IntroductionRadon (Rn-222), a radioactive gas with a half--life of 3.82 days, is emitted naturally (0.72-1.2 atoms·cm −2 ·s −1 ) from ice-free, unsaturated terrestrial surfaces [1,2]. The Rn-222 concentration in the near-surface air layer is undoubtedly strongly infl uenced by local conditions, which are connected with the environment of the gas emissions to the atmosphere (inter alia: geological formation, soil characteristics -radium content, permeability and porosity, temperature vertical profi le, soil heat fl ux, and humidity) and microclimatic conditions occurring in the near-ground air layer [3]. Research on Rn-222 in the context of meteorology dates back to the 1920s [4] and has developed intensively since the 1960s. In most publications, Rn-222 has been used to study the atmosphere as a natural tracer of the processes of transport and dispersion of gases [5][6][7][8][9][10][11]. Rn-222 has also been used as an indicator of vertical mixing processes resulting in the atmospheric stability regime [12][13][14][15][16][17]. The temporal variability of the Rn-222 concentration with reference to the meteorological variables (such as wind velocity and direction, atmospheric pressure, air tempera- Abstract. A unique, highly time-resolved, and synchronous three-year dataset of near-surface atmospheric radon-222 as well as soil heat fl ux and air temperature measurements at two sites (rural and urban) in Central Poland are investigated. The recognition of temporal variability of Rn-222 and selected meteorological variables in the urban and rural areas served to create two statistical models for estimation of this radionuclide concentration at 2 m a.g.l. The description of the relationships between the variables for individual months was established on the basis of an exponential function and an exponential function with time derivative of predictor to account for the hysteresis issue. The model with time derivative provided better results. The weakest fi tting of modelled data to empirical ones is observed for the winter months. During subsequent seasons, air temperature as well as QG-driven (soil heat fl ux) models exhibited very high agreement with the empirical data (MBE, RMSE, MAE, and 'index of agreement' by Willmott were used to evaluate the models). A restriction in the use of QG for Rn-222 concentration was observed only in winter in the case of snow cover occurrence, which reduces the daily QG variability.