Global IWV trends and variability in atmospheric reanalyses and GPS observations

Abstract

Abstract. This study investigates the means, variability, and trends in integrated water vapour (IWV) from two modern reanalyses (ERA-Interim and MERRA-2) from 1980 to 2016 and ground-based GPS data from 1995 to 2010. It is found that the mean distributions and inter-annual variability in IWV in the reanalyses and GPS are consistent, even in regions of strong gradients. ERA-Interim is shown to exhibit a slight moist bias in the extra-tropics and a slight dry bias in the tropics (both on the order of 0.5 to 1 kg m−2) compared to GPS. ERA-Interim is also generally drier than MERRA-2 over the ocean and within the tropics. Differences in variability and trends are pointed out at a few GPS sites. These differences can be due to representativeness errors (for sites located in coastal regions and regions of complex topography), gaps and inhomogeneities in the GPS series (due to equipment changes), or potential inhomogeneities in the reanalyses (due to changes in the observing system). Trends in IWV and surface temperature in ERA-Interim and MERRA-2 are shown to be consistent, with positive IWV trends generally correlated with surface warming, but MERRA-2 presents a more general global moistening trend compared to ERA-Interim. Inconsistent trends are found between the two reanalyses over Antarctica and most of the Southern Hemisphere, and over central and northern Africa. The uncertainty in current reanalyses remains quite high in these regions, where few in situ observations are available, and the spread between models is generally important. Inter-annual and decadal variations in IWV are also shown to be strongly linked with variations in the atmospheric circulation, especially in arid regions, such as northern Africa and Western Australia, which add uncertainty in the trend estimates, especially over the shorter period. In these regions, the Clausius–Clapeyron scaling ratio is found not to be a good humidity proxy for inter-annual variability and decadal trends.