In this paper

a composite model based on regression analysis and time series analysis is constructed to improve forecast accuracy and deal with problems of data deficiencies in mid-to-long-term precipitation forecast.Generally speaking

hydrologic time series may be split into the trend

cyclic

and random components.Under this guideline

firstly

precipitation sequence is split into the three components mentioned above by using regression analysis and time series analysis methods;three submodels are respectively constructed for the three components;the independent random sequence is stabilized to modify traditional model structure;and the three submodels are added by linear superposition to establish a new precipitation forecast model(the 3rd type).Later

the paper gives accuracy assessment indexes.Results show that errors of the modified model are lower in rain spell

i.e.

its accuracy is higher than other two models.So

using historical data

the model can work well in precipitation forecast and get a proper accuracy.It is proved to be a practical model for forecast.