Matching Trees To Sites By Remote Sensing Diagnosis

Matching trees to sites is a classic forestry issue in describing tree species and site conditions,and is the basic principle of afforestation.The principle of suitable tree placement can maximize the potential of tree species and sites.It is necessary to carry out research to characterize the response relationship between tree species and site to guide tree planting and afforestation accurately and to improve forest quality.It is difficult to fully support the afforestation planning in complex areas at the regional scale with traditional survey methods.Therefore,it is urgent to introduce high technology to solve the problem of tree species and site condition selection in the afforestation covering the entire regional scale.With the widespread application of remote sensing technology in forestry,the benefits of forestry planning and design,forest resource management and forest pest control have been greatly improved.The research results of forest types and structural parameters based on multi-source remotely sensed data inversion have laid a solid foundation for the research on remote sensing diagnosis methods of suitable trees.Therefore,this study has profound scientific significance and practical value in quantitatively scientifically supporting the effective implementation of afforestation.This work aims to scientifically describe the response relationship between tree species and site factors,and to construct a remote sensing diagnostic model based on the potential suitability index and growth status index of tree species that were estimated by remote sensing,so as to provide a scientific solution for remote sensing diagnosis for solving the bottleneck problem in forestry.the research results substantially play a scientific supporting role for the forestry department in guiding tree planting and forest management.In this study,the suitable sites of six typical northern tree species including Pinus tabulaeformis,Larix spp.,Populus spp.,Betula spp.,Quercus,and Armeniaca sibirica in Chifeng city were diagnosed using multi-source remotely sensed data.The main contents and findings of this study are described as follows:(1)Development of remote sensing methodology for spatial distribution of multi-target tree species.Based on the multi-temporal Sentinel-1 backscatter coefficient data and Sentinel-2 surface reflectance images,a remote sensing feature collection with three dimensions and seven categories including spectrum,texture and time was constructed by using the powerful computing power of Google Earth Engine platform.Combining the feature collection with slope and aspect features,the Random Forest model was used to extract the spatial distribution of each tree species and map tree species’spatial distribution with an overall accuracy of 77%and a kappa coefficient of0.71 at a spatial resolution of 10 m.The results showed that Pinus tabulaeformis,Larix spp.and Populus spp.were mainly distributed in the south of Chifeng,while Betula spp.,Quercus mongolia and Armeniaca sibirica were mainly distributed in the north of Chifeng.(2)Development of remote sensing inversion algorithm for tree species’structure and function parameters.First,a vegetation index envelope algorithm(BEVIs)was developed based on the spectral index of satellite images to extract vegetation and bare soil endmembers at regional scale.Combined with dimidiate pixel model,the canopy closure with R~2 of 0.60 and RMSE of 0.13 was derived based on Landsat-8 and the canopy closure with R~2 of 0.81 and RMSE of 0.09 was derived based on Sentinel-2images,respectively.The research results prove that the algorithm is reliable and robust.In addition,based on the spectral characteristics and topographic characteristics of the satellite images,a sample plot expansion algorithm for forest volume was proposed to realize the expansion of a small number of locally distributed samples in the whole space.Combined with the Random Forest model,K-Nearest Neighbor model,and Long Short-Term Memory model,the forest volume with R~2 of 0.62 and RMSE of 43.64m~3/ha,the forest volume with R~2 of 0.62 and RMSE of 43.69 m~3/ha,and the forest volume with R~2 of 0.70 and RMSE of 34.52 m~3/ha were estimated,respectively.The results showed that the expanding algorithm can achieve reliable accuracy with different models.Finally,biomass expansion factors and carbon content factors were used to estimate the carbon density of tree species based on the forest volume inversion results.(3)Remote sensing diagnosis index calculation of tree species suitability.Based on the spatial distribution information of tree species extracted from remote sensing,the actual distribution sample collection of tree species was constructed.Combined with the maximum entropy model,the spatial distribution of potential suitability indices of the six tree species was simulated.The results showed that the potential distribution areas of Pinus tabulaeformis,Larix spp.and Populus spp.were mainly in the southwest of Chifeng,while the potential distribution areas of Betula spp.,Quercus mongolia and Armeniaca sibirica were mainly in the northwest of Chifeng.In addition,the growth status indices of each tree species were calculated based on the canopy closure,forest volume,and carbon density retrieved from remote sensing.The results showed that pinus tabulaeformis grew well mainly in the south of Chifeng,and the other five species grew well in the southwest and northwest of Chifeng.(4)Study on the thresholding remote sensing diagnosis of suitable sites for trees.Based on the potential suitability and growth status of tree species,a remote sensing diagnosis model of suitable sites for tree species was constructed.The spatial distribution patterns of suitability of six tree species,their dominant driving factors and suitable site conditions were thresholding diagnosed.The study found that(1)Populus spp.had the strongest ability to adapt to the opposite conditions.(2)The suitable areas of each tree species showed a significant spatial aggregation distribution pattern.The suitable areas of larch,Pinus tabulaeformis and Populus spp.were mainly distributed in the southwest of Chifeng,while the suitable areas of Betula spp.,Quercus mongolia and Armeniaca sibirica were mainly distributed in the northwest of Chifeng,and the suitable areas of Pinus tabulaeformis,Armeniaca sibirica,and Larix spp.were more extensive than the other three tree species.(3)The main climate factor affecting the spatial distribution of adaptation was temperature,soil factors were organic carbon content and soil particle density,and topographic factors were elevation and slope.(4)The elevations of suitable areas of Betula spp.,Quercus mongolia and Armeniaca sibirica were higher than that of Pinus tabulaeformis,Larix spp.and Populus spp.The main innovations of this study were as follows:(1)remote sensing extraction method of tree species spatial distribution with 10 m spatial resolution was proposed based on the GEE platform.(2)A spectral index-based endmember extraction algorithm for vegetation and bare soil was developed.(3)A regional-scale forest volume expanding algorithm based on spectral and topographic features was proposed(4)A thresholding remote sensing diagnosis model of suitable sites for trees was proposed.