Abstract: Temperature advection is a basic physical variable in the weather prediction. It directly leads to the change of atmospheric thermal structure and further brings about the alteration of atmospheric physics field, and can reflect the development of weather systems better. In recent years, the technology of wind profile radar (WPR) improves greatly, and WPR data has both high precision and high spatial-temporal resolution and could continuously provide the distribution of horizontal winds over time. Based on the concept of temperature advection and the principle of thermal wind, retrieving the temperature advection with high precision and spatial-temporal resolution is feasible by algorithms established.The temperature advection information is retrieved from WPR data at Yanqing Station of Beijing, whose results are compared to temperature advection prediction products of Global Medium Range Forecast System of T639L60 (hereinafter, T639L60 model). A cold air invasion process on 15 Nov 2014 in the night is analyzed in detail and 6 more samples occurring from Sep to Nov in 2015 are statistically analyzed for evaluation too. Results show that the horizontal wind of the WPR is highly consistent with the wind prediction data of T639L60 model within a certain amount of prediction time (about 6 to 12 h). The order of magnitude is consistent and the value is close between the retrieval of temperature advection by the WPR data and the initial data of T639L60 model. The quality of retrieving temperature advection depends on the quality of horizontal wind profile data. Temperature advection prediction products of T639L60 model and the real-time retrieval products are accordant within a period of 6-12 h depending on the temporal scale of different weather system. The derivation gradually increases as the prediction time lengthens, sometimes even reverse prediction advection attribute may appear. Temperature advection with real-time, continuous attribute and high precision can be provided through WPR observations whose time and vertical resolution is 6 min and 120 m, respectively. And some prediction results can be calculated using the measured results by linear extrapolation in a short period. Thus, the precision of numerical forecast can be improved with WPR data assimilated in the future. Preliminary results show that the real-time 3D products and forecast products of temperature advection with high precision, spatial and temporal resolution could be generated through high quality WPR network data assimilated.