Balancing Accuracy and Efficiency of Atmospheric Models in the Northern Adriatic During Severe Bora Events

Abstract

In process‐oriented studies, accurate representation of severe bora rotor dynamics in the northern Adriatic is known to require the use of model resolutions of the order of 100 m. In regional climate studies, computation time and numerical cost are, however, minimized with resolutions of the order of 10 km. The latter is not accurate enough to drive the coastal dense water formation and the long‐term Adriatic‐Ionian thermohaline circulation resulting from these events. This work leverages the capacity of kilometer‐scale atmospheric models to balance accuracy and efficiency in coupled atmosphere‐ocean climate studies in the Adriatic Sea. The sensitivity of severe bora dynamics and air‐sea interactions to atmospheric model resolution is thus tested within the Adriatic Sea and Coast (AdriSC) modeling suite as well as with the best available reanalysis. The Weather Research and Forecasting (WRF) model at 15‐km, 3‐km, and 1.5‐km resolution, and ERA5 at 30‐km resolution, are compared for an ensemble of 22 severe bora storms spanning between 1991 and 2019. It is found that (1) ERA5 reanalysis and WRF 15‐km model highly diverge (up to 43% for the wind speed) from WRF 3‐km results while (2) WRF 3‐km conditions converge toward the WRF 1.5‐km solution for both basic bora dynamics (differences below 6% for the wind speed) and air‐sea interactions (differences 5 times smaller than with WRF 15‐km results). Consequently, kilometer‐scale atmospheric models should be used to reproduce properly the dense water formation during severe bora events and the long‐term thermohaline circulation of the Adriatic‐Ionian basin.

Performance of the Adriatic Sea and Coast (AdriSC) climate component – a COAWST V3.3-based coupled atmosphere-ocean modelling suite: ocean part

“…computational resources and running time)depending of the temporal and spatial scales of the studied processesremains one of the major issues of the climate modelling community. For example, (1) the RCMs of the Med-CORDEX community have already been proven to largely underestimate the dense water budget of the Adriatic Sea (Dunić et al, 2019), while (2) the recently developed MEDSEA ocean re-analysis at approximately 4-5 km (Escudier et al, 2020) is forced by the ERA5 atmospheric re-analysis known to underestimate the extreme bora events (Denamiel et al, 2021a). To properly capture the Adriatic thermohaline circulation triggered by the dense water formation in the northern Adriatic Sea, the reliability of MEDSEA in the Adriatic Sea thus largely depends on the data assimilation and not the physics.…”

“…Therefore, to quantify the impacts of climate change in the Adriatic, it is crucial to obtain an adequate representation at climate scales of the atmosphere-ocean interactions during extreme events which are, for example, driving the formation of dense water within the basin. Atmospheric RCMs generally fail to provide such a representation, especially in the northern Adriatic where they cannot be used to study the extreme bora dynamics (Denamiel et al, 2020b(Denamiel et al, , 2021a. Additionally, it has also been recently demonstrated that the latest higher resolution ECMWF reanalysis datasetthe ERA5 product (Hersbach et al, 2018), cannot be used either as a reference for climate model evaluation nor as a forcing for ocean models during bora events in the northern Adriatic as it also strongly underestimates the extreme bora speeds (Denamiel et al, 2021a).…”

“…Atmospheric RCMs generally fail to provide such a representation, especially in the northern Adriatic where they cannot be used to study the extreme bora dynamics (Denamiel et al, 2020b(Denamiel et al, , 2021a. Additionally, it has also been recently demonstrated that the latest higher resolution ECMWF reanalysis datasetthe ERA5 product (Hersbach et al, 2018), cannot be used either as a reference for climate model evaluation nor as a forcing for ocean models during bora events in the northern Adriatic as it also strongly underestimates the extreme bora speeds (Denamiel et al, 2021a). Consequently, in the recent study of Liu et al (2021) which investigated the BiOS variability using a regional ocean circulation model at 9-km resolution driven by the ERA-20C atmospheric forcing (Poli et al 2016) with a 101-year long simulationthe impact of the bora events not properly represented by the atmospheric forcing was mimicked by artificially setting up the 2-m air temperature and the dew point temperature over the entire Adriatic Sea to 0 °C in January, February, November and December.…”

Performance of the Adriatic Sea and Coast (AdriSC) climate component – a COAWST V3.3-based one-way coupled atmosphere–ocean modelling suite: ocean results

“…The high-resolution atmosphere-ocean Adriatic Sea and Coast (AdriSC) climate model (up to 3 km in the atmosphere and 1 km in the ocean) was thus implemented, and a 31-year-long evaluation simulation was performed for the 1987-2017 period. In this work, the performance of the AdriSC ocean coastal model is evaluated, while the skill assessment of the AdriSC atmospheric kilometre-scale model is done in a separate study (Denamiel et al, 2021b). In general, a proper evaluation of a high-resolution climate model is not a trivial task and most often the biggest challenge turns out to be the availability, incompleteness, and scarcity of observational data, as well as the imperfections of the observing systems, which set further limits on the evaluation process (Horak et al, 2021).…”

Performance of the Adriatic Sea and Coast (AdriSC) climate component – a COAWST V3.3-based coupled atmosphere–ocean modelling suite: atmospheric dataset

“…Rinke etal. 2011;Nikulin et al, 2012;da Rocha et al, 2014;Huang et al, 2015;Ruti et al, 2016;Zou and Zhou, 2017;Di Virgilio et al, 2019). Specifically, in the Mediterranean Sea, several RCMs have been developed within the Med-CORDEX initiative (e.g.…”