Other Data Paper Zone II Versions EN1 Vol 4 (2) 2019
Grassland production dataset in central and eastern Mongolia based remote sensing (2006–2015)
: 2018 - 11 - 24
: 2018 - 12 - 18
: 2019 - 06 - 25
1398 7 0
Abstract & Keywords
Abstract: Grassland production is an important basis for scientific use of grassland resources and management decision-making concerning grass-livestock balance. An accurate timely understanding of the spatio-temporal distribution of grassland production and its changes is of great significance for the sustainable development of grassland ecological environment. This dataset used MODIS remote sensing data and meteorological data, in combination with grassland sampling data, to construct a grassland production estimation model suitable for high altitude and arid environment in Mongolia. An optimal model was obtained based on an accuracy evaluation of the models, which was used to obtain the spatial and temporal distribution data of grassland production in the six provinces of central and eastern Mongolia from 2006 to 2015. The results of data quality evaluation showed that exponential model was more suitable for the estimation of grassland production in this area than linear model or multi-linear model. The exponential model based on MSAVI had the best simulation effect, R2 was 0.72, RMSE was 279.09 kg/hm2, and the simulation accuracy was 78%. This data set is in TIF format with a data volume of 113 MB.
Keywords: Mongolia; grassland; MODIS; grassland production; spatio-temporal distribution
Dataset Profile
English titleA dataset of grassland production in central and eastern Mongolia based on remote sensing(2006–2015)
Corresponding authorWang Juanle (wangjl@igsnrr.ac.cn)
Data author(s)Li Ge, Wang Juanle
Time range2006–2015
Geographical scope42°38'N-50°30'N,101°27'E-113°51'E; specific areas include: Tov, Ulaanbaatar, Hentiy, Govisumber, Dundgovi, Overhangay
Spatial resolution250 mData volume113 MB
Data format*.Geotiff
Data service system<http://www.sciencedb.cn/dataSet/handle/691>
Source(s) of fundingStrategic Priority Research Program (Class A) of the Chinese Academy of Sciences grant number XDA2003020302; the 13th Five-year Informatization Plan of Chinese Academy of Sciences grant number XXH13505-07.
Dataset/Database compositionThe dataset is composed of two subsets:
(1) Grassland production data of 6 provinces in the central and eastern Mongolia (2006–2015). Each data document is recorded as YYYY.tif, and YYYY represents the year. Data volume is 89.6MB. (2)Spatial distribution map of grassland production of 6 provinces in the central and eastern Mongolia (2006–2015). Each data document is recorded as YYYY.jpg. Data volume is 24MB.
1.   Introduction
Grassland is a renewable natural resource and a major component of terrestrial ecosystems. It plays an important role in climate regulation, soil and water conservation, and wind and sand fixation[1][2] . Grassland biomass is an important indicator of grassland productivity and the most direct reflection of grassland ecological status. Located within the continent of Asia, Mongolia is an important part of the Mongolian Plateau. Its largest biome is grassland, accounting for 83.4% of the total land area[3]. However, in recent years, due to the poor stability of grassland ecosystems and the vulnerability to external factors and under the multiple effects of natural and human factors such as climate change and overgrazing, the grassland ecological environment in Mongolia has deteriorated, and grassland in some areas has shown a degradation trend, which is manifest by a decline of grassland productivity, soil erosion, and other ecological barriers. The role of grassland is also weakening, which poses a serious threat to the ecological environment of the Mongolian Plateau and the sustainable development of agriculture and animal husbandry. Timely and accurate monitoring of the temporal and spatial distributions of grassland production are of great significance for the scientific use and management of grassland is also an important condition for the sustainable development of grassland animal husbandry and the ecological environment[4][5] [6] .
Most domestic and foreign scholars till date have focused mostly on grassland monitoring in Inner Mongolia or the Xinjiang Autonomous Region of China. There have been few studies on grassland production in Mongolia, and there is a lack of continuous monitoring of long time series[7][8][9][10] . In this study, six provinces in central and eastern Mongolia were selected as study areas. Based on field data and satellite remote sensing data, the grassland production in this area from 2006 to 2015 was estimated, and the spatial and temporal distributions and the changing dataset of grassland production in central and eastern Mongolia through the past 10 years were obtained.
2.   Data collection and processing
2.1   Study area
The six provinces in central and eastern Mongolia selected by this study are Tov Province, Ulaanbaatar City, Hentiy Province, Govisumber Province, Dundgovi Province, and Overhangay Province. The location and sample distribution of the study area are shown in Figure 1.

Figure 1   Location and sample distribution of the study area
The study area is located in the hinterland of the continent of Asia, 42°38'–50°30' N, 101°27'–113°51' E. It is a typical arid and semiarid region, with a continental temperate grassland climate, large seasonal temperature difference, and large daily temperature difference. There is little precipitation, which is mainly concentrated in summer. The terrain, with an average elevation of approximately 1400 m, is high in the west and low in the east. The grassland features are obvious, and the main grassland types are meadow grassland, typical grassland, and desert grassland. The study area has an area of 307,600 km2, accounting for 19.6% of the area of Mongolia, and has a population of 1.78 million, accounting for more than half of the national population. Grassland animal husbandry is developed. The number of livestock is 19.29 million, approximately accounting for 30% of the country's total livestock.
2.2   Data source
2.2.1   Remote sensing data
Remote sensing data were selected from the Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Net Photosynthesis (PSNnet) of the MODIS sensor land standard products MODIS 13Q1 and MODIS 17A2H, and the Modified Soil-Adjusted Vegetation Index (MSAVI) was obtained through calculation. The NDVI, EVI, and MSAVI have a spatial resolution of 250 m and a time resolution of 16 days. PSNnet has a spatial resolution of 500 m and a time resolution of 8 days. The data source is the National Aeronautics and Space Administration's MODIS website (https://modis.gsfc.nasa.gov/).
2.2.2   Ground-measured and auxiliary data
The ground-measured data in this study were obtained through the joint field survey of the research group and the Mongolian National University. The sampling were collected in August 2013 and 2014, and the grassland production was collected by ground-level mowing. The sample plot size was 10 m × 10 m, and three 0.5 m × 0.5 m sample squares were selected randomly. The coordinates of the sample points were recorded by GPS, and the fresh grass was weighed and recorded. The average of the three samples was the grassland production of the sample plot. After eliminating some samples through a screening process, 29 sample data were finally obtained. Socioeconomic data were downloaded from the Mongolia National Bureau of Statistics[11] (http://www.1212.mn/). Mongolian administrative boundary data, meteorological data, and grassland type data were provided by the Mongolian National University.
2.3   Data processing
MODIS data preprocessing: The MOD13Q1 and MOD17A2H products were in HDF format, and the data were spliced by the MODIS Reprojection Tool (MRT) software batch processing function. The projection method of the MODIS data products was sinusoidal projection. For convenience of data viewing and post-processing, the data were re-projected using MRT software and converted into a Universal Transverse Mercator (UTM) projection.
Vegetation index acquisition: The MODIS data were converted to TIF format data files. Clip tools under the Analysis Tools module in ArcGIS were used to overlay the administrative boundary data of the study area, and then the re-projected raster data were clipping processed. NDVI, EVI, and PSNnet were processed by the maximum value composite method, and the MSAVI was calculated from the RED band and the NIR band. According to the Guide of Effective Range of Remote Sensing Images in the MODIS product user manual, invalid values beyond the range were assigned to null values. The ground sample data were vectorized, and the precipitation data formed by ArcGIS spatial interpolation were superimposed on the remote sensing raster data to extract the corresponding sample pixel values. Eighty percent of the sample data was selected randomly for modeling, and the remaining 20% was used for accuracy verification.
The calculation methods of the remote sensing indices are shown in formulas (1)–(4):
where is the near-infrared radiation reflectance, is the red band reflectance, is the blue band reflectance, GPP is the gross primary productivity, is the leaf autotrophic respiration consumption, and is the root autotrophic respiration consumption.
Estimation model building: Based on the preprocessed ground measured data and precipitation data, the linear, exponential, and multivariate fitting relationships between them and the NDVI, EVI, MSAVI, and PSNnet were established, and 12 regression models were obtained. Model accuracy was evaluated by the determination coefficient, average relative error (REE), root mean square error (RMSE), and other indicators (see formulae (5) and (6)). The exponential model based on MSAVI was determined to be the optimal model, and the grassland production in the study area was estimated by the optimal model.
In the above, N is the number of samples, Yi is the measured grassland production (g/m2), is the estimated grassland production (g/m2), and is the average measured grassland production (g/m2).
3.   Sample description
After data processing, the spatiotemporal distribution and the changes on the datasets of the grassland production in the six provinces of central and eastern Mongolia (2006–2015) were obtained. The spatial resolution of the data was 250 m, and the unit was kg/hm2. The data were reclassified using the Reclassify tool under the Spatial Analyst module in ArcGIS and divided into six levels to describe the distribution of grassland production. Figure 2 shows the spatial and temporal distributions of grassland production of the study area through the past 10 years.




Figure 2   Spatial distribution of grassland production in the study area from 2006 to 2015
4.   Quality control and assessment
4.1   Quality control
To improve the accuracy and quality of the dataset, this study carried out strict quality control in the processes of sample data collection, data preprocessing, model building, and accuracy verification. The sampling route had good coverage for different types of grassland. Randomness and representativeness were considered in the sampling, and the sampling method followed the relevant survey technical specifications. The remote sensing data products were processed by radiation correction,, geometric correction, and the remote sensing index was processed by referring to the MODIS data guide published by MODIS Web. Remote sensing data with two kinds of spatial and temporal resolutions were used for grassland production estimation, and the two kinds of data were modeled separately. By analyzing the accuracy indexes of 12 models of three types, linear, exponential, and multivariate, the optimal model based on the 250 m resolution vegetation index was selected to estimate the grassland production.
4.2   Quality assessment
There were 29 field samples, from which 20% were selected randomly for accuracy verification and 80% were used for modeling. According to the validation results of the grassland production estimation data (Table 1), all models passed the extremely significant test (Sig < 0.01). The determinant coefficient (R2) and the RMSE were used to assess model fit. Larger R2 and smaller RMSE indicate a better fitting effect of the model. The exponential model had the highest accuracy among all models (72.75%), followed by the multivariate linear model (70.25%) and the one-dimensional linear model, which had the lowest accuracy (69.75%). The correlation coefficient of each model showed that the exponential model R2 was 0.67, which was significantly higher than the value of 0.59 of the multivariate linear model and the value of 0.56 of the one-dimensional linear model. The results showed that the exponential model was more suitable for grassland production estimation in this area. Further analysis of each exponential model showed that the model based on MSAVI had the best simulation effect. Its R2, RMSE, and simulation accuracy were 0.72, 279.09 kg/hm2, and 78%, respectively. This model can be used to study the spatial and temporal distributions of grassland production in this area.
Table 1   Validation results of remote sensing estimation of grassland production
ParameterModel TypeInversion ModelR2Sig.RMSE
EVILinearY = −19.490+384.791×X10.460.000423.5566
ExponentialY = 4.257×exp(12.647×X1 )0.630.000466.5363
MultivariateY = −16.655+436.870X1 −0.049X0.470.002423.9166
MSAVILinearY = −27.370+165.323×X20.570.000370.0071
ExponentialY = 3.594×exp(5.224×X2 )0.720.000279.0978
MultivariateY = −18.070+239.539X2 −0.177X0.610.000383.4070
NDVILinearY = −13.940+204.158×X30.560.000335.3773
ExponentialY = 5.728×exp(6.300×X3 )0.680.000306.6876
MultivariateY = −3.192+264.166X3 −0.119X0.590.000308.7675
MOD17A2H PsnNetLinearY = 0.482+0.403×X40.680.000389.2169
ExponentialY = 12.701×exp(0.010×X4 )0.660.000322.8274
MultivariateY = 16.944+0.481X4 −0.106X0.700.000375.4270
Note: X1, X2, X3, and X4 are EVI, MSAVI, NDVI, and PsnNet values, respectively; X is the annual average precipitation.
5.   Usage notes
This dataset is a publicly shared data product for the temporal and spatial distributions and change of grassland production in the six provinces of central and eastern Mongolia in 2006–2015. The data are in TIF format and have a spatial resolution of 250 m. The dataset, based mainly on the production of software platforms such as ArcGIS, is reliable. Users can open it through ArcGIS software or ENVI image processing software. If other formats of the data files are needed, the aforementioned software can be used to convert the data. Users can calculate the raster image data of different areas by superimposing the administrative boundary data and the grassland type data of the study area to obtain the annual variation of grassland production of each province and grassland type. Due to the difficulty of sampling in Mongolia, the number of samples in this dataset is limited. However, it is expected that the number of sample data will increase and that the quality of inversion will improve through continuous accumulation, which will provide a relevant reference for grassland production ,,vegetation distribution and change in Mongolia and the Mongolian Plateau.
The production process of this dataset was strongly supported by the Mongolian National University. We sincerely thank Professors Davaadorj and Sonomdagwa of the Mongolian National University for their help in the field sampling.
[1] Xie G, Zhang Y, Lu C, et al. Study on valuation of rangeland ecosystem services of China [J]. Journal of Natural Resources, 2001, 16(1): 47-53.
[2] Xu B , Yang X , Tao W, et al. MODIS-based remote sensing monitoring of grass production in China[J]. International Journal of Remote Sensing, 2008, 29(17-18): 5313-5327.
[3] World Resources Institute. A Guide to World Resources 2002–2004: Decisions for the Earth Balance, Voice, and Power [EB/OL]. (2003-02-05)[2018-05-20]. http://www.wristore.com/.
[4] Luo L, Wang Z, Ren C, et al. Models for estimation of grassland production and spatial inversion based on MODIS data in Songnen Plain [J]. Transactions of the Chinese Society of Agricultural Engineering, 2010, 26(5): 182-187.
[5] Yang X, Xu B, Zhu X, et al. Models of grass production based on remote sensing monitoring in northern agro-grazing ecotone [J]. Geographical Research, 2007, 26(2): 213-221.
[6] Shi P, Li B, Li Z, et al. A research on the production estimation by remote sensing for large area grassland [J]. Acta Agrectir Sinica, 1994, 2(1): 9-13.
[7] Li J, Jiang P, Dai R. Advances in study on the remote sensing technology and GPS and GIS integration systems in estimation grassland yield application in the north of Xinjiang, China [J]. Acta Ecologica Sinica, 1998, 18(5): 504-510.
[8] Xie Y , Sha Z , Yu M , et al. A comparison of two models with Landsat data for estimating above ground grassland biomass in Inner Mongolia, China [J]. Ecological Modelling, 2009, 220(15): 1810-1818.
[9] Jin Y, Xu B, Yang X, et al. Remote Sensing Dynamic Estimation of Grass Production in Xilinguole, Inner Mongolia [J]. Scientia Sinica Vitae, 2011, 41(12): 1185-1195.
[10] Zhang Y, Niu J, Zhang Q, et al. A discussion on applications of vegetation index for estimating aboveground biomass of typical steppe [J]. Acta Prataculturae Sinica, 2012, 21(1): 229-238.
[11] National Statistics Office of Mongolia. Social and economic situation of Mongolia[EB/OL]. (2017-04-17)[2018-05-24]. http://www.en.nso.mn/index.php.
Data citation
Li G and Wang J. Grassland production dataset in central and eastern Mongolia based remote sensing (2006–2015). Science Data Bank, DOI: 10.11922/sciencedb.691(2019).
Article and author information
How to cite this article
Li G and Wang J. Grassland production dataset in central and eastern Mongolia based remote sensing (2006–2015). China Scientific Data 4(2019). DOI: 10.11922/csdata.2018.0081.zh.
Li Ge
Contributions: basic data processing, space–time distribution calculation of grassland production, and writing of the paper.
male, master’s degree student, research directions of remote sensing and mapping.
Wang Juanle
Contributions: dataset design and technical guidance.
male, Ph.D., researcher, research directions of resource and environmental data integration and sharing and GIS and remote sensing applications.
Strategic Priority Research Program (Class A) of the Chinese Academy of Sciences grant number XDA2003020302; the 13th Five-year Informatization Plan of Chinese Academy of Sciences grant number XXH13505-07.
Publication records
Published: June 25, 2019 ( VersionsEN1
Released: Dec. 18, 2018 ( VersionsZH1
Published: June 25, 2019 ( VersionsZH2