Jingjiang Reach of the Yangtze River, which is an important area for fishery resource conservation, is located at the connected region of lower reaches and estuary of Yangtze River. From 2002 and 2006, there were 3514.84 kg fish collected by stow net at Jingjiang coastal area. Average monthly fish catches added up to 58.59 kg in these 5 years, the maximum was 108.61 kg in June while the minimum was 31.57 kg in December which presented a typical non-stationary time-series. One stable random series could be obtained as monthly fish catches converted by natural logarithm and a seasonal difference. Model structure was determined according to criteria of residual un-correlation and model goodness was determined on the basis of Akaike Information Criteria (AIC) and Schwarz Bayesian Information Criteria (SBC). SPSS V13.0 software was used to construct the ARIMA model based on the monthly fish catches from Jan. 2002 to Dec. 2006 with consideration of residual un-correlation and concision．Akaike's information criterion (AIC) and Schwarz’s Bayesian criterion (SBC) were used to confirm the fitness of mode1. These results showed that ARIMA(1，0，0)(0，1，1)_l2 model was confirmed and the predicting error for the equation of (1-0.327B)(1-B¹²)Lny^t=(1-0.825B¹²)e_t was white noise (P>0.05). The fitting precision of ARIMA model was 71.49%-83.28% during 2003-2006, which could ideally fit the past monthly fish catches. Relative precision of the forecasting for gradual month in 2007 was 58.64%-99.44% as well as the forecasting for 2007 is 81.60%. It indicated that ARIMA Model can be well used to forecast the fish catches of coastal wetland at the estuary of the Yangtze River.