CAJ | 학술논문

利用大兴安岭林区外业调查的72块样地数据、遥感影像数据及地形数据，建立森林地表可燃物载量估算模型，推算林区可燃物载量。初步选出19个自变量因子，包括从校正后的林区遥感图像上计算的各种植被指数（如：坡度、高程等），采用SPSS 统计软件分析这些变量与对应样地可燃物载量的相关性，分析可燃物载量与各类遥感变量间的相互关系，提取相关性高的自变量建立常规多元统计模型、线性与非线性偏最小二乘模型来估算可燃物载量。结果表明：多元统计模型逐步回归法建立的模型相关系数0．797，决定系数0．6346，拟合精度73．62％，预测精度70．2％，均方根误差6．9 t/hm2。线性偏最小二乘模型的0．7575，拟合精度78．98％，预测精度76．3％，均方根误差2．49 t/hm2。非线性偏最小二乘模型决定系数为0．8325，拟合精度83．82％，预测精度82．67％，均方根误差24．5 t/hm2。可见，偏最小二乘回归法要优于逐步回归法，非线性偏最小二乘法优于线性偏最小二乘法。
이용대흥안령림구외업조사적72괴양지수거、요감영상수거급지형수거，건립삼임지표가연물재량고산모형，추산림구가연물재량。초보선출19개자변량인자，포괄종교정후적림구요감도상상계산적각충식피지수（여：파도、고정등），채용SPSS 통계연건분석저사변량여대응양지가연물재량적상관성，분석가연물재량여각류요감변량간적상호관계，제취상관성고적자변량건립상규다원통계모형、선성여비선성편최소이승모형래고산가연물재량。결과표명：다원통계모형축보회귀법건립적모형상관계수0．797，결정계수0．6346，의합정도73．62％，예측정도70．2％，균방근오차6．9 t/hm2。선성편최소이승모형적0．7575，의합정도78．98％，예측정도76．3％，균방근오차2．49 t/hm2。비선성편최소이승모형결정계수위0．8325，의합정도83．82％，예측정도82．67％，균방근오차24．5 t/hm2。가견，편최소이승회귀법요우우축보회귀법，비선성편최소이승법우우선성편최소이승법。
With 72 pieces of sample data in the Daxing’ an Mountains forest region field, remote sensing image data and the ter-rain data, we established the forest surface fuel loads estimation model and estimated the forest fuel loads.We selected 19 preliminary independent variable factors including the indexes from the forest after correcting remote sensing image on all kinds of vegetation indexes (slope and elevation) by SPSS to analyze the correlation of these variables with the correspond-ing sample fuel loads and the relationship between fuel loads and all kinds of remote sensing variables, to extract the high correlation between independent variables to establish conventional multivariate statistical model and the linear and nonlin-ear partial least squares models to estimate the fuel loads.The multivariate statistical model by regression method is with the correlation coefficient of 0.797, the decision coefficient of 0.634 6, the fitting accuracy of 73.62%, the prediction ac-curacy of 70.2%, and the root mean square error of 6.9 t/hm2 .Linear model of least squares is 0.757 5 with the fitting accuracy of 78.98%, the prediction accuracy of 76.3%, and the root means square error of 2.49 t/hm2 .Nonlinear partial least squares model decision coefficient is 0.832 5, the fitting accuracy is 83.82%, and the prediction accuracy is 82.67%with the root mean square error of 2.45 t/hm2 .The partial least squares regression method is superior to the stepwise re-gression method, and the nonlinear partial least squares is superior to linear partial least squares.