Forest Above-ground Biomass Estimation for Rugged Terrain by Using ESAR Polarization Data
ZHANG Haibo, WANG Changcheng, ZHU Jianjun, FU Haiqiang
2018, 47(10):
1353-1362.
doi:10.11947/j.AGCS.2018.20170120
Asbtract
(
)
HTML
PDF (3967KB)
(
)
References |
Related Articles |
Metrics
The influence of the ground slope on radar backscatter has been proven to be greater for lower radar frequencies due to deeper canopy penetration. In order to solve this problem and obtain accurate estimation of forest above-ground biomass (AGB) in the region of rugged terrain, the analytic model integrating the topographic factors was presented based on the modified water-cloud model (WCM) and the relationship between different backscattering coefficients and the forest AGB using the airborne P-band full polarimetric SAR (PolSAR) data acquired by E-SAR. In this study, genetic algorithm (GA) was used to determine the optimal parameter values for the model, the terrain slope was divided into three grades (0~5°、5°~10°、≥ 10°). Then we analyzed the reliability and stability of the model under the condition of different slope. Meanwhile, in order to determine advantage of the water-cloud analysis model in evaluating AGB, we used common models include linear model、logarithm model、exponential model、quadratic model to comparison and analysis. Through the comparative analysis, we found that when the forest AGB at lower level, the variational trend of backscatter coefficients (HH、HV、VV) kept the same with the vatiational trend of AGB. With the increase of AGB values, this consistency in HV backscatter coefficient values to keep alone, therefore, HV polarization was the best to estimate biomass in the complex terrain region. The terrain has a great impact on estimating forest AGB, a phenomenon was that the correlation of backscatter coefficients and forest AGB decreased with the increase of ground slope. The capabilities of estimate biomass in the five models were different, from strong to weak was that water-cloud analysis model > quadratic model > logarithm model > exponential model > linear model. Meanwhile, through comparing the change of the determination coefficients (R2), these models were found that have different stabilities to estimate forest AGB in different slope levels. When the slope changed from 0~5° to 5°~10°, the stability from strong to weak was water-cloud analysis model > quadratic model > logarithm model > exponential model > linear model. With the slope from 5°~10° to ≥ 10°, this sequence became that water-cloud analysis model > exponential model > linear model > quadratic model > logarithm model. In addition, between 0~5° to ≥ 10°, this sequence was water-cloud analysis model > quadratic model > linear model > exponential model > logarithm model respectively. Although, there was different sequence in five models, the stability of the water-cloud analysis model was higher than other models. So, we tried to use water-cloud analysis model to estimate forest AGB for the study area. The result showed that the R2 between the field AGB and estimated AGB was 0.597, the root mean squared error (RMSE) was 30.876 t/hm2, the overall accuracy was 77.40%.