Abstract
Since much of the current researches have focused on daily, monthly or annual near-surface (2 m) temperature lapse rate (NSTLR), there is little guidance on best estimation practices and analyses of time-varying characteristics for the hourly NSTLR. To estimate hourly NSTLR and identify its time-varying characteristics accurately and objectively, this study proposed a robust estimation strategy based on IGGIII equivalent weight using multiple linear regression models. The accuracy and reliability of the proposed method was verified. The results show that the robust estimation strategy can further improve the hourly NSTLR solution accuracy relative to the least square (LSQ) method, especially in the time period of relatively high temperature. The hourly NSTLR was positively correlated with temperature, with a 24-h average maximum of 0.604 °C/100 m at universal time coordinated (UTC) 7.2 h and minimum of 0.284 °C/100 m at UTC 20.5 h, respectively. Throughout the year, the NSTLR was the largest from June to August, with an average median of around 0.492 °C/100 m. However, from November to the following January, the NSTLR value was the smallest, with a mean median of about 0.323 °C/100 m. In addition, the hourly NSTLR values were essentially less than the constant value of 0.65 °C/100 m. When the hourly NSTLR estimated based on the proposed method was applied to the temperature interpolation, the interpolation accuracies at the highest altitude (1545 m) and other meteorological stations (below 310 m) can increase by 22.4 % and 8.1 %, respectively, relative to the hourly NSTLR calculated by the LSQ method, and increased by 55.6 % and 13.0 %, respectively, relative to the no-NSTLR correction. The results are important for the fine establishment of high spatiotemporal resolution temperature fields and for the study of climatic phenomena characterized with rapid spatiotemporal variation.