Archive

  • 2019 Volume 21 Issue 10
    Published: 25 October 2019
      

  • Select all
    |
  • HONG Biwen, CAO Qing, ZHANG Ling, LONG Yi, Kou Xuan
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Natural language and maps both have the ability to express geospatial information. Compared with maps, natural language is more abstractive and acceptable to people. But a map has irreplaceable advantages to show the spatial morphological characteristics of geographic entities intuitively. The description of geographic entities in natural language is usually not completely quantitative. Except for limited quantitative descriptions of distances and sizes, there are likely also qualitative descriptions of colors, shapes, and etc. Transforming spatial information described by natural language to maps can release the burden of working memory and promote discovery, inference, and insight. Yet, it remains a challenge in the fields of spatial cognition and map symbols regarding how to convert qualitatively described geographic entities in natural language, which are usually fuzzy or semantics missing, into quantitative graphical symbols. In recent studies on "natural language to map", simple geometrical shapes and common icons were used to express geographic entities. However, much information of geographic entities, which represents the spatial cognition results of human to a certain extent, are lost. This paper proposed a method of geographic entity expression based on morphological description by natural language. By analyzing the semantic information of geographic entities described in natural language, the spatial shape information (e.g., shape, size, and distribution of geographical entities) and other properties necessary for entity expression (e.g., color) were approximately expressed in the form of map. First, the definition and connotation of natural language morphological description were introduced, and the classifications of morphological description and semantic fuzziness were given. Second, combined with the map symbol theory, the morphological description-driven geographic entity simulation expression strategies were studied. Five strategies were proposed for shape, size, color, orientation, and combination. Third, the geographic entity simulated expression with different semantic ambiguity was analyzed and designed, including geographic entity symbol design for single morphological type and different morphological types. Finally, an experiment was implemented for evaluating the validity and quality of the simulated expression of geographic entities. The experiment took Yihe Park, Beijing, China as an example and its tour commentary was applied to convert to a map. In addition, an evaluation method of spatial entity similarity was applied to assess the converted map. Our findings suggest that the graphic design method of this paper can achieve better expression results and has the potential of facilitating better conversion from natural language to maps.

  • JIN Cheng, AN Xiaoya, CUI Haifu, ZHAO Yujun, WANG Hui
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The simplification of linear elements is very important to improve the efficiency of data transmission and visual expression in vector tile map services. Most classical simplification algorithms do not consider the consistency of curves' spatial relations before and after the simplification, which leads to abnormal problems such as sharpening the results of simplification, missing local extremum points, generating line intersection. Considering consistency will affect the efficiency of simplification. In this context, an improved Visvalingam algorithm was proposed according to the application requirements of the vector tile map service.The algorithm applies the minimum heap technology to solve the problem of low efficiency of minimum weight value search, and uses the judgment strategy of self-intersecting topological relation of the line to consider the influence of other points on the current point from the global perspective. In so doing, we can solve the problem of consistency preservation of the topological relationship before and after the simplification of linear elements. The improved Visvalingam algorithm was compared with the original Visvalingam algorithm in terms of topological relationship, geometric features, position accuracy, and simplification efficiency. Results show that the improved Visvalingam algorithm accounted for the topological relations of linear elements and ensuredthe consistency of the overall morphology and topologicalrelationship before and after thesimplification.Our findings suggest that the prosposed Visvalingam algorithm can be applied to the online vector tile map service more efficiently.

  • HUANG Zongcai,QIU Peiyuan,LU Feng,WU Sheng
    Download PDF ( ) HTML ( )   Knowledge map   Save

    News have important application value in especially detecting environmental pollution event perceptions. However, due to the "domino effect" of environmental pollution incidents, news corpora often have mixed descriptions of multiple types of pollution incidents, and existing event detection methods easily lead to text classification errors. This paper proposed a new method for detecting environmental pollution events in news corpora based on joint theme features, which accounts for the global features and theme distribution characteristics. In this method, a joint topic feature vector,which combines TF-IDF (Term Frequency-Inverse Document Frequency) and theme distribution feature vector of the document, is constructed as the input of the supervised classification model to detect environmental pollution events. Using joint topic feature vector as the input of SVM (Support Vector Machine) method, the experimental results show that the average F1 value of event classification detection was 15% higher than that of global feature and 36% higher than that of topic feature.Our findings suggest that the proposed method supports the detection of pollution event types and the extraction of information and helps reveal their spatiotemporal statistical characteristics and the temporal trends.

  • DONG Xiaogang,QIAO Qinghua,ZHAI Liang,SUN Li,ZHEN Yunpeng
    Download PDF ( ) HTML ( )   Knowledge map   Save

    More and more natural environment has been replaced by buildings during the rapid urbanization. However, with the increasing living standards, residents have been paying growing attention to the leisure recreation function of plaza park. Therefore, the accessibility of plaza park has gradually become one of the important indicators for modern people to make their housing choices. But how to objectively and realistically evaluate the accessibility of plaza park is one of the long-term concerns for scholars. Based on previous studies, this article used the gravity model to analyze the influence of the distance friction coefficient on the accessibility evaluation. Importantly, the sensitivity of the evaluation was assessed with the step size of β changing with 0.2 from 0.6 to 2.2. Meanwhile, to further determine accurately the distance friction coefficient β, the idea of tourism domain model was applied for the first time in the accessibility study, for estimating the distance friction coefficient β in the gravity model. Lastly, the accessibility of plaza park was evaluated by a case study in downtown Wuhan. Results show that the distance friction coefficient β is an important parameter in the gravity model. A small change in the friction coefficient β has a great impact on the result of accessibility evaluation. As β increased, the maximum value of the evaluated accessibility became larger and larger, the minimum value became smaller and smaller, and the standard deviation also became larger and larger. It is very important to select the appropriate distance friction coefficient in the gravity model; different β values determine whether the evaluation results are good or not. Therefore, the method of estimating the distance friction coefficient β was constructed based on the idea of tourism domain model method, this paper conducted five experiments to estimate the distance friction coefficient β in the gravity model. Finally, the distance friction coefficient β was obtained by calculating the mean value of different experimental groups. The result of accessibility evaluation is more reliable when β=1.70. The proposed methodology was demonstrated through a case study of the central city plaza park in Wuhan. The results show that the optimized model can fully reflect the change of the attraction of plaza park in distance attenuation. The model fully considers the scale of plaza park and the influence of population scales, which can make a true and objective evaluation of the accessibility of plaza park.

  • MA Yongming,ZHANG Lihua,ZHANG Kang,ZHU Zhiru,WU Zongfan
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Distributions of river systems are the basic data for studying hydrology and water resources, landform evolution, ecological environment, and water and soil management in a basin; so, high-precision extraction of river systems is very important for watershed research. Different digital elevation models (DEMs) are the basic data to extract river systems and delineate watershed. Taking Jiang River Basin in western Hubei Province of China as the study area, this paper extracted the river systems and discussed the extraction precision using different elevation model data including AW3D30 DSM, SRTM1 DEM and ASTER GDEM2 with a spatial resolution of 30 m. The slope flow simulation algorithm and ArcSWAT model were used to extract the river systems, and the relationship between catchment area threshold and stream network density were examined. Meanwhile, the match error and relative error of river system extraction, Google map hydrological data, and the blue line of stream network were used to evaluate the extraction accuracy. Results show that: (1) The threshold of catchment area is the key parameter to determine the accuracy of extracting stream network. The relationship shows that the larger the threshold, the smaller the density of the extracted stream network. The change rate of stream network density is equal to the change rate of catchment area threshold. The inflexion point which corresponds to the optimal catchment area threshold is determined by the power function of the second derivative tangent to the line, hence the optimal threshold of catchment area for AW3D30 DSM, SRTM1 DEM, and ASTER GDEM2 is 70.2 ha, 60.2 ha, and 50.8 ha respectively. (2) The matching error and relative error of the extracted stream network system from AW3D30 DSM data are the lowest and can accord best with the actual river system. (3) The narrower and deeper riverbed profile with a “V” shape has a high extracted precision for the DEMs. (4) The extraction of stream network density, drainage area and river length from AW3D30 DSM is very closer to real value, and can truly reflect the development degree of the river system in the study basin. (5) The topographic fluctuation and slope standard deviation of AW3D30 DSM data are the largest which is helpful to extract the river system in study area. Our findings indicate that, in general, the AW3D30 DSM data is more suitable for extracting river systems in montane basins.

  • HONG Yalan,XUE Cunjin,LIU Jingyi,LIU Xing,SUN Qiang,WU Chengbin
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Driven by multiple marine environmental factors, global marine NPP (net primary productivity) shows different spatiotemporal variations in different ocean areas, and the variations are more remarkable in ENSO events. This paper improved the Dual-constraint SpatioTemporal Clustering Approach (DcSTCA) by adjusting threshold attribute parameters to explore the spatiotemporal evolution clusters of global marine NPP using the satellite remote sensing dataset from January 1998 to December 2017, and analyzed the relationships between the marine NPP spatiotemporal evolution clusters and ENSO events. Results show that: (1) During the period of Eastern Pacific (EP) El Niño events, the spatiotemporal evolution clusters with an abnormal decreased intensity were mainly located in either the equatorial mid-eastern Pacific Ocean (PO) or the equatorial eastern PO, while the spatiotemporal evolution clusters with an abnormal increased intensity were mainly located in the equatorial western PO and the central South Pacific. During the period of Central Pacific (CP) El Niño events, the spatiotemporal evolution clusters with an abnormal decreased intensity were mainly located in the equatorial central PO, and the spatiotemporal evolution clusters with an abnormal increased intensity were mainly located in the equatorial western PO and the central South Pacific. (2) During the period of EP La Niña events, the spatiotemporal evolution clusters with an abnormal increased intensity were located in the equatorial central or the central eastern PO, the equatorial Atlantic ocean and the equatorial Indian ocean, while the spatiotemporal evolution clusters with an abnormal decreased intensity were located in the east-central South Pacific. During the period of Central Pacific (CP) La Niña events, the spatiotemporal evolution clusters with an abnormal increased intensity were mainly located in the equatorial central PO, and the spatiotemporal evolution clusters with an abnormal decreased intensity were mainly located in the mid-western South Pacific. (3) The distribution and spatial movements of marine NPP spatiotemporal evolution clusters in the tropical PO showed some regularity. The spatiotemporal evolution clusters had more significant variation characteristics in EP ENSO events as compared with in CP ENSO events. During the period of EP ENSO events, the spatiotemporal evolution clusters presented a trend of moving to the east. During the period of CP ENSO events, the spatiotemporal evolution clusters presented a tendency to arise and disappear in the central equatorial PO. (4) The area of spatiotemporal evolution cluster had a strong correlation with MEI in ENSO events.

  • XIONG Junnan,LI Jin,ZHU Jilong,CHENG Weiming,GUO Liang,WANG Nan,ZHANG Xiaolei
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The spatiotemporal distribution of mountain torrents and its influencing factors are the key issues for disaster research. Based on the historical torrent disaster data of Chongqing from 1950 to 2015, the spatiotemporal distribution of torrential disasters in Chongqing was analyzed by mean center, standard deviation ellipse, kernel density estimation, and M-K mutation detection. Subsequently, the correlation between mountain torrents and various influencing factors were qua.pngied. Results show that: (1) From 1950 to 2015, the occurrence frequency of the historic mountain torrents in Chongqing presented a trend of stabilization at first and then increase. Mountain torrents mainly occurred from May to September. The occurrence frequency of mountain torrents presented an exponential growth trend on the whole according to the interdecadal trend. (2) The occurrence of mountain torrents in Chongqing had an obvious agglomeration, and the frequency of mountain torrents in adjacent counties was similar. The density of mountain torrents in Jiulongpo, Nanan, Beibei, and Bishan was more than 50 times per 1000 km 2, which belonged to high risk regions. (3) The spatial distribution pattern of mountain torrent disasters was "scattered in southwest-northeast and concentrated in northwest-southeast." Before 2010, the gravity center of mountain torrents mainly concentrated in the vicinity of Fuling; After 2010, the distribution of mountain torrent disasters inclined to northwest, the heart moved to Zhongxian county, the degree of accumulation decreases, and the occurrence of mountain torrents increased randomly; (4) 2002 was the year of a sudden change of mountain torrents in Chongqing, which mainly increased in Tongliang, Bishan, Jiulongpo, Banan, Pengshui Miao Autonomous County, and Kai County. (5) Both the elevation and the density of river networks were positively correlated with the density of mountain torrents, while vegetation coverage was negatively correlated. Short-duration heavy rainfall can stimulate the occurrence of mountain torrents. Our findings are of great significance for flood prevention and disaster mitigation in Chongqing.

  • FU Li,WANG Yong,ZENG Biao,MAO Yong,GAO Min
    Download PDF ( ) HTML ( )   Knowledge map   Save

    In recent years, China's health industry has made rapid progress, but there are still gaps between different regions. As one of the basic public services, medical services is closely related to the quality of resident’s life. However, there are still many problems in getting medical services with high quality in some areas, e.g., inconvenient transportation, lack of medical facilities, poor medical services, and so on. Therefore, it is critical to evaluate the rationality of the distribution of medical resources in a region. Spatial accessibility of medical facilities is an important index to evaluate the rationality of medical service distribution. Among a wide range of methods in measuring the spatial accessibility of facilities, the two-step floating catchment area method (2SFCA) is most popular. In this study, we analyzed the spatial accessibility of medical facilities in Beibei District, Chongqing, by using the modified two-step floating catchment area method and GIS spatial analysis technology. The modified two-step floating catchment area method takes the scale of hospital grade and the distance between supply and demand points into account, and adds Multi Catchment Sizes and Gaussian distance decay to make up for deficiencies of the traditional two-step floating catchment area method, so it is more widely used in spatial accessibility analysis. The spatial accessibility of medical facilities in Beibei were visualized by spatial interpolation. Moreover, the cluster of spatial accessibility was analyzed by Hot Spot Analysis. The basic unit of analysis was administrative villages. The results show that: (1) The results obtained by the original/unmodified and the modified two-step floating catchment area methods have different characteristics, but the modified takes into account the attraction of hospital scale to residents and the influence of distance attenuation factors to residents travel intention, it has higher sensitivity in ide.pngying high accessibility regions with internal differences and low marginal accessibility regions, so its results can better reflect the spatial accessibility of medical facilities. (2) Overall, the spatial accessibility of medical facilities in Beibei District is high, illustrating that the medical services are more accessible to local residents. Meanwhile, the spatial accessibility of medical facilities in Beibei gradually decreased from central areas to surrounding areas. (3) The spatial accessibility of medical facilities in Beibei District varies greatly with obvious polar differences. The high-value areas are mainly concentrated in Dongyang Street, Chaoyang Street, Tiansheng Street, Beiwenquan Street, and Longfengqiao Street, while the low-value areas are mainly concentrated in marginal areas such as Jindaoxia Town, Liuyin Town, Sansheng Town, Fuxing Street, and Jingguan Town, etc. Our findings can provide reference for the relevant departments to make more informed decision-making.

  • XIAO Chiwei,RAO Didi,LIU Yiyuan,FENG Zhiming,LI Peng
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Since the 1990s, driven by the geoeconomic cooperation mechanisms, Laos, the only land-locked country in mainland Southeast Asia, has been experiencing rapid land use/cover change (LUCC) dominated by expansion of construction land, especially in its borderlands with adjacent countries (e.g. China). Taken the Mohan-Boten Port between China and Laos as the study area, a total of ten Sentinel-2 A/B 10m spatial resolution images (with none or little cloud coverage) from February 2016 to November 2018 were used to map the construction land through object-oriented classification and post-classification visual adjustments. The spatio-temporal change pattern and national difference of construction land in a 15 km buffer area of the Mohan-Boten Port, were analyzed and revealed via spatial and statistical analyses with ArcGIS 10.x software. Results showed that: (1) The area of construction land increased rapidly from 1098.8 hm 2 in early 2016 to 2238.8 hm 2 by the end of 2018 in the Mohan-Boten Port area. In particular, about 50.8% of the newly established construction land were located at elevations between 800 m and 1000 m and 80.9% on slopes below 20°. (2) During 2016-2018, the area of construction land increased from 695.4 hm 2 to 1226.7 hm 2 in the Mohan Port, with a growing rate of ca. 6.7%, while the proportion of the total area of construction land declined from 63.3% to 54.8%. (3) The construction land significantly increased by 1.5 times in the adjacent area of Boten Port, i.e., from 403.4 hm 2 to 1012.1 hm 2 in the same period and grew at an average rate of 11.1%. Moreover, the proportion of construction land increased from 36.7% to 45.2%, increased by 8.5% in the study area. We concluded that the developing geoeconomic cooperation is a dominant factor for land use change in the borderland between China and Laos, especially in the port area.

  • SUN Jiayu,ZUO Zhengkang,SUN Yiyuan,ULLAH Sana,ZHANG Ruihua,ZHAO Haimeng
    Download PDF ( ) HTML ( )   Knowledge map   Save

    With the promotion of 3D reconstruction technology in various scenarios, especially in 3D reconstruction from remote sensing images, the demand for accurate 3D point clouds are becoming more and more intense. Image matching is an important procedure in 3D reconstruction, and its result directly affects the accuracy of subsequent procedures such as bundle adjustment and orthore.pngication. In feature-intensive areas such as towns and farms, the feature distance is small enough to match well. Based on high confidence, the precision of image matching is high. However, in featureless areas such as grassland and desert, the large feature distance may cause mismatch between images. If the feature point matching method is used, it is difficult to obtain correct pose under high thresholds. If a sufficient number of matching pairs is promoted, the matching pairs will contain many mismatching ones, which will cause failure to image matching, and the sparse point cloud will not be evenly distributed. In this scenario, a forward-looking image matching algorithm based on dynamic polar coordinate parameterization was presented. First, a polar transform was designed to a pair of normal images, a dynamic strategy was designed to solve the problem of uneven sampling in the polar axis direction. The obtained polar coordinate image pair was matched with the least squares method in the rotation and translation directions. After the rotation and translation parameters were calculated, we used the result to compared with the result of the SIFT algorithm. In this paper, two sets of experiments were designed to obtain poses between images: one is to use simulation datasets with known rotation and translation parameters; the other is to use true scene datasets with unknown rotation and translation parameters. In each set of the experiments, we used 3 different pair of images, and 3 different known rotation and translation parameters in the first set. With the same computer hardware, and with two images with a resolution of 7360 pixels * 5400 pixels and depth of 32 bit, the proposed method took about 57% less than the time took by the SIFT algorithm, and the rotation and translation error in the two methods was usually less than 1%. Our findings suggest that the proposed algorithm gets image poses with an accuracy similar to that of the SIFT algorithm, but its time consumption is significantly less. Our algorithm shows good practical application value.

  • YANG Jinyi,XU Weiming,WANG Chengjun,WENG Qian
    Download PDF ( ) HTML ( )   Knowledge map   Save

    The following problems exist in the change detection of high-resolution remote sensing imagery: lack of the semantic information of low-level features, the "salt and pepper" phenomenon in the detection results based on pixel-level methods, and the low degree of sample labeling automation in supervised classification. In this paper, we proposed a change detection method based on super-pixel Bag-of-Words features and active learning. Firstly, we used the entropy rate segmentation algorithm to obtain the segmentation objects of superimposed images. Secondly, we extracted the features of the Neighborhood Correlation Images (correlation, slope, and intercept) and the change intensity features of texture (mean value, variance, homogeneity, and dissimilarity) while considering neighborhood context information between pixel pairs of the studied two phases of images, and then combined them as the low-level features of pixel pairs. Followingly, based on these low-level features, we constructed the expression of Bag-of-Words features in the super-pixel regions by the Bag-of-Words (BOW) model, and we adopted an improved annotation strategy to annotate automatically the samples with large information from the unlabeled sample pool. Finally, we conducted the change detection using the trained classification model. By choosing two groups in different parts of GF-2 imagery and Worldview-Ⅱ imagery as a data source for experiments, the experimental results show that the F1 scores of the two groups of data sets are 0.8714 and 0.8554, the precision is 0.9148 and 0.9022, the missed detection rate is 0.1681 and 0.1868, and the false detection rate is 0.0852 and 0.0978, respectively. The results demonstrate that our proposed method can effectively ide.pngy the variation area, improve the accuracy of change detection. In addition, the comparison of the learning curves between the traditional active learning method and the active learning method with improved annotation strategy shows that the improved annotation strategy can effectively improve the automation degree of sample annotation at a lower precision loss.

  • FENG Shanshan,FAN Fenglei
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Impervious surface is considered as a major indicator of the degree of urbanization and also an important indicator of environmental quality. Currently, impervious surface extraction is usually based on remote sensing data, including different resolution remote sensing data sources. In extracting high-precision impervious surface, there would be great differences in extraction accuracy caused by different spatial scales. Therefore, it is necessary to explore impervious surface extraction characteristics with different remote sensing data sources. This paper used Landsat/OLI spectral data and VIIRS/DNB nighttime data to extract impervious surface and compare their extraction accuracy difference. The primary objective of this study was to determine the optimal data sources for estimating Impervious Surface Percentage (ISP) for regions with different density of impervious surface distribution. Firstly, Linear Spectral Mixture Analysis (LSMA) was used to extract impervious surface with Landsat/OLI data, and Large-scale Impervious Surface Index (LISI) was used to estimate ISP with VIIRS/DNB data. Then, accuracy of the impervious surface extraction results from these two data sources was assessed respectively, based on the Root Mean Square Error (RMSE), Systematic Error (SE), and coefficient of determination (R 2). The accuracy results showed that the overall ISP accuracy based on the Landsat/OLI data was slightly better than that based on the VIIRS/DNB data, with the overall RMSE being 0.18 and 0.21, SE 0.12 and 0.13, and R 2 0.76 and 0.67, respectively. The accuracy assessments from different density results of impervious surface indicated that the extraction capabilities of Landsat/OLI data and VIIRS/DNB data were greatly different for regions with different density of impervious surface distribution. In the region of low-density impervious surface distribution, the extraction accuracy of impervious surface results based on VIIRS/DNB data was better than based on Landsat/OLI data, because the impervious surface information can be effectively distinguished based on light brightness of VIIRS/DNB data. The impervious surface extraction results from Landsat/OLI data had better accuracy in the areas of medium and high-density impervious surface distribution, because the spatial details of high-density urban impervious surface can be extracted more effectively by the spectral differences of Landsat/OLI data. In future studies, more research is needed to explore the impervious surface extraction characteristics with remote sensing data at different spatial scales and to determine the optimal data sources for effectively and accurately estimating impervious surface.

  • ZHANG Qinyu,LI Zhe,XIA Chaozong,CHEN Jian,PENG Daoli
    Download PDF ( ) HTML ( )   Knowledge map   Save

    GaoFen-6 satellite (GF-6), with the advantages of wide coverage, various resolutions, and multiple bands, provides valuable information to interpret remote sensing data. The objective of this study is to explore the application of the new bands of GF-6 in the ide.pngication of forest tree species. GF-6 WFV image data that covers the whole Alongshan Forestry Bureau was used as the material. Based on the feature space optimization (FSO) algorithm and maximum likelihood classification method, the spectral features and Gray-Level Co-occurrence Matrix (GLCM) features were applied to the first 4 bands and all bands (8 bands) of GF-6 WFV to classify main forest trees (including Larix gmelinii (Rupr.) Kuzen., Betula platyphylla Suk, Pinus sylvestris var mongolica Litv, Salix and Populus davidiana) by utilizing the bands texture characteristics. Furthermore, to study the amount of information provided by the new bands in the classification of tree species, several new band features were added individually to the whole classification features to determine the contribution rate rank of each band. To ensure that the classification of the first 4 bands and 8 bands is comparable, the spectral features and three texture features of each bands were used for experimental classification of tree species. Results show that, for the 34 features of 4 bands and 68 features of 8 bands, the preferred texture features were mainly the GLCM Homogeneity, the GLCM Mean, and GLCM Angular second moment. The classification accuracy after adding the preferred texture features increased 13.23% and 24.63% than only using 4 bands and 8 bands, respectively. Regarding the use of different bands for classification, the classification accuracy of using 8 bands was 11.88% higher than the using first 4 bands when based on spectral features while 23.24% higher when both spectral and texture features were applied. The tree species based on 8 bands spectral and texture features has the highest classification accuracy of 68.74%. These findings suggest that the new bands of GF-6 WFV could improve the accuracy of tree species classification. Moreover, the contribution rates of the new bands rank as follows: Band 6 (Red Edge 2)> Band 5 (Red Edge 1)> Band 8 (Yellow)> Band 7 (Purple). This indicates that red edge band contributes most to tree species classification in northern China.

  • BIAN Zenggan,WANG Wen,JIANG Yuan
    Download PDF ( ) HTML ( )   Knowledge map   Save

    Timely and accurate acquisition of cropping structure is crucial for regional water resource management and crop yield estimation. Remote sensing of cropping structure based on multiple-temporal imagery can make full use of temporal featuresfor especially regions with complicated cropping structure.However, how to select remote sensing data and classifier remains a challenge.In this paper, to map the crop distribution and planting characteristicsin the middle reaches of the Heihe River Basin, we proposed a new method by combining multi-temporal remote sensing imagery and multiple classifiers. Eighteen images available from the Gaofen-1(GF-1) satellite in 2018 were applied to construct the time series of normalized differential vegetation index (NDVI) according to the following hierarchical principles: from easy to difficult, from specific to general. The level-1 and level-2 land cover categories were interpreted successively. Specifically, in interpreting level-1 land cover categories, the decision tree classification method was used to ide.pngywater bodies based on NDVI; and then, the object-oriented classification method was used to interpret construction land by NDVI texture information after zoning; and finally, the random forest classification method was used to classify cropland, forest land, grassland, bare land, and wetland. To further classify cropland, the decision tree classification method was used firstly to interpret alfalfa which has special phenology regularity and is easy to distinguish, and then to interpret wheat categories which possess large phenological differences from other categories, and finally to interpret corn, vegetables, and other crops that have similar phenological conditions. The overall accuracy of the level-1 land cover classification and cropping structure interpretation in the study area is 97.24% and 86.58%, respectively, with the kappa coefficient being 0.96 and 0.80, respectively. In addition, four factors affecting the interpretation accuracy of the research area in the middle reaches of the Heihe river basin are analyzed: definition of landcover categories, mixed pixels, selection of basic image in image segmentation and selection of classification method. By comparing different classification methods, Compared with methods which employ only a single classifier (e.g., maximum likelihood classification method, support vector machine classification method, and random forest classification), the proposed method has higher interpretation accuracy.

  • SHI Lan,HE Qiquan,YANG Jiao,WAN Yibo
    Download PDF ( ) HTML ( )   Knowledge map   Save

    In view of present difficulties in providing high-precision precipitation information, this paper constructed three models for downscaling the GPM IMERG precipitation product, based on Geographically Weighted Regression (GWR) and two Ordinary Least Square (OLS) methods. Using these three models, the spatial distribution of precipitation in each month in 2015 was downscaled by integrating the original GPM IMERG product, the ground measured precipitation data, MOD05 water vapor data, and Vegetation index data. The resolution of the precipitation product was downscaled from 0.1° to 1km. Validation results showed that the goodness of the GWR-based model was 102.9% and 93.9%, higher than the goodness of the two OLS models. More specifically, the GWR model has exhibited better stability and less monthly variations. Of the two OLS models, the one that incorporated water vapor exhibited better model fitting goodness in eight months. Compared with the GPM precipitation product, the GWR-based downscaled product, in addition to increase the spatial resolution, decreased the relative error and root-mean-square error by 42% and 32%, respectively. Our findings suggest that the proposed GWR-based model has good potential in downscaling the GPM IMERG precipitation product.