GLOBAL SENSITIVITY ANALYSIS OF PROGNOSTIC METEOROLOGICAL FIELDS TO CHANGES IN SURFACE PARAMETERS OF SAINT PETERSBURG

Global sensitivity analysis of individual prognostic fields computed by the WRF-ARW numerical weather prediction model to the change of the major urban surface parameters describing the urban surface of Saint Petersburg is carried out. Numerical experiments were held for several dates of the 2015 warm season, when urban heat island was observed. The Morris method for global sensitivity analysis was used to quantify the most influential parameters. The quantitative assessment of the individual and joint contribution of the urban surface parameters was carried out using the Sobol method and machine learning to build surrogate models. As a result, it turned out that the surface temperature field is most sensitive to the changes in albedo in the daytime and in surface roughness and thermal inertia at night; the wind speed field is most sensitive to the changes in surface roughness at any time of a day; the relative humidity field is strongly influenced by the changes in soil moisture in the daytime and in surface roughness at night.