).

1.2 Comparison with observational estimates

Figure 8c shows an estimate of the decadal variability from the CRU TS3.0 observational dataset (Mitchell and Jones 2005) for 1901–2006, calculated in a similar way to the GCM estimates above. The observations consist of land-only annual mean data, have been regridded onto a 10° × 10° grid, and detrended with a fourth order polynomial. Note that the scale on Fig. 8c is twice that of Fig. 8a,b. The regions identified as having relatively large decadal variability are similar to the GCMs, but this analysis suggests that the observed decadal variability in precipitation is larger than the GCM estimates, although it must be noted that the observational estimate will have considerable uncertainties. Zhang et al. (2007) also found that for land-only zonal means, the variability in precipitation was significantly underestimated in GCMs when compared to observations, but Boer (2009) found that the internal variability of the GCMs agreed well with an observational estimate of zonal mean precipitation. There is therefore a suggestion, but certainly no consensus, that GCMs may underestimate the magnitude of decadal variability in precipitation.

Appendix 2

In Figure 2 we show the proportion of the total standard deviation due to each type of uncertainty. This has been estimated by considering that the total variance in the projections (T ²) is the sum of the variance due to internal variability (I ²), model uncertainty (M ²) and scenario uncertainty (S ²),

$$ T^2 = I^2 + M^2 + S^2. $$

(2)

When considering the total standard deviation, T, we would like,

$$ T = I'+M'+S' = {\frac{I} {F}} +{\frac{M} {F}} + {\frac{S} {F}}, $$

(3)

where the primes denote scaled versions of I, M and S. The common scaling factor, F, is then,

$$ F = {\frac{I+M+S} {T}} $$

(4)

and the boundaries between the different coloured sections in Fig. 2 are at ${\frac{\pm I} {F}}, {\frac{\pm(I+M)} {F}}$ and ${\frac{\pm(I+M+S)} {F}}$ .