Abstract <p>L-band Interferometric Synthetic Aperture Radar (InSAR) is advantageous in measuring surface deformation over densely vegetated areas as the L-band radio waves scattered by the forest are more coherent than C- and X-bands. One of the critical disadvantages of the L-band InSAR is its vulnerability to ionospheric disturbances. The phase advance due to ionospheric disturbances can be identified (and removed) by the split-spectrum method (SSM), which utilizes the dispersive nature of the ionosphere. Under severe conditions, however, the performance of SSM strongly depends on the user’s experiences because the robust way of application has not been established. Here, we propose a robust SSM processing flow applicable to severe conditions. In a series of SSM procedures, we focus on two steps: (1) outlier removal and (2) image filtering. Outliers can be detected efficiently using a polynomial surface, and should be removed prior to the filtering of the ionospheric phase correction. To smooth the ionospheric phase, the use of a Gaussian filter is favored, after the removal of a polynomial surface, reintroduced afterwards. Those improved steps enable us to suppress the random noise without smearing out the meaningful signal of crustal deformation. Through a systematic investigation, the proper ranges of the parameters required for the improved methods are obtained, and we also find a trade-off between the look number and the filter radius. Following the improved processing flow, we apply SSM to the ScanSAR images of the Nepal Himalayas taken by ALOS-2 just after the 2015 Gorkha earthquake. Despite the narrow bandwidth (14&#xa0;MHz) and steep topography, we successfully obtained the post-seismic deformation following the Gorkha earthquake. The processing flow we established should serve as a practical recipe for applying the split-spectrum method to ALOS-2 ScanSAR data, which represents a valuable archive starting in 2014 and is now freely available from JAXA.</p> Graphical Abstract <p></p>