Post–processing of WRF model for precipitable water using meteorological satellite data

Precipitable water (PW) is an important part of the water cycle in the atmosphere. This variable plays a significant role in hydrological and meteorological studies to understand the behavior of water vapor and the involved processes. One of the management tools in this area is knowledge of the Total Precipitated Water content (TPW) in the atmosphere. This quantity can also be used as a meteorological index for accurate prediction of atmospheric parameters including rainfall and runoff estimates. PW is the amount of vertically integrated water vapor and can be expressed in g/cm2, or as the height of an equivalent column of liquid water in centimeters. The traditional technique for PW measurement is to launch radiosondes, normally twice a day. This method is not only expensive but also poor in both spatial coverage and temporal resolution. Temporal changes in atmospheric water vapor occur rapidly and water vapor measurements by radiosondes do not satisfy the needs of research for a variety of variation scales of atmospheric water vapor. In past years, the emerging ground-based Global Positioning System (GPS) has opened the possibility of improved monitoring of atmospheric humidity. Beginning from the 1990s, an observational technique based on the Global Navigation Satellite System (GNSS), which is sensitive to the spatial and temporal distribution of the water vapor content in the atmosphere, has made it possible to retrieve precise and continuous estimates of PW. In recent years, studies have been conducted on the use of satellite data and GPS data to estimate this parameter. The purpose of this research is to evaluate the post-processing of WRF model for PW using meteorological satellite data, neural network and Kalman filter. This parameter cannot be easily calculated without the use of equipment such as GPS, satellite, radiosonde station and radar. In this research, we attempt to eliminate the systematic error of a PW parameter obtained through the WRF numerical model in the absence of measurement data. In the first step, we compare the rainfall water for two stations in Iran including Mashhad and Tehran with the TPW parameter measured by MSG1(IODC) and FY-2E satellites for about two years. Then, the WRF model results are post-processed based on the observational data to educate the neural network model using the ANFIS implementation with genetic algorithm training. In second step, the Kalman filter was applied to the satellite data modified by genetic algorithm. The results were improved and the correlation coefficient was to 0.987 and the RMSE error reduced to 1.1. Results show that by implementing this approach and converging on the correlation values of the selected satellite and the post-processed model to the correlation values of the modified model, the results showed that the meteorological satellite data METEOSAT8 (IODC) can be used as an alternative of radiosonde data in places without high atmosphere station to teach the proposed model for post-processing of PWV parameter in WRF numerical model.