Neural Network-Based Models for Estimating Weighted Mean Temperature in China and Adjacent Areas

The weighted mean temperature (Tm) is a key parameter when converting the zenith wet delay (ZWD) to precipitation water vapor (PWV) in ground-based Global Navigation Satellite System (GNSS) meteorology. Tm can be calculated via numerical integration with atmospheric profile data measured along direction, but this method not practical most cases because it easy for general users get real-time data. An alternative obtain an accurate value establish regional or global models on basis of its relations surface meteorological elements as well spatiotemporal variation characteristics Tm. In study, complex between and some essentially associated factors including geographic position terrain, pressure were considered develop models, then non-meteorological-factor model (NMFTm), single-meteorological-factor (SMFTm) multi-meteorological-factor (MMFTm) applicable China adjacent areas established by adopting artificial neural network technique. generalization performance new was strengthened help ensemble learning method, accuracies compared several representative published from different perspectives. results show that all exhibit consistently better than competing under same application conditions tested within study area. NMFTm superior latest non-meteorological has advantages simplicity utility. Both SMFTm MMFTm higher accuracy listed study; particular, about 14.5% first-generation network-based (NN-I) model, best so far terms root-mean-square error.