Aggravation of flood has been observed in Xinjiang, China. However, no reports are available addressing changing properties and related mechanisms. In this paper, flooding frequency in the Tarim River Basin, the largest inland arid river basin in China, was analyzed by using the Kernel smoothing technique and bootstrap resampling method. Besides, the flood frequency and extreme precipitation events were analyzed by using the POT method. Both stationary and non- stationary models were conducted by using GAMLSS (Generalized Additive Models for Location, Scale and Shape) to model the flood frequency and its relations with explanatory variables (e.g. time, climate index, precipitation and temperature). The results indicated that: 1) The flood occurrence in the Tarim River Basin clustered in two periods, i.e. ~1960s and ~1990s with around 2- 3 fluctuations. 2) The change of flood frequency is nonstationary process. Persistent increase of flood frequency can be observed since 1990s and the flood frequency gradually reached its peak value, implying evident aggravation of floods in terms of magnitude and frequency. 3) Winter AMO and AO are the principle influencing factors of the change of flood frequency, while winter NAO and SOI are two critical climate indices influencing flood frequency in the Tarim River Basin. The results of this study can provide great scientific and theoretical support for the management of floods and mitigation of flood hazards.