Estimating standing stocks of the typical conifer stands in Northeast China based on airborne lidar data.

To promote the application of lidar technology in estimating standing stocks of the typical conifer stands in Northeast China, i.e., spruce-fir forest, larch forest, Korean pine forest, Pinus sylvestris var. mongolica forest, we combined the point cloud data obtained by airborne lidar with the data of 800 ground plots and established models of standing stocks for the four conifer stands by stepwise regression and partial least square. Partial least squares method was better than stepwise regression method (R2=0.05-0.15, RRMSE=2.6%-4.2%). Among the three types of feature variables involved in modeling, height variable (selected for 26 times) is more important than others (selected for 12 times and 11 times, respectively). With respect to the accuracy of models established based on the means of the partial least square, they worked best for Korean pine forest (R2=0.79, RMSE=60.92, RRMSE=22.9%) and larch forest (R2=0.76, RMSE=28.39, RRMSE=25.8%), followed by spruce-fir forest (R2=0.81, RMSE=46.96, RRMSE=27.7%) and P. sylvestris var. mongolica forest (R2=0.50, RMSE=55.49, RRMSE=30.4%). This study provi-ded an effective way to estimate standing stocks of four typical conifer stands in Northeast China.

[Temporal stage and method selection of tree species classification based on GF-2 remote sensing image]. / åŸºäºŽé«˜åˆ†äºŒå·é¥æ„Ÿå½±åƒæ ‘ç§åˆ†ç±»çš„æ—¶ç›¸åŠæ–¹æ³•é€‰æ‹©.

It's important to master tree species composition and distribution in forests for the study of forest ecosystems. To promote the application of domestic Gaofen data in the classification of tree species and to explore the effects of different combining images, classification features and classifier on tree species classification results, three kinds of single temporal data and four kinds of multi-temporal data were constructed. Based on three GF-2 images, according to the multi-scale segmentation, C5.0 feature optimization as well as two classifiers including support vector machine (SVM) and random forest (RF), we finished the object-based classification of eight tree species of different temporal and feature dimensions respectively, and finally achieved good results with overall accuracy between 63.5% and 83.5% and the Kappa coefficient between 0.57 and 0.81. The results showed that the choice of temporal stage would affect the classification results. The results based on multi-temporal were generally better than that on single temporal stage. There were obvious precision differences between different image combinations of multi-temporal as well as different single temporal stage. It is notable that feature optimization played a positive role in the improvement of classification accuracy. SVM was relatively stable across different temporal and feature dimensions, which should be given priority when single temporal and classification features are difficult to distinguish tree species directly, while it should be noted that SVM was easy to overfit. RF was not easy to overfit, but it was more dependent on the quality of classification features and would get good results under favorable image combination.

[Estimation of forest vegetation carbon storage in Hunan Province, China based on k-NN method and domestic high-resolution data]. / åŸºäºŽåŸºå‡†æ ·åœ°æ³•å’Œå›½äº§é«˜åˆ†æ•°æ®çš„æ¹–å—çœæ£®æž—æ¤è¢«ç¢³å‚¨é‡ä¼°æµ‹.

To promote the application of domestic high-resolution satellite data in large-scale carbon storage estimation and measurement, a total of 206 high-resolution remote sensing images covering Hunan Province were used as the data source, and the estimated minimum unit was fixed as a 0.06 hm2 square composed of multiple pixels. Through the establishment and purification of the interpretation marks, in the extraction of forest information, the pixel-based method and object-oriented classification method were used to compare. In the estimation of carbon storage of arbor forest, the robust estimate, partial least squares method and k-NN estimate were used to compare. Finally, we estimated forest carbon storage in Hunan Province and generated the distribution map of carbon density levels. The results showed that the interpretation mark based on the automatic extraction of plots could increase the extraction accuracy of arbor forest after purification. For the estimation of forest carbon storage at large-scale, the k-NN algorithm embodied a large advantage in forest information extraction and arbor forest carbon storage modeling. The average classification accuracy of the 206 scene images was 76.8%, the average RMSE was 8.95 t·hm-2, the average RRMSE was 19.1%, and the total carbon stock in Hunan Province was 22.28 Mt. The results provided effective reference for the estimation and measurement of forest carbon storage at the provincial and national scales.

[Changes of Forest Canopy Spectral Reflectance with Seasons in Lang Ya Mountains].

The physiological mechanism and ecological structure of forest trees can change with the changes of years. In a certain extent, the changes were expressed through the canopy spectral features. The mastery of changing rules about spectral characteristics of trees over the years is benefit to remote sensing interpretation and provide scientific basis for the classification of different trees. The study adopted high-resolution spectrometer to measure the canopy spectral characteristics for seven major deciduous trees and seven evergreen trees to gain the spectrum curve of four different ages and calculate the first derivative curve. The analysis of changing rules about spectral characteristics of different deciduous trees and evergreen trees and the comparison of changes about spectrum of various trees in the visible and infrared band could find the best year and best band for identification of trees. The results showed that the canopy spectral reflectance of deciduous and evergreen trees increases with the increase of age. And the spectral changes of two species were most obvious in the near infrared band.

[Advances in researches on hyperspectral remote sensing forestry information-extracting technology].

The hyperspectral remote sensing technology has become one of the leading technologies in forestry remote sensing domain. In the present review paper, the advances in researches on hyperspectral remote sensing technology in forestry information extraction both at home and abroad were reviewed, and the five main research aspects including the hyperspectral classification and recognition of forest tree species, the hyperspectral inversion and extraction of forest ecological physical parameters, the hyperspectral monitoring and diagnosis of forest nutrient element, the forest crown density information extraction and the hyperspectral monitoring of forest disasters were summarized. The unresolved problems of hyperspectral technology in the forestry remote sensing applications were pointed out and the possible ways to solve these problems were expounded. Finally, the application prospect of hyperspectral remote sensing technology in forestry was analyzed.

[Tree species information extraction of farmland returned to forests based on improved support vector machine algorithm].

The difference analysis of spectrum among tree species and the improvement of classification algorithm are the difficult points of extracting tree species information using remote sensing images, and are also the keys to improving the accuracy in the tree species information extraction in farmland returned to forests area. TM images were selected in this study, and the spectral indexes that could distinguish tree species information were filtered by analyzing tree species spectrum. Afterwards, the information of tree species was extracted using improved support vector machine algorithm. Although errors and confusion exist, this method shows satisfying results with an overall accuracy of 81.7%. The corresponding result of the traditional method is 72.5%. The method in this paper can achieve a more precise information extraction of tree species and the results can meet the demand of accurate monitoring and decision-making. This method is significant to the rapid assessment of project quality.

[Tree-crown information extraction of farmland returned to forests using QuickBird image based on object-oriented approach].

The improvement of segmentation algorithm and the optimization of feature space are the key factors of improving the accuracy of tree-crown information extraction, and are also the urgent problems of tree-crown information extraction using high resolution images. In the present study, the spectral threshold method was used on the first-class segmentation of QuickBird multi-spectral image to obtain vegetation regions. On the second-class segmentation, the improved algorithm based on edge wa used to segment the panchromatic image, which was processed by the non-linear filtering. Afterwards, the feature space consisting of spectrum, shape and texture features was selected to extract tree-crown information. Finally, 300 random samples and an error matrix were applied to undertake the accuracy assessment of identification. Although errors and confusion exist, this method shows satisfying results with an overall accuracy of 84.67% and a KAPPA coefficient of 0.7953. The corresponding results of the traditional method are 67.67% and 0.6273. The method in this paper can achieve a more precise information extraction of the tree-crown and the results can meet the demand of accurate monitoring and decision-making.