Abstract: Typhoon In-fa lands on the coast of Zhoushan, Zhejiang Province on 25 July 2021, bringing severe wind and rain to the coastal areas for a long time. Before landing, it maintains in the sea, east of Ryukyu Islands and then suddenly turns northward. The official subjective and numerical model forecasts both produce serious errors on the landing time and location, which has a certain impact on the decision of disaster prevention.Using the best track data of China Meteorological Administration (CMA) and the water vapor channel data of FY-4A satellite cloud image, the characteristics of Typhoon In-fa, especially the causes of its stagnation and northerly bend are investigated. Deterministic model forecast results from the regional mesoscale typhoon numerical forecast system (CMA-TYM), the fine-grid numerical forecast product of European Center for Medium-range Weather Forecasts (ECMWF) and National Center for Environmental Prediction Center (NCEP), and ECMWF ensemble forecast data are also analyzed. The fifth generation global climate reanalysis data set ERA5 is used for the analysis of the real situation field and physical quantity field. The center of the upper-tropospheric cold low (UTCL) is determined by referring to the water vapor channel of the satellite cloud image and the horizontal flow field is obtained according to the wind field data of ERA5. The center of the cyclonic circulation over 200 hPa is set as the center of UTCL.Through FY-4A water vapor cloud image, it is found that in the stagnation stage of Typhoon In-fa, there is a UTCL system on the north side. By analyzing the vertical distribution of circulation situation field, steering flow, and relative vorticity, it is found that the position of the subtropical high system is to the east and north, and typhoon guiding effect on the typhoon is weak. Therefore, the main weather systems that affect the change of its track are the UTCL and westerly trough system in the upper troposphere. There are also differences in the interaction between UTCL and typhoon at different intensities and distances. By analyzing the deterministic and ensemble prediction of ECMWF and CMA-TYM models, it is found that the prediction errors and deviations of UTCL are important reasons for the track predictions. There are still large errors and uncertainties of ensemble prediction in the prediction of UTCL, especially for long lead time forecast products. The UTCL slows down Typhoon In-fa and makes it bend to the left, which indirectly leads to the northerly bend of Typhoon In-fa.In the future, it is necessary to further study the influence of UTCL on typhoon track and intensity, pay more attention to the prediction performance of UTCL, and carry out the prediction and inspection of UTCL in the model, so as to support the interpretation and improvement of the model product.