Large-scale co-variations of freshwater fluxes and storages on land can critically regulate green (vegetation) and blue (hydrosphere) water balances, land-atmosphere interactions, and hydroclimatic hazards. Such essential co-variation patterns still remain largely unknown over large scales and in different climates around the world. To contribute to bridging this large-scale knowledge gap, we synthesize and decipher different data time series over the period 1980-2010 for 6405 hydrological catchments around the world. From observation-based data, we identify dominant large-scale co-variation patterns between main freshwater fluxes and soil moisture (SM) for different world parts and climates. These co-variation patterns are also compared with those obtained from reanalysis products and Earth System Models (ESMs). The observation-based datasets robustly show the strongest large-scale hydrological co-variation relationship to be that between SM and runoff (R), consistently across the study catchments and their different climate characteristics. The predominantly strongest large-scale SM-R co-variation relationship, however, is also the most misrepresented by ESMs and reanalysis products, followed by that between precipitation and R. Comparison between corresponding observation-based and ESM results also shows that an ESM may perform well for individual hydrological variables, but still fail in representing the patterns of large-scale co-variations between variables.

Access to spatio-temporally consistent precipitation data is a key prerequisite for hydrological studies, especially in data-scarce regions. Different global precipitation products offer an alternative way to estimate precipitation over areas with inadequate gauge distributions. However, before use of the datasets, the accuracy of these global estimations must be carefully studied at local and regional scale. This study evaluated 14 global precipitation products against gauge observations 2003-2012 in Karun and Karkheh basins in southwest Iran. Different categorical and statistical indices, including Kling-Gupta Efficiency (KGE), bias, correlation coefficient, and variability ratio, at varying spatial and temporal resolution were used to evaluate the products. KGE results at both daily and monthly time steps suggested that TMPA-3B42V7.0 and MERRA-2 outperformed all other products, while CMORPH-BLDV1.0 and PERSIANN-CDR was the best-performing product at daily and monthly time steps, respectively. ERA5-Land showed the highest positive bias compared with in-situ observations, particularly for mountainous southeastern parts of Karun basin. Overall, bias-adjusted products obtained by merging ground-based observations in the estimations outperformed the unadjusted versions. The spatial distribution of statistical error metrics indicated that almost all products showed their greatest uncertainties for mountainous regions, due to complex precipitation processes in these regions. These results can significantly contribute to various horological and water resources planning measures in the study region, including early flood warning systems, drought monitoring, and optimal dam operations.