The Remote Sensing based Ecological Index （RSEI） has been widely used since its publication and was modified recently. In this paper， the differences between RSEI and the Modified Remote Sensing Ecological Index （MRSEI） are analyzed and compared based on the principle of the principal component analysis and an application case. The results show that the MRSEI index unreasonably adds the second principal component （PC2） and the third principal component （PC3） into the first principal component （PC1）， as PC2 and PC3 have no clear ecological meanings. The addition also reduces the weight of PC1. Therefore， the MRSEI does not improve the original RSEI， but reduces the value of RSEI as the added principal component components can cancel each other. Therefore， the modification made in MRSEI lacks rationality. This paper also analyzes and discusses some issues that users encountered in calculating and applying the RSEI index. The RSEI should be calculated using surface reflectance data rather than the top of atmospheric reflectance data or Digital Numbers （DNs）. Also， the imagery should be acquired in plant growing seasons. When there is large-area open water in study images， the water must be masked in advance. The "1–PC1" procedure can only be performed when the loadings of the greenness and wetness indicators in PC1 have negative signs.
The technology of small satellite atmospheric microwave sounder is currently a research hotspot of microwave remote sensing technology. In this paper， a small Satellite Atmospheric Microwave Sounder （SAMS） has been developed and planned to be carried on commercial small satellites to achieve rapid measurement of atmospheric temperature and humidity profiles， extreme weather and rainfall. This paper introduces the characteristics of SAMS application design， combined with the current sensitivity test data， analyzes the on-orbit application performance， especially provides a new kind of flexible means of detection for sea surface pressure and atmospheric humidity path delay. At the same time， this paper also summarizes and analyzes the conventional detection capabilities of atmospheric temperature and humidity profile retrieval， laying a foundation for the application and development of satellite data.
With the development of the remote sensing technology， the high-resolution satellite data is gradually enriched， and the information extraction of landslide disaster is further promoted. The current emergency investigation of the landslide disaster mainly focuses on the visual interpretation and field investigation， which is time-consuming， laborious， and difficult to meet the urgent need of the rescue after disaster. The single-phase landslide information extraction methods by using remote sensing based on the pixel-oriented or object-oriented have problems of over-recognition or mis-recognition of landslides. Therefore， the multi-temporal landslide information extraction method is worth studying and is expected to achieve good results， especially through the notable NDVI change in landslide. First， multi-temporal remote sensing images before and after the landslide are used as the data source. The landslide pre-selection area is determined using the pixel-oriented NDVI change detection. Then， the object-oriented geometric rules are used to complete the fine identification of landslides. This method based on the combination of the change detection and geometric rules effectively eliminates non-landslide parts which are with the spectral characteristics similar to landslides， such as roads， buildings， and bare land. Taking Jiuzhaigou landslide as the study case， the Gaofen-1 multi-spectral images of August 1， 2015 （before Jiuzhaigou earthquake） and the images of August 16， 2017 （after the earthquake） are used as data sources to conduct landslide identification experiments. The experimental results show that the multi-phase method has high accuracy in landslide identification. Compared with the object-oriented single-phase method， the former method has a mapping accuracy of up to 88.80% and the user accuracy up to 81.19%， both of which greatly exceed the accuracy of the object-oriented single-phase method. Moreover， the omission error and the mis-classification error decreased by 23.22% and 11.72%， respectively. This method determines landslides through the change of NDVI and has high timeliness in landslide identification， which does not need to consider the restrictions of excessive topographic and geomorphic factors and can be applied to most areas. It is believed that our method can provide a reliable basis for the effective organization of rescue and reconstruction work after landslide disaster.
Outfalls into rivers are the last checkpoint for man-made pollutants flowing into rivers. Accurate investigation of them plays an important role in the protection of water resources and the prevention and control of water pollution. Firstly， the progress of large-scale investigations of outfalls into rivers in the past 30 years were reviewed and the four aspects of manual field survey， GIS accounting system construction， satellite remote sensing monitoring and Unmanned Aerial Vehicle （UAV） investigation are introduced. Secondly， after analyzing remote sensing monitoring techniques for outfalls into rivers， which are based on direct visual interpretation， water environment parameters inversion and ground targets classification and other methods commonly used， the limitations of the application of the above methods on UAV images are discussed. And then， through introducing briefly the principle of the object detection method based on deep learning， the application status and key techniques of the deep learning-based object detection method implemented on the UAV remote sensing investigation of outfalls into rivers are discussed. Finally， analyzing the application prospect of deep learning on the recognition of outfalls into rivers using UAV imagery and looking forward the research emphasis of monitoring complex geographical objects including outfalls into rivers based on UAV remote sensing technique.
In this study， different types of NPP VIIRS 375 m active fire product （ACF，NPP-VIIRS active fire/hotspot data） in Handan were analysed based on heavy industry heat sources detection model base on long-term data. （1） 81 heavy industry heat sources worked between 2012 and 2020 were detected based on 89 249 ACF data. They were mainly distributed in the hilly area （the west of Beijing Guangzhou line and G4 Expressway）， especially in Shexian， Wu'An and Fengfeng mining areas. It was consistent with the distribution of mineral resources in Handan city. And， compared the num of working heavy industrial heat sources between 2020 and 2013， 32 （42.67%） industrial heat sources were shut down， 19 （25.33%） industrial heat sources which thermal abnormal points were significantly reduced， also 5 new enterprises appeared. And heavy industrial heat sources detected were maily belong to iron and steel and foundry， coal chemical and coking， cement plants， accounting for 53.24%， 28.16% and 0.08% respectively. （2） Betwwen 2012 and 2020， the average annual proportion of thermal anomaly points from industrial heat sources was 91.61%， especially more than 94.34% in 2016. There was 42.73% of ACF data on the cultivated land surface coverage belonging to industrial thermal anomaly points， 5.5% data on the artificial surface coverage belonging to non industrial thermal anomaly points. And， the most of industrial heat anomaly points in Handan were on the artificial land （accounting for 89.87%）， while the most of non industrial heat anomaly points were on the cultivated land （accounting for75.93%）.There was obvious "double peak" phenomenon （with a larger peak point in June / October） for the monthly statistical chart， which has disappeared since 2018.
Scientific publications between 2010 and 2019 with remote sensing backgrounds that are indexed by the Science Citation Index and Social Science Citation Index are retrieved from the Web of Science core collection as data sources. With techniques including statistical analysis， co-occurrence matrix and spatial centroid models， the spatial-temporal dynamics， subject distribution and topic hot spots of global remote sensing publications are analyzed. The results show that the authors of global remote sensing research are concentrated in Europe， North America and eastern Asia. During the last decade， the gravity center of both the output and influence of remote sensing publications has a prominent eastward shift. However， the gravity center of publication output show a significantly larger shift distance than the gravity center of publication influence. The top five productive countries including China， the United States， Germany， Italy and the United Kingdom show clear differences in the main interdisciplinary studies. The United States has balanced performance in all the 13 main interdisciplinary categories. China， however， has relatively low output in many interdisciplinary subjects such as astronomy and astrophysics， as well as ecology. There are also differences in the thematic hot spots for the five productive countries. Chinese scholars are concerned about global change and the Qinghai-Tibet Plateau， while American scholars have comprehensively explored the Mars and the Moon with remote sensing technology. In recent years， research on climate change， urbanization and change detection has attracted broad attention. Research on interdisciplinary application can be carried out comprehensively with multi-source remote sensing data. Combining remote sensing big data with artificial intelligence algorithms to promote the construction of a smart earth.
The water surface level is essential for the assessment of fresh water resources， disaster prevention， and highly related to the understanding and response to climate change and its impact on water cycle. With the development of remote sensing technology， observation of water level based on satellite platforms provides an alternative way of river water level monitoring featured by automated， long-time， and low-cost river monitoring solution. The principle， characteristics and accuracy of satellite-based river observations are the basis for applications. In this paper， the characteristics and accuracy of three major satellite river water level datasets， Hydroweb， DAHITI， GRRATS are summarized verified with in-situ water level measurements from gauge stations in China. Taking water level time series derived from the Jason mission， we evaluated the accuracy of different water level retrieval algorithms employed by the three datasets. The global accuracy of the Hydroweb dataset （average RMSE 0.70 m） is higher than the other two sources （average RMSE 1.29 m and 3.21 m for the DAHITI and GRRATS）， and that is owing to the usage of a large number of Sentinel-3 observations which are characterized by smaller footprints and Synthetic Aperture Radar （SAR） and the on-board tracking system in open-loop. The accuracy of river water level derived from the Sentinel-3 mission （with average RMSE of 0.51 m） is significantly higher than that of ENVISAT （with average RMSE 3.34 m） and Jason （with average RMSE 1.69 m for Hydroweb and 2.96 m for GRRATS）.Generally， the three datasets can capture reliable river water level changes at some stations （with RMSE < 1.2 m and R2 > 0.8）， but their performances vary considerably among different stations （with RMSE > 2 m for majority of the evaluated stations）. Among all the stations， the Gaocun virtual station from the DAHITI dataset shows the highest accuracy （RMSE 0.22 m）. In addition， the variation of river water level in dry and wet seasons and the small lakes， ponds and seasonal water around rivers pose significant influences on the accuracy of retrieved water level. This study provides guidance for future applications of relevant data sets， and also highlights the challenges of accurate water level retrieval over land surface conditions in China as well as the necessity of algorithm improvement in the future.
In recent years， environmental pollution problems caused by straw burning and industrial emissions have become more serious. The use of satellite thermal abnormal products to analyze the temporal and spatial distribution of thermal abnormalities plays an important role in environmental monitoring. Based on MODIS standard products from 2001 to 2018， the temporal and spatial distribution characteristics of thermal anomalies in China and seven major regions are studied. The results showed that： in terms of spatial distribution， thermal anomalies are mainly distributed in most areas except Northwest and East China. In terms of inter-annual trends， the number of thermal anomalies continued to increase from 2001 to 2014 years， with an average annual growth rate of 15.01%， 2015 years After that， it decreased year by year， with an average annual decline rate of 14.96%. On month and season scales， thermal anomalies occur most frequently in spring and autumn （spring： 551 716， autumn： 416 698）， Spring and autumn are relatively most distributed in Northeast China （spring： 164 898， autumn： 186 727）. The highest in October （118 274）； the lowest number of hot anomalies in summer （290 793）， mostly distributed in East China （120 455）， the average monthly number in East China is the highest in June （76 465）； the number in winter is 358 483， South China has the most distribution （108 209）， and South China has the highest monthly average number in January （37 770）. This research is helpful to master forest and grassland fires in typical regions of China， as well as changes in thermal abnormalities caused by straw burning and industrial emissions， and then provide technical support for regional disaster prevention and environmental monitoring.
Land surface evapotranspiration has an important impact on the simulation of earth’s hydrological processes and hydrological balance， especially in the arid and semi-arid regions. Where the heterogeneous land surface conditions pose new challenges to the simulation of evapotranspiration using the remote sensing models. In this paper， the satellite data collected from Landsat and MODIS were used as driving data of TSEB model， respectively， to obtain the temporal and spatial distribution pattern of land surface evapotranspiration in the downstream of the Heihe watershed. Then the surface heat fluxes measured by the large aperture scintillator with a source area of several kilometers and the eddy covariance with a source area of several hundred meters were used to evaluate the model outputs at the multiple spatial scales. The results showed that the TSEB estimated surface sensible heat flux using Landsat and MODIS data is comparable with the observations from the large aperture scintillator， with Root Mean Square Error （RMSE）values of 48.47 W/m2 and 58.57W/m2， respectively. While comparing with the observation of eddy covariance， the Landsat data-driven TSEB estimates produced RMSE value of 89.37W/m2. Therefore， it can conclude that the TSEB model has a better agreement with ground measurements when using the finer satellite remote sensing data as model inputs. In addition， large aperture scintillator observation with a source area of several kilometers can partially solve the spatial mismatch issues between the remote sensing products and the ground measurements.
The Heihe River Basin is the second largest inland river basin in China. The Heihe River basin has been studied as a representative basin in arid area. In order to explore the characteristics of carbon fluxes of wetland， cropland and grassland ecosystems in Heihe River Basin， the Net Ecosystem Productivity （NEP）， Ecosystem Respiration （Reco）， Gross Primary Productivity （GPP） and meteorological factors were observed and analyzed. To further understand the carbon source/sink effect and its climate regulation mechanism， in this study， the correlation between carbon fluxes （NEP， GPP and Reco） and driving factors were calculated at different time scales. The result indicated that： （1） The temporal variation of NEP in wetland， grassland and cropland ecosystems in the Heihe River Basin was a single peak "inverted U" on daily scale， and the carbon fluxes of grassland reached the peak at 12：00 am， the carbon fluxes of wetland and cropland reached the peak at 13：00 pm； （2） On seasonal scale， the temporal variation of carbon fluxes of wetland， cropland and grassland ecosystems showed patterns of single peaks. During the growing season which from June to September， carbon absorption reached peak in July and the peak value of carbon absorption was farmland>wetland>grassland. （3） The average annual NEP for Shidi site （reed）， Daman site （cropland）， Arou site （grassland） and Dashalong site （grassland） are 627.51 gC/m2/a， 648.90 gC/m2/a， 228.15 gC/m2/a and 307.89 gC/m2/a， respectively. （4） The results showed that NEP， Reco and GPP of wetland ecosystem were significantly correlated with LE， Tair， VPD and Rg， NEP， Reco and GPP of cropland and grassland ecosystem were greatly affected by LE， Tair and Tsoil. However， there was no significant correlation between NEP， GPP， Reco and environmental factors on annual scale. The annual carbon fluxes of Arou （grassland） and Daman （cropland） were positive correlated with Tair and Ms（R>0.4）， and negative correlated with LE and Rain （R<-0.4） annual carbon fluxes of Dashalong （grassland） was negative correlated with Tair， the annual carbon fluxes of Shidi （reed） was mainly negative controlled by LE. In addition， the inter-annual carbon fluxes variation at Arou and Shidi were significantly correlated with NDVI and EVI（0.6<R<0.8）.
As an important part of the East Asia–Australia Flyway （EAAF）， wetlands along the Yellow Sea play important roles in global biodiversity conservation. Here we examined the wetland landscape dynamics and their driving forces in the western Korean Peninsula by Landsat series images for supporting the coastal ecosystem conservation and management and regional sustainable development. Using a method combining object-oriented and decision-tree classification to obtain the multi-temporal wetland datasets from 1980 to 2018 the wetland landscape pattern， the difference in wetland changes and driving forces were compared between North Korea and South Korea. The results revealed that natural wetlands dominate the coastal wetland landscapes in the western Korean Peninsula with 41.1% of the total area of study area. Tidal flat is the main natural wetland type. During the recent four decades， the natural wetland in the western Korean Peninsula experienced a consistent areal decline with the total wetland loss of 1 094.4 km2， while the human-made wetland got a rapid increase in area with a percentage of 45.1%. Due to the differences in national system， politics， population， and economy， most of the natural wetlands in North Korea were converted into cultivated land， while the natural wetlands in South Korea were mainly converted into artificial surface. Human activities are the driving forces of wetland changes in this zone and thus require the improvement in controlling anthropogenic threats and sustainable usage of coastal resources.
The Latent Heat Flux（LHF） is an essential indicator for measuring energy and water vapor exchange between the air and sea. Satellite-based surface turbulent fluxes are widely used due to their wide coverage and high timeliness advantages. However， there are still problems with non-synchronous observations and low accuracy of latent heat flux estimation. Since the sea surface air specific humidity is the primary error source in satellite remote sensing of latent heat flux， the air specific humidity retrieval algorithm is improved based on the Fengyun-3 Micro-Wave Radiation Imager（MWRI） data. Compared with the in-situ measurements from moored buoys， the inversion results have been significantly improved. In view of satellite’s relatively fixed overpassing time of satellites， the intraday variation process of latent heat flux is analyzed using the in situ data. Then a daily average latent heat flux estimation model is established. The Fengyun-3/MWRI data are used to calculate the global air-sea latent heat flux by the COARE3.6 algorithm. The bias， Root Mean Square Difference（RMSD）， and correlation coefficient（R2） between satellite and buoy are 3.50 W/m2， 32.96 W/m2， and 0.79.
Sun-Induced Chlorophyll Fluorescence （SIF） is an effective probe characterizing vegetation photosynthesis. Various global SIF satellite remote sensing products have been derived and widely used. However， the performance and consistency of different products in China are still unclear. CSIF， GOSIF， SIFoco2-005， SIFLUE （SIFLUE_JJ/SIFLUE_PK）， and SIF005 were validated using the ground observations of four farmland sites from ChinaSpec， and their consistency was analyzed. Results showed that： （1） The performance of different products is different in different places and overestimation occurred in three farm stations. SIFLUE performs well as a whole， followed by GOSIF， SIF005 and SIFoco2_005. （2） The spatial pattern of the annual mean SIF and annual maximum SIF in 2016 based on different products is highly consistent， but with different magnitude. The national averaged value of annual mean SIF for SIFLUE_PK， SIF005， SIFoco2_005， SIFLUE_JJ， GOSIF andCSIF is 0.21， 0.17， 0.15， 0.13， 0.11 and 0.08 W m-2 μm-1 sr-1， respectively. The national averaged value of annual maximum SIF for SIFLUE_PK， SIF005， SIFoco2_005， GOSIF， CSIF and SIFLUE_JJ is 0.48， 0.44， 0.36， 0.33， 0.31， and 0.30 W m-2 μm-1 sr-1， respectively. （3） The annual mean value of GOSIF/CSIF and SIFLUE_PK/SIF005 significantly increased and decreased in 23.6%/18.6% and 16.3%/14.7% areas respectively from 2008 to 2016. The national averaged trends of GOSIF， CSIF， SIFLUE_PK， and SIF005 is 0.001 6， 0.001 2， -0.003 6， and -0.001 4 W m-2 μm-1 sr-1 a-1， respectively. SIFLUE_JJ showed no significant change. （4） The annual maximum value of CSIF/GOSIF and SIFLUE_PK/SIF005 significantly increased and decreased in 10.1%/9.9% and 16.9%/22.3% areas respectively during the period of 2008 to 2016. The national averaged trends of GOSIF， CSIF， SIFLUE_PK， and SIF005 is 0.002 9， 0.002 2， -0.008 0， and -0.008 1 W m-2 μm-1 sr-1 a-1， respectively. SIFLUE_JJ showed no significant change.
Terrestrial Evapotranspiration （ET）， defined as the sum of water lost to atmosphere from soil through evaporation and plant transpiration， is a primary process driving the energy and water exchange among the atmosphere， hydrosphere and biosphere. Facing a significant warm-wet change in the Three-River Headwater Region （TRHR）， accurate ET information is of great importance for a wide range of applications including water resources management， hydrometeorological predictions and ecological protection. However， due to the complex topography and sparse distribution of ground-based meteorological observations， the accurate estimation of ET over the TRHR is always not easy. The traditional surface temperature-vegetation index triangular/trapezoidal characteristic space was transformed from regional to pixel scale based on land surface energy balance principle， so daily ET over the TRHR from 2011 to 2019 could be retrieved continuously from a series of MODIS （Moderate-resolution Imaging Spectroradiometer） products. Then we analyzed the temporal-spatial distribution characteristics of ET and its influencing factors over the study region with special focus on ET difference over a variety of land cover types. Comparison between our estimation with other remote sensing-based ET products shows that the accuracy of our algorithms has reached a comparable level， which lays a good basis for further analysis. Results show that ET in recent nine years over the whole TRHR decreased first and then increased with annual average value of 420.04 mm. Controlled by altitude and precipitation， the distribution of ET varied significantly in space with the high values in the southeast and low values in the northwest. ET with the elevation between 3 194 m and 4 620 m increased first and then deceased with altitude. The Pearson correlation coefficient （r） between annual precipitation and ET at site scale was 0.71. The ET statistics of natural ecosystems varied with different land use/cover maps， but all statistics show clearly that ET per unit area followed the order： forest land > shrubland > grassland > bare land. The r between vegetation coverage and annual ET was as high as 0.77 at pixel scale.
Under the framework of the "Belt and Road" initiative， the China-Myanmar Economic Corridor has gradually moved from planning to substantive construction. Understanding the spatial pattern and distribution characteristics of land cover in Myanmar is of great strategic significance for the rational exploitation and utilization of resources and the planning of economic corridor construction. In this paper， a 30m resolution land cover product of Myanmar in 2015 （hereinafter referred to as the MyanmarLC-2015） was produced using Landsat8 OLI remote sensing images， based on the Object-oriented Iterative Classification method based on Multiple Classifiers Ensemble （OIC-MCE）. Besides， the accuracy validation of the MyanmarLC-2015 was conducted by using samples obtained from high-resolution Google Earth imagery. The verification results show that the overall classification accuracy of MyanmarLC-2015 product is 89.05%， the Kappa coefficient is 0.87， which can accurately reflect the spatial distribution characteristics of land cover in Myanmar. According to statistics， forest is the major land cover class in Myanmar， accounting for 56.15% of the total land area of Myanmar. The cultivated land area followed， accounting for 27.01% of the total land area. Combined with topographic factors， we know that with the increase of altitude， the appearance pattern of the typical land cover classes is tree wetlands， paddy fields， dry lands， deciduous shrublands， deciduous broadleaf forests， evergreen shrublands， evergreen broadleaf forests， and evergreen needleleaf forests. From the perspective of vegetation productivity， the NPP of vegetation is the largest in the eastern， northeastern and southeastern parts of Myanmar， while it is lower of vegetation in the central arid region and the southern Irrawaddy Delta. In 2015， the average of Net Primary Productivity for vegetation in Myanmar was higher in evergreen forest than deciduous forest， broad-leaved forest higher than shrub forest， and the productivity of dry land in cultivated land higher than paddy field.
The leaf and wood separation of the terrestrial laser scanning tree point cloud data is an important prerequisite for the accurate estimation of above-ground biomass and leaf area index， and it is also an important step for three-dimensional modeling of a tree. However， complex trees in mountain areas have large crowns and complex structures， resulting in mutual occlusion between leaves and branches. Therefore， it is difficult to obtain high-quality point cloud data. At present， it is still difficult for complex trees to separate leaf and wood components. High-resolution point clouds were acquired with Faro Focus3D X330. This paper proposes a method for leaf and wood separation of tree point cloud based on network graph. First， the LeWos model is used to perform preliminary leaf and wood separation on the point cloud， and separate the wood and leaf point cloud. On this basis， the path retrace detection algorithm is used to finely separate the leaf and wood for the mixed point clouds. As the retrace steps increases from 10 to 100， the wood points continue to increase， the leaf points continue to decrease， the accuracy， wood F-score decreases， leaf F-score and the Kappa coefficient first increases and then decreases. By comparing with LeWos model， Tlseparation model and Gaussian mixture model， it is found that the research method in this paper has the better precision， with an accuracy of 91.97%. Moreover， the wood F-score and leaf F-score of the proposal method are both greater than 85%， which means that the proposal method has a good balance when classifying wood and leaf. The proposal method only uses the retrace steps and does not consider the geometric characteristics， so the accuracy of branch and leaf separation of coniferous trees is greatly improved. At the same time， the method in this paper has a good effect on the separation of wood and leaf components of point clouds with different densities and different species. Therefore， the method in this paper is more robust. Accurate wood and leaf separation of tree point clouds is of great significance to forest resource management and biodiversity research.
Flood disaster has great harm and poses a great threat to regional people's lives and property and economic development. Therefore， the continuous high-time resolution remote sensing monitoring will be conducive to more objective and accurate detection of the temporal and spatial variation characteristics of flood risk areas. This research takes Chaohu Lake Basin as the experimental area， based on Google Earth Engine platform， collects Sentinel-1 Synthetic Aperture Radar （SAR） images， uses the flood inundation identification method combining spectral relationship and threshold segmentation to map the flood range of Chaohu Lake Basin from 2015 to 2020， and combines land use data， The impact of flood on farmland and residential areas represented by construction land in Chaohu Lake Basin is analyzed. The results show that： （1） the accuracy of this method is 3%~7% higher than that of single band threshold method and simple index method， and can quickly extract the flood inundation range of watershed over the years by using remote sensing data； （2） From 2015 to 2020， two major floods and one small-scale flood were monitored， and the inundation scope was concentrated in Hangbu River， Yuxi River， Zhaohe River and other river areas； （3） Farmland accounts for 86.47%~95.35% of the flooded area， and construction land accounts for 4.47%~5.36%. The residential areas affected by the flood are mainly grass-roots villages and towns. The research shows that the application of SAR satellite data in flood monitoring can effectively monitor the impact range of flood on farmland and rural residential areas， which is very key to formulate relevant planning strategies in the future， strengthen rural flood control in the basin and ensure personnel and food security.
River runoff is one of the most important hydrological elements on land. Accurate access to runoff information plays an important role in regional water resources evaluation and ecological restoration. Based on the Sentinel-1 data and Sentinel-2 data provided by the Google Earth Engine cloud platform， combined with digital elevation model， the hydraulic parameters such as river length， river width， roughness， slope， river depth and velocity were estimated by remote sensing. Then， the relationship fitting method and improved Manning formula method were used to inverse the runoff of the reach near Tangnaihai station in the source area of the Yellow River. The influence of the length difference of the reach on the runoff inversion accuracy is discussed. By establishing the river width relationship between the station reach and the upper and lower reaches， the runoff monitoring time series of the station reach can be extended and supplemented. The results show that the two models can effectively simulate and estimate the runoff， and the Nash efficiency coefficient is above 0.80； the root mean square error of the relationship fitting method and the improved Manning formula method are 233.431 m3s-1 and 271.704 m3s-1 respectively， and the relative root mean square error are 16% and 24% respectively. The inversion accuracy of relationship fitting method is better than that of the improved Manning formula method. Through the comparative analysis of runoff inversion results of different lengths of river reaches， it is found that the river width estimation of braided river core beach has great uncertainty in flood season， which affects the accuracy of runoff inversion， and should be avoided in the selection of river reach； there is a strong correlation between the average river width of the station reach and the upstream and downstream reaches， and the correlation coefficient is above 0.96. The data can provide an important supplement for the runoff inversion of the station reach and realize the intensive monitoring of the runoff of the target river section.
Accuracy assessment is important for the application of land cover products. In practice， there are many differences for sample size and spatial distribution of the reference data used for accuracy assessment， and the impact of this difference on accuracy assessment results is not very clear. This paper validated the ESA CCI-LC using GlobeLand30 as the reference data in China. We test the impact of sample size， sampling model， and sample unit on the overall accuracy and various types of accuracy. The results are as following. Firstly， the sample size makes a difference to the accuracy assessment results. The overall accuracy based on point sample is close to that of the theoretical sample size while the sample size in half （about 600 sample points）， however， the sensitive degree of different classes of precision on the sample size is different， classes with low area weight are more sensitive to sample size， especially stratified random sampling used in the case of small sample size. Secondly， based on the theoretical sample size， the sampling model has little effect on the overall accuracy whether it is the accuracy assessment of point sample or cluster sample， but it has an impact on the accuracy of each class， especially the class with a small area weight. Finally， the stability of accuracy assessment of simple random sampling based on cluster sample is worse than the accuracy assessment based on point sample. The increase of the sample unit will increase the uncertainty of the accuracy assessment result on the basis of the theoretical sample size， with the increase of the sample unit， although the volatility of the accuracy assessment result becomes smaller， the result is still large compared to the point sample. Therefore， in order to reduce the uncertainty of the accuracy assessment results of land cover remote sensing products， the sample size can be increased for point sample， and the number of sample units can be increased or the size of sample units can be appropriately reduced for cluster sample.
Moderate Resolution Imaging Spectroradiometer （MODIS） MCD12Q2 dataset and MOD17A3HGF NPP （Net Primary productivity） data were used as data sources to study the spatial distribution and change trend of annual mean of phenological period of grassland and woodland in Hebei Province. The phenological period included Start Of growing Season （SOS）， End Of growing Season （EOS） and Length Of growing Season （LOS）. The correlation between phenological period of grassland and woodland and NPP was studied by correlation analysis method， and the significance of change trend and correlation coefficient at α=0.05 level was judged by significance test. The results showed that SOS of grassland in Hebei Province mostly appeared on 108~153 DOY （day of year）， EOS mostly appeared on 273~304DOY， most of LOS was 128~190 days. while SOS of woodland mostly appeared on 107~128 DOY， EOS mostly appeared on 282~306 DOY， most of LOS was 162~194 days. SOS of grassland and woodland in Hebei Province showed early trend， EOS showed delayed trend， and LOS showed growing trend； the phenological period of grassland and woodland was mainly moderately correlated to NPP， and SOS was mainly negatively correlated to NPP， while LOS and EOS were mainly positively correlated to NPP.
Using Landsat remote sensing images from 1980， 1990， 2000， 2010 and 2020 as data sources， combined with field measurements and Google Earth's high-resolution images， we obtained the land cover changes of Liaohekou National Nature Reserve from 1980 to 2020 by using the object-oriented decision tree classification method. The spatial and temporal dynamics of the artificial types of land in the Liaohekou National Nature Reserve in the past 40 years were studied by combining the land use transfer matrix， landscape pattern analysis and equal sector orientation analysis. The results show that the natural wetlands in the study area decreased by 270.12 km2 between 1980 and 2020， and were mainly transformed into artificial land cover types such as arable land， oil wells， construction land and transportation land. Due to the large disturbance by human activities， the landscape in the study area tends to be fragmented and balanced， and the landscape heterogeneity is reduced. In the past 40 years， the artificial land cover types in the study area have expanded mainly along the north-northwest direction. National policies and economic factors have greatly influenced the evolution of wetlands in the Liaohe estuary， and human activities such as farmland reclamation， urban construction， oil field development and mariculture are the main driving forces for the evolution of natural wetlands.
As the most widely distributed vegetation type on earth， grassland plays an important role in the terrestrial carbon cycle. Grassland productivity is the basis for estimating grassland yield. Grasping the temporal and spatial variation of grassland productivity is of great significance for rational utilization of grassland resources and protection of grassland ecological environment. This thesis taking the productivity of grassland in northeastern China as core， constructing and validating light use efficiency model based on eddy covariance flux data， remote sensing， and climate data， explored the spatiotemporal patterns on this basis. The research results are as follows： in the northeastern China steppe LUE model， FPAR was represented by NDPI， water stress factor was represented by LSWI + 0.5. Based on the flux data of four grassland stations， the R2 of the northeastern China steppe LUE model was 0.855， which was higher than that of MODIS GPP （R2 = 0.719）， and slightly higher than VPM GPP （R2 = 0.848）. MAE and RMSE of the northeastern China steppe LUE model were 0.374 gCm-2 and 0.735 gCm-2，respectively，which were lower than that of MODIS GPP（MAE=0.562 gCm-2， RMSE = 1.026 gCm-2） and VPM GPP products （MAE = 0.667 gCm-2， RMSE = 1.339 gCm-2）. VPM GPP product generally overestimated the flux GPP； MODIS GPP product significantly overestimated typical steppe GPP in dry years， and significantly underestimated meadow steppe GPP. Although the northeastern China steppe LUE model was higher than the typical steppe flux GPP in the dry years， its overestimation degree is less than that of MODIS GPP and VPM GPP products. The northeastern China steppe LUE model is superior to MODIS GPP and VPM GPP products in terms of model accuracy and dynamic consistency， and the fitting accuracy of the annual scale is much higher than MODIS GPP and VPM GPP. The modified of water stress and FPAR was the reason for the improvement of LUE model accuracy， and the relative contribution of water stress is greater. This study demonstrates that it is necessary to use the improved light energy utilization model to simulate grassland productivity in northeastern China.
Solar-Induced Chlorophyll Fluorescence （SIF） is the spectral signal （650~800 nm） emitted by plants in the process of photo-synthesis under sunlight conditions. SIF is more direct than vegetation index and other parameters. Reflecting the relevant infor-mation of vegetation photosynthesis， it brings a new way for large-scale Gross Primary Productivity（GPP）estimation. However， the current satellite SIF data may have insufficient resolution or discontinuity in the data space， which is difficult to apply to the estimation of continuous GPP on a large scale. OCO-2 SIF data has high spatial resolution， but it is spatially discrete data. In response to the above problems， this paper focuses on the method of con-tinuous prediction of discrete OCO-2 SIF data to generate a high-precision continuous SIF data set of the China-Mongolia grassland ecosy-stem. The results are as follows： Through the Cubist regression tree algorithm， combined with MODIS reflectance data， meteorologi-cal data and land use types， a continuous SIF data set with a resolution of 0.05° every 8 days is established， and the prediction accuracy is R2= 0.65 and RMSE = 0.114. Among them， the accuracy of crop SIF prediction is the highest， with R2= 0.71 and RMSE= 0.117； the second is the prediction of forest and grassland， with R2 and RMSE of 0.64/0.123 and 0.60/0.112 respectively.
Leaf area index （LAI） is an important parameter to characterize the vegetation condition， which is closely related to the growth and change of vegetation. The investigation of the spatiotemporal pattern of LAI in Inner Mongolia grassland over a long time series and the influence of water and heat conditions on LAI can provide data to support an accurate understanding of the differences in the distribution and growth conditions of Inner Mongolia grassland， meanwhile， it is helpful for understanding the spatial distribution characteristics of the production capacity of Inner Mongolia grassland.Based on the GEOV2 LAI product dataset from 2000 to 2019， three indicators， namely slope of variation， coefficient of variation and correlation coefficient， were selected to analyse the grassland LAI in Inner Mongolia in combination with the data of temperature and precipitation. The results show that the LAI of Inner Mongolia grassland decreases from northeast to southwest with a multi-year mean value of 1.34 m2/m2， and among different grassland types， desert grassland （0.28 m2/m2）<typical grassland（0.96 m2/m2）<meadow grassland（2.27 m2/m2）<meadow（2.60 m2/m2），and is inversely proportional to the coefficient of variation，with desert grassland showing the sharpest inter-annual fluctuations. Over the past 20 years，the LAI of Inner Mongolia grassland showed an increasing trend in fluctuation（0.02 m2/m2/a），67.08% of regional grassland LAI was significantly correlated with annual precipitation，and only 4.98% of regional gras-sland LAI was significantly correlated with annual mean temperature. These indicate that the spatial distribution of grassland LAI in Inner Mongolia has zonal characteristics， and there are significant differences between different grassland types， and precipitation is the main influencing factor of grassland LAI in Inner Mongolia.
The livestock’s distribution across space is essential to the research on food safety， agricultural society economy， environmental influence assessment and zoonosis. In this study， an approximation model of livestock's distribution across space was constructed on the basis of Random Forest （RF） regression algorithm to combine remote sensing data and statistical data. In order to test and validate the proposed method， statistics for sheep in 87 counties of Gansu Province was collected in 2010 and 11 environmental factors were considered in this scheme. Finally， the spatial distribution information of sheep on the scale of 1 km×1 km in Gansu Province is obtained by the model. As is indicated by the results， the grid model of livestock’s spatial distribution based on the RF regression has included the advantages of both remote sensing data and statistical data. It is able to estimate the spatial distribution situation of sheep on the scale of 1 km×1 km with certain accuracy. The correlation coefficient （R） between estimated results and statistical data reached 0.88， the Root Mean Square Error （RMSE） was 0.24， and the Relative Root Mean Square Error （RRMSE） was 15.1%. Sheep in Gansu Province are mainly distributed in the Gobi area of the Hexi Corridor， the grassland and meadow area of the Gannan Plateau， the southwestern part of the hilly area of the Loess Plateau， and the northern part of the gully area of the Loess Plateau. The environmental factors that have a greater impact on the spatial distribution of sheep are： percentage of cultivated land， altitude， surface temperature， and slope.
Grassland biomass is an important parameter to evaluate the grassland ecosystem function. To estimate the grassland aboveground biomass rapidly， accurately and effectively， six vegetation indices （GNDVI， LCI， NDRE， NDVI， OSAVI and EVI） were selected and calculated based on UAV multi-spectral images and satellite remote sensing （Sentinel-2） images， combined with the ground measured biomass data. The vegetation index regression model was established， and the precision was verified by the left one method. The results showed that the accuracy of LCI-biomass regression model （RRMSE = 18%， the measured and predicted R2 = 0.70） and NDRE-biomass model （RRMSE = 18%， the measured and predicted R2 = 0.71） based on UAV multi-spectral images was higher than that of other vegetation -biomass models. The biomass-vegetation index models based on UAV multi-spectral images （RRMSE lower than 22%） have better simulation accuracy than Sentinel-2 biomass-vegetation index models （RRMSE higher than 25%）， which can more accurately retrieve the aboveground biomass of Hulunbuir grassland. The results can provide scientific methods and basis for accurate retrieval of grassland biomass.
