Atmospheric delay effect is unavoidable and sometimes serious when using spaceborne repeat-pass synthetic aperture radar interferometry（InSAR） to generate Digital Elevation Models（DEMs）. InSAR atmospheric correction methods come in various types. However

they are rarely used in InSAR DEM generation. The method based on atmospheric numerical models offersunique advantages.It can estimate atmospheric water vapor field at the SAR data acquisition time

and it is not affected by clouds. Therefore

this work studies the atmospheric correction strategy and method during the process of InSAR DEM generation based on the atmospheric numerical model called the Weather Research and Forecasting（WFR） model to improve the accuracy of the generated InSAR DEM.First

atmospheric correction strategies are discussed. These strategies include the proper settings of WRF Preprocessing System（WPS）to make the temporal and spatial scales of WRF results match those of SAR data

as well as the selection of a proper atmospheric correction timing to improve the accuracy of InSAR DEMs. One of the key issues of InSAR is phase unwrapping

and atmospheric correction removes the residual atmospheric phase from the interferometric phase of SAR data. Thus

there are two possible atmospheric correction timings: before phase unwrapping and after phase unwrapping. Atmospheric correction before phase unwrapping is theoretically helpful for phase unwrapping because the contribution of the atmosphere is removed. Atmospheric correction after phase unwrapping is commonly used in the field of Differential InSAR（D-InSAR）

and its efficiency has been validated. The topographic phase in the process of D-InSAR is removed before phase unwrapping

which makes phase unwrapping easier than that in the process of InSAR DEM generation. A method of atmospheric correction based on WRF results is then introduced. The direct output results of WRF are not atmospheric water vapor field

which is needed to calculate the residual atmospheric phase. Thus

a method for transforming the direct output results of WRF to integrated water vapor（IWV） andthen transforming the IWV to the residual atmospheric phase is introduced. The original coordinates of the WRF results are transformed to the coordinates of SAR data. A workflow of atmospheric correction during the process of InSAR DEM generation is proposed. Experiments are carried outwith Terra SAR-X data to validate the efficiency of the proposed methods. The accuracies of the generated InSAR DEMs with residual atmospheric phase corrected at each possible timing are compared. The compared qualitative and quantitative results prove that the atmospheric corrections work at both timings and that the atmospheric correction before phase unwrapping performs well. However

the atmospheric correction before phase unwrapping may not always work or may worsen the accuracy of the DEM resultin regions where interferometric qualities are poor. The proposed methods are then applied to the fusion of multi-baseline and multi-frequency InSAR results

with the experimental resultsalso proving the efficiency of the methods and the good performance of atmospheric correctionin regions of good interferometric quality.