A multi-instrument and multi-model assessment of atmospheric moisture variability over the western Mediterranean during HyMeX

International audienceThe Hydrological cycle in the Mediterranean eXperiment (HyMeX) was held in fall 2012 during which part of the observational effort has been established on Menorca Island to characterize the upwind marine low-level flow. The ground-based Water vapour Raman Lidar (WALI), the airborne water vapour lidar LEANDRE 2 and boundary layer pressurized balloons were implemented during the first Special Observing Period and contributed to characterize water vapour variability in the vicinity of the Balearic Islands, together with the existing network of Global Positioning System receivers. Furthermore, analyses from regional and global numerical weather prediction (NWP) models (AROME-WMED, the ECMWF/IFS NWP system and the Weather Research and Forecasting (WRF) model) were also available over large domains encompassing part or the entire Western Mediterranean basin. We assess the consistency of water vapour mixing ratio (WVMR) profiles and integrated water vapour contents (IWVC) derived from the different data sets by comparing them to a common reference, the ground-based lidar WALI. We use consistency indicators such as root mean square errors, biases and correlations. Comparison between WVMR profiles from ground-based and airborne lidars (ground-based lidar and boundary layer pressurized balloons) leads to a root mean square error lower than 1.6 g kg−1 (1.3 g kg−1) when the closest possible air masses are sampled. We observed a good agreement between the vertical WVMR profiles derived from WALI and the numerical models with correlations higher than 0.7 and root mean square errors lower than 2 g kg−1. Regarding IWVCs, the models exhibit biases less than 2 kg m−2, root mean square errors lower than 2.3 g kg−1 and correlations higher than 0.86 when compared to WALI. Finally, AROME-WMED 48-h forecasts were compared with WALI data composited over eleven 48-h periods. The quality of the forecast does not visibly degrade within the 48-h period from the initial analyses