AbstractLandslides are often considered to be a natural hazard that frequently occurs in mountainous, precipitation‐abundant regions worldwide; however, they can also occur in dry climates resulting from anthropogenic infuences, such as irrigation farming. Generally less affected by precipitation, landslides in arid regions are ideal for examining impacts of non‐precipitation controls on landslide motion, such as basal geometry. Here, we focused on 25 large, irrigation‐triggered, slow‐moving and catastrophic landslides that occurred in a desert valley near Hanford, Washington along the Columbia River. Optical images in submeter resolution from 1996 to 2020 and Sentinel‐1 interferometric synthetic aperture radar (InSAR) deformation measurements from 2016 to 2020 were utilized to map 12 catastrophic landslides and obtain time‐series deformation of 13 actively slow‐moving landslide complexes. Our results show that the catastrophic landslides usually completed their life cycle from initiation to final deposition within minutes to days, whereas the slow‐moving ones have lasted approximately 40 years, exhibiting rapid acceleration and a subsequent deceleration period that is almost 12 times longer. Though varying in rates, both types of landslides facilitate sediment transport into the Columbia River, which is of concern for aquatic habitats. Motion dynamics of the four riverside, slow‐moving landslides were strongly modulated by the Columbia River height subject to seasonal dam water release, whereas the eight river‐distant landslides primarily respond to the infiltrated irrigation water from the farmlands. We observed an apparent “early deceleration” phenomenon for the riverside Locke landslide, where the landslide starts to decelerate despite the threshold groundwater conditions being exceeded. We interpret that the early decelerations result from the forced water circulation near the basal‐surface asperities, because river bathymetry data suggest that basal surfaces of the Locke landslide (originally the Columbia River bed) are highly irregular. This interpretation was also supported by numerical simulations, which show that incorporating the forced water circulation may help accurately predict the InSAR‐observed maximum velocity and the timing of deceleration. Additionally, our analyses of basal profiles of 10 slow‐moving and catastrophic landslides in the study region reveal that basal geometry could impart a bifurcation between slow and catastrophic landslide movements. Gentle basal surfaces are more likely associated with large, slow‐moving landslides, whereas steep ones are often involved with relatively small, catastrophic landslides due to rapid kinetic energy gains while moving downslope. In general, the examined impacts of basal geometry in this study are widely applicable to natural landslides worldwide for their kinematics characterization and forecasting.