Fog events and local atmospheric features simulated by regional climate model for the metropolitan area of São Paulo, Brazil

• Simulation reproduces the climatology of fog events over Metropolitan Area of SP.
• The use of model fine resolution improves the representation of the local climate.
• Synoptic patterns during fogs are well characterized in fine resolution simulation.

This study aims to evaluate the impact of the horizontal resolution of a RegCM3 (Regional Climate Model version 3) in order to reproduce local features of climate, particularly fog, in the Metropolitan Area of São Paulo (MASP), Brazil. RegCM3 simulations with horizontal resolutions of 50 km (R50) and 20 km (R20) are conducted for 2003 and 2004 austral winters, June to September (JJAS). Simulations are compared with both gridded analysis and station observations. An objective criterion, is used in the simulations (applied at 09 and 21 local time) to search fog events based on relative humidity and precipitation thresholds in order to identify mainly radiation fog formation. The overall result shows improvements in order to reproduce the local climate using the finest resolution simulation, since R20 simulates better the observed spatial patterns. The evaluations of the daily mean air temperature, relative humidity, and pressure show that both R50 and R20 simulate mean values comparable to the local observations of MASP. Firstly, there is high time correlation and small root mean square error for air temperature (mean, maximum and minimum) using R20 simulation. Additionally, R20 reduces the rainfall overestimation of R50 due to the more realistic representation of topography and coastlines. Both R20 and R50 are able to simulate the fog event frequency similar to the observed ones during 2003 and 2004 JJAS periods, with overestimation below 10%. Local observation and simulations (R50 and R20) present a higher frequency of fog in 2003 than 2004. Regarding fog characteristic, it is found that: (1) stations located in the valleys as well as coastal areas have different patterns with the first ones presenting higher frequency of fog events; (2) the atmospheric variable errors during fog events are clearly smaller than for the whole JJAS period; (3) it is simulated properly for the relative humidity increase, temperature decrease and small specific humidity increase during fog events; and (4) the synoptic pattern during fog event can be clearly characterized by R20 simulation. Therefore, R20 and R50 indicate the modeling ability in order to characterize fog events over MASP.