Remote Sensing Based Estimation of Forest Aboveground Biomass at County Level

In this paper, Xixia County in Henan Province was chosen as the case study area, and Landsat 8 image in 2013 and 217 fixed plot data of forest resources continuous survey in the same period were collected as the main information to estimate forest above ground carbon in the study area. Four remote sensing based models namely multivariate linear regression (MLR), classification and regression tree (CART)， bagging (Bagging) and random forest (RF) were established by using 9 vegetation index and three terrain variables. Five indicators of correlation coefficient (COR), mean absolute error (MAE), root mean squared error (RMSE), relative absolute error (RAE), root relative squared error (RRSE) were figured out to evaluate the performance of the four models by using 10 fold cross validation method. Then the model with the best performance was applied to predict forest aboveground biomass in 2013. Results showed that: Among the four models, the performance of random forest was the highest, followed by bagging method, while the performance of multiple linear regression was the lowest; The terrain factors including elevation and slope, soil conditions (e.g. brightness, wetness), the vegetation index (vertical vegetation index, effective leaf area index) were the six enforcing variables impacting regional forest carbon; In 2013, the unit forest biomass in study area was 3856 t/hm2, in which the percentage of low (<40), medium (40-60) and high (>60) was 5992%, 2430% and 1578%, respectively; The places with higher forest above ground biomass in study area was mainly distributed in the northern rocky mountains with inconvenient traffic conditions, high forest cover and less human disturbance, while places with lower forest biomass was located in the southern Guan River valley with good traffic conditions, high population density and gentle slope.