The predictive way of two-dimensional fields of surfaoe salinity off Ohangjiang(Yangtze River) mouth with empirical orthogonal functions, herein referred to as eigen.vectors, is discussed. The eigenvectors are derived from monthly measurements of sea-level salinity at 35 points off the area during April 1975 to June 1978. The 35 eigenvectors only depend on selected 35 stations, and they are independent on time. The observed salinity deviation vectors (SD)_i ( i=1,2,……,35) can be expressed as a linear combizzation of the 35 eigenvectors. The elamet Q_ik of the matrix Q is referred to as the coefficient associated with the k-th eigenvector far the i-th observed salinity deviation vector.The coefficient is only dependent on time. These coefficients are used to find selfregreesion model of the prediction.It is shown that the empirical orthogonal function analysis has a larger potential and some advantages.