关键词: 生物量转换与扩展因子, 空间自回归模型, 林分平方平均胸径, 落叶松人工林

Abstract:

Objective: Based on the national forest inventory sample plot data of larch plantation in northeast China, the best model form of biomass conversion and expansion factor(BCEF) were discussed, and the spatial autoregressive BCEF model was established for larch plantation in northeast China. The model is useful for accurate stand biomass estimations. Method: Selecting a variety of model forms to establish BCEF general regression model, from which the best fitting model is selected. The two spatial autoregressive methods, spatial error model(SEM) and spatial lag model(SLM), were used to renew the BCEF model. The determination coefficient(R²), root mean square error (RMSE) and relative root mean square error(rRMSE) were used to evaluate the model. Moran index(MI) was applied to test the spatial autocorrelation of all variables and BCEF model residuals. Result: 1) There is obvious spatial autocorrelation in BCEF data. When the spatial distance is small, the BCEF attributes of stands within a province are similar. The differences of BCEF attributes among provinces are gradually appeared with the increase of spatial distance, and tend to random distribution finally. 2) The fitting results of allometric model, logarithmic model and hyperbolic model are better than those of other regression models, and the optimal models varied with independent variables. Stand quadratic mean diameter (D_g) is the best variable for interpreting BCEF. The R² of the effective model with D_g as an independent variable is between 0.945 and 0.958. Followed by stand mean height(H) and volume(V), the R² of the effective model is more than 0.60. The explanatory ability of stand average age is slightly lower than that of D_g, H and V, and the R² of its effective model is only about 0.50. Stand basal area(BA) and density(N) are poorly to explain the variance of BCEF with R² less than 0.50. The residuals of the general regression model showed spatial autocorrelation. 3) The spatial autoregressive model of hyperbolic function with D_g as an independent variable is the best one with SEM better than SLM. Compared with the corresponding ordinary regression model, the R² of SEM is increased by 3%, and the RMSE and rRMSE are reduced by 33% and 35%, respectively. The MI of the model residual is less than 0.02, which eliminates the spatial autocorrelation. Conclusion: The hyperbolic model is the most stable model for BCEF, and D_g is the best independent variable. It is recommended to adopt the hyperbolic function with the inclusion of spatial error model and with D_g as the predictor to estimate BCEF.

Key words: biomass conversion and expansion factors(BCEF), spatial autoregressive model, stand quadratic mean diameter, larch plantations