Abstract
•Thirty-one years of radiosonde data from seven sites were used to calculate the weighted mean temperature.•Variations of the weighted temperatures have been studied and characterized.•Relationship between the weighted temperature surface temperature for each site and from all the sites was established.•The predictability of 14 models to the measured data was conducted.
In this study, 107,776 radiosonde observations from seven sites in Saudi Arabia for the period 1985–2016 were used to calculate the atmospheric weighted mean temperature (Tm) and investigate its variations. The results showed that the Tm exhibited a seasonal cycle, with a maximum Tm during summer and a minimum Tm during winter. For each site, and for the combined data from all sites, a simple empirical model was developed and tested to predict the Tm based on surface temperatures. The predictability of fourteen Tm-Ts (surface air temperature) models proposed by several investigators was carried out. The predictions of these models differ from one site to another. While some models give adequate predictions for one site, they fail in others. However, some models present reasonable estimates when the measured data from all sites are combined. Comparisons between measured (from radiosonde observations) and calculated precipitable water vapour (PWV) values for some sites using different Tm values were carried out. By using the site specific Tm-Ts models, the mean bias error (MBE) and root mean square error (RMSE) were below one 1 mm, and between 4 and 5 mm, respectively. On the other hand, the regional developed model developed by this study gives sub-mm MBE values and about 12 mm RMSE. Using seven of the Tm-Ts models found in the literature, the PWV values were calculated for the same selected sites. The accuracies of these models are comparable with those developed in this study for each site.