Mountainous areas provide diverse ecosystem services in many aspects,including biodiversity maintenance,regional climate regulation and water conservation.The retrieval of biophysical variables in mountainous areas faces more challenges in technology and theory compared to that in flat terrain.These challenges derive from the significant surface heterogeneity and the lack of appropriate models for radiative transfer and ecological process.The related research progresses and key theory and techniques in montanic biophysical variables retrieval were reviewed in this paper.Further improvement of retrieval accuracy calls for paradigm shift in modeling,observation and data processing in mountainous areas.
Remote sensing is a key technology for the monitoring of land cover and land use change and the retrieving of bio\|physical parameters of vegetation in mountainous area.However,due to the obscuring by terrain relief,the mountainous shadows in the optical remote sensing images have brought great difficulties for the remote sensing applications in mountain area.In this paper,a new method for the restoration of information obscured by mountain shadows was proposed.This method took advantage of the characteristics of MODIS Aqua and Terra,whose time of passing territory are morning and afternoon separately,to compose the information sources for the shadow area on high resolution images.Then the homogenous pixels in both MODIS and TM were selected based on the angle filter,cloud filter and homogenous filter.Based on the hypothesis that the statistics relationship exists for the homogenous pixels between MODIS and TM,the proposed method further built up a regression tree model for the sun\|light area,and then used the built model to predict information of the shadow area in Landsat TM images.According to the comparison of the restoration result between the proposed method and SCS+C model,the method in this paper can better reflect the detail information of the shadow area and simultaneously preserve the information of sun\|light area well.With the development of new high spatial resolution sensors such as Sentinel\|2A/B and Landsat 8 Operational Land Imager,using multi\|sources data to restore information of mountainous shadow area is a new development tendency,and the proposed method in this paper can be used as a reference for those similar satellite datasets.
Sampling design is a prerequisite for field campaigns,which is especially important for the supporting of bio\|physical production validation.A good sample should get a proper compromise between representation of heterogeneous surface and implementation cost.Rugged terrain in mountainous areas complicates the sampling designing,resulting from its more heterogeneous and inaccessible characteristics.A novel sampling strategy was proposed based on constrained Latin Hypercube Sampling to address this problem.Results of a comparison experiment,implemented in Hailuogou,China,show that this sampling strategy can get a sample which is representative in both feature and geographical space,and with a relatively low implementation cost.
Mountain is the major distribution areas of forest.However,the accuracy of forest types mapping in this region by remote sensing technology is affected by various factors directly or indirectly,such as heterogeneous landscape patterns,conspicuous topographic effects and frequent cloud containmination of satellite images.Temporal signature contained in the multi\|source and multi\|temporal satellite images is one of the important information to improve the accuracy of land cover product.A case study was conducted at the upper reaches of Minjiang River,and the native HJ CCD images and Landsat TM images were taken as main input data.Five controlled experiments with different satellite images (single growing season satellite images,single non\|growing season satellite images,multiple growing season satellite images,multiple non\|growing season satellite and all\|temporal satellite images) were designed to validate the contribution of multi\|source and multi\|temporal infomation for automatically mapping of forest types in moutainous area.Comparsion result shows that the multi\|temporal information combined with growing season and non\|growing season can significantly improve the mapping accuracy of forest types in mountainous area compared with single\|temporal or multi\|temporal images of single season,and can simplify the classification rule sets.
The leaf area index (LAI) estimation from remotely sensed data is one of hotspots in quantitative remote sensing of vegetation.Monitoring the spatial and temporal changes of LAI is very significant for carbon cycle of terrestrial ecosystem,global changes and other related studies.The paper selected ten 50 km×50 kmsampling regions as our study area,including five forest regions,three crop regions and two grassland regions.The several parameters,such as leaf area index (LAI),canopy density,biomass,were measured in these regions.Taking leaf area index as a case,this study applied the partial least\|squares regression method to build the estimation model of LAI combining remote sensing with in situ data and considering topographic effects for different vegetation types.Then,the cross\|validation approach was used to test model accuracy.The results indicated that the forest LAI inversion models taking topographic effects (altitude,aspect and slope) into accout is superior to those that topographic effects were not considered (R2 increased from 0.30~0.75 to 0.50~0.80;RMSE decreased from 0.52~0.93 to 0.48~0.89 m2/m2).For all vegetation types,the model validation R2 and RMSE changed between 0.40~0.80,0.22~0.89 m2/m2,respectively.The method regarding LAI estimation from remotely sensed observations developed in this paper can help to understand topographic effects on LAI retrieval,and further provide scientific proof for monitoring vegetation growth status over mountain areas.
Aboveground biomass is an essential parameter for estimating the growth trend and ecosystem services of grassland.It is important to correctly evaluate biomass for grassland ecosystem carbon budget and sustainable development of resources.In this study,the field measurements and synchronous observations from UAV were combined to estimate the aboveground biomass of grassland.Based on the “valley\|peak\|valley” morphological features of green vegetation in the visible band reflectance,multiple types of vegetation indices were constructed.Then,the regression models between the fresh weight of grassland and vegetation indices were established.Accuracy analysis showed that visible vegetation indices had the potential to distinguish grassland from others and the correlation between biomass of grassland and vegetation indices was strong.The highest estimation accurate model was provided by NGRDI which derived from red and blue band,and the RMSE and R2 were 124 g/m2 and 0.856,respectively.The prediction accuracy of the regression model with the same band combination was stable,regardless of the fitting methods Among the different regression models.The results of this study was anticipated to support the research of carbon budget and validation of remote sensing products researches in Zoige.
Land surface temperature (LST) plays an important role in land surface energy budget.It is also a key parameter related to land surface water and heat transfer processes.With the fast development of LST retrieval algorithm by remote sensing methods,LST products are provided by coarse resolution satellite observations such as MODIS and AVHRR,which are commonly used for global or regional study.However,it is still a big challenge for their application over mountainous area due to the high resolution demanded and terrain effect.In order to investigate the influences of topographic factors such as elevation,slope,and aspect on LST,a systematic study was conducted based on LST estimations in the typical mountain environment with Landsat 8 thermal infrared data.The results indicated that there are significant effects from topographic factors on LST distribution.Generally,LST decreases with the increase of elevation and slope.LST in south face usually is higher than that in north face.To discuss the scale problem over mountainous area,the estimated LST was analyzed at 1 km scale,close to MODIS LST spatial resolution,at the high mountain area and hilly area respectively.It was found that the LST at 1 km scale shows higher spatial heterogeneity with elevation and land cover at the complex terrain area.Therefore,high spatial resolution LST data is necessary for land surface water and heat transfer studies over mountain area to consider the impacts from topographic changes.
Chlorophyll is one of the significant biochemical parameters during the crop growth.It has important applications for identifying crop growing condition,diseases and insects,predicting the crop mature date.This paper introduced the basic principles,methods of crop chlorophyll content estimation models with remote sensing data,and summarized main research achievements in this field.Based on the domestic and foreign research,these models were divided into three categories including statistic model,physical model and coupled model.And this paper discussed the advantages,deficiencies,the scale of application and a large number of application examples of these three kinds of models in detail.On the basis of the four aspects——measured data,remote sensing data,the principle and establishment of models,some problems were raised and indicated that the key issue of the crop chlorophyll content estimation models with remote sensing data is the improvement of the physical model method.On the other hand,the priori knowledge is also an important factor in the process of model improvement.Finally,according to some existing results and problems,the method and application prospect of crop chlorophyll content models were put forward.
Although traditional D\|InSAR is widely utilized in surface deformation measurement,its accuracy is limited by many factors,such as decorrelation,atmospheric impact,DEM errors.In order to overcome these limitations,the time dimension is introduced into the D\|InSAR technique.Pixels less affected by decorrelation are selected from a time series of SAR images,and then deformation extraction is conducted on these pixels,based on a phase model.After continuous modifications of PS\|InSAR since its appearance,it plays an important role in monitoring long\|time slow surface deformation.The development history,technical improvements and application fields of PS\|InSAR are summarized here;the facing problems and corresponding research of PS\|InSAR are also pointed out.
The spatial distribution and multi\|year change in land surface temperature (LST) over the Qinghai\|Tibet Plateau (QTP) was analyzed using MODIS LST products.Firstly,the null pixels were reconstructed by integrating temporal and spatial information.The validity ratio of LST was showed to be over 97% after reconstruction.Secondly,a sine and linear piecewise function was utilized to fit the four instantaneous observations to mean daily LST.Ground observation of 0 cm soil temperature was employed to validate the fitting results,which the accuracy is within 1 K.Lastly,a cosine function model was built to describe the seasonal fluctuation of LST.The mean annual LST,the amplitude and the peak LST date were subsequently extracted by this model.Results showed that the mean annul LST was highly correlated to attitude,latitude and the type of underlying surface.Amplitude of LST was showed to rise from the southeast to northwest.The peak date of water was obviously delayed compared with other land cover types.Slope analysis showed that the mean annual LST over the QTP was increasing with a velocity of 0.015 K per year.Amplitude was elevated by 0.076 K per year,which indicated that the probability of extreme weather was larger than before,due to the greenhouse effect and climate change.The peak date of LST appeared earlier to some extent.
Flood disaster extent is a key physical parameter to characterize flood disaster temporal and spatial distribution,and it is the major content and basis of disaster losses assessment.It is a preceding study to perform space\|airborne\|ground\|field integrated flood extent dynamics monitoring using remote sensing technique,ground observation network data and filed surveys.A flood disaster extent extraction model using HJ\|1 CCD data was developed on the basis of region growing algorithm.This model was used to improve the efficiency of flood disaster extent real time dynamic evolution monitoring which resulted from the data acquisition and information extraction efficiency.First,spatial distributions and spectral response characteristics of ground targets were analyzed.Then flood disaster in Fuyuan county of Heilongjiang province was chose as an example to use the region growing algorithm for identification and determination of the disaster extent.The results indicated that the stable and high\|frequency acquisition capacity of HJ\|1 CCD data can support flood disaster extent identification and dynamic evolution monitoring during the flood season.The region growing method can extract concentrated water with high efficiency.Which cost 10 minutes to finish the work,thus that could cost 5 hours by visual interpretation and artificial sketch in the past.The extraction method plays an important role in a large area of flood disaster dynamic monitoring.
In this study,the sea surface wind speed and wind direction data from five microwave radiometers including SSM/I,SSM/IS,TMI,AMSR\|E and WINDSAT,and two microwave scatterometers including ASCAT and QUIKSCAT,are compared with collocated buoy data.The results shows that the accuracy of wind speed from microwave radiometers is about 1 m/s,which meet requirements of most applications.Low frequency products of microwave radiometers are more accurate than the middle frequency products.However,the middle frequency products are recommended to be used in offshore applications for its higher horizontal resolution,and low frequency products are recommended for applications in the open ocean.In the five microwave radiometers,AMSR\|E and WINDSAT give better wind speed products than TMI,while TMI preforms better than SSM/I and SSM/IS.The microwave scatterometers are slightly better at monitoring sea surface wind speed than the microwave radiometers,but the microwave radiometers are better at monitoring larger wind speed.While WINDSAT is the only microwave radiometer able to get wind direction information,microwave radiometers perform much better on monitoring sea surface wind direction than WINDSAT.Only when the wind speed exceeds 6 m/s,sea surface wind data provided by WINDSAT could meet the requirements of applications.And when wind speed is larger than 8 m/s,wind direction from WINDSAT shows similar accuracy with that from microwave scatterometers.On the basis of these results,this study pointed out the needs of improvement of wind remote sensing under high wind conditions.Meanwhile,the accuracy of wind data from buoy is also preliminarily analyzed and based on the analysis,directions of improvements of buoys are proposed.
Landscape index analysis is one of the important methords to monitor the Spatial pattern of land use change by remote sensing.And because of different methods of remote sensing classification,land change patterns that revealed by the same landscape which calculated by different classification methords which may be quite different.Taking Miyun Lounty of Beijing as Qur study Qrea.Object\|oriented classification and pixel\|based supervised classification were Applied to the same image separately.And landscape index was calculated by classification results.The difference of landscape index was calculatedby different classification methords which was analyzed by paired samples t\|test method.The results show that the difference of some landscape indexes which include Percent of landscape (PLAND) and Largest patch index (LPI) is not significant.The difference of Edge density (ED) is slightly significant.And the difference of some landscape indexes include Number of patches (NP),Mean patch size (AREA_MN),Aggregation index (AI) and Perimeter\|Area Fractal Dimension (PAFRAC) is significant.
Combined with bio\|optical model of the Yangtze River Estuary and adjacent water,this study utilized underwater light filed simulation model Hydrolight to obtain remote sensing reflectance spectra of various water types.The sensitivity of remote sensing reflectance to suspended particulate matter(SPM)and the influence on four kinds of chlorophyll\|a(Chla)retrieval algorithms(two\|band algorithms,three\|band algorithm,fluorescent light height(FLH)algorithm,synthetic chlorophyll index(SCI)algorithm)by SPM were analyzed.Results showed that the RSME between the Hydrolight simulated remote sensing reflectance and that of in situ measurements were smaller than 0.01 sr-1,with higher simulation accuracy of remote sensing reflectance from 550 nm to 725 nm.The impact of SPM on remote sensing reflectance decreased with an increase of Chla concentration,which means a decrease of sensitivity.The two\|band and three\|band algorithms were suitable for chlorophyll\|a retrieval with low SPM concentration.FLH was highly influenced by SPM when retrieving Chla,while SCI exhibited better potential of dismissing the influence of SPM when retrieving Chla in highly turbid waters.
Accurate soil moisture retrieval of large area is of great significance to the management and protection of the plateau pasture.Using fully polarimetric Radarsat\|2 Synthetic Aperture Radar(SAR) images at C\|band,this paper carried out the study of soil moisture inversion in the country of Gangcha,Qinghai province,which is a part of Qinghai Lake watershed.Based on water\|cloud model and Chen model,an algorithm was developed for soil moisture inversion.Elimination of vegetation cover and soil surface roughness effect for backscattering was achieved by the algorithm.Through field measurement validation,the developed algorithm gained reliable results.The results of R2,RMSE and RPD value(0.71,3.77%,1.64) show that the developed algorithm can meet the requirement of soil moisture inversion in study region.In the future,if the vegetation cover and soil surface roughness effect for backscattering could be described in more detail,the accuracy of soil moisture inversion is expected to be further improved.
Sea surface brightness temperature is the key to the sea surface salinity inversion.Starting from the relationship between the different sea surface brightness temperature parameters and sea surface salinity,the paper analyzes the relationship between the sea surface salinity with SMOS satellite in different ways and incident polarized brightness temperature parameters from data fitting,significance test,partial correlation analysis and generalized additive models,using the L1C data of SMOS satellite and Argo salinity data in the Northwest Pacific region on July 8,2014.And obtained the following conclusions:there are strong correlation between the incidence angle with horizontally polarized brightness temperature,vertically polarized brightness temperature,the first Stokes parameters and the second Stokes parameters of four brightness temperature parameters.And the horizontally polarized brightness temperature and the first Stokes parameters have good correlation with sea surface salinity.12.5 ° incidence angle of the first Stokes parameter is the best brightness temperature parameters of the sea surface salinity inversion.
Building height retrieval from single SAR image is quite difficult due to complexity of urban scene and geometric distortion of SAR imaging system.Aiming at this problem,a new method for building height retrieval was proposed by matching between geometric model and strong backscattering features in dual\|aspect SAR images,because strong backscattering features received by SAR sensor formed by layover,double bounce scattering,and strong odd scattering is very distinct and sensitive to orientation.Scattering characteristics of building in SAR image,as well as its sensitivity to SAR imaging orientation was first analyzed.Then geometric model of strong backscattering features for building in dual\|aspect SAR images was constructed.And then matching function was defined based on backscattering mean,probability density function of backscattering,and boundary information.Finally building height was derived using multi\|population genetic algorithm to optimize matching function.The experiments based on simulated and airborne dual\|aspect SAR images showed that the average error for building height retrieval using the proposed method is smaller than 1 meter,thus it can effectively improve the building height retrieval results from SAR data.
Accurately acquiring rural residential land is significant for rational utilization of land resources,improvement of the rural ecological environment and acceleration of urbanization.According to polarization scatting characteristic and spectral property of rural residential land,this paper proposed a method for extracting rural residential land based on polarization scatting characteristics of POLSAR and normalized difference index of optical images.This paper analyzed the inapplicability of polarimetric correlation parameters in distinguishing between rural residential land and artificial forests by experiments.The proposed method in this poper can extract rural residential land precisely,with 91.7% in user accuracy and 95.2% in producer accuracy.Compared with the H/α/Wishart iterative classification,the user accuracy and producer accuracy of our method improved 34.9% and 14.4%,and the overall accuracy improved 16.2%.Compared with the supervised classification based on NDVI and NDBI,the user accuracy of our method improved 24.3%.
In the process of classification in remote sensing data,acquiring greater refinement of the land cover type can deliver undoubtedly more information and further deepen the comprehension and interpretation for remote sensing data.With the support of point clouds elevation data,the method of refined classification in remote sensing image is proposed and achieved out.In order to gain high accuracy of subdividing the same kind of land cover type,four factors are taken into consideration,which includes registration,supplementary data source,first echo and point clouds density and image spatial resolution,and the focus is placed on dealing with the problems of mismatch between point clouds density and image spatial resolution.Decision tree is developed to improve remarkably the classification quantity of buildings and vegetation in this study,which represents superiority of classification of fusing point clouds and imagery and achieves the desired goal of the unity of classification accuracy and quantity.
Remote sensing images play a more and more important role in the present social life and economic construction.For the organization and management issues of massive multi\|source and heterogeneous remote sensing images,this paper designs and implements an efficient visualization management system for massive remote sensing images based on the characteristics of remote sensing images.This study included the organizational structure of the system and enhanced the operability of the system and the visual experience by the technology of three\|dimensional globe.Combining xml file with database,we designed an efficient and unified metabase,which achieved an efficient query,retrieval and positioning function for massive,multi\|resolution remote sensing images.Through low\|resolution thumbnails,multi\|resolution pyramid model and high concurrent thread pool,the efficiency of loading and visual display was significantly improved.Finally,using the above techniques,we completed a three\|dimensional globe\|based efficient visualization management system,which achieved a variety of management modes and multi\|level needs.On this base,this system provides strong technical support for the distribution of massive remote sensing images.
Hyperspectral remote sensing is a cutting edge field in remote sensing.It offers the fine detection of objects by its spectral response characteristics in various spectral bands,and has superiority to multispectral remote sensing in fine extraction.However,due to high\|dimensional feature space and limited training samples of the huge data of hyperspectral images,it is difficult for conventional statistical pattern identification methods to classify hyperspectral images.Thus this paper explores the basic principle of support vector machine classifier and employs Binary Decision Tree Support Vector Machine (BDT\|SVM) classification algorithm based on EO\|1 Hyperion hyperspectral imagery.And this study proposes a new definition of the class separation against the long training time and low classification efficiency of existing multi\|class SVM algorithm and generates a modified BDT\|SVM algorithm.On the basis of theoretical analysis,this paper completes the object classification experiments on Hyperion hyperspectral imagery of the test area and verifies the high classification accuracy of the method.Experimental results show that the effect of hyperspectral image classification based on the modified BDT\|SVM algorithm is apparently better than other multi\|class classification methods,which total classification accuracy is up to 90.96% and kappa coefficient is 0.89.The algorithm also solves the problem of non\|separable region,which may be present in the classic SVM multi\|class classification methods.
The segmentation quality of high\|spatial resolution remote\|sensing images has a significantly impact on image classification accuracy.The better image segmentation results will facilitate to get higher classification accuracy.Therefore,it’s necessary to obtain optimal segmentation results by assessing image segmentation quality.A method for assessing the segmentation of high\|spatial resolution remote\|sensing images is proposed by measuring both area and position discrepancies between reference image and segmented image in this paper.This method has been applied to assess the segmentation result quality of image from GeoEye\|1 satellite.The experimental results show that this new method can objectively assess the image segmentation quality,and with the high matching degree of boundary between the obtained optimal segmentation results and reference image is benefit for subsequent image classification.
Conventional remote sensing image classification is usually based on the spectral information and can’t perform well in the classification of high spatial resolution imagery.This study presents an approach to improve the classification accuracy by combining spectral information with spatial texture features to extract from high spatial resolution bands.The KZ\|1 image and Landsat\|8 image of a portion of Shihezi City,Xinjiang were acquired and preprocessed in the study.Object\|oriented image segmentation and grey\|level co\|occurrence matrix texture analysis were used to create image objects and extract textural features for object\|oriented classification.An optimal window size and threshold values were first determined for the image segmentation operation,and the support vector machine algorithm was used to perform the classification procedure.Eight textural features such as Mean,Variance,Homogeneity,Contrast,Dissimilarity,Entropy,Angular Second Moment,and Correlation were extracted from the KZ\|1 and Lansat\|8 OLI panchromatic bands and used to create several different feature sets to conduct the SVM classifications.Classification results from these various image feature sets indicate that due to its high spatial resolution the KZ\|1 image containing abundant textural information in the study area which can achieve the classification accuracy with overall accuracy of 90.06%,and Kappa coefficient of 87.93%.Compared with conventional spectral classification,the overall accuracy of the textural classification with KZ\|1 imagery is increased by 8.02%,and Kappa coefficient increased by 9.65%.The proposed approach is valuable for combing remotely sensed imagery from different satellite platforms to extract urban expansion information quickly and accurately in Xinjiang.
