As an accompanying product of the photosynthesis of leaves， solar-induced chlorophyll fluorescence contains abundant photosynthetic information， so it is considered as a fast and non-destructive indicator that can well reflect the photosynthesis of plants. Chlorophyll fluorescence plays a unique role in studying plant stress， monitoring plant diseases and insect pests， and also estimating the gross primary production. Gross Primary Production （GPP） is an important part of the researches on global climate， carbon cycle change and the global ecosystem. Grasping the spatial and temporal distribution characteristics of GPP accurately and timely is conducive to an in-depth understanding of the interactions between biosphere and atmosphere. It can provide corresponding suggestions and policies for the ecological process management of global climate change mitigation. Compared with vegetation index， chlorophyll fluorescence is more sensitive to photosynthesis， which has been proved to be a more direct estimation method of GPP. The chlorophyll fluorescence model has significant advantages over other traditional estimation methods. It is of profound importance to discuss the basic principle， methods， uncertain， latest breakthrough， the challenges and future trend of solar-induced chlorophyll fluorescence in the field of remote sensing estimation of GPP.
Development and validation of Leaf Area Index （LAI） product from satellite remote sensing data is a crucial research topic in vegetation remote sensing. Over the past decade， a number of global and national LAI products， such as GLOBALBNU， GLASS， GLOBMAP， MuSyQ， and FSGOM have been developed in China from MODIS and AVHRR observations. These products have been widely used in home and abroad. At the same time， Chinese scholars have carried out extensive product validation studies at global and regional scales. This paper summarizes the current status and future development trends in LAI product development and validation in China. During the past years， significant progresses have been made in theory， technology and method studies in this field. The accuracy and continuity of domestic LAI products are on par with the advanced international level. However， there are still some drawbacks， such as heavily relying on data sources from abroad， unclear algorithm uncertainties， discontinuous product， and lack of sufficient validation， which greatly limit the breadth and depth of the product application. For future research， new satellite data， especially domestic satellite data， should be fully harnessed. The development of remote sensing models and inversion algorithms should be strengthened， and applications broadened in order to generate high quality LAI products to meet the research needs in Earth system sciences. In LAI product validation， current field measurement infrastructure should be improved， more extensive validation sites be developed， international collaboration be facilitated， and product usage broadened. The product market should be improved through more interactions and feedbacks with product users. With the increasing funding opportunities in this field， it is expected that the next two decades will see China's LAI remote sensing production and validation studies transit from a “following” role to a “parallel running” and even a “leading” role internationally.
Estimation of forest Leaf Area Index （LAI） using waveform LiDAR has been performed in many studies. However， the LAI estimation from waveform LiDAR was affected by terrain slope. Terrain slopes can blur the boundary between ground and canopy returns in a waveform LiDAR， and it is difficult to obtain accurate ground return and canopy return. In order to estimate the LAI under different terrain slopes， a slope-adaptive method was used to process the airborne LVIS and spaceborne GLAS waveform data. First， the ground peak position was obtained by slope-adaptive method. Subsequently， the ground return and canopy return were separated based on the height threshold. Finally， the energy ratios were calculated for LAI estimation. For the LVIS and GLAS data， LAI of different forest sites was estimated and validated with the field LAI. The result shows that forest LAI was successfully estimated with waveform LiDAR data， and the slope-adaptive method can overcome the effect of terrain and improve the accuracy of LAI estimation. For airborne LVIS， the accuracy of LAI in New England is R2 = 0.77 and RMSE = 0.21. For spaceborne GLAS， the accuracy of LAI in the Saihanba is R2 = 0.81 and RMSE = 0.28. No matter on the airborne or spaceborne data， the proposed method indicates high accuracy and shows potential for LAI estimation over complex topography.
Terrestrial Gross Primary Production （GPP） is a key component of the carbon cycle， which represents the ability of plants to absorb and fix CO2 in the atmosphere. Light Use Efficiency （LUE） model is commonly used in regional simulation of GPP. Leaf Area Index （LAI） is a key input data in TL-LUE model. There are great spatial and temporal difference between various LAI data. Difference in spatial and temporal patterns between GPP simulations derived with different LAI needs to be investigated further. In this study， three satellite-derived LAI data， MCD15， GLASS and GlobMap， were used to simulate GPP in China from 2003 to 2017. Firstly， three LAI data were compared to investigate the difference in the spatial and temporal patterns. Then， GPP simulated by three LAI data were compared to investigate the difference. Results showed that spatial and temporal patterns of LAI differed substantially among different LAI data， and there were great differences in forest regions. Averaged annual value of three LAI data showed significant increasing trends from 2003 to 2017（p<0.01）. However， the interannual variation of the annual mean value of different LAI data were obviously different. The GPP simulated by GLASS LAI had high correlation with EC GPP. Mean annual total GPP in China simulated with different LAI data has great difference， varied from 6.39 Pg C a-1 （GlobMap） to 7.46 Pg C a-1 （GLASS）. Annual total GPP in China simulated by three LAI data showed significant increasing trends from 2003 to 2017 （p<0.05）. However， the interannual variation of different annual total GPP were obviously different. The spatial and temporal patterns of GPP differed substantially among different simulated GPP， and there were great differences in forest and crop regions. This study was helpful to assess the uncertainties of regional GPP simulation derived from input data.
In order to evaluate the temporal and spatial uncertainty of the trend of the Leaf Srea Index （LAI） in China caused by the update of the versions， four sets of long-term LAI datasets are used （MODIS LAI， GLOBMAP LAI， GLASS LAI and GIMMS LAI3g）， and the trends in total， spatial distribution and different vegetation types between the former and last versions were compared. Results showed that the positive trend in GLASS LAI is higher than that in GLOBMAP LAI and GIMMS LAI during the past 30 years but the discrepancy between versions is not obvious. However， the uncertainty due to update of the versions occurs after 2000. The trend of last version （12.1±2.1×10-3m2/（m2·year1）） is higher than former version （7.9±2.0×10-3 m2/（m2·year1））， and the higher trends value mainly appear in the southeast coastal areas， the northeast China， the Yunnan-Guizhou Plateau and the Loess Plateau. The differences in trends and the net change of LAI are concentrated in the vegetation types of cropland， grassland， shrubland and evergreen coniferous forests. The study quantifies the uncertainty of China’s trend of LAI among versions and offers references of data choice for further LAI researches in China.
In order to clarify the trend of vegetation growth and its response to the changing environment factors in China during the past 30 years， three sets of long-term satellite-based LAI （Leaf Area Index） datasets and eight process-based ecosystem models are used to analyze the trend of LAI and its attribution. In the total， the trend of satellite-based LAI datasets （9.8×10-3m2/（m2·year1）） during 1982~2015 is much higher than ecosystem process-based models LAI datasets （4.2×10-3m2/（m2·year1））， in which CO2 is the dominant contributor （3.5×10-3m2/（m2·year1））. In the spatial pattern， the satellite-based LAI datasets show that about 79.5% of area in which LAI has a significant increasing trend， while about 33.1% of the area in which LAI simulated by process-based ecosystem models shows a growth trend. Except for grassland， the other vegetation types all shows that the LAI from models underestimates the growth of vegetation. The response to the changes of precipitation is too sensitive in models and models’ insufficient ability to simulate human activities are important sources of uncertainty in the model’s simulating the trend of LAI in China. The study quantitatively analyzes the change of LAI and its drivers of various vegetation types in China in the past 30 years， and conducted attribution analysis. This study also explains the underestimation of vegetation growth in process-based ecosystem models， which provides a reference for subsequent research on vegetation in China.
Leaf Area Index （LAI） is the key indicator for ecological monitoring and application in agricultural production. Retrieve precision improved LAI using quantitative algorithms has been a comprehensive work for the ecological research. The paper developed a time series LAI inverse method by using Data-Based Mechanistic（DBM） modeling method and time series multi-angular remote sensing observations. Based on radiative transfer theory， the work used RossThick-LiSparse-Reciprocal（RTLSR） and Scattering by Arbitrarily Inclined Leaves with Hotspot（SAILH） model to extract the vegetation canopy bidirectional reflectance character. The Anisotropic Index （ANIX） derived from MODIS BRDF product was used to express the directional reflectance signature of vegetation canopy， and the MOD09GA multi-angular remote sensing observation and MOD15A2 LAI products data were used together in time series LAI modeling and estimation. Typical vegetation sites data are used to make validation of the LAI inversion. The basic inversion results shows that： （1） Time series multi-angular observation data combined with DBM LAI inversion method can be used to improve the integrity of LAI estimation in time series. The developed method can reduce the disturbance from observation data noise in DBM modeling and estimation. （2） Anisotropic index data enriched the vegetation canopy directional reflectance signature. It not only works for improving the time series LAI inversion but also provides the surface bidirectional reflectance properties for the other relative land surface parameters retrieved. （3） The preliminary results are superior to the MODIS LAI product in time series integrity and data value stable.
The rapid development of satellite remote sensing technology makes it possible to obtain global large-scale Leaf Area Index （LAI）. However， it is difficult to estimate high-resolution LAI based on existing algorithms and data. In this paper， four typical research areas consist of three research areas for modeling validation and an independent research area for applicability validation including grassland， farmland and woodland were selected. Field data and correspondent satellite of the areas were then collected. The empirical model of NDVI vegetation index， the BP neural network model and the BP neural network based on simulated annealing algorithm were established and 30 m resolution LAI data were estimated with all models. Estimated results were validated with the Field data in three main research areas. The results indicated that NDVI empirical model has the worst accuracy in the three main research areas selected in this paper. The estimation accuracy of BP neural network model based on simulated annealing algorithm is higher than that of BP neural network model. The determinant coefficients of estimation results of farmland， grassland and woodland sites are 0.899， 0.858 and 0.863 respectively. The determinant coefficients of BP neural network model were 0.763， 0.710 and 0.742 respectively， while the determinant coefficients of NDVI empirical model were 0.622， 0.536 and 0.637 respectively. In order to verify the applicability of SA-BP neural network， an independent research area was selected for further verification. The results show that the validation accuracy is high，R2 is 0.842 0， and RMSE is 0.689 5， which shows that the model has good extrapolation ability. This study proves that the BP neural network model based on simulated annealing algorithm improves the generalization ability of the model， effectively prevents the BP neural network model from sliding into the local minimum， and it is an effective method for LAI estimation in high spatial resolution.
Terrain correction is an important approach to improve the accuracy of remote sensing quantification of surface parameters in complex terrain areas. The widely used remote sensing Leaf Area Index（LAI）productsalwayshave certain terrain error. It has a great importance to eliminate the influence of terrain and improve LAIproducts’ accuracy. Taking the Qianyanzhou area of Jiangxi Province as the research area， the paper aims to establish a terrain correction model which takes terrain error into account to promote the accuracy of GLOBMAP LAI product. Based on the measured LAI data， Landsat TM data， GLOBMAP LAI product and elevation data， the model achieved terrain correction by establishing the index relationship between elevation standard deviation and LAI product values. The terrain correction model of GLOBMAP LAI product was established ， and then used to correct the product in the study area. The results indicated that the corrected leaf area index was closer to the ground measured data， and the RMSE between the LAI product and the ground measurement decreases from 2.11 to 2.04. The standard deviation of the corrected LAI dataset was reduced from 2.08 to 1.69， which meant the terrain error could be eliminated. The method in this paper had well completed the terrain correction of LAIproduct. The model is meaningful to improve the accuracy of LAI product.
This study used Sentinel-2A and Landsat-8 images of Fuzhou in Fujian， Nima in Tibet， China and French Island in Australia to assess the performance of three commonly-used water indices， i.e.， Modified Normalized Difference Water Index （MNDWI）， Automated Water Extraction Index （AWEIsh and AWEInsh） and Water Index 2015 （WI2015）. The objective threshold value， i.e.， 0 threshold， and random forest importance assessment method （Gini coefficient） were adopted to do the comparison with different water types （river， lake， and ocean）. Among the water enhanced images of the three indices， MNDWI-enhanced water image has the highest contrast and rich information， whereas AWEI and WI2015 have relatively low contrast and are less informative. Accuracy validation shows that the water features extracted by the three indices all have high accuracy. Nevertheless， the average overall accuracy of MNDWI in the three areas is slightly higher than that of WI2015 and AWEI， which are 91.83 %， 91.16 % and 90.07 %， respectively. In addition， MNDWI can detect small water bodies and remove mountain slope shadows more effectively than the other two indices. The importance assessment revealed by the Gini coefficient of random forest further shows that MNDWI has the strongest importance in the separation of water with non-water features， especially shown in Sentinel-2A images， while WI2015 and AWEI have a relatively lower importance.
Fire is a global natural hazard. Effective methods of active fire monitoring would significantly contribute to disaster risk reduction as well as the studies on climate change. Based on the MERSI data from a new generation polar-orbiting FY-3D satellite in China， we proposed an improved algorithm for potential fire pixels identification. Then， dynamic threshold and a contextual fire detection algorithm are combined to carry out the fire monitoring experiment. FY-3C VIRR data， MOD14A1 fire products， and Landsat8 OLI data are used to validate and analyze the detection results. The results show that the improved algorithm can effectively detect fire spots including small fires， which provides a method for the effective hazard monitoring.
Analyzing the eco-hydrological structure of the lake wetland by remote sensing is of great significance for maintaining its ecological service function. However， the available high-resolution remote sensing images at specific water levels may be absent due to the influence of atmospheric conditions， and the spatial-temporal fusion technology in remote sensing is an important approach to compensate for this deficiency. Shengjin lake wetland in Anhui province was used as the research area in our study. The high spatial-temporal resolution remote sensing images were simulated by the Enhanced Spatial and Temporal Adaptive Reflectance Fusion Model （ESTARFM）， and the numerical accuracy of simulated images were evaluated by comparing with real Landsat8-OLI images. Moreover， five water indices were evaluated and the optimal water index was selected to extract the water information. Finally， the high-resolution remote sensing images at specific water level were simulated to extract the water information and analyze the eco-hydrological structure of Shengjin lake wetland. The results showed that： （1） ESTARFM could effectively simulate high-resolution remote sensing images. The correlation coefficients between fusion images and real images in near-infrared band and short-wave infrared band reached 0.93 and 0.91 respectively， and the Root Mean Square Error（RMS） were 0.06 and 0.036 respectively. Additionally， the shorter the date interval between the input images and the fusion images is， the higher the simulation accuracy will be； （2） The water extraction results of lake wetland were evaluated by different water indices and the New Combined Water Index （NCWI） had the highest accuracy with Kappa coefficient of 0.95 and overall accuracy of 96.78%； （3） The NCWI was adopted to extract water body information in High-resolution remote sensing images at different water levels. According to the analysis of Eco-hydrological structure of Shengjin lake wetland， the wetland central area， appropriate activity area and inappropriate activity area were approximately accounted for 32.8%， 12.1% and 55.1% of the total wetland area respectively.
To overcome the low classification accuracy problems in complex land use regions， a case study is carried out to develop a new classification method based on two traditional classification methods and Rapid Eye remote sensing imagines in the eastern part of ecological conservation region in Beijing City. Firstly， the land use classification system is developed and these land such as cultivated land， water body， build-ups， forest， shrub， mine lot and quarry are included. Secondly， the imagines are segmented into 37 100 polygons using object-oriented technology according to different spectral features， structural features and morphological features. Thirdly， the land use types are identified using Decision Tree method and the Nearest Neighbor method. The overall accuracy values are 75% and 71% for the Decision Tree method and the Nearest Neighbor method， respectively. The Kappa coefficient values are 0.69 and 0.71 for the Decision Tree method and the Nearest Neighbor method， respectively. The results show that the Decision Tree method is with higher accuracy in the regions with distinct spectral characteristics such as water body， vegetation and cultivated land， while the Nearest Neighbor method is with higher accuracy in the regions with similar spectral characteristics such as shrubs and forests. Fourthly， a new optimized combination classification method is proposed based on these two methods with the overall accuracy of 90% and the Kappa coefficient of 0.9. Finally， the land use changes are analyzed in ecological conservation area in Beijing from 2010 to 2018 based on the new method. The results show that the ecological damage zone has being repaired and the areas for mine lot and quarry have being declined. These results could provide technical support to explore the evolution process and the disruption characteristics in the ecological conservation region.
In order to solve the problems of low precision of cultivated land information extraction due to complex vegetation types， complex terrain and broken plots in southern China， a method of arable land area extraction under complex conditions of object-oriented and cart decision tree is proposed. Taking Longan County and Wuming County of Nanning City， Guangxi as the study area， using Sentinel-2A image， combining digital elevation data DEM and normalized vegetation index NDVI and other multi-source data， using object-oriented segmentation technology to identify plot information， and then using CART decision tree classification method， according to the shape and spectral characteristics of different land types， the cultivated land in the study area is extracted. The results show that the overall precision and Kappa coefficient of the object-oriented CART decision tree classification method are 96.1% and 0.94， respectively. Compared with the total accuracy of cultivated land information extraction of cart decision tree without object-oriented segmentation， the kappa coefficient is increased by 0.54. The object-oriented segmentation method is beneficial to reducing the influence of complex background on the extraction of cultivated land. Based on the object-oriented CART decision tree classification method， the extraction of the cultivated land information in the research area is better than the traditional method， and the extraction precision of the cultivated land information can be improved.
Canopy density is an important parameter reflecting the quantity and quality of forest， and also one of the important factors of forest survey. This paper is based on the airborne LiDAR point cloud data obtained from the experimental area of Gaofeng Forest Farm in Guangxi Zhuang autonomous region， the canopy height model （CHM） and three-dimensional point cloud were used to estimate forest canopy density. The accuracy of canopy density estimation results were evaluated by using 105 field samples as reference data. The results showed that R2=0.388 and RMSE=0.17 between canopy density estimation and measured values based on canopy height model（CHM）.Two methods are used to estimate canopy density based on three-dimensional point cloud： In the first method， canopy density was estimated by the ratio of point cloud density of vegetation at a height of more than 2 meters after normalization to density of all point clouds after normalization， and R2=0.467 and RMSE=0.13 between the estimated results and measured values. In the second method， the density of vegetation point cloud in the first echo at a height of more than 2 meters after normalization and the density ratio of all point cloud in the first echo after normalization were used to estimate canopy density. R2=0.478 and RMSE=0.12 between the estimated results and the measured values. The accuracy of the two methods based on three-dimensional point cloud is better than that based on Canopy Height Model （CHM）. Among the methods based on three-dimensional point cloud， the accuracy of the second method is better than that of the first method.
The geological structure of Beishan area in Gansu Province is complex， and the magmatic activity is intense. Due to the low level of work in this area， the 1∶200 000 and 1∶50 000 geological maps have been used to delineate the lithology （such as medium-acid intrusive rocks） of the Beishan， Gansu， but the boundaries are not accurate enough. Taking Baixiani Mountain in Beishan as the research area， the ETM multi-spectral image and the ZY3 panchromatic high spatial resolution image were fused by color spatial transformation （IHS）， Brovey and other methods to obtain the high-resolution image with both ETM spectral resolution and ZY3 spatial resolution. Then the raw image was enhanced by ratio， principal component analysis and false color synthesis to highlight the lithological differences. The images processed by various methods were combined with Digital Elevation Model（DEM） data to construct 3D images for comprehensive interpretation. Based on the field verification of the interpretation results， sample thin section identification and reflection spectrum characteristic analysis， the results are modified to obtain the geological map of remote sensing interpretation in the study area. The results show that the existing geological maps can be updated by using multi-source remote sensing data fusion in western areas with good outcrop of bedrock， which can provide reference for subsequent mapping and ore-prospecting.
Tobacco is a special crop and the extraction of tobacco plays an important role in its statistics. Aiming at the difficulty of extracting tobacco plants， a tobacco fine extraction method in Unmanned Aerial Vehicle image combined with multi-features and superpixels is proposed. Firstly， the image is segmented by simple linear iterative clustering algorithm； secondly， the Mean value， Brightness， Length/Width， Shape index， Red， Green and Blue band value and custom vegetation index of super pixel are counted； thirdly， fine extraction of tobacco by superpixel features combination and features threshold selection； finally， the extracted information are satisficed and analyzed. The experimental results shown that the method can effectively extract tobacco trees， and the accuracy is 99% and 98.6%， respectively. Using this method， it provides an effective reference in calculating tobacco production， saving most of the human and financial resources.
Fine spatial and temporal resolution remote sensing images are of great significance for the study of fine scale land use and land cover change. However， the presence of cloud noise poses challenges to image interpretation and analysis. Therefore， cloud detection plays a very important role in image interpretation and analysis. The QA60 has been widely recommended as the cloud detection product for Sentinel-2 （S2） images. However， our recent research found that there was a obvious cloud noise omission in the cloud detection results based on the QA60 product. To improve the ability of the cloud noise detection in S2 satellite images， this study developed cloud segmentation algorithms based on two cloud-related bands （B1 and B9） and four products （QA60， AOT， MSK_CLDPRB and SCL products） of 2A-level （L2A） data with the help of Google Earth Engine （GEE） platform. By taking three typical regions as cases， we investigated the spatial patterns and differences of different cloud detection results from the perspectives of image band characteristics and cloud microphysics. Further， we also evaluated accuracies of the different cloud detection reults. Results showed that： （1） From the perspective of cloud detection algorithm， the dynamic threshold segmentation algorithms used in B1 and B9 bands presneted the good robustness. And the detection results could largely match characteristics of corresponding bands and reasonably captured the cloud patterns. （2） For the spatial distributions of cloud， a relatively poor performance was observed in AOT product. The reliabilities of B9 band and QA60 product were relatively low. By contrast， the cloud detection potentials of B1， SCL and MSK_CLDPRB were much stronger. （3） B1 band gave the best cloud detection effect， and its sensitivity was much stronger than those of other cloud-related bands/products. Also user’s accuracy， product’s accuracy， overall accuracy and F1 score were all greater than 0.90， implying the robustest performance. This study estimated the accuracy， robustness and sensitivity of B1 （i.e. aerosol band） for cloud detection of S2 images. These findings are expected to provide some new references for further optimizing the cloud detection of satellite images.
As an important physical indicator parameter， urban impervious surface has important significance in urban management planning， environmental assessment， disaster prediction， energy consumption， urban heat island and climate change. Mapping impervious surface timely and accurately at a regional scale is of great value for urban development planning and scientific research. The paper takes DMSP （Defense Meteorological Satellite Program） and MODIS （Moderate Resolution Imaging Spectroradiometer） as the main data source to constructing Vegetation-Adjusted Impervious Surface Index （VAISI）. Using nonlinear machine learning model： supports vector regression （SVR） achieved 22 years urban impervious surface remote sensing mapping in arid area of China from 1992 to 2013 and completed the cluster analysis on the impervious surface processing changes. The research we have done suggested that in the past 20 years， the urban impervious surface of arid area of China has shown a significant expansion trend. In 2000， expansion trend reaches its maximum. The distribution of city is very dispersed in arid area， and urban impervious surface expansion rate is unbalanced in different scale levels. According to the urban expansion dynamics and the urban impervious surface area， the expansion process of the arid regions city in China can be divided into： high-speed expansion， rapid expansion， medium-speed expansion and low-speed expansion.
To solve the problems of complex data processing， subjective model recognition and complex internal mechanism in the study of ecological vulnerable human-land system， a model analysis framework based on cloud platform and big data methods was proposed. Remote sensing and socio-economic cloud platform are used to collect and process data. Self-organizing mapping neural network clustering （SOM） method is used to recognize model without prior knowledge. The trajectories was analyzed from the perspective of social-economic development and ecological friendliness by using perceptual map， and the laws between social economy and ecological environment was selected by using association rules.The experimental analysis was carried out in 65 Belt and Road countries. The experimental results effectively divided 65 countries into 10 models， and analyzed the trajectories and relationship rules of 10 models. The results show that the framework can perform the functions of data acquisition and processing， multi-model recognition of human-land system， trajectories visualization and rules detection. It effectively makes up for the deficiencies in the multi-model study of human-land system.
Urban heat island not only affects local and regional climate of the city， but also has significant adverse effects on urban air quality， energy consumption and human health. By using long time series of remote sensing data， the systematic analysis of the temporal and spatial characteristics of the heat island in megacities can provide helpful information for the formulation of policies for urban heat island effect mitigation， and thus is of great importance for the urban sustainable development in the Belt and Road region. Based on the MODIS land surface temperature products from 2001 to 2017 and land use classification data from Landsat， the spatiotemporal changes of the surface urban heat island effect were analyzed from the seasonal and inter-annual perspectives in the coastal mega cities by using the Surface Urban Heat Island Intensity （SUHII） as an indicator. The analysis results showed that during 2001~2017， the urban areas of mega cities all experienced an expanding process， and the high intensity urban heat island was mainly distributed in the densely populated core areas of the cities. Secondly， among the 10 megacities， the annual average SUHII of Karachi was strongest with the value of 3.02 ℃， and the SUHII of Chennai showed a significant upward trend （0.07 ℃/a， P<0.1）. Finally， there were seasonal differences in the urban heat island among the mega cities. In summer， the average SUHII of Istanbul was strongest with the value of 2.88 ℃. In winter， the average SUHII of Karachi was the strongest with the value of 4.45 ℃.
In the background of rapid urbanization， there is a great significance to optimizing the urban-rural land use structure of metropolises and urban-rural spatial integration development by master the change characteristics of construction land and impervious surface in urban expansion period. Based on the remote sensing monitoring dataset and the internal impervious surface dataset of urban-rural construction land since the 21st century， this research analyzes the structure and impermeable land proportion of urban and rural construction land in Harbin from 2000 to 2015. The purpose is to explore the urban expansion patterns， regional differences， construction land use intensity， and urban-rural differences. The results show that： ①From 2000 to 2015， the urban-rural construction land expanded by 158.32km2 rapidly， the trend of annual gradient and dynamic degree were firstly increased and then decreased. In the same period， from the core area of the city to the far suburbs， the scale of expansion increased in turn， and the construction focus continues move towards to the urban periphery， which shows a spatial heterogeneity obviously. ②The area and proportion of urban construction land and independent industrial and mining land increased year by year， and the sources of expansion were mainly cultivated land. The proportions of rural residential areas decreased by 13.14% from 2000 to 2015， while the structural characteristics of urban-rural construction land changed significantly. ③From 2000 to 2015， the area and proportion of impervious surface in urban-rural construction land increased by 145.32 km2 and 10.04% respectively. The land use intensity of urban construction reached a high level， because of the land use intensity of rural residential areas increased rapidly， and the gap between urban and rural areas is narrowing. The proportion of impervious surface was decreasing continuously along the direction of the urban core area to the far suburbs， but the potential for development and utilization was greater in the same direction， because the increment， proportional increment， proportional growth rate and expansion intensity of impermeable surface area was generally increasing. In general， there is a similar trend between the area of impermeable surface and the scale of urban and rural construction land， which can reveal the urban expansion track to a certain extent.
Exploring the impact of land use change on urban heat flux has important significance for urban land use planning and urban heat island mitigation. Using the pixel component arranging comparing algorithm， four Beijing surface instantaneous heat fluxes in Septembers of 2004， 2009， 2014 and 2017 are estimated by the surface parameters retrieved from Landsat series data and meteorological reanalysis data， and the spatiotemporal variation of heat fluxes in Beijing is analyzed with the change of land uses during the same period. Results show： （1） the distribution of surface temperature and heat flux in Beijing has obvious spatial heterogeneity， and the difference between mountainous areas and plains and among different land use types in plains is obvious； （2） the order of surface temperatures or heat fluxes between different land use types has consistency at these four moments. For latent heat flux， the highest is 347.85~546.95 W/m2 for forest land， followed by cultivated land and grassland， and the minimum is 225.23~349.03 W/m2 for construction land. For sensible heat flux and surface temperature， the order is reversed， the highest for construction land is 94.06~189.28 W/m2 and 25.18~32.25 ℃， followed by cultivated land and grassland， the lowest is 28.15~102.55 W/m2 and 19.25~28.38 ℃ for water body； （3） in terms of change in urban heat fluxes caused by land use transformation， when natural surface is converted to construction land， latent heat flux is reduced and sensible heat flux increases. The latent heat flux of the arable land around the city is increased by the influence of urban heat radiation， and the urban green space can effectively alleviate urban heat island effect.
Chestnut forest is widely distributed in Europe， Asia and North America， and provides notable ecological and economic benefits. Chestnut is an important economic tree species in China， with its production ranks first in the world. The method of extracting the spatial distribution of chestnut forest based on remote sensing image can provide quantitative data for its scientific management. However， the classification of tree species is difficult in remote sensing classification and there are few reports on extraction of chestnut forest based on remote sensing data. Taking Kuancheng county of Hebei province as the research area， this paper integrates MODIS high temporal resolution observations and Landsat high spatial resolution images to select the optimal time phase and classification characteristics， and then chestnut forest was mapped based on multi-temporal Landsat OLI images using Support Vector Machine. The results showed that： （1）the spectral differences were the largest among different vegetation types from April to June， followed by September， which are the key phenological periods for chestnut forest extraction， and January helps to distinguish chestnut forest and evergreen forest； （2）Reflectances in blue， green， red， near-infrared and short-wave infrared bands are the effective bands of classification. NDI， NDVI， NDWI， RSI and RVI vegetation indexes enhance the information of vegetation growth state and coverage， which are effective classification features； （3）In the classification with single temporal image， the accuracy was highest in early growing season in July， followed by late growing season in September， and poor in non-growing season in January； （4）Integrating the images of June， September and January perform best， and the mapping accuracy and user accuracy of chestnut are up to 89.90% and 87.25%. The accuracy can reach 93.45% when compared with the statistics data of chestnut forest area of local forestry bureau in 2018.
