Aburrá Valley's weather forecast system: Implementation and improvements using radar and radiometer data assimilation

The Aburrá Valley is a highly urbanized region located in the Andean region of Colombia (northern South America) where heavy rainfall occurs throughout the year, causing flooding, landslides and several hazardous events. As part of the risk reduction strategy, the region's early warning system (SIATA) has implemented an operational numerical weather prediction model (WRF) in the valley and nearby region to forecast precipitation, temperature, wind velocity among other atmospheric variables. Even though the results on a daily scale show a good agreement with observations, it is necessary to obtain better results on a subdaily scale to used it as input of the hydrologic model used for flood forecasting. To improve the forecasts data assimilation techniques have been implemented. In particular, both reflectivity and radial velocity from a Doppler radar as well as temperature and relative humidity profiles from a microwave radiometer have been assimilated into the model using the three-dimensional variational (3DVAR) technique. This is the first time this type of observations are ingested into a weather forecast model in the study region, which represents great challenges for the implementation of state of the art data assimilation techniques and observations. The results show that the assimilation of both reflectivity and radial velocity from the radar has a large impact on the timing and location of the precipitation, whereas the effect of the radiometer profiles is not as clear. Since the general purpose is to improve quantitative precipitation forecasts (QPF) to use it as part of the early warning system of the valley, the results are promising and show that assimilating local data into the operational weather forecast model has the potential to improve the precipitation forecast for the region.