In the course of analyzing and forecasting regional groundwater depth, its variation process complexity was ignored in the past, so that it was difficult to realize scientific management of groundwater resources. Aiming at the aforesaid defect, taking Jiansanjiang branch bureau as research platform, this paper diagnosed the complexity of each monthly groundwater depth series through using complexity measurement methods which include wavelet theory, fractal theory, Approximate Entropy and so on. The comprehensive measurement analysis results show that the complexity order of monthly groundwater depth series in the eighth district of Farm Nongjiang is highest in four farms of northern region. On that basis, the most complexity groundwater depth series combination stochastic forecasting model based on wavelet transform in northern region of Jiansanjiang branch bureau was set up. The model forecasting results show that local annual mean groundwater depth will be in a state of fluctuant lifting, and have considerable gap in comparison to groundwater ideal depth in the future. So, local government should reinforce the scientific management of groundwater. The research achievements break through the traditional study paradigm of hydrological forecasting, provide a research mode for studying regional hydrologic process, and provide scientific evidence for sustainable utilizing regional groundwater resources.