The necessity of continuous improvement of the mesometеoроlogical models in coastal areas is imposed by the fact that the physical and geographic variety of climatic and meteorological conditions in coastal zones is considerable. Furthermore, the interest of society in weather and climate in coastal zones is defined by the growing population and industrial infrastructures at the coasts of oceans, seas, or large lakes. Over the last two-three decades, a number of scientific experiments have been conducted with GBRS instruments, high meteorological masts, ground, and aerological measurements. The potential of the ground-based remote sensing (GBRS) data within the framework of the LEARNED project is planned to be used for the mesometeorological Weather Research and Forecasting (WRF) model evaluation. For this purpose, typical regimes of the lower part of the coastal atmospheric circulation with sufficient data availability in time and space are searched in the developed specific database

Local circulation (night and day breeze cells fell within the range of sodar) is one of the challenges in numerical modeling.

The ability of the mesoscale numerical weather prediction system Advanced Research WRF (ARW) to represent the observed evolution of the coastal ABL vertical structure is planned to be examined. Initialization with US National Center for Environmental Prediction Final Analyses (FNL) of 1° × 1° spatial and 6 h temporal resolution will be used for the model simulations. For the purposes of the numerical simulations, a Lambert projection will be used with at least 3 nested domains. Тhe outermost one will cover the territory of the Balkan Peninsula, and the innermost one will be centered in relation to the Ahtopol Municipality. The goal will be to descend to a spatial resolution of at least 1 km. The upper boundary of the model domains will be set at 50 hPa and at least 40 vertical levels will be defined, most of which below 2000 m. Different parameterizations of the physical processes are planned to be studied with 24 categories applied for land surface according to USGS (US Geological Survey). The model data will be extracted with 10 min temporal resolution to correspond to the time resolution of the sodar data. The numerical simulations for each period will be initialized with an additional 12 hours of spin-up.