WATER experiment is a simultaneous airborne,satellite-borne and ground-based remote sensing experiment taken place in the Heihe river Basin,a typical inland river basin in northwest China.Many teams contributed to the WATER through their participation.In total,28 institutions and more than 280 scientists,students,engineers and aircrews participated in the field campaigns of WATER.This paper introduces the research progresses since the intensive observation period of WATER,which was completed in 2008.The most important outcome of the WATER is its multi-scale,high-quality,and integrated observational data sets,which was officially released and opened to public from July 2010.It offers supports for the development,improvement and validation of a series of ecological,hydrological,and quantitative remote sensing models.The designed objective to break through the data bottleneck in integrated watershed study has been accomplished.In addition,the WATER has achieved abundant research results in ① retrieving snow parameters,surface soil freeze/thaw state,forest structure parameters,evapotranspiration,soil water content and biogeophysical parameters from remote sensing;② hydrometeorology observation;③ and development of scaling methods,hydrological models and data assimilation system.The highlights of WATER include improvement of remote sensing models on estimating evapotranspiration,application of airborne LiDAR,and development and application of a multi-angle thermal infrared remote sensing sensor.It is also illustrated that the airborne remote sensing is still playing an irreplaceable role in obtaining high-quality and high-resolution data.
This paper reviewed and summarized the progresses on the remote sensing-based inversion and estimation of hydrological and ecological variables/parameters,within the framework of the Watershed Allied Telemetry Experimental Research (WATER) project.We make progresses in remote sensing of hydrological variables as follows:The basin-scale precipitation observation with high accuracy are carried out with a truck-mounted dual polarized Doppler radar in the upstream and midstream of the Heihe River basin,aiming to obtain the quantitative relationship between the precipitation rate,radar reflectivity and its polarization information.The substantial developments and improvements of remote sensing estimation models of evapotranspiration are achieved with the aid of multi-source observations.A retrieval algorithm of snow depth in the mountainous area is developed by using the K and Ka band airborne microwave radiometry.The method to eliminate the influence of surface roughness on soil moisture remote sensing is proposed by using the multi-angles SAR data.We also succeed in the remote sensing estimation of ecological-process variables/parameters as follows:Fine land surface classification method is developed by combining information from airborne laser radar and high-resolution optical images.The C3/C4 vegetation functional type classification is realized by integrating the vegetation fluorescence under solar light condition extracted from the hyper-spectral airborne remote sensing images,and NDVI information.The methods using multi-angles and multi- spectrums remote sensing information to retrieve the LAI are improved,especially exploiting the potential of LiDAR to obtain the vegetation vertical structure,and the scale conversion of remote sensing based LAI is also explored.Other progresses include developing a new method to retrieve light-use efficiency using fluorescence information from hyperspectral data,proposing an inversion model of FPAR taking the soil reflectance and canopy structure into considerations,improving the remote sensing estimation model of ecosystem productivity and developing a method to obtain chlorophyll content and chlorophyll fluorescence intensity by using hyper-spectral remote sensing data.
The global time series satellite remote sensing product has caught worldwide attention since it is developed.It is widely used in the global, intercontinental and regional dynamic monitoring of the surface features.It is also used to analyze the global changes by integrating with some climate change parameters (e.g.air temperature,precipitation).The contents and application areas of the time series remote sensing product extends greatly as the time series increases gradually and new sensors are constantly emerging.The development status of the currently international popular satellite remote sensing products (visible,near infrared,short wave infrared and thermal infrared bands) are introduced in this paper.The satellite sensors mainly include AVHRR,VEGETATION,and MODIS.The early researches mainly focused on the basic data preparation,which concentrates on the band information of visible-near infrared band reflectance and thermal infrared bright temperature,vegetation indexes.The thematic products,such as leaf area index and land surface temperature,are retrieved and estimated by these basic data products at present.The research progress and development trends of the further data processing,analysis,and application of the data product are introduced in detail.The data processing includes time series reconstruction,comparison and conversion,validation.The data application includes dynamic monitoring of the surface features, information extraction of phenology and planting structure,modeling application.
The SSS(Sea Surface Salinity) is the important parameter of study ocean circulation and global climate and factors which determine the essential properties of seawater.Satellite microwave remote sensing satisfies the salinity research needs of extensiveness and continuous observations.International chose L-band,with a central frequency of 1.413 GHz is the band of the first choice for salinity remote sensing.At the moment,there are two main inversion algorithms of SSS remote sensing by microwave oversea:Algorithm of estimate SSS with sea surface emissivity and inversion algorithm based on Bayesian.Main factors affecting the accuracy of the salinity inversion are Space radiation,Faraday rotation in ionized layer,atmospheric and sea surface roughness,and the surface roughness have a great impact on the salinity inversion.The surface roughness model can be divided into three categories:theoretical algorithm,empirical algorithm,semiempirical algorithm.Because Aquarius/SAC-D satellite and SMOS satellite were successfully launched,salinity can be retrieved with an accuracy of 0.2 psu by the two satellites,the inversion accuracy is expected to be higher by improving the inversion algorithm.
Linear spectral unmixing has been intensively studied for more than ten years,now it is a commonly accepted approach to spectral unmixing,and a number of algorithms have been developed,but the objective evaluation of these algorithms is the foundation to the wide use of them.As the acquisition of reliable ground-truth data is difficult and expensive,the use of real scenarios is limited,the simulated data have been used widely in the evaluation of these algorithms.The approach based on Dirichlet distribution is one of the common used methods for the simulation of hyperspectral data,it is more flexible and the Dirichlet density is suited to model abundance fractions.However,the parameters of the Dirichlet distribution affect the simulated results,and improper parameters will cause the simulated data to lose the property that the hyperspectral data should preserve.This paper analyses the influence of the Dirichlet parameters on simulated results,and discusses the restrictions of the parameters' value,and provides the recommended parameters range (0,1.5).
Compared two sub-pixel mapping methods which are based on the spatial correlation characteristics.One method is based on Hopfield Neural Network(HNN) and the other is based on geometric theory.A study with TM imagery data is employed to evaluate the performance measured by accuracy,visual effects and time consuming on the two methods.The comparisons illustrate that the availability of the spatial correlation incorporated into the two means is perfectely expressed in the results,and five conclusions are provided for the sequential researches of sub-pixel mapping as well.
The texture features of high-resolution spaceborne SAR image is of great significance to image interpretation and classification.As the texture features of high-resolution spaceborne SAR image is different from other objects which appear in SAR images,a novel method for built-up areas extraction using both grey-scale and texture features is proposed in this paper.Firstly,reducing the speckles of the SAR image,then eight texture feature statistics of the built-up areas and non-building areas are calculated by using GLCM.Secondly,the best window size of GLCM is discussed and the texture features are selected according to the Bhattacharyya distance.Then,two principal components of the richest information are selected as the best texture components based on the principal component analysis method,and combine with the original image.Finally,the new image is classified with K-means classification method and the built-up areas are extracted after some post classification process.The proposed method has been tested by COSMO-SkyMed SAR image.The results show that this method can effectively extract built-up areas in the high-resolution spaceborne SAR image and is better than the method without using the texture features.
The improvements of earth observation technology and the growing accumulated multi-dimensional spatial-temporal field data.It is already a research hotspot of the integration and expression of these data.The modeling and simulation of global geo-processes also require new demands on the modeling,retrieval and analysis methods of these data.In this paper,the building framework of spatial-temporal field data model is constructed based on the tensor structure.Data from different dimensions are intergrated as different perspectives of the tensor structure.For implementations,a spatial-temporal cube model is constructed for data organization and storage,and the corresponding data manipulation and data interface are defined.After that,we designed a layered spatial-temporal field data indexing mechanism and the spatial-temporal field data analysis methods,such as princple tensor decomposing,based on operators of tensor operation.The framework is validated based on satellite altimetry data,which suggest that our model can effectively support the presentation,retrieval and analysis of the multi-dimensional space-time field data and be a useful exploration on high-dimensional spatial-temporal field modeling and data analysis.
The types of urban ground objects and their spatial distribution are complex.And the ground objects are multi-scale,different types of urban ground objects have different texture scale.The paper uses Principal Component Analysis(PCA) to deal with high\|resolution remote sensing images in order to reduce the quantity of data,suppress the noise,and highlight important information.On this basis,this paper extracts the texture features from the first principal component of PCA on basis of Gray Level Co-occurrence Matrix,and chooses the best combination of multi-scale textures to decision tree classification.The results show that the method of the decision tree classification based on PCA and multi-scale texture can extract the types of ground objects effectively.The precision of classification is 82.4%and Kappa coefficient is 0.78.
Similar shape object recognition is widely used in automatic target recognition system of remote sensing and weapon guidance.A hierarchical method of shape feature extraction and selection is proposed to increase the recognition efficiency and rate.Learning from human visual perception,multi-scale features are extracted.Global features are used to make a quick classification,and local features are used to distinguish targets with similar shape.To achieve the feature selection,fuzzy criterion is introduced which improves the matching processing and increases the recognition rate.Experimental results show this method is an effective and general way in recognizing targets with similar shape,and the feature selection improves the recognition rate by 6.9% than before.
The Range-doppler model is a basic model for SAR image locating without GCP.So obtaining the solution of RD equations accurately and efficiently is an essential question for SAR image location without GCP.It is also the key step of SAR automated image location.In this paper,the RD model has been simplified to quartic equation with one variable based on the characteristic that the objects velocity equals to zero in Earth Centered Rotating.The paper deduces an analytical solution of Range-doppler model by solving the quartic equation.This paper compares the analytical solution with numerical solution by simulation.The results show the analytical solution is feasible and validity.The new solution has three advantages:no setting initial value,easy to programming realization and high robustness.
The forest is the largest carbon sink in terrestrial ecosystems.It has an irreplaceable role in adjusting the global carbon balance and slowing down CO2.Forest biomass is the most important parameters in the terrestrial ecosystem carbon cycle and becomes more and more universal concern by the scientists in forest biomass estimation and changes and the focus of carbon cycle research.Taking the ecologically sensitive areas in Northwest Yunnan Shangri-La County as a study area,in the support of survey data in the wild forest,combined with 3S technology, geography, ecology, meteorology and other related knowledge,the variable of 9 vegetation index,2-band gray data,growth season precipitation,growth season accumulated temperature,growth season total radiation,elevation,slope,aspect,slope position and soil organic matter content were selected,which combine the layer of remote sensing integrated factors and the layer of geographic comprehensive factor with water,light,heat into the variables,and then establish the regional forest biomass estimation model and be tested.The statistic of model R,R2,aR2 and F is 0.809,0.655,0.661 and 101.436.The linear regression equation constant (a) and regression coefficient (b) which established by sample measured data and model is 0.09,1.021.The independence test of estimation model had done by 22 biomass data of field sample;the average estimation accuracy is 76.43%.The result shows that the estimation accuracy of the model is generally stable,and basically meets the accuracy requirements of biomass estimation.It can be used to the study of estimating the forest biomass in this area.
In this paper,firstly planck function is used to calculate the radiation intensity about the two bands(ASTER13、14) from the range of 253~333 K.Secondly the scatter plot between the two is set up.Thirdly,get a index formula based on regression analysis,which is applied to deduce split algorithm on land surface temperature inversion by ASTER data.On the estimates of atmospheric parameters,we use cubic polynomial to fit the relationship between atmospheric transmittance and atmospheric water vapor content in order to improve fit accuracy.Finally,the simulation data method is used to validate accuracy of the algorithm.The results indicate that if the error is under 1 K no matter what ground temperature and atmospheric water vapor content,the higher accuracy the algorithm have,the smaller the inversion error is.And then the sensitivity of parameters for the algorithm is analyzed,which demonstrates the parameters (atmospheric transmittance and atmospheric water vapor content) are not sensitive to the authors algorithm.
The use of land surface albedo products from satellite observations is based on the evaluation of their inversion quality.In this paper,MODIS surface shortwave albedo products are compared with in-situ measurements from ChinaFLUX.Eight sites over forest,grassland and cropland are chosen in this study.Corresponding MODIS data in 2004 are extracted from surface albedo products(MCD43B3).The results show MODIS shortwave albedo products match ground measurements well except some cases with snow cover.The error of all the eight sites is about 0.002,and root-mean-square error (RMSE) can reach 0.028 without snow cover situations.This paper analyses the factors that cause the errors and proposes some suggestions for improvement of inversion quality.
Leaf Area Index(LAI)is one of the most important parameters,which controls biological and physical processes associated with vegetation on the Earth's surface,such as photosynthesis,respiration,transpiration carbon and nutrient cycle,and rainfall interception.However,field measurement of LAI is very difficult for a large area.Five Vegetation Indexes (VIs) of Linzhi in Tibet are estimated from satellite TM image respectively.Then the five VIs are regressed against field-measured LAI using the linear and nonlinear regression models.The regression model with lowest error was found from the five VIs,and has a significant correlation (R2=0.653).Finally,the determined model was used to map LAI of Linzhi in Tibet.This study demonstrated the remotely sensed TM image can be used in large-scaled LAI estimation and can provide data for ecological research.
Monitoring the extent and duration of snow cover in Tianshan mountains is significant for fully utilizing water resources and understanding the regional impact of climate change.Daily MODIS snow cover data can provide a basis for large scale snow cover mapping.However,the high cloud obscuration limits its applications.This study uses a four-step approach to remove cloud from the daily data.Using improved MODIS snow cover data and DEM analyse the spatial and temporal variability of snow cover extent and duration from 2002 to 2009 in the Tianshan Mountains of China.Snow cover extent is closely related to terrain factors.The snow cover frequency as a whole becomes higher with increasing altitude.The ratio of snow cover is higher at the northwest,north and west aspect than that at the south and southeast aspect.The snow cover extent of the whole area is small from 2006 to 2008.Maximum snow cover extent increases yearly on average.The regions snow cover duration exceeds 200 days,which are mainly in the high-altitude perennial snow cover areas,and their Coefficient of Variation (COV) is low.The snow cover duration in the southern Tianshan mountains and other low-altitude areas is very short,and change substantially due to seasonal climate situations.
The AVHRR and MODIS satellites have played a vital role in monitoring vegetation phenology responses to climate change at a global scale.It is important to examine whether the derived phenological parameters from AVHRR are consistent with from MODIS.In this paper,a dynamic threshold method was applied to extract the phenological metrics based on GIMMS AVHRR NDVI and MODIS 13A2 NDVI dataset smoothed with HANTS in 2005,such as Start of Season (SOS),End of Season (EOS),and Duration of Season (DOS).Then,the comparative analysis was performed on the phenological metrics between AVHRR and MODIS.The results showed the SOS appeared between 100th and 140th days,the EOS occurred between 260th and 300th days,the DOS mainly existed from 130th to 180th days in most regions.The phenological variation along latitude based on AVHRR was consistent with based on MODIS,a trend of later SOS,an earlier EOS and a shorter DOS with increase of latitude were observed.The SOS,EOS and DOS from AVHRR and MODIS data were quite consistent with a correlation coefficient larger than 0.9 for the deciduous forests and grasslands in Eurasia and North America.
This paper compares the L3 product of QuikSCAT 10 m Scatterometer Winds with the fishing vessels wind data.The statistical biases between the two sets of data show that:① in general,the fishing vessels wind speed of offshore Chile is higher than QuikSCAT,and the wind direction of fishing vessels is on the left side of the QuikSCAT ;② the wind speed bias between the fishing vessels and QuikSCAT wind data is concentrated on the range of -1~1 m/s,and the wind direction bias is mainly in the range of-60°~-10°,with minor range of10°~60°and-10°°~10;③ the wind speed deviation characteristic value in the day time is better than that at night,and there is distinct difference of wind direction mean bias between day and night ,but the absolute bias and root mean square bias of wind direction are similar;④ and the biases between the fishing vessels' winds and QuikSCAT wind data in 2008 are larger than the average of other years,especially for the wind speed bias of high wind speed section.
For the problem of low classification accuracy of sandy lands based on spectral feature of remote sensing images,a methodology by applying object-oriented method for extracting sandy lands information was studied by Landsat-5 TM image data in this paper.First,the mult-scale image segmentation was conducted to obtain homogeneous areas of objects.Then,based on field survey data,a variety of feature diagrams of different land surface types were made to select the features of target objects and establish the decision tree for classification of sandy and non-sandy lands.Finally,the fuzzy image classification with the decision tree was implemented,accuracy of classification was validated by ground truth data.The result showed that the overall accuracy reached 84.89% and the Kappa coefficient was 0.8077,which indicated that the object-oriented method for extracting sandy lands information could provide a foundation for further study on extraction of sandy land information.
Crop inundated time is one of the most important index for accurate evaluation of crop reduction in flooded area.However,it is difficult to get inundated time directly due to the crop distribution,terrain differences,and other factors.Crop inundated time is consist of the period in rainstorms and after.The current extraction method of crop inundated time after rainstorms applies the ratio between the flood accumulation and the drainage capacity in a watershed to estimate the inundated time.This method application is limited because of the difficulty of getting the basins drainage data,what's more,the drainage capacity is usually replaced by observation points which not only can not represent the watershed characteristics effectively but also can not describe the spatial distribution of crop after rainstorms.Therefore,a new method is proposed based on time-series MODIS data to extract crop inundated time after rainstorms in large scale,which constructs a robust and consistency inundated time index combined with the characteristics of flood receding process ,and then obtains the right time point by the optimal two-part segmentation algorithm.Finally,Huaihe flood in 2003 and Liaohe flood in 2010 are taken as examples to testify this new method.The results showed that the method is reliable.At the same time,this method is simple to implement and intuitive spatially,and applies widely.
Remote sensing image information extraction research is one of the key problems of remote sensing research,it is also one of the hot and difficult points in remote sensing research.With the aid of the GIS spatial analysis and statistical analysis method,the Haloxylon Aammodendron forest Mean NDVI time series characteristic curve were reconstructed in the Gurbantunggut Desert by using the NASA/ MODIS-NDVI16 days synthetic data (from 2000 to 2010) and phenology record.Analysis of phenological and Mean NDVI time series,the result showed that the Ephemeral plants (or short-lived plants) under the Haloxylon Aammodendron forest growth period is earlier than Haloxylon Aammodendron.Research of Mean NDVI time series curve,showed that there is a feature point that obviously different from other features in the curve,which can be used as the “diagnosis point” of Haloxylon Aammodendron.A model of Haloxylon Aammodendron forest features index(HFFI) was developed based on the “diagnosis point” characteristics.Retrieved the information of Haloxylon Aammodendron forest in the Gurbantunggut Desert from HFFI.And utilizing the data of the ground practical observation validation,the results indicate that the classification accuracy reached 83%.
River Main-stream is the key factor of river regime,which is a very important information for flood control decision.It is a good approach to get the main-stream information from remote sensing images in time.According to the field observation on water surface phenomena of river main-stream,water surface characteristics on main-stream and non main-stream were analyzed on river transect,and a river main-stream detection algorithm based on wavelet transform and its multi-scale peak value analysis was proposed.The proposed algorithm was verified by TM images on the lower reaches of Yellow River,and the precision was evaluated through comparing with manual-surveyed line on the same reaches,therefore the results validated that the proposed method can be used to river main-stream detection in the lower Yellow River.
Studing the relationship between changes of land surface parameters and heat island effect which has significance for city planning and city sustainable development.This paper adopts three periods of Landsat ETM+ and remote sensing images of 2000,2005,2009 to retrieve the impervious surface area,vegetation and water distribution in Shanghai.It analyzes three parameters such as NDISI(Normalized Difference Impervious Surface Index),IVI(Index-based Vegetation Index) and MNDWI(Modified Normalized Difference Water Index) to get the retrieval results.Then from the angles of time,space and regression analysis,this paper analyzes the changes of Shanghai s urban surface parameters and urban heat island effect of the 9 a.The study shows that,during the 9 a,the area of impervious surface area increased significantly with the cost of vegetation and water reduction which is also the main cause of urban heat island effect.The urban heat island effect in Shanghai trends to strengthen at first and then slowly weakened and the distribution of urban heat island is changing from centralization to decentralization.
