Abstract: The following 10 curves were chosen as the models for species-area curves: S = b + aA (1) S= b+alnA (2) S=(b+alnA)c (3) S=aln(A+1) (4) S=aln(bA+1) (5) S= aAb (6) S=aA/(1+bA) (7) S=c/(1+ae-bA) (8) S=c-ae-bA (9) S=a(1-e-bA) (10) The algorithms were given to calculate the initial values of the parameters in the 7 nonlinear models, and Gauss-Newton and Marquardt algorithms were used to solve the nonlinear problems. Four indices were chosen to assess the fittness of the models, which are residual standard error (RSE), correlation index (CRI), average of absolute deviation (AAD), and average of absolute relative deviation(AARD). The results show that :l) the algorithms of calculating the initial values of the parameters in the 7 nonlinear models were suitable, and the solutions from nonlinear least squares were better than those from linear least squares according to the 4 assessment indices; 2) the fittness of the 10 models were high because 71.5% of the 200 values of CRI were greater than 0.9 and 89% were greater than 0.8. The calculation results of curves (3) and (9) were the best two; the next three were (5), (6) ,and (2) ;and the worst two were (1) and (10); 3) rank correlation analyses show that there existed extremely strong positive rank correlations between any two of the three indices of RSE, AAD, and AARD.