Forest Stands Delineation Using Airborne Multi-source Remote Sensing Data

Sub-compartments are forest areas or stands with basically the same internal characteristics,which are distinctly different from adjacent areas and require the same management measures.Sub-compartments are the basic units for forest resource planning and design survey,statistics and management,and play an important role in the statistics and mapping of forest biomass,carbon stock.Forest sub-compartments are usually manually delineated based on high-resolution images and field surveys.This work can be time-consuming,labor-intensive,and somewhat subjective.The results of automatic delineation can reflect the consistency of properties within a stand and the heterogeneity of different stands better.In this study,we use the canopy height model extracted from the airborne laser scanning data,the tree species classification results of hyperspectral images and high-resolution aerial images to delineate forest sub-compartments.The work was carried out as follows:(1)Supported by tree species derived from hyperspectral image,the 1 m and 5 m resolution CHM were used for multiresolution segmentation and optimization.We tested three methods to add tree species information: tree species vector map by manual sub-compartments,direct vectorization of tree species map and tree species vector map by CHM segments.By combining and optimizing the 1 m and 5 m segmentation results,we got the final sub-compartment delineation.The manual sub-compartments,the sub-compartments delineated based on 0.1 m resolution DOM and the logging sub-compartments were used as reference data.The Ultimate Measurement Accuracy rule was used to validate the accuracy of the boundary of automatic subcompartments in its consistency with the reference sub-compartments.And the overlapping ratio was also introduced to quantify the consistency between the automatic sub-compartments and reference sub-compartments.The explained variance of mean canopy height,mean height,mean DBH were calculated to validate the internal consistency and external variability accuracy of automatic sub-compartments.The automatic sub-compartments are similar to the reference subcompartments in shape,area characteristics and are most similar to the manual subcompartments.The result with highest precision between the three methods of adding tree species was picked.The proportions of automatic segments whose overlapping ratio were more than 85% with manual sub-compartments,the DOM sub-compartments and the Logging subcompartments were 43%,34%,and 42% respectively.The explained variances of mean DBH and mean height of automatically delineated sub-compartments were 97% and 98%,the explained variances of mean canopy height was 84.81%,which was 1.77% higher than the manual sub-compartments.The results indicated high internal consistency and distinct differences between adjacent sub-compartments of automatic sub-compartments.The results show that the sub-compartments boundary delineated by multi-scale CHM segmentation has obvious advantages in the accuracy and internal consistency of the boundary,which is more consistent with the distribution of forest trees at the boundary of sub-compartments and can be helpful in fine forest management planning.The CHM data has a greater potential to be applied to forest delineation.(2)A method of segmentation and merging based on features was developed to automatically delineate sub-compartments.The 5 m resolution CHM was over segmented to generate segments.Then some features about sub-compartments were calculated in each segment based on CHM and the tree species,such as canopy density,mean tree height,dominant tree species and the percentage of it.The area of over-segments and the length between adjacent segments were calculated as well.We developed a rule to merge homogeneous segments and a rule to eliminate all the segments smaller than the threshold of forest stand standard using the above features.With all these steps,we finally got the fine sub-compartments.Three types of reference forest were used as reference data for accuracy verification.The explained variance of mean tree height,mean DBH and mean canopy height was calculated using the field data and CHM.And the automatically delineated sub-compartments were compared with reference subcompartments by these indexes.The proportions of automatic segments whose overlap ratio was more than 85% with manual sub-compartments,the DOM sub-compartments and the logging sub-compartments were 22%,41%,and 59% respectively.The explained variances of mean DBH and mean height of automatically delineated sub-compartments were 97% and 99%,the explained variances of mean canopy height was 87.40%,which was 4.36% higher than the manual sub-compartments.And 2.59% higher than the multiresolution segmentation results,indicating the high internal consistency and distinct differences with adjacent sub-compartments of automatic sub-compartments.The results showed that this method can generate good subcompartments and the thresholds can be adjusted according to the purpose of the division.