Weighted Mean Temperature Models for Estimating Global Navigation satellite system precipitable water vapor over East Africa

Abstract

The weighted mean temperature ( Tm ) is an important parameter in deriving accurate and reliable Global Navigation Satellite System (GNSS) Precipitable Water Vapor (PWV) values especially in tropical regions. However, the accuracy of GNSS - PWV in the East African region is limited by the availability of precise and reliable meteorological data for computing accurate Tm values. To address this limitation, this study developed site-specific and regional Tm models for East Africa (E.A) using radiosonde and Ozonesonde profiles from 1971 to 2019. The models were validated using radiosonde profiles from two stations 63741 (Dagoretti) and 63894 (Dares-salaam) based on Root Mean Square Error (RMSE) and Mean Bias Error (MBE) values of Tm . The Tm values from the established models were compared with the two global (Bevis and Yao) and one Tropical regional (Raju) Tm models in order to assess their performance. At station 63741, the RMSE of the E.A model, the site-specific Tm model, the Bevis, Yao, and Raju Tm models are 2.01 K, 1.76 K, 1.79 K, 2.54 K, 2.74 K, and 2.47 K, respectively. At station 63894, the RMSE of the E.A model, the site-specific Tm model, the Bevis, Yao, and Raju Tm models are 1.66 K, 1.63 K, 1.70 K, 2.52 K and 2.64 K, respectively. At both stations, the site-specific Tm models have higher accuracy followed by the E.A model as compared to Raju Tm model and the two global Tm models (Bevis and Yao). The study further evaluated the accuracy of Tm models in the retrieval of PWV by computing PWV values at three GNSS sites (ARSH, DODM, and RCMN). The GNSS - PWV values computed from different Tm models were compared with PWV values derived using Tm from ERA5 reanalysis dataset from 2013 to 2020. The use of GNSS-PWV from ERA5-Tm as the reference values was due to the limited Tm values from Radiosonde during the same periods of available GNSS data required for PWV estimation. In terms of RMSpwv, the site-specific models exhibit values ranging from 0.107 mm to 0.11 mm, with a mean value of 0.109 mm, which is the lowest among the five models. This is followed by, the E.A regional model with mean value of 0.127 mm, Yao Tm with a mean value of 0.129 mm, Bevis with a mean value of 0.130 mm, and Raju with a mean value of 0.148 mm.The results show that the E.A regional Tm model and site-specific models perform better than the global and other tropical regional Tm models. Therefore, the established Tm models are the preferred choices for estimating Tm for GNSS - PWV retrieval in the East African region. Overall, the study provides improved Tm models for accurate estimation of GNSS - PWV in the East African region, which could improve weather and climate forecasts.