Reconciling land / ocean moisture transport variability in reanalyses with P-ET in observationally-driven land surface models.

Vertically-integrated atmospheric moisture transport from ocean to land, VMFC, is a dynamic component of the global climate system but remains problematic in atmospheric reanalyses with current estimates having significant multi-decadal global trends differing even in sign. Regional VMFC trends over continents are especially uncertain. Continual evolution of the global observing system, particularly step-wise improvements in satellite observations, has introduced discrete changes in the ability of data assimilation to correct systematic model biases, manifesting as non-physical variability. Land Surface Models (LSMs) forced with observed precipitation, P, and near-surface meteorology and radiation provide estimates of evapotranspiration, ET. Since variability of atmospheric moisture storage is small on interannual and longer time scales, VMFC = P-ET is a good approximation and LSMs can provide an alternative estimate. However, heterogeneous density of rain gauge coverage, especially the sparse coverage over tropical continents, remains a serious concern. Rotated Principal Component Analysis (RPCA) with pre-filtering of VMFC to isolate the artificial variability is used to investigate artifacts in five reanalysis systems. This procedure, though ad hoc, enables useful VMFC corrections over global land. P-ET estimates from seven different LSMs are evaluated and subsequently used to confirm the efficacy of the RPCA-based adjustments. Global VMFC trends over the period 1979-2012 ranging from 0.07 to -0.03 mmd-1 decade-1 are reduced by the adjustments to 0.016 mmd-1 decade-1, much closer to the LSM P-ET estimate (0.007 mmd-1 decade-1). Neither is significant at the 90 percent level. ENSO-related modulation of VMFC and P-ET remains the largest global interannual signal with mean LSM and adjusted reanalysis time series correlating at 0.86.