Assessment of changes in ocean precipitation at multi-decadal time scales is very difficult owing to the lack of reliable observation based data sets prior to the satellite era. The few studies available rely on reconstruction techniques. Remote sensing based precipitation observations from the Global Precipitation Climatology Project (GPCP) for 1979–2003 have been used by Smith et al. (2009, 2012) to reconstruct precipitation for 1900–2008 (over 75°S to 75°N) by employing statistical techniques that make use of the correlation between precipitation and both SST and sea level pressure (SLP). Each of the reconstructions shows both centennial and decadal variability in global ocean mean precipitation (Figure 3.7). The trend from 1900 to 2008 is 1.5 mm per month per century according to Smith et al. (2012). For the period of overlap, the reconstructed global ocean mean precipitation time series show consistent variability with GPCP as is to be expected (Figure 3.7). Focusing on the Tropical Ocean (25°S to 25°N) for the recent period 1979–2005, Gu et al. (2007) have identified a precipitation trend of 0.06 mm day–1 per decade using GPCP. Concerns have been expressed in the cited studies over the need for further work both to determine the most reliable approach to precipitation reconstruction and to evaluate the remotely sensed precipitation data sets. Given these concerns, confidence in ocean precipitation trend results is low.
Evaporation and precipitation fields from atmospheric reanalyses can be tested for internal consistency of different components of the hydrological cycle. Specifically, the climatological mean value for E – P averaged over the global ocean should equal both the corresponding mean for P – E averaged over land and the moisture transport from ocean to land. Trenberth et al. (2011) find in an assessment of eight atmospheric reanalyses that this is not the case for each product considered, and they also report spurious trends due to variations in the observing system with time. Schanze et al. (2010) examine interannual variability within the OAFlux evaporation and GPCP precipitation data sets, and find that use of satellite data prior to 1987 is limited by discontinuities attributable to variations in data type. Thus, it is not yet possible to use such data sets to establish whether there are significant multi-decadal trends in mean E – P. However, regional trends in surface salinity since the 1950s do suggest trends in E – P over the same time (see Section 3.3.4).