【Objective】 This study investigated drought characteristics at annual and inter-annual scales in the Yellow River basin to provide reference for the assessment of extreme climate in arid and semi-arid regions.【Method】 Grid cells with 0.5°*0.5° were set up based on VIC model. Combined with the multiply years precipitation data, the average runoffs of all sub basins were calculated, and the plotting position Gringorten algorithm was used to construct a non-parametric multivariate standardized drought index (NMSDI).【Result】(1) On 6-months scale,NMSDI was similar as standardized precipitation index (SPI) in early drought and similar as standardized streamnow index (SSI) in end drought. (2) The drought in the upper reaches and north of Yellow River was more severe than other regions, and drought risk in spring and summer was higher than in autumn and winter. (3) The annual and inter annual NMSDI indexes of the Yellow River basin decreased, while R/S analysis showed that the trend in future would reduce. (4) In the Yellow River basin, annual NMSDI series would tend still, only mutations occurred in the Wei River basin in 1991 as a result of climate change and human activities.【Conclusion】 Compared with traditional drought index, NMSDI does not rely on assumptions about the distribution of function and avoids the lack of different indicators. It has good applicability in the study watershed.