Using the monthly precipitation and temperature dataset from 59 statio ns over Shanxi Province, and the 500hPa height data in North hemisphere and sea surface temperature over 10°S～50°N、120°E～80°W in North Pacific during 196 1-2005, first of all, the index which represent the drought grade in Shanxi Prov ince were defined; secondly, the climatic variability and anterior signals of dr ought in Shanxi were analyzed by using the triple splinefunction and power spectrum. The results show that the drought exhibits a statistically significant interdecadal variability and interannual timescale variabilities with near 2 ye ar, 3.5 year and 5.6 year periods. Based on this analysis, a Huabei predicting drought model on an integration of multiscale was improved and modulated and a series of prediction results were p rooftested in months and seasons. By analyzing the forecast of drought grade in months and seasons during 2000-2005, it shows that the modeling techni que catches the change in drought well and exhibits a higher prediction skill.