| Forest is one of the important parts of global terrestrial ecosystem.To acquire accurate information of regional forest tree type and its spatial distribution information,not only is the security and important premise for area forest area,the spatial location precision and dynamic changes of forest management and monitoring indicators,also directly affect the stand volume advantages of different tree species,biomass,species diversity and ecosystem services such as the accuracy of the model results.In the past 20a,remote sensing technology has gradually replaced artificial census as the main method for forestry departments and related scholars to obtain forest type information,However,due to the limitations of canopy sizes of different dominant tree species and the spatial and temporal resolution of observation images,it is difficult to identify dominant tree species using remote sensing data with limited time phases or medium and low spatial resolution.Therefore,it has not been able to effectively break through the technical bottlenecks such as the same spectrum of foreign bodies in the identification of regional dominant tree species,nor can it meet the general needs of fine identification and spatial mapping of regional forest types.According to the fact that the effect of both spatial and temporal scale could not be ignored in recognition results of forest types generated by multi-resolution images,besides the influence of adding texture information and vegetation phenological characteristics from remote sensed data on the accuracy of forest trees species recognition at different spatial resolutions has not been addressed clearly so far.To clarify this situation,we studied the Wangyedian forest farm in Chifeng city,southeast Inner Mongolia,China,by using quasi-synchronous and geographical coordinate matched multi-resolution satellite observations(six spatial resolution levels:1 m,2 m,4 m,8 m,16 m,and 30 m)to investigate any possible correlations between spatial resolution and the recognition result,besides the influence of adding texture information,and examined the recognition results of the dominant forest trees species obtained by using the up-scaling algorithm.In addition,we also used a total of 36 scenes covering the whole year medium-high resolution satellite observations(16 m spatial resolution)which were supported with GF-1 WFV(wide field view)to extract various time series of NDVI reflectance data(include single season,every quarter,month-to-month and every ten-days).The data contains all the seasonal phases and phenological growth stages of different tree species and is propitious for the fine recognition of forest types.Based on these multi-source remote sensing we analyze the different scales of forest information complementary characteristics,according to the regional dominant species growth characteristics,canopy texture and a variety of forest information phenological response characteristics of different image data.Five dominant forest types of Pinus tabulaeformis,Larix gmelinii,Populus davidiana,Betula platyphylla,and Quercus mongolica forest were classified and recognized using Support Vector Machine(SVM)classifier in satellite images of different spatial and temporal scales and other auxiliary data(mainly included the sufficient field plots and forest inventory data of 2017),to provides a new method reference and technical support for remote sensing identification and spatial mapping of dominant tree species in regional forest.The main contents and results of this research are as follows:1.Based on the analysis of the spatial scale effect of the remote sensing identification,we discussed the spatial scale variation rule and the influence of texture characteristic parameters of the remote sensing identification result of the dominant tree species in the regional forest emphatically,and the difference of forest tree species recognition results based on the method of scale up-conversion image was also tested.The results showed that overall classification accuracy of tree species was significantly influenced by the spatial resolution of images(P<0.05),with the highest accuracy at the 4m resolution(P<0.05),and the accuracy decreased to a minimum as the resolution reduce to 30 m.The addition of texture information increased classification accuracy using multispectral imagery with resolutions from high resolution(1 m spatial resolution)to medium-high resolution imagery(8 m spatial resolution;P<0.05),for the overall accuracy of dominant tree species recognition created after adding texture information was 2.0-3.6%higher than that from results of spectral information alone in the study area.However,the accuracy improvement does not appear to hold for medium resolution imagery(16 and 30m spatial resolution;P>0.05).In addition,there was a significant difference between the multi-scale classification results provided by up-scaled images and that obtained from native remote-sensing images for each of the same spatial scale(P<0.05),which indicated that when multi-spatial-scale remote sensing observations or applications are studied,the real satellite images should be used whenever possible to ensure the accuracy of the related results.2.By constructing different time scales reflect the characteristics of vegetation growth dynamic of NDVI time series spectrum,we combined vegetation phenology and image texture feature parameters,to fully exert different advantages of forest tree canopy spectral dynamic difference,and analyzd and compared tree species identification results based on different temporal resolution of image.The results showed that the vegetation temporal information contained in time series images is very important to distinguish different forest species,and could the largely improved recognition accuracy of forest tree species in comparison to single season data across all different seasons(P<0.05).Compared with the single data of spring,summer and autumn,the overall accuracy(OA)based on the every quarter data improved,which increased by 7.67%,6.64%and 3.6%respectively,and autumn is the best single season to identify the dominant tree species in the study area(P<0.05).Besides,the results of spectral recognition based on month-to-month and every quarter data were significantly lower than those based on every ten-days(P<0.05),which showed that the denser time series spectral information is more beneficial to the improvement of the accuracy of regional tree species identification(P<0.05).In addition,combined appropriately with spectrum differential transformation,phenological and temporal texture features increased classification accuracy using time series multispectral imagery(P<0.05),for the overall accuracy of tree species created with the combined data was higher than that from results of time series NDVI spectral alone in the study area.Among them differential transformation of NDVI time-series spectrum and combination of time-series texture and vegetation phenological characteristics at every ten-days could achieve the best recognition results(P<0.05).3.Based on the improved ESTARFM spatio-temporal data fusion model,we constructed a high-resolution data fusion application model for forest remote sensing applications.By combining the high temporal resolution GF-2 PMS data with the med-high spatial resolution GF-1 WFV NDVI reflectance data,the annual 4 m spatial resolution NDVI time-series images in the study area were reconstructed,to make full use of the rich spectral information contained in the time-series remote sensing observation data with high spatial resolution to identify the dominant tree species.The results showed that the improved ESTARFM algorithm is suitable for NDVI reflectivity image fusion between two kinds of Gaofen products,and the time series changes of canopy spectral with high resolution images can reflect the growth differences of different dominant tree species during the phenological period.Among the different recognition methods,the best result is obtained by differential transformation of NDVI time-series spectra combined with vegetation phenological characteristics(P<0.05).Based on these reconstructed images,the spatial and temporal information and variation characteristics of different dominant tree species can be better displayed,thus effectively improving the ability of multi-spectral temporal sequence data to identify dominant tree species in the forest(P<0.05). |
PDF Full Text Request