Evaluating global precipitable water vapor products from four public reanalysis using radiosonde data

Precipitable Water Vapor (PWV) is a crucial parameter for short-term climate warnings and long-term climate monitoring. Due to their seamless spatiotemporal coverage, atmospheric reanalysis products provide excellent PWV datasets for related research. At present, multiple research institutes continue to release long-term reanalysis PWV products, maintaining a steady output over time. However, comprehensive evaluation reports comparing these products are lacking. This paper therefore uses global radiosonde (RS) data from 2019 as a reference to first compare and evaluate the accuracy of four widely-used reanalysis datasets: the European Centre for Medium-Range Weather Forecasts reanalysis version 5 (ERA5), the Japanese 55-year Reanalysis (JRA55), the Second Modern-Era Retrospective Analysis for Research and Applications (MERRA2), and the global atmospheric reanalysis by the China Meteorological Administration (CRA40). The evaluation results show that the PWV values from all four reanalysis products are highly consistent with RS PWV. The RMSE and correlation coefficients for ERA5, CRA40, MERRA2, and JRA55 were 2.09 mm/0.99, 2.31 mm/0.98, 2.86 mm/0.97, and 2.80 mm/0.97, respectively. ERA5, JRA55, and CRA40 products exhibited a negative bias, while only the MERRA2 overestimated PWV. Overall, the product accuracy ranks from highest to lowest are ERA5, CRA40, MERRA2, and JRA55. All four products are affected by altitude and season, with lower accuracy observed at low altitudes and during high-precipitation seasons. Notably, CRA40, the first global reanalysis product from the China Meteorological Administration, has demonstrated comparable accuracy to the latest ERA5 in the Chinese mainland. Given its strong potential, CRA40 merits greater attention and wider adoption.