NOAA IMS （Interactive Multisensor Snow and Ice Mapping System） is a blended snow and ice product based on active and passive satellite sensors， ground observation and other auxiliary information， and it is most widely used for large-scale snow cover detection and relevant climate research， providing daily cloud-free snow cover extent in the northern hemisphere and having promising application prospects in snow cover monitoring and research in the Tibetan Plateau（TP）.In this study， Landsat-8 OLI images are used to evaluate and validate the accuracy of IMS 4km-resolution snow and ice product in snow cover monitoring on the TP. The results show that （1） average overall accuracy of IMS 4km snow and ice products is 76.0% and average produce’s accuracy is 88.3%， which presents that IMS 4 km snow-ice product has good accuracy in snow cover monitoring and can be used for large-scale snow cover detection on the TP. （2） The average commission rate is 45.4% and omission rate is 11.7%， which shows that IMS 4 km products overestimate the actual snow area， and the higher the proportion of snow-covered area， the lower the probability of omission rate and the higher the probability of commission rate.（3） The mapping accuracy of IMS 4 km snow cover on the TP generally is higher in the high altitudes， and the commission and omission errors of snow cover monitoring increase with the decrease of elevation. （4） Compared with less regional representativeness of ground observation data， the spatial characteristics of snow cover based on high-resolution remote sensing data are much more detailed， and more accurate verification results can be obtained. The study also shows that overall accuracy and produce’s accuracy based on the reference image instead of classified image can better reflect the overall monitoring accuracy of IMS 4km snow-cover product on the TP in comparison with other assessment indicators.
High spatial resolution snow depth data is very important to study regional climate changes and hydrological cycle. We have developed a spatial dynamic downscaling snow depth retrieval algorithm （SDD）， which retrieves the snow depth by fusing AMSR2 L1B brightness temperature data （10 km） and MODIS daily cloud-free fractional snow cover data （500 m）. By using the SDD algorithm， snow depth data （SDDsd） with a spatial resolution of 500 m in the northern Xinjiang region was obtained， and the results were verified and evaluated by comparison with snow depth provided by 30 meteorological stations and field work. The results show that： the snow depth derived from SDD algorithm and in situ snow depth are in good agreement，the R2 is 0.74， and the RMSE is 3.47 cm. After further analysis， it is found that the snow depth inversion accuracy of different land cover types is different. The accuracy of grassland is the best， followed by urban and built-up Lands， and cultivated land is relatively poor. The accuracy of snow depth is also affected by the terrain， and it decreases as the slope increases. Compared with the results of direct re-sampling of microwave remotely sensed snow depth data， the SDD algorithm effectively improves the accuracy of snow depth in shallow snow areas， the spatial distribution of snow is also more perfectly reflected. SDDsd data provides reliable data support for understanding regional climate change and hydrological cycle. In addition， with the global-scale production of long-time series of cloud-free fractional snow cover products， combined with the existing long-time series global-scale AMSR2 data， the SDD algorithm is expected to produce global downscale snow depth products.
Using the AVHRR daily cloud-free snow cover products of China's long-term series from 1980 to 2019 and the measured snow depth data from weather stations to calculate snow phenology parameter such as snow cover days， snow start days， melt out of snow cover days，length of Snow Period ，snow depth， etc. to study the temporal and spatial distribution of snow phenology. At the same time， combined with ECMWF-ERA5 reanalysis data and GIMMS NDVI3g data set to extract meteorological factors （temperature， precipitation） and vegetation factors （The start of growing season， the end of growing season，and the length of growing season）， and explore the response of snow phenology changes in northern Xinjiang to meteorological factors and vegetation factors. The results show that the average snow cover days in northern Xinjiang in the past 40 years is 81.62 days/year， 73% of the area is stable snow cover， the snow start days is in November， the melt out of snow cover days is in March， and the length of Snow Period is early November every year to the end of March and early April of the following year； the spatial distribution is uneven， of which the Altay Mountains， the Tianshan Mountains， most of the Tacheng Basin and the Irtysh Valley are the main snow-covered areas. From 1980 to 2019， the proportion of snow cover， snow cover days and the length of Snow Period in northern Xinjiang decreased year by year，the snow start days remained basically unchanged， but the snow start days was significantly advanced.ECMWF-ERA5 reanalysis data showed that there was no significant change in precipitation during the snow cover period in northern Xinjiang from 1980 to 2019， but the significant decrease in the proportion of snow cover indicates that the snow depth in the snowfall area may increase， which is consistent with the gradual increase in snow depth measured by the Northern Xinjiang Meteorological Station.The average temperature is highly correlated with the proportion of snow cover during the snow cover period， the snow cover days， and the length of the snow period， showing a significant negative correlation. The precipitation during the snow cover period is positively correlated with snow cover phenology parameters；snow phenology and its climatic effects have caused the start of growing season to be significantly earlier， and the length of growing season is extended in Northern Xinjiang.
The taiga-tundra ecotone is sensitive to climate change， and location determination helps to understand climate change in the Arctic region. At present， most of the ecotone extraction methods are manual and semi-automatic， and it is difficult to distinguish forest， ecotone and tundra from moderate resolution satellite data， which have similar spectra. The taiga-tundra ecotone reference data was generated using high-resolution aerial images and Canopy Height Model （CHM）. After that， four Random Forest （RF） classification models based on Landsat images and LiDAR data were constructed， named RF_Spring， RF_Summer， RF_Spring_Las and RF_Summer_Las. The salt-and-pepper patches were removed from classification results， and the ecotone boundaries were extracted using the connected region boundary point extraction algorithm in MATLAB. In addition， the position accuracies of ecotone boundaries were tested. The Kappa coefficients of RF_Spring_Las model and RF_Summer_Las model are 0.92 and 0.98 respectively， and the overall accuracies are 0.95 and 0.94 respectively. The classification accuracies of the two models are much higher than those of RF_Spring model and RF_Summer model. Based on the classification results of RF_Spring_Las model and RF_Summer_Las model， the extraction results of ecotone boundaries have high precision. The position errors of the lower boundaries of ecotones are 25.13 m and 25.11 m respectively， and the position errors of the upper boundaries of ecotones are 43.11 m and 44.80 m respectively. Thus， the taiga-tundra ecotone in northern Finland can be extracted from spring and summer Landsat8 images and Airborne LiDAR data， which can help long-term monitoring of the taiga-tundra ecotone.
Tidal flats are precious natural resources， and the high-precision inversion of the topography has important scientific value. However， the existing technologies and methods have great limitations. UAV LiDAR technology can quickly obtain high-precision and high-density three-dimensional point cloud data of large-area tidal flats， which is one of the important technologies for tidal flat terrain inversion. How to perform high-precision filtering of the tidal flat vegetation in the point cloud data is a technical difficulty to be solved in terrain inversion. Particularly， the universality and robustness of the filtering algorithm should be considered when the tidal flat is covered with dense heterogeneous vegetation （e.g.， different types and geometric forms）. In this paper， a tidal flat of Chongming Xitan in Shanghai is selected as the research area. Three typical vegetation coverage areas （grass， shrub and tall tree）， are selected. Three typical point cloud filtering algorithms （slope filtering， progressive mathematical morphology filtering， and cloth simulation filtering） are used to process the point cloud data， and the results are compared to analyze the applicability of the three methods. The results show that the total error of cloth simulation filtering for the three typical areas is 1.57%， 0.16% and 0.23% respectively， and the kappa coefficient is 96.74%， 98.70% and 99.30% respectively. Compared with the other two algorithms， the accuracy of the cloth simulation filtering is higher， and it is more suitable for multi-type vegetation covering tidal flat areas. Therefore， the cloth simulation filtering is used to process the entire study area. A satisfactory filtering result is obtained， which is in good agreement with the real topography. Finally， the high-precision topographic data of the entire study area is obtained through kriging interpolation.
To improve the accuracy of extracted DBH of forest trees in backpack laser scanning point clouds. Taked three mountain plantations as research objects， point clouds of tree trunk with a certain thickness in the1.3 m away from the ground was selected as the DBH slices， and the slice thickness were 0.2 m， 0.4 m and 0.6 m respectively. The slice point clouds were divided into point cloud intervals based on the intensity of point clouds to obtain variety of DBH slices. The processed slice point clouds were mapped to the 2D plane， and the DBH of the 2D points was extracted by least square method. The result shows that the best result is obtained by extracted DBH from the slice thickness of 0.6 m and the intensity interval of ［5，10］. RMSE of the three plots is 0.46 cm， 0.83 cm and 1.03 cm respectively， MAE is 0.37 cm， 0.66cm and 0.81cm respectively， relative accuracy is 97.03%， 94.73% and 96.73% respectively. Compared with the slices of complete under the same conditions， RMSE decreases by 61.34%， 25.90% and 61.71% respectively， MAE decreases by 68.91%， 31.96% and 65.97% respectively， relative accuracy increases by 6.10%， 1.95% and 5.8%， respectively. Furthermore， the number of point clouds used decreases by 97.63%， 97.25% and 97.83% respectively， the time of used decreases by 98.5%， 97.6% and 96.36% respectively. By using the point clouds in the best intensity interval to extract DBH， it can not only save time by reduce the number of point clouds， but also improve the accuracy of extracting DBH， and provide the reference for extracting other parameters.
Time series point clouds registration of transmission corridor area is an important task and problem in airborne LiDAR inspection application. This paper presents a multi-level registration method combining PCA transformation and improved ICP algorithm according to the key elements of transmission corridor. It is based on the characteristics that the power towers are not easy to deform. Firstly， PCA algorithm is used to calculate the three principle axis vectors of the corresponding power tower point cloud. By correcting the direction of the main axis， the approximate pose transformation of two power towers point clouds can be obtained. After the coarse registration， the ICP method with improved search and convergence strategies is used to achieve fine registration. Finally， the transformation parameters are applied to the total registration to achieve fast and accurate registration of the transmission corridor point cloud. The experiment shows that the processing efficiency is improved and the average point-to-point spacing distance is reduced by more than 94% after registration， which meets the demands of subsequent applications and has practical application significance.
Due to the irregular spatial distribution， various density of point cloud data and the complexity of power scenarios， the application requirements of "what you see is what you get" in practical applications and higher requirements are put forward for automatic point cloud classification. In this paper， PointNet++ algorithm of deep learning is applied to the classification of airborne LiDAR point cloud in transmission corridor， and the end-to-end automatic point cloud classification is achieved. At the same time， the effect of sample weighting on classification accuracy is analyzed. Two test datasets are used to verify the accuracy and efficiency of the proposed algorithm， and compared with the results from random forest algorithm. The experimental results show that the algorithm based on sample weighted-PointNet++ is suitable for transmission corridor point cloud classification and reaches 87.14% on the macro average F score. Moreover， the classification performance and time-consuming are better than that of random forest.
Precise spatial position of power pylon is the basis for the transmission line point cloud segmentation， power pylon points extraction， and change monitoring in airborne Light Detection and Ranging （LiDAR） power inspection application. In order to improve the algorithm efficiency of its automatic positioning and the accuracy and robustness， an automatic positioning method for the high-voltage transmission line of complex terrain is proposed. Firstly， according to the analysis of the relative height， vertical and horizontal distribution characteristics of the airborne point cloud of the transmission line， a grid preprocessing is used to remove low-level point grid and a grid cluster analysis is applied to determine the candidate clusters， and then based on the grid vertical continuous distribution coefficient， elevation distribution coefficient， convex hull coefficient and so on， the grids where the power pylon points are located， are identified and the adjacent grid center serves as the horizontal position of the power pylon. The experimental results show that compared with the previous methods， the accuracy of the proposed algorithm has increased by 11.7%， the precision and recall rate has increased by 50% and 25% respectively， especially when the terrain is rough and discontinuous， it has better robustness.
Remote sensing is an essential means to obtain DEM data in areas lacking information. However， due to the scarcity of field elevation measurements in alpine mountains， it is difficult to verify multi-source remote sensing DEM data uniformly. New remote sensing elevation data such as ICESAT-2 also lack the corresponding accuracy evaluation in alpine mountainous areas. To solve this problem， we take the Binggou Basin on the northeast margin of the Tibetan plateau as the research area. Applying the wide range of LiDAR DEM data acquired by airborne airborne remote sensing to the new product ICESAT-2 ATL06， ALOS DEM 12.5 m， the new version of SRTM V3 and ASTER GDEM were verified， and the relationship between terrain factor and RMSE was analyzed. The results show that ICESat-2 ATL06 can reach 0.747 m in RMSE in alpine mountains. Because of its high precision， it can be used to verify other remote sensing elevation in the data shortage area. The accuracy of other remote sensing elevation is relatively low. The RMSE of ALOS 12.5 m data is 5.284 m. RMSE of ASTER GDEM V3 version is 9.903 m. The five kinds of remote sensing elevation data used in this study have a high correlation with airborne LiDAR DEM， with the correlation coefficient between 0.997 and 1.000. Our study also reveals that slope is the main factor affecting the accuracy of remote sensing DEM. Except ICESat-2 ATL06， RMSE of other elevation data decreases first and then increases with the increase of slope， and there is an optimal slope value for all of them. In view of the typical characteristics of the alpine mountains on the Tibetan plateau， the verification conclusions of multi-source remote sensing DEM data in this region are representative， which can provide an important knowledge supplement for the application and evaluation of laser remote sensing DEM data in similar areas.
Heavy metals in soil are among the most harmful environmental pollutants due to their toxicity. Detecting and mapping the distribution of heavy metal using remote sensing technique is inexpensive and efficient. In this study， Sentinel-2 multispectral data and field spectroscopy were adopted to estimate soil copper （Cu） concentrations of the tailing reservoir of Tongkuangyu Copper deposit， Shanxi Province， China and the surrounding farmland soil. Sixty-eight soil samples were collected and their reflectance spectra were used to estimate Cu concentration in soil. Spectral index applicable to the prediction of Cu contents in soil was derived， united with piecewise partial least square regression （P-PLSR）， the soil Cu contents were estimated. The coefficient of determination （R2） and residual prediction deviation （RPD） for the model developed using lab-measured spectra were 0.89 and 2.81. The model was applied to the Sentinel-2 multispectral data and the spatial distribution map of Cu content was predicted with relatively high R2 （0.83） and RPD （1.56）. The result could facilitate the development of remediation strategies in terms of environmental protection. Sentinel-2 multispectral data， due to its high spatial resolution （10 m， 20 m and 60 m）， and large swath width （290 km）， could provide an alternative method for large-scale soil environment monitoring through reasonable selection of sensitive bands.
The importance assessment of ecological protection is an important component of the "double evaluation". It can reflect the importance of ecological functions in different environments in the region， and protect the ecological security of the district which includes the ecologically fragile areas and the key areas of ecological service functions. This paper proposes an ecological protection importance assessment method， which uses remote sensing data as the main driving data and combines ecological service model and ecological sensitivity index to quantitatively evaluate the ecological protection level in the whole land area of ??Qingdao. The method comprehensively considers the important ecological service function area， fragile area and protection factors， delineate the importance of ecological protection in Qingdao. The assessment results are verified according to the ecological protection red line. The results show that the proposed method is reliable for ecological protection importance assessment， and the area of ??Qingdao's key ecological protection is 1 125.57 km2， accounting for 10.34 %， mainly distributed in the area where the ecological service functions are richer and the ecological sensitivity are more vulnerable， including Laoshan mountain region in the east， the Jiaonan mountain region in the southwest， and the Dazeshan region in the north. In this paper， remoting sensing data as the main driving data provides a good technology support for the "double evaluation"， and the results have important reference for the scientific and rational compilation of urban development planning and construction of ecological civilization city.
The percent cover of impervious surfaces has been widely used as an indicator to quantify the urbanization level and urban environmental quality， and is essential to understand the interactions between human and the environment. The indicator 11.3.1 proposed by the United Nations-The ratio of land consumption rate to population growth rate （LCRPGR） requires effective monitoring of the relationship between land urbanization and population urbanization. In the light of the existing problems at present， including the lack of high-resolution and high-precision urban land products， as well as few researches on urban sustainable development in low latitude areas. Based on the Google Earth Engine platform， a method of multi-source （SAR and optical） data fusion was proposed to extract India impervious surface information with 10-m resolution in 2015 and 2018. In addition， the city scope was determined according to the population grid， and the urban impervious surface area was coupled with urban population to calculate the index. The results show that： （1） The overall accuracy of impervious surface mapping in this paper is higher than 91%， and the average Kappa coefficient is higher than 0.82， and values of R2 are 0.85 and 0.86， respectively， the overall accuracy is high. Comparised with the details of other products， the effectiveness of the method was further proved. （2） The average LCRPGR of cities is 0.76， indicating that the population growth rate of cities is higher than that of land expansion in India， and urban sustainable development faces challenges. Combined with spatial analysis， there are differences in the level of sustainable development of Indian cities from north to south， east to west， and coastal and inland.
To meet the technical requirements of building monitoring in forest districts by using remote sensing images， The Southern Sichuan Bamboo Sea is selected as the study area to form the application method of building recognition from GF-2 data. According to image characteristics of the building in the selected area， a building recognition method that combines pixel-based and object-based methods in the forest district has been proposed. First， Random Forest-Recursive Feature Elimination is used to perform feature selection on the pre-processed GF-2 images. By comparing the results of the buildings identified by using SVM classifier and RF classifier， the building in the study area obtained by SVM classifier has been selected as the pixel-level building recognition result. Then the image objects are obtained using multiresolution segmentation method， and the building targets in the study area are identified by fusing both the pixel-level building result and the image objects. The results show that the correctness， completeness and quality of the building recognition result using SVM classifier are higher than RF classifier in the pixel-level. The proposed building recognition method combining pixel-level and object-level that not only retains the advantages of simplicity and ease of use， but also avoids the phenomenon of salt and pepper. The correctness， completeness and quality of the method are better than the pixel-level or the object-level method and the quality has been improved by 0.20 and 0.13， respectively. This method can provide technical support for the superior authorities to effectively supervise illegal buildings in forest districts.
The reclamation activities of rapid urbanization process is a significant factor to cause land subsidence. This study has focused on land subsidence along with the coastal reclamation activity over Nansha district in Guangzhou city. A total of 34 Synthetic Aperture Radar （SAR） images acquired by Sentinel1 between June 6， 2015 and April 2018 are used to monitor the surface deformation and find the spatial and temporal variations of land subsidence by employing a small baseline subset Interferometric Synthetic Aperture Radar （SBAS-InSAR） technique. The results show that： （1） the Nansha district shows a trend of continuous subsidence of the whole.，but the land subsidence rate is highly polarization. The average settlement rate is 3.2 mm/a， while the center layer and the outermost layer are 2.6 mm/a and 26.8 mm/a， respectively；（2） The land subsidence shows spatial heterogeneity. The mainly distributed in the east and south， which Wanqingsha area and Longxue-Island in the south have the most serious land subsidence， with the maximum annual subsidence rate exceeding 60 mm/a （up to -68.9 mm/a）. And the land subsidence rebound phenomenon also is found from June to September， 2015. （3） Cross-validation was conducted with different Sentinel-1 polarization modes. The average values of VV polarization and VH polarization monitoring results were 2.09 mm and 1.01 mm， respectively， and the root-mean-square errors are 1.12 mm and 2.65 mm， respectively. The results show that SBAS-InSAR technology is effective and reliable in extracting land subsidence information in the reclamation area and provides scientific basis for better monitoring land subsidence in coastal areas.
The identification of geothermal anomaly areas along the Sichuan-Tibet Railway is helpful to the construction and later management and maintenance of the project. Taking The Qamdo-Nyingchi section of Sichuan-Tibet Railway as the research area， based on the landsat-8 thermal infrared image data， the surface temperature was inverted and the planetary geostationary experiment was carried out to obtain the corrected geotherm value. Focusing on the genesis and distribution of geothermal anomalies， six influencing factors， namely， formation assemblage entropy， fault buffer distance， fault line density， surface temperature， water buffer distance， and peak ground motion acceleration， were selected as the evaluation indexes of geothermal anomaly areas and the independence of factors was tested. An information quantity model was built for quantitative prediction， and the recognition results were finally divided into 5 sub-regions. The results show that the high anomaly area and the middle anomaly area account for 9.14% and 28.57% of the total area of the study area respectively， and the spatial distribution of geothermal high-temperature points is basically consistent with the evaluation results of geothermal anomaly area. The research results can provide reference for the design and construction of Sichuan-Tibet railway.
The changes of sea-land boundary have indication effect on the changes of the intertidal areas and wetland system， and also influence sailing activities. The sea-land boundary is regarded as the transient boundary between land （including tidal flat） and sea （including rivers and ditches）. Firstly， pre-processing such as color and contract enhancement and bilateral filtering is applied on false color images combined with near-infrared， green and blue bands. Secondly， the Sea-land Boundary Extraction Index （SBEI） is constructed. Then， the automatic threshold analysis method is used to divide the images into ocean and continent， combining ocean areas on land by mistake with continent afterwards. In order to solve the problem of landward extending shoreline caused by the W width rivers， a specific erosion and dilation algorithm with the matrix of W+1 width is applied. Finally， the Canny edge detection operator is used to extract the boundary on images. To verify the effectiveness of this algorithm， GF-2 satellite images of three different areas with a resolution of one meter were used to conduct the experiments. The results show that the extraction accuracy by the proposed method is about 5 pixels， which is improved at least 10 pixels compared with NDWI and CV model.
Vegetation cover is a crucial determinant of ecological environment in big cities. But the spatial-temporal dynamics of vegetation cover in the inner city and peri-urban areas in the process of rapid urbanization are still unclear and need to be researched in combination with remote sensing data. This study estimated the distribution of Fraction Vegetation Cover （FVC） of Beijing by using Landsat images， and calculated moving window mean value and standard deviation of FVC， which were respectively used as proxies for local vegetation coverage and FVC heterogeneity. Then the moving windows with significant change trend were identified by Mann-Kendall test and the slope of change was estimated by Sen’s Slope. And on this basis， we analyzed the change trend of FVC of Beijing. The results showed that during 1984~2014 the areas with significant increasing trends of vegetation coverage were mainly distributed in the urban center and the north and the west mountainous areas， and the areas with significant decreasing trends of vegetation coverage were mainly distributed in the northeast， east， southeast， south and southwest suburbs. Besides， the areas with significant increasing trends of FVC heterogeneity were mainly in flatlands while the areas with significant decreasing trends of FVC heterogeneity were mainly in the north mountainous areas.
The common rock forming minerals are found to have more obvious spectral characteristics in the thermal infrared spectrum than in the visible and near-infrared spectrum， which makes them easier to be identified and classified in the former case. Consequently， how to effectively identify and extract lithological types from thermal infrared hyperspectral images becomes a hot and difficult issue. The traditional lithologic classification is only based on the spectral shape characteristics and ignores the detailed characteristics of the spectrum. In order to resolve this problem， a new integrated method of Spectral Angle Mapper-Spectral Characteristic Parameters （SAM-SCP） is proposed. It not only utilizes the shape characteristics of the spectrum， but also makes full use of the detailed characteristics of the spectrum， which avoids the poor lithology recognition effect due to the similar shape of the spectrum curve and effectively improves the classification accuracy. The artificial and real thermal infrared hyperspectral data are respectively used for experiments with SAM-SCP. In the process of setting SCP， the weights of primary and secondary valleys were adjusted to obtain the best classification effect. The final results showed that SAM-SCP can effectively classify the thermal infrared hyperspectral images， and can get better classification results than the other traditional classification methods.
The Aerosol Optical Depth （AOD） derived from remote sensing imageries has been widely used in estimating ground-level PM2.5 concentrations in large areas. Previous studies that focused on PM2.5 estimation have reported high predictability of PM2.5 concentrations when using AOD and the advanced statistical model （i.e.， Linear Mixed Effects model （LME））. However， the interpretation ability of the LME model was lowered， as it introduced many meteorological and land use variables in the model， and the importance of each variable to PM2.5 concentrations was hard to interpret. Therefore， this study developed two nonparametric machine learning methods， i.e.， Support Vector Machine （SVM） and Random Forest （RF）， to estimate the ground-level PM2.5 concentrations. The eastern Yangtze River Delta-Fujian-Guangdong （i.e.， YRD-FJ-GD） region in China was employed as our study case， and we also compared the predictability of these two models with the LME model. The results showed that the overall R2 between estimated and observed PM2.5 concentrations exceeded 0.6 for three models， where RF received a R2 of 0.9， i.e.， 13% and 30% higher than SVM （R2=0.79） and LME （R2=0.64） model， respectively. The RMSE values were 9.07， 17.29 and 19.09 μg/m3 for RF， SVM and LME model， respectively. In addition， the spatial distribution of PM2.5 concentrations estimated from the optimal model （i.e.， RF） illustrated high annual PM2.5 in YRD （>46 μg/m3）， and GD ranked the second. FJ exhibited a relatively low annual PM2.5 （<37 μg/m3）. The seasonal PM2.5 concentrations presented a distribution pattern as winter （6.32 μg/m3） > spring （38.80 μg/m3） > autumn （36.15 μg/m3） > summer （30.16 μg/m3）. Our results revealed that the AOD and RF model could be a good proxy for estimating PM2.5 concentrations in YRD-FJ-GD region.
In this paper， we used the UAV hyperspectral images of the mangrove reserve at Qinglan Harbor， Wenchang， Hainan Province， and then preferentially selected vegetation spectral features and texture feature variables using Recursive Feature Elimination-Random Forest （RFE-RF）. We further used the Random Forest （RF） and Support Vector Machine （SVM） algorithms to classify the mangrove tree species in the study area， and further the results of the classification model parameters on the overall accuracy were analyzed and evaluated. The results showed that the overall accuracy of RF classification was 92.70% and the Kappa coefficient was 0.91. Compared with the traditional SVM classification method， RF improved the producer accuracy and user accuracy of five types of tree species， which could effectively classify mangrove tree species and provide technical support for germplasm resource planning and ecological environmental protection.
High spatial and temporal resolution data is of great significance for dynamic monitoring of vegetation productivity and ecological environment assessment. In this paper， Xiong’an New Area was taken as the research area to build a high spatial-temporal resolution NDVI data set based on our improved ESTARFM fusion model. Combined with the improved CASA model， the spatial-temporal variation characteristics of regional vegetation NPP from 2000 to 2018 were simulated and analyzed， and the impacts of temperature and precipitation on NPP were discussed. The results showed that ：（1） the improved ESTARFM fusion model predicted better performance. （2） The distribution of NPP in the study area was spatially closely related to land cover. （3） The change trend of NPP from 2000 to 2018 was not significant， but it had obvious characteristics of periodic fluctuation， which is mainly affected by urbanization development and improvement of agricultural technology level. （4） As the regional climate change causes vegetation water stress， precipitation had a more significant impact on vegetation NPP than air temperature. In short， the improved ESTARFM fusion method performs well in Xiong’an new area. The improved CASA model， which takes into account the change of vegetation cover in different periods， can simulate the NPP in the study area relatively accurately. This research can provide some scientific basis and reference significance for the sustainable development assessment of Xiong’an New Area and other similar areas.
As an important symbol of China's urbanization process， high-rising buildings have important social and economic functions and unique geometric form. We proposed to use the fully convolutional network and Sentinel-2 multi-spectral data to extract high-rising buildings within the sixth Ring Road of Beijing， furthermore， we analyzed the spatial distribution and traffic accessibility of the high-rising buildings with the vector data of ring roads， township boundaries and rail transit stations. The results show that the proposed fully convolutional network based method can efficiently and effectively extract high-rising buildings from Sentinel-2 images in Beijing. The overall accuracy is above 90%. The total area of high-rising buildings within the sixth ring road is about 192 km2. The density of high-rising buildings between the second ring road and the fourth ring road is the densest and spatially uniform. Within the second ring and between the fourth and fifth rings is the secondary group. The density is the lowest between the fifth and sixth rings. High-rising buildings in the counties of the sixth Ring Road show obvious flake gathered characteristics， the largest area of density is in Chongwenmenwai Street， Donghuashi Street and Jianguomenwai Street et al.， and they are followed by Financial street Street， Zhongguancun Street and Wangjing development Street et al.. The density of high-rising buildings in counties near the sixth Ring Road and The Forbidden City is rather low. The accessibility of rail transit has obvious spatial coupling with distribution of high-rising buildings. The lower the accessibility， the fewer high-rising buildings. The area within 1 km of subway station is about 92.62 km2， while the area at 6 km away is only 2.04 km2. Our results provide a new perspective for urban construction and ecological landscape protection in Beijing.
Combining with the kernel density analysis， the method for transforming data into regular grids and mapping double factors， we take the main urban area of Dongguan as study area， and use the nighttime light data， POI data and Lacation-based data in 2019 as data source， to obtain the same or different spatial coupling relationship of these three data and compare their relationship with the urban spatial structure. Research demonstrates that the overall spatial distribution trends of these three data types are generally consistent.， but there are different couplings in partial areas： （1） Influenced by the factors such as traffic， functional areas， and “spillover” effect of nighttime light data， the coupling of nighttime light data and POI data are different in the roads， commercial districts， and public service areas in the urban scopes； the coupling of nighttime light data and Lacation-based data are different in logistics industrial parks， schools， and suburban parks. （2） The differences of the spatial distribution of job-housing places， causing the different spatial coupling of POI data and Lacation-based data. The public services and business districts where have complete infrastructure， the density of POI data is higher than Lacation-based data； the infrastructure construction of the residential areas is relatively weak， but the population distribution is concentrated， making the POI density lower than the Lacation-based data. The integration of these three types of spatial data can effectively reflect the spatial structure of the cities and the existing problems.
