Zone II • Versions EN1
Abstract: Land cover change is one of the most direct manifestations of global environmental change, which influences global environment. An analysis of regional land cover dynamics lays a basis for researches on global change, ecosystem assessment and the human-environment interaction In this study we take northwest China as research area, including Xinjiang Uygur Autonomous Region, Qinghai Province, Gansu Province, Ningxia Hui Autonomous Region, Shaanxi Province and Alxa League of Inner Mongolia Autonomous Region. We collect Landsat TM image data of the area in 1990, 2005 and 2010, and Landsat ETM+ image data in 2000. Through the eCognition platform, we extract land cover data of the four periods using the object-oriented automatic classification and artificial visual interpretation methods. Finally, data accuracy is verified through three methods, including field sample point survey, high-resolution image recognition and Google Earth sample point identification. This dataset can be used as the background data for regional eco-environment assessment, regional planning, climate change research, environmental modeling, biodiversity studies, and ecosystem carbon research.
Keywords: northwest China; land cover; object-oriented classification method; eCongnition; Landsat
|English title||A dataset of land cover in northwest China from 1990 to 2010|
|Corresponding author||Author’s name (firstname.lastname@example.org)|
|Data author(s)||Xie Jiali, Yan Changzhen, Chang Cun|
|Time range||Year of 1990, 2000, 2005 and 2010|
|Geographical scope||31°35′–49°52’N, 73°56’–111°47’E|
|Spatial resolution||200 m||Data volume||574 MB|
|Data service system||http://www.crensed.ac.cn/portal/metadata/215ea67d-cfa5-4636-8a12-dec526332224|
|Source(s) of funding||Construction of Ground-Remote Sensing Data Platform in Northwest China (KFJ-EW-STS-006); National Science and Technology Infrastructure Platform project – “Shared Service Platform of Special Environment and Function Observation and Research Station” (Y719H71006); Chinese Academy of Sciences informatization special project – “Environmental Evolution Research in Cold and Arid Regions: ‘Technology Field Cloud’ Construction and Application” (XXH13506).|
|Dataset/Database composition||This dataset of land cover in northwest China from 1990 to 2010, contains 25 files, including land cover of Xinjiang Uygur Autonomous Region in 1990, 2000, 2005 and 2010, land cover of Qinghai Province in 1990, 2000, 2005 and 2010, land cover of Gansu Province in 1990, 2000, 2005 and 2010, land cover of Ningxia Hui Autonomous Region in 1990, 2000, 2005 and 2010, land cover of Shaanxi Province in 1990, 2000, 2005 and 2010, land cover of the Alashan League of Inner Mongolia Autonomous Region in 1990, 2000, 2005 and 2010 respectively, and a document of the classification system.|
Land cover refers to the observable complex of natural and artificial landscape on the land surface under the joint action of natural process and human activities. The land cover changes in long time series have both natural and social attributes. As an issue closely intersecting natural and human processes, land use/land cover change has become the focus of attention of all parties in the study of global environmental change1.2. . As early as 1995, the International Geosphere-Biosphere Program (IGBP) and the International Human Dimension Programme on Global Environmental Change (IHDP) jointly proposed the scientific research plan for land use/land cover change1.2.3. . Land cover change is not only the most direct and main manifestation of global environmental change, but also can affect the global environment. Therefore, the analysis of regional land cover change dynamics is the basic work to carry out the studies of global change, ecosystem assessment and the interaction between human and environment. The rapid development of remote sensing technology in recent years provides a strong support for obtaining land cover data sources. The reconstruction of regional land cover database using the remote sensing monitoring has become a new way to understand the changes of landscape pattern.
Northwest China, located in the arid and semi-arid area, has a typical temperate continental climate. The climate in this region is dry with little precipitation, cold winter and hot summer, and the daily and annual temperature range is very broad. Most rivers in this region are inland rivers, and the water supply is mainly from precipitation in mountainous areas and melting water from snow and ice. An inland river basin is a complete water system in which surface water, groundwater and atmospheric water are interrelated, and a complete ecological functional unit system with mountain, plain and desert4 . . Due to the weak infrastructure and water resources shortage in northwest China, coupled with the continuous development of oasis agriculture since the 1950s, there are many problems in the ecological environment in this region, including vegetation degradation, soil erosion, river and lake drying up, and land desertification. In recent years, some ecological projects has carried out such as returning farmland to forests and grassland, protecting natural forests, controlling sand sources, returning grazing land to grassland, and ecological water diversion, which lead to obvious land cover changes. On the basis of land resource, land use/land cover carry a broad and lasting human activities. Both of the resources and environment research and social sustainable development strategy need to be based on land use/land cover data, reveal its characteristics of spa-temporal change, then analysis its influence factors, and make a reasonable plan of land resources use to realize an efficient land use5.. Therefore, accurate and effective measurement and disclosure of land use/cover change can provide a reliable basis for the prediction of change trend in the future in the study area6.7. .
The vision and actions on promoting the jointly building of the Silk Road Economic Belt and the 21st Century Maritime Silk Road were released on March 28, 2015, which marked that China's ‘One Belt And One Road’ initiative had entered the comprehensive construction stage. As an important domestic and foreign juncture of the Silk Road Economic Belt, the northwest region obtained many development opportunities in its construction. Therefore, it is particularly important to strengthen the monitoring and construction of ecological environment in northwest China. Due to the land cover is the most intuitive manifestation of regional ecological environment, it is crucial to interpret and monitor land cover in a long time series, which can provide background data for regional ecological environment protection and sustainable social and economic development.
In this study, we take the northwest China as the study area, including Xinjiang Uygur Autonomous region, Qinghai Province, Gansu Province, Ningxia Hui Autonomous Region, Shaanxi Province and Alxa League of Inner Mongolia Autonomous Region, use a series of Landsat data as the main data sources, and generate the land cover change data set in the period of 1990-2010 (1990, 2000, 2005 and 2010) using an integration method of remote sensing monitoring, and validation through ground investigation and the interpretation of satellite imagery with high resolution.
The production of this data set uses the object-oriented classification method based on the platform of eCognition and data processing method based on ArcGIS platform. For the time phase of remote sensing images, the period from mid-June to late September are selected in northwest China for the cloud-free images. This is because the land use monitoring needs to understand and analysis the vegetation information, and remote sensing data of vegetation growth stage is required to capture more vegetation information5.. In addition, the images in non-growth season and other time-phase that can reflect the differences of ground features, as well as DEM, vegetation coverage and other characteristic indexes are used.
In the previous land use classification research, different classification systems had been adopted in order to meet different needs, such as research and management of land resources, and there are mainly three classification systems, including the land use classification systems established by the National Agricultural Zoning Committee, and by the Chinese Academy of Sciences combining with the characteristics of remote sensing, and the national standard of the land-use status classification formulated by the Ministry of Land and Resources8.9.10. . The classification system adopted in this set of data sets is developed from the perspective of ecology and combined with the characteristics of remote sensing through 19 indicators, such as material composition, structure, arrangement and seasonal characteristics, etc., and the data can be used for the ecological system carbon budget estimation and national ecological environment monitoring11..
2.1 Data source and preprocessing
The data sources are Landsat TM and ETM+ data (Table 1), and most of them are obtained from the United States Geological Survey (USGS, https://glovis.usgs.gov/), and a few from China Geospatial Data Cloud (http://www.gscloud.cn/). The cloud cover of remote sensing image is required to be less than 10%, and the time phase requires that one scene of each orbital image must be in the growth season of vegetation, with the period from July to September being the best. Secondly, the threshold value is set according to the land cover type, and the non-growing season images and other time-phase images are added, which can reflect the differences of different land features. Then the image is preprocessed, including band synthesis, projection conversion and cropping5..
2.2 The extraction process of land cover data
2.2.1 The processes of land cover data classification
The object-oriented classification algorithm and the idea of decision tree are used to classy the land cover data12.. Object-oriented classification method is more and more widely used in the information classification based on remote sensing data. After image segmentation, the homogeneity pixels form different sizes of objects13.14. , and are then classified based on their spectral information and other information such as texture, shape, spatial topological relations. Due to the different scales of different land cover types, multi-scale object segmentation is adopted in the classification process, and it is a local optimization process15.. Successful image segmentation is the necessary premise of object-oriented information extraction method, and the scale and accuracy of segmentation directly affect the accuracy of classification16.. The detailed process is shown in Figure 1.
Finally, in order to meet the requirement of the mapping scale of 1:700,000, we resample the regional land cover data interpreted based on 30 m resolution image to obtain land cover data products with a resolution of 200 m in northwest China.
2.2.2 Main characteristic indexes
When extracting land cover data using the object-oriented classification method, different parameter information is required to set thresholds for different ground features. Besides the band and texture information of the image, DEM, slope, slope direction, and some indexes representing specific ground features are as follows:
(1) Cloud Index (CI)
When classifying land cover, the images without clouds or with cloud cover less than 10% are selected, and the images with clouds need to be processed. In this study, the substitution method is adopted to deal with areas with high cloud coverage17., and the cloud index is used to detect the cloud coverage area. The calculation formula is as follows:
CI = (TM1 +TM2 +TM3 )/3 (1)
Where, TM1 , TM2 and TM3 are the blue band, green band and red band of Landsat data respectively.
(2) Normalized Difference Vegetation Index (NDVI)
The normalized difference vegetation index can directly reflect the growth state of vegetation and vegetation cover, and the information of seasonal changes and human activities can be obtained indirectly according to the change curve of NDVI with time18.. The calculation formula is as follows:
NDVI = (TM4 −TM3 )/ (TM4 + TM3 ) (2)
Where, TM4 is the near-infrared band of Landsat data.
(3) Fractional Vegetation Cover (FVC)
Fractional vegetation coverage refers to the percentage of the vertical projection area of vegetation on the ground in the total area of the statistical area. The commonly used estimation method is pixel binary model19.. The calculation formula is as follows:
FVC = (NDVI−NDVIsoil )/ (NDVIveg +NDVIsoil ) (3)
Where, NDVIsoil is the NDVI value of bare soil region and NDVIveg is the NDVI value of pure vegetation pixel.
However, due to the arid climate, sparse vegetation and simple group structure in northwest China, there are not the typical characteristics of healthy vegetation on the spectral curve in this region. In addition, the spectral information of vegetation obtained from remote sensing images is extremely weak due to the strong interference of ground soil background information. According to existing literature and field sampling verification, the improved three-band gradient method is adopted to calculate the FVC of arid areas, and the formula is as follows20.21. :
Where, TM5 is the short-wave infrared band of Landsat data; λ3 , λ4 and λ5 are red, near-red, short-wave infrared wavelength; d is the pixel gradient, and dmax is the maximum gradient of pixel.
(4) Modified Normalized Difference Water Index-Blue, NDVI-B)
NDWI is used to extract open water surface information22.. In addition, since the reflectivity of water is the lowest in the near infrared band compared with other ground objects, and its reflectivity decreases greatly from blue to near infrared band, the normalized difference water index based on blue band is adopted in this study23. The calculation formula is as follows:
NDWI -B = (TM1 – TM4 )/ (TM1 + TM4 ) (5)
(5) Normalized Difference Built up Index (NDBI)
The normalized building built up index was proposed for automatic remote sensing extraction of urban area information24.. The reflected electromagnetic wave of built up land has obvious differences in the 5 band and 4 band of TM. The calculation formula is as follows:
NDBI = (TM5 - TM4 )/ (TM5 + TM4 ) (6)
(6) Normalized Difference Snow Index (NDSI)
The normalized difference snow index is the extension of vegetation index in remote sensing monitoring of glacier. It normalizes the visible strong reflection band and the mid-infrared low reflection band of the glacier to highlight the characteristics of the glacier. The calculation formula is as follows25.:
NDSI = (TM2 −TM5 )/ (TM2 + TM5 ) (7)
The land cover data sets in northwest China in the period of 1990-2010, cover Xinjiang Uygur Autonomous region, Qinghai Province, Gansu Province, Ningxia Hui Autonomous Region, Shaanxi Province and Alxa League of Inner Mongolia Autonomous Region in 1990, 2000, 2005 and 2010. The spatial resolution is 200 m with the format of SHP, and they are named as “northwest China XX province 1:7 00000 land cover data sets”. Based on the FAO (Food and Agriculture Organization of the United Nations) classification system, the land cover data classification system in this study redefines the classification system suitable for regional characteristics and related application requirements, including 6 primary classes and 33 secondary classes (Table 2)11.. The final coordinates and projection parameters of all data are secant conic projection with equal product and double standard weft. The data results are shown in Figure 2.
|First level code||First level classification||Second level code||Second level classification||Indicators|
|1||Forest lands||102||Deciduous broadleaf forest||Natural or semi-natural vegetation, H=3-30m, C>0.2, deciduous leaves, broadleaf|
|103||Evergreen needleleaf forest||Natural or semi-natural vegetation, H=3-30m, C>0.2, evergreen, needleleaf|
|104||Deciduous needleleaf forest||Natural or semi-natural vegetation, H=3-30m, C>0.2, deciduous leaves, needleleaf|
|105||Broadleaf and needleleaf mixed forest||Natural or semi-natural vegetation, H=3-30m, C>0.2, 25%<F<75%|
|106||Evergreen broadleaf shrubland||Natural or semi-natural vegetation, H=0.3-5m, C>0.2, evergreen, deciduous leaves|
|107||Deciduous broadleaf shrubland||Natural or semi-natural vegetation, H=0.3-5m, C>0.2, deciduous leaves, broadleaf|
|109||Tree orchard||Artificial vegetation, H=3-30m, C>0.2|
|110||Shrub orchard||Artificial vegetation, H=0.3-5m, C>0.2|
|111||Tree garden||Artificial vegetation around the artificial surface, H=3-30m, C>0.2|
|112||Shrub garden||Artificial vegetation around the artificial surface, H=0.3-5m, C>0.2|
|2||Grasslands||21||Meadow||Natural or semi-natural vegetation, K>=1, Soil water saturation, H=0.03-3m, C>0.2|
|22||Steppe||Natural or semi-natural vegetation, K=0.9-1.5, H=0.03-3m, C>0.2|
|23||Tussock||Natural or semi-natural vegetation, K>1.5, H=0.03-3m, C>0.2|
|24||Lawn||Artificial vegetation around the artificial surface, H=0.03-3m, C>0.2|
|3||Wetlands||33||Herbaceous wetland||T>2 or wet soil, H=0.03-3m, C>0.2|
|34||Lake||Natural water surface, calm|
|35||Reservoir/Pond||Artificial water surface, calm|
|36||River||Natural water surface, flowing|
|37||Canal/Channel||Artificial water surface, flowing|
|4||Croplands||41||Paddy field||Artificial vegetation, land disturbance, aquatic crops, harvesting process|
|42||Dry farmland||Artificial vegetation, land disturbance, xerophytic plants, harvesting process|
|5||Built-up lands||51||Settlement||Artificial hard surface, residential building|
|52||Industrial land||Artificial hard surface, production building|
|53||Transportation land||Artificial hard surface, linear feature|
|54||Mining field||Artificial excavated surface|
|6||Other lands||61||Sparse forest||Natural or semi-natural vegetation, H=3−30 m, C=4％−20％|
|62||Sparse shrubland||Natural or semi-natural vegetation, H=0.3−5 m, C=4％−20％|
|63||Sparse grassland||Natural or semi-natural vegetation, H=0.03−3 m, C=4％−20％|
|602||Bare rock||Natural, hard surface|
|604||Bare soil||Natural, loose surface, loamy|
|605||Desert/Sand||Natural, loose surface, sandy|
|606||Salina||Natural, loose surface, high salinity|
|607||Permanent ice/snow||Natural, solid water solid|
Note: C, Cover degree/Canopy density (%); F, the ratio of needleleaf forest to broadleaf forest; H, height of vegetation; T, the time covered with water in a year (month); K: humidity index.
4.1 Data validation sample survey
Generally, there are three sample survey methods to verify the quality of land cover data, including field sample point survey, sample point identification based on high-resolution image and sample point identification based on Google Earth. Most of the field sample points are distributed on both sides of the road, and the survey route is along the roads with good accessibility and passes through various land cover types as much as possible. In order to ensure that the acquired landscape photos can reflect the land cover, the average distance between GPS fixed-point survey is no more than 20 kilometers, which can be appropriately relaxed for contiguous single ground features. GPS survey points locate in the areas with good sight, significant changes of ground objects or interlaced ground objects. After positioning with handheld GPS, we take landscape photos of land cover in different directions (no less than 4 directions) to record landscape features, including the shooting direction angle, main shooting content and surrounding environment, fill in the field survey, and digitize the data obtained from the field survey. For the areas not reached by field survey, high-resolution images and Google Earth are used to verify sample points. In order to ensure the authenticity of verification points, many interpreters independently identify the target points of the image, and the type of consensus reached by three people is the effective verification point.
Due to space limitations, this paper only takes Guanzhong plain as an example (Figure 3). The figure can clearly reflect the change of land cover in the region with Xi 'an as the core. With the growth of population, urbanization develops rapidly, and the area of artificial surface continues to expand, resulting in the decrease of surrounding farmland, grassland and other areas.
4.2 The accuracy assessment of land cover data
Since most of the field verification sample points are distributed in areas accessible to roads, the land cover data of the fourth phase are randomly checked through re-interpretation of high-resolution images and Google Earth to ensure the accuracy of the extracted information. The method of random sampling is adopted to select samples at a sampling rate of 5% for all pattern spots. The land cover type of the verified sample points and the vector land cover data of the corresponding position are superimposed spatially to judge the error rate of the spots one by one. The verification accuracy includes first-level classification and second-level classification accuracy of land cover data. The accuracy can be calculated by comparing the classification results with the verification sample points. For a certain sample i, the formula of classification accuracy (yi) is:
Where, pa is the number of spots correctly interpreted; P is the number of all pattern spots contained in sample i.
After verification and modification of the data, the overall accuracy of land cover data products can meet the requirements of 95% (Table 3).
Land use/cover change is a key element representing the intensity of human activity and global environmental change, and an important input parameter for simulating climatic effects and biogeochemical effects. The measurement, simulation and understanding of dynamic mechanism of spatial and temporal processes of land cover have become the forefront of scientific concern26.. In this study, we use the Landsat image as the main data source, combining with the ground survey validation, and verification through interpretation of high resolution imagery and identification of sample points in Google Earth, take northwest China as the study area with a total area of 325 x 104 km2, and produce the land cover data sets of four period from 1990 to 2010. It verifies the adaptability of the object-oriented information extraction method in the land cove interpretation in the arid region in a larger spatial and temporal scale, and provides important basic data for the ecological environment assessment, sustainable development and ecosystem carbon storage estimate in in northwest China.
The land cover data sets from 1990 to 2010 in northwest China were freely access on the website of data service system platform (http://www.crensed.ac.cn). In order to standardize the use of data sharing, data use application form should be filled out when downloading data on the platform. The data sets were stored in vector SHP format. ArcGIS, ArcView, ENVI, ERDAS and other common GIS and remote sensing software can support the reading and operation of this data. The spatial overlay analysis of land cover data in the fourth period can obtain the spatial and temporal distribution and trend of land cover changes in the region from 1990 to 2010. Combined with regional meteorological factors and human activities, we can carry out the assessment of regional ecological environment change and its driving force analysis, as well as the research on regional sustainable development.
TURNER II BL, SKOLE D, SANDERSON S, et al. Land-Use and Land-Cover Change, Science/Research Plan [R]. IGBP Report No. 35/HDP Report No. 7 , Stockholm, Sweden, and Geneva, Switzerland, 1995.
Li X. A review of the international researches on land use/land cover change. Acta Geographical Sinica 6 (1996): 553-558.
Yu X, Yang G, Wang Y. Advances in researches on environmental effects of land use/cover change. Acta Geographica Sinica 5 (2004): 627-633.
Zhang Z, Zhao X, Wang X, et al. Remote sensing monitoring of land use in China. Beijing: Planet Map Publishing House, 2012.
Kong J, Yang R, Su Y, et al. Effect of land use and cover change on carbon stock dynamics in a typical desert oasis. Acta Ecologica Sinica 21 (2018): 1-11.
Yang F, Yang X, Qi W, et al. A land use/land cover change detection method integrated spectrum and terrain factors. Remote Sensing Information 3 (2018): 116-123.
National Agricultural Zoning Committee. Technical regulations for land use investigation. Beijing: Surveying and Mapping Publishing House, 1984.
Liu J. Macro-investigation and dynamic research on remote sensing of resources and environment in China. Beijing: Science Press, 1996.
General Administration of Quality Supervision, Inspection and Quarantine of the People's Republic of China, National Standardization Administration, National Standard of the People's Republic of China - classification of land use status (GB/T 21010-2007). Beijing: China Standards Press, 2007.
Zhang L, Wu B, Li X, et al. Classification system of China land cover for carbon budget. Acta Ecologica Sinica 24 (2014): 7158-7166.
LAWRENCE RL, WRIGHT A. Rule-based classification systems using classification and regression tree (CART) analysis. Photogrammetric Engineering and Remote Sensing 10 (2001): 1137-1142.
Chen Y, Feng T, Shi P, et al. Classification of remote sensing image based on object oriented and class rules. Geomatics and Information Science of Wuhan University 4 (2006): 316-320.
MYINT SW, GOBER P, BRAZEL A, et al. Per-pixel vs. object-based classification of urban land cover extraction using high spatial resolution imagery. Remote Sensing of Environment 5 (2011): 1145-1161.
BENZ UC, HOFMANN P, WILLHAUCK G, et al. Multi-resolution, object-oriented fuzzy analysis of remote sensing data for GIS-ready information. ISPRS Journal of Photogrammetry and Remote Sensing 3 (2004): 239-258.
Qian Q, Xie R, Zhang L, et al. Land cover extraction based on object oriented analysis. Remote Sensing Technology and Application 3 (2005): 338-342.
Xu Z, Gong C, Liu J, et al. Cloud detection and removal technology in remote sensing image based on object-oriented and substitution method. Agricultural machinery journal 6 (2013): 210-214.
Zhao Y. Principles and methods of remote sensing application analysis. Beijing: Science Press, 2003.
Li M. The method of vegetation fraction estimation by remote sensing. Beijing: Chinese Academy Science, 2003.
Tang S, Zhu Q, Zhou Y, et al. A simple method to estimate crown cover fraction and rebuild the background information. Journal of Image and Graphics 11 (2003): 1304-1308.
Guli J, Chen X, Bao A. Coverage extraction and up-scaling of sparse desert vegetation in arid area. Chinese Journal of Applied Ecology 12 (2009): 2925-2934.
MCFEETERS SK. The use of the normalized difference water index (NDWI) in the delineation of open water features. International Journal of Remote Sensing 7 (1996): 1425-1432.
Qu W, Lu J, Li L, et al. Research on automatic extraction of water bodies and eetlands on HJ Satellite CCD images. Remote Sensing Application 4 (2011): 28-33.
Zha Y, Ni S, Yang S. An effective approach to automatically extract urban land-use from TM imagery. Journal of Remote Sensing 1 (2003): 37-42.
Dou W. Research on glacier extraction based on multisource RS imagery and changes monitoring of eastern Tienshan Mountains. Jiaozuo: Institutes of Technology of Henan, 2014.
Xie J,Yan C and Chang C.A dataset of land cover in northwest China from 1990 to 2010. Science Data Bank, DOI: 10.12072/casnw.057.2018.db(2019).
How to cite this article
Xie J,Yan C and Chang C.A dataset of land cover in northwest China from 1990 to 2010. China Scientific Data 4(2019). DOI: 10.11922/csdata.2018.0047.zh.