A study of ABL processes that increase the uncertainty in temperature forecasting

In this study, we analyze the atmospheric boundary layer (ABL) processes contributing to the increase in uncertainty in two-meter temperature forecasting. For temperature prediction, we employ the Weather Research and Forecasting (WRF) mesoscale model with a one-kilometer horizontal resolution, using a numerical configuration calibrated for forecasting temperatures in the Basque Country. This region, located in the north of the Iberian Peninsula, is characterized by complex topography and is highly influenced by the Atlantic Ocean. Specifically, in this contribution, we assess parameterizations (surface layer, boundary layer, microphysics, etc.) and their combinations, as well as the number and distribution of vertical levels and other dynamic options such as 6th-order diffusion, to best replicate the evolution of two-meter temperatures during fair weather and high-pressure conditions

We conduct statistical analyses for specific periods of interest to understand how temperature bias correlates with variables or factors such as surface thermal profile, mechanical turbulence, surface and 850 hPa winds, cloud cover, and temperature drops associated with sea breeze onset in the warmest months. Depending on variations in error or bias across different time ranges (nighttime, daytime) and meteorological scenarios (e.g., calm winds, clear skies, strong surface heating and cooling), we estimate uncertainty for each temperature prediction

Our analysis reveals that high pressure conditions often coincide with significant errors in temperature prediction, primarily attributed to mesoscale model challenges in simulating night time stable conditions, boundary layer morning and evening transitions, and the accurate representation of thermally driven mesoscale flows (such as sea breezes) and their impacts on hourly temperature forecast.