With climate change, an increase in the frequency and intensity of extreme precipitation is expected in the Western Mediterranean (WMed). Simulating this high impact weather remains a challenge, in part due to errors in the representation of moist processes in relation to heavy precipitation. These errors in representing the moisture distribution in atmospheric models are likely to affect and deteriorate the final precipitation simulation.
In the recent years, we have experienced relevant advancements in measurement techniques and supercomputing for atmospheric prediction. With the arrival of Global Positioning System (GPS) meteorology, sampling of Integrated Water Vapour (IWV) has been made possible up to frequencies of a few minutes only, with high quality. Further, high-resolution atmospheric simulations have enabled the explicit representation of deep convection, reducing the need to rely on mathematical simplifications (parameterizations) to represent this phenomenon.
One of the core objectives of the project Predictive Models for Extremes and High-impact Weather under Climate Change (PREMIUM), carried out within the Extreme Weather in a Changing Climate group, is to advance the understanding and model representation of extreme precipitation using sub-hourly GPS observations. We have performed novel simulations assimilating GPS measurements every 10 minutes, using the nudging method in model resolutions ranging from 7 km to 500 m. The autumn period of 2012 was selected for this experiment because this is the season with the largest probability of heavy precipitation in the north WMed region and it coincides the Special Observation Period (SOP) 1 of the Hydrological Cycle in the Mediterranean Experiment (HyMeX). HyMeX collected a unique group of observational data sets available for model validation and process studies. During HyMeX, an unprecedented GPS network, with over one thousand GPS receivers, was jointly reprocessed for the SOP1 for research purposes.
The impact of assimilating GPS data using the nudging technique revealed an improvement in the simulation of atmospheric humidity and convective processes. A reduction of the root mean square error for IWV and of the humidity overestimation in heights above 1 km were observed. The combination of GPS nudging and sub-kilometre (500 m) simulations reduced the model differences with the observations. Relevant impacts on atmospheric instability, and moisture transport improved the simulation of precipitation in its daily cycle, the probability of extreme values and the representation of its structure.
Figure 1 shows an example of the good performance by the GPS assimilation in the representation of IWV in the autumn 2012 period using a 2.8 km horizontal resolution. The mean absolute bias is reduced between 20 % and 60 % over the Iberian Peninsula (IP), France (FR) and Italy (IT) due to GPS nudging. Nevertheless, still significant differences over mountain areas persist.
These innovative experiments demonstrate the benefit of assimilating minute-frequency GPS information. They show that GPS alone can bring important corrections in aspects of heavy precipitation representation such as the daily cycle or the probability distribution. These new insights can potentially lead to improvements in weather forecasting situations where the observation systems fail or in regions where in-situ stations are scarce.
Figure 1. IWV differences (observations - model) for the September to November 2012 period between GPS measurements and the simulations using COSMO on a 2.8 km grid. The differences are obtained at the location of the GPS receivers (blue and red dots) where blue colours stand for a wet bias in the model. The IWV differences are averaged in the study period. The black lines denote three investigation areas for the Iberian Peninsula (IP), France (FR) and Italy (IT) and the red lines the simulation domains.
Working group: Extreme Weather in a Changing Climate (https://www.imk-tro.kit.edu/english/6760.php)
Contact: Alberto Caldas-Alvarez (alberto.caldas-alvarez∂kit.edu)