Iceberg calving, one of the key process of Antarctic mass balance, has been regarded as an important variable in fine monitoring the changes of ice shelves. The authors used multi-source remote sensing data near early August of each year from 2005 to 2020, including ENVISAT ASAR (WSM) images from 2005 to 2011, Terra/Aqua MODIS 7-2-1 band composite images from 2012 to 2014, Landsat-8 OLI 4-3-2 band composite images from 2013 to 2020, and Sentinel-1 SAR (EW) images from 2015 to 2020, to generate annual circum-Antarctic image mosaics after pre-processing. Next, combining MEaSUREs ice velocity dataset, grounding line, ice thickness dataset (Bedmap 2 and Bedmachine), spatial calculation and map digitization techniques were applied to extract all annual calving events larger than 1 km² that occurred on the Antarctic ice shelves from August 2005 to August 2020. Also, their area, thickness, mass and calving recurrence cycle were calculated to derive the annual iceberg calving dataset of the Antarctic ice shelves (2005-2020). This dataset contains the distribution of 15-year annual calving events, along with the attributes of each individual calving event including calving year, length, area, average thickness, mass, and recurrence interval. This dataset can directly reflect the magnitude characteristics and distribution of Antarctic iceberg calving in different years, which fills the gap of fine monitoring dataset of iceberg calving and provides fundamental data for subsequent research on calving mechanism and mass balance of Antarctic ice shelf-ice sheet system.
QI Mengzhen, LIU Yan, CHENG Xiao, HUI Fengming, CHEN Zhuoqi
Land surface temperature (LST) is one of the important parameters of the interface between the earth's surface and atmosphere. On the one hand, it is the direct reflection of the interaction between the surface and the atmosphere, on the other hand, it has a complex feedback effect on the earth atmosphere process. Therefore, land surface temperature is not only a sensitive indicator of climate change and an important prerequisite for mastering the law of climate change, but also a direct input parameter of many models, which has been widely used in many fields, such as meteorology, climate, environmental ecology, hydrology and so on. With the deepening and refinement of Geosciences and related fields, there is an urgent need for all weather LST based on satellite remote sensing. The generation principle of this dataset is a satellite thermal infrared remote sensing reanalysis data integration method based on a new land surface temperature time decomposition model. The main input data of the method are Aqua MODIS LST products and GLDAS data, and the auxiliary data include vegetation index and surface albedo provided by satellite remote sensing. This method makes full use of the high-frequency and low-frequency components of land surface temperature and the spatial correlation of land surface temperature provided by satellite thermal infrared remote sensing and reanalysis data, and finally obtains a high-quality all-weather land surface temperature dataset. The evaluation results show that the land surface temperature data set has good image quality and accuracy, which is not only completely seamless in space, but also highly consistent with MODIS LST products widely used in the current academic circles in amplitude and spatial distribution. When MODIS LST is used as reference, the mean deviation (MBE) of the data set is 0.08k to 0.16k, and the standard deviation of deviation (STD) is 1.12k to 1.46k. Compared with the AATSR LST products released by ESA, the MBE and STD of AATSR LST products are -0.21k to 0.25k and 1.27k to 1.36k during the day and night. Based on the measured data of 15 stations in Heihe River Basin, Northeast China, North China and South China, the test results show that the MBE is -0.06k to -1.17k, and the RMSE is 1.52k to 3.71k, and there is no significant difference between clear sky and non clear sky. The time resolution of this data set is twice a day, the spatial resolution is 1km, and the time span is from 2000 to 2019, which will be updated continuously. The spatial range is China land and its surrounding areas (72 ° e-135 ° e, 19 ° n-55 ° n). This dataset is abbreviated as trims LST China (thermal and reality integrating medium resolution spatial sealing LST – China) for user's convenience. It should be noted that the spatial subset trims lst-tp (trims lst-tp; 2000-2019) V2 of trims LST China has also been released in the national Qinghai Tibet Plateau scientific data center to reduce the workload of data downloading and processing for relevant users.
ZHANG Xiaodong, ZHOU Ji, TANG Wenbin, DING Lirong, MA Jin
1) Significance: construction land is one of the highest performance of human activities. The consumption of natural resources and the change of ecological environment can be closely linked with the development of construction land. This data reflects the evolution of high-precision construction land with 30 m spatial resolution from 1990 to 2019 in 7 provinces/municipalities directly under the central government of China, which are also important areas for rapid urbanization. 2) Data sources: Landsat series satellite data; China regional surface meteorological element driven data set (1979-2018) 3) Processing method: supervised classification method is adopted, random forest algorithm and Fourier transform are used to process characteristic bands, and control points are classified based on visual interpretation. 3-1) Obtaining spectral features: First, screen out Landsat images with transport volume <20%, and superimpose these images in units of 3 years, and then take the median of each superimposed pixel as the target pixel for pixel stitching. Obtain cloud-free images of the entire study area. This method can also better remove the banding influence of Landsat7 data. 3-2) Acquisition of time features: each pixel that has been superimposed for 3 years is screened for cloud cover, and discrete Fourier transform is performed following the minimum mean square error fitting principle to obtain the time latitude of each pixel. "Crest", "Trough" and "Phase". This method can better eliminate the influence of “bare land” on the extraction of construction land, because bare land may be covered by vegetation in spring and summer, and its time characteristics are quite different from construction land. 3-3) Extraction of meteorological and terrain features: The meteorological features are calculated from the China Regional Ground Meteorological Elements Driven Data Set (1979-2018): the data set is superimposed at the same time interval as Landsat, and each image is obtained The average value of yuan is used as the meteorological feature (due to the lack of meteorological data for 2019, the meteorological feature of the last period only calculates the average value of 2017 and 2018). Topographic features (elevation, slope) use SRTM-30m data. The detailed method and code can be seen as follows: https://github.com/wangjinzhulala/North_ China_ Plain_ GEE_ Organized 4) Data quality: the overall accuracy of all years is better than 94%. 5) Application prospects: Simulation of regional urban expansion; estimation of environmental impact of urbanization; quantification of food security and sustainable development.
Gwadar deep water port is located in the south of Gwadar city in the southwest of Balochistan province, Pakistan. It is 460km away from Karachi in the East and 120km away from Pakistan Iran border in the West. It is adjacent to the Arabian Sea in the Indian Ocean in the South and the Strait of Hormuz and Red Sea in the West. It is a port with strategic position far away from Muscat, capital of Oman. This data includes the median values of 343 landsat8 data in each 30 meter grid of Gwadar Port Area and its surrounding area from 2014 to 2015. The data includes 12 bands with a spatial resolution of 30 meters, of which the thermal infrared band is 100 meters and the resampling resolution is 30 meters.
Remote sensing image refers to the film or photo recording the electromagnetic wave size of various ground objects, mainly divided into aerial photo and satellite photo. The 1-5m remote sensing data set of Yangon deep water port area is from gaofen-2 satellite, with the highest resolution of 1m and the lowest resolution of 5m, including a total of 7 regional images. There are four images in each region, which are band composite images of 5m level and 1m level. The accuracy of 5m level image can meet the needs of most research purposes, and the amount of data is smaller; the accuracy of 1m level image is higher, which can be used for synthesis, verification and other purposes, but the amount of data is larger than 5m level data. In practical use, we can choose 5m or 1m images according to the needs of researchers.
GE Yong, LI Qiangzi, LI Yi
The meter resolution remote sensing image data of hanbantota area is composed of data fusion and splicing of different satellites. Multispectral remote sensing images with resolution between 0.5 m and 1 m from 2018 to 2019 are selected, and cloud free data with similar time are selected, and the result data set is formed by cutting and splicing according to the research area. The spatial resolution of the data is about 0.6 meters. The data is mainly used to study the high-precision extraction of disaster bearing body elements, such as port facilities, roads and so on. The extracted thematic elements will be used as the basic data of storm surge exposure and vulnerability analysis.
High Asia is very sensitive to climate change, and is a hot area of global change research. The changes of temperature and precipitation will be reflected in the freezing and thawing time of ice and snow. Satellite microwave remote sensing can provide continuous monitoring ability of ice and snow surface state in time and space. When a small part of ice and snow begins to melt, micro liquid water will also be reflected in active and passive microwave remote sensing signals. In the microwave band, the dielectric constant of ice and liquid water is very different, so it provides a basic theory for the microwave remote sensing monitoring of ice and snow melting. In the case of passive microwave, when ice and snow begin to melt and liquid water appears, its absorption and emissivity increase rapidly, so its emissivity, brightness temperature and backscatter coefficient will also change rapidly. This data set is the initial time of ice and snow melting in the high Asia region retrieved by using the satellite microwave radiometer and scatterometer observations from 1979 to 2018. The passive microwave remote sensing data are SMMR on satellite (1979-1987) and SSM / i-ssmis radiometer on DMSP (1988 present). The active microwave remote sensing data is the QuikSCAT satellite scatterometer (2000-2009).
XIONG Chuan, SHI Jiancheng, YAO Ruzhen, LEI Yonghui, PAN Jinmei
This data includes the daily average water temperature data at different depths of Nam Co Lake in Tibet which is obtained through field monitoring. The data is continuously recorded by deploying the water quality multi-parameter sonde and temperature thermistors in the water with the resolution of 10 minutes and 2 hours, respectively, and the daily average water temperature is calculated based on the original observed data. The instruments and methods used are very mature and data processing is strictly controlled to ensure the authenticity and reliability of the data; the data has been used in the basic research of physical limnology such as the study of water thermal stratification, the study of lake-air heat balance, etc., and to validate the lake water temperature data derived from remote sensing and different lake models studies. The data can be used in physical limnology, hydrology, lake-air interaction, remote sensing data assimilation verification and lake model research.
The long-term evolution of lakes on the Tibetan Plateau (TP) could be observed from Landsat series of satellite data since the 1970s. However, the seasonal cycles of lakes on the TP have received little attention due to high cloud contamination of the commonly-used optical images. In this study, for the first time, the seasonal cycle of lakes on the TP were detected using Sentinel-1 Synthetic Aperture Radar (SAR) data with a high repeat cycle. A total of approximately 6000 Level-1 scenes were obtained that covered all large lakes (> 50 km2) in the study area. The images were extracted from stripmap (SM) and interferometric wide swath (IW) modes that had a pixel spacing of 40 m in the range and azimuth directions. The lake boundaries extracted from Sentinel-1 data using the algorithm developed in this study were in good agreement with in-situ measurements of lake shoreline, lake outlines delineated from the corresponding Landsat images in 2015 and lake levels for Qinghai Lake. Upon analysis, it was found that the seasonal cycles of lakes exhibited drastically different patterns across the TP. For example, large size lakes (> 100 km2) reached their peaks in August−September while lakes with areas of 50−100 km2 reached their peaks in early June−July. The peaks of seasonal cycles for endorheic lakes were more pronounced than those for exorheic lakes with flat peaks, and glacier-fed lakes with additional supplies of water exhibited delayed peaks in their seasonal cycles relative to those of non-glacier-fed lakes. Large-scale atmospheric circulation systems, such as the westerlies, Indian summer monsoon, transition in between, and East Asian summer monsoon, were also found to affect the seasonal cycles of lakes. The results of this study suggest that Sentinel-1 SAR data are a powerful tool that can be used to fill gaps in intra-annual lake observations.
ZHANG Yu, ZHANG Guoqing
This data set uses SMMR (1979-1987), SSM / I (1987-2009) and ssmis (2009-2015) daily brightness temperature data, which is generated by double index (TB V, SG) freeze-thaw discrimination algorithm. The classification results include four types: frozen surface, melted surface, desert and water body. The data covers the source area of three rivers, with a spatial resolution of 25.067525 km. It is stored in geotif format in the form of ease grid projection. Pixel values represent the state of freezing and thawing: 1 for freezing, 2 for thawing, 3 for deserts, 4 for water bodies. Because all TIF files in the dataset describe the scope of Sanjiangyuan National Park, the row and column number information of these files is unchanged, and the excerpt is as follows (where the unit of cellsize is m): ncols 52 nrows 28 cellsize 25067.525 nodata_value 0
The Qinghai Tibet Plateau is a sensitive region of global climate change. Land surface temperature (LST), as the main parameter of land surface energy balance, characterizes the degree of energy and water exchange between land and atmosphere, and is widely used in the research of meteorology, climate, hydrology, ecology and other fields. In order to study the land atmosphere interaction over the Qinghai Tibet Plateau, it is urgent to develop an all-weather land surface temperature data set with long time series and high spatial-temporal resolution. However, due to the frequent cloud coverage in this region, the use of existing satellite thermal infrared remote sensing land surface temperature data sets is greatly limited. Compared with the daily 1 km spatial resolution all-weather land surface temperature data set (2003-2018) V1 in Western China released in 2019, this data set (V2) adopts a new generation method, namely satellite thermal infrared remote sensing reanalysis data integration method (RTM) based on the new land surface temperature time decomposition model. The main input data of the method are Aqua MODIS LST products and GLDAS data, and the auxiliary data include vegetation index and surface albedo provided by satellite remote sensing. This method makes full use of the high frequency and low frequency components of land surface temperature and the spatial correlation of land surface temperature provided by satellite thermal infrared remote sensing and reanalysis data. The evaluation results show that the land surface temperature data set has good image quality and accuracy, which is not only completely seamless in space, but also highly consistent with MODIS LST products widely used in the current academic circles in amplitude and spatial distribution. When MODIS LST was used as the reference value, the mean deviation (MBE) of the data set in daytime and nighttime was -0.28 K and -0.29 K respectively, and the standard deviation (STD) of the deviation was 1.25 K and 1.36 K respectively. The test results based on the measured data of six stations in the Qinghai Tibet Plateau and Heihe River Basin show that under clear sky conditions, the data set is highly consistent with the measured LST during the day / night, with R2 of 0.93 ~ 0.97 / 0.93 ~ 0.98; MBE of -0.42 ~ 0.25 K / - 0.35 ~ 0.19 K; RMSE of 1.03 ~ 2.28 K / 1.05 ~ 2.05 K; under non clear sky conditions, the MBE of the data set during the day / night is -0.55 ~ 1.42 K / - 0.46 ~ 1.27 K. The RMSE was 2.24-3.87 K / 2.03-3.62 K. Compared with the V1 version of the data, the two kinds of all-weather land surface temperature show the characteristics of seamless (i.e. no missing value) in the spatial dimension, and in most areas, the spatial distribution and amplitude of the two kinds of all-weather land surface temperature are highly consistent with MODIS land surface temperature. However, in the region where the brightness temperature of AMSR-E orbital gap is missing, the V1 version of land surface temperature has a significant systematic underestimation. The mass of trims land surface temperature is close to that of V1 version outside AMSR-E orbital gap, while the mass of trims is more reliable inside the orbital gap. Therefore, it is recommended that users use V2 version. The time span of this data set is from 2000 to 2019, and it will be updated continuously; the temporal resolution is twice daily (corresponding to the two transit times of aqua MODIS in the day and night respectively); the spatial resolution is 1 km. In order to facilitate the majority of colleagues to carry out targeted research around the Qinghai Tibet Plateau and its adjacent areas, and reduce the workload of data download and processing, the coverage of this dataset is limited to Western China and its surrounding areas (72 ° e-104 ° e, 20 ° n-45 ° n) with the Qinghai Tibet Plateau as the core. Therefore, this dataset is abbreviated as trims lst-tp (thermal and reality integrating medium resolution spatial seam LST – Tibetan Plateau) for user's convenience.
ZHANG Xiaodong, ZHOU Ji, TANG Wenbin, DING Lirong, MA Jin
The Tibetan Plateau Glacier Data –TPG2017 is a glacial coverage data on the Tibetan Plateau from selected 210 scenes of Landsat 8 Operational Land Imager (OLI) images with 30-m spatial resolution from 2013 to 2018, among of which 90% was in 2017 and 85% in winter. Therefore, 2017 was defined as the reference year for the mosaic image. Glacier outlines were digitized on-screen manually from the 2017 image mosaic, relying on false-colour image composites (RGB by bands 654), which allowed us to distinguish ice/snow from cloud. Debris-free ice was distinguished from the debris and debris-covered ice by its higher reflectance. Debris-covered ice was not delineated in this data. The delineated glacier outlines were compared with band-ratio (e.g. TM3/TM5) results, and validated by overlapping them onto Google Earth imagery, SRTM DEM, topographic maps and corresponding satellite images. For areas with mountain shadows and snow cover, they were verified by different methods using data from different seasons. For glaciers in deep shadow, Google EarthTM imagery from different dates was used as the reference for manual delineation. Steep slopes or headwalls were also excluded in the TPG2017. Areas that appeared in any of these sources to have the characteristics of exposed ground/basement/bed rock were manually delineated as non-glacier, and were also cross-checked with CGI-1 and CGI-2. Steep hanging glaciers were included in TPG2017 if they were identifiable on images in all other three epochs (i.e. TPG1976, TPG2001, and TPG2013). The accuracy of manual digitization was controlled within one half-pixel. All glacier areas were calculated on the WGS84 spheroid in an Albers equal-area map projection centred at (95°E, 30°N) with standard parallels at 15°N and 65°N. Our results showed that the relative deviation of manual interpretation was less than 3.9%.
This dataset was derived from long-term daily snow depth in China based on the boundary of the three-river-source area. The snow depth ranges from 0 to 100 cm, and the temporal coverage is from January 1 1980 to December 31 2020. The spatial and temporal resolutions are 0.25o and daily, respectively. Snow depth was produced from satellite passive microwave remote sensing data which came from three different sensors that are SMMR, SSM/I and SSMI/S. Considering the systematic bias among these sensors, the inter-sensor calibrations were performed to obtain temporal consistent passive microwave remote sensing data. And the long-term daily snow depth in China were produced from this consistent data based on the spectral gradient method.For header file information, refer to the data set header.txt.
Gf-2 satellite is the first civil optical remote sensing satellite independently developed by China with a spatial resolution better than 1 meter. It is equipped with two high-resolution 1-meter panchromatic and 4-meter multi-spectral cameras, and the spatial resolution of the sub-satellite can reach 0.8 meters. This data set is the remote sensing image data of 6 jing gaofen-2 satellite in 2017.The folder list is: GF2_PMS1_E100.5_N37.2_20171013_L1A0002678101 GF2_PMS1_E100.5_N37.4_20171013_L1A0002678097 GF2_PMS1_E100.6_N37.6_20171013_L1A0002678096 GF2_PMS2_E100.3_N37.4_20170810_L1A0002534662 File naming rules: satellite name _ sensor name _ center longitude _ center latitude _ imaging time _L****
China Centre for Resources Satellite Data and Application
Sentine-1 SAR data were used to monitor the permafrost of Biuniugou in Heihe River Basin of Qinghai-Tibet Plateau. Based on the Sentine-1 SAR image of Bison Valley from 2014 to 2018, the active layer thickness in the study area was estimated by using the small baseline set time series InSAR (DSs-SBAS) frozen soil deformation monitoring method based on distributed radar target, combined with SAR backscattering coefficient, MODIS surface temperature and Stefan model. The results show that the thickness of active layer is between 0.8 m and 6.6 m, with an average of about 3.3 M. It is of great significance to carry out large-scale and high-resolution monitoring.
The “Long-term series of daily global snow depth” was produced using the passive microwave remote sensing data. The temporal range is 1979~2017, and the coverage is the global land. The spatial resolutions is 25,067.53 m and the temporal resolution is daily. A dynamic brightness temperature gradient algorithm was used to derive snow depth. In this algorithm, the spatial and temporal variations of snow characteristics were considered and the spatial and seasonal dynamic relationships between the temperature difference between 18 GHz and 36 GHz and the measured snow depth were established. The long-term sequence of satellite-borne passive microwave brightness temperature data used to derive snow depth came from three sensors (SMMR, SSM/I and SSMI/S), and there is a certain system inconsistency among them. So, the inter-sensor calibration was performed to improve the temporal consistency of these brightness temperature data before snow depth derivation. The accuracy analysis shows that the relative deviation of Eurasia snow depth data is within 30%. The data are stored as a txt file every day, each file is a 1383*586 snow depth matrix, and each snow depth represents a 25,067.53m* 25,067.53m grid. The projection of this data is EASE-Grid, and following is the file header which describes the projection detail. File header: ncols 1383 nrows 586 xllcorner -17334193.54 yllcorner -7344787.75 cellsize 25,067.53 NODATA_value -1
CHE Tao, LI Xin, DAI Liyun
Global warming and human activities have led to the degradation of permafrost and the collapse of permafrost, which have seriously affected the construction of permafrost projects and the ecological environment. Based on high-resolution satellite images, the permafrost of oboling in Heihe River Basin of Qinghai Tibet Plateau is taken as the research area, and the object-oriented classification technology of machine learning is used to extract the thermal collapse information in the research area. The results show that from 2009 to 2019, the number of thermal collapse increased from 12 to 16, and the total area increased from 14718.9 square meters to 28579.5 square meters, nearly twice. The combination of high spatial resolution remote sensing and object-oriented classification method has a broad application prospect in the monitoring of thermal thawing and collapse of frozen soil.
This dataset is the data of human activities in the key areas of Qilian Mountain in 2018, spatial resolution 2m. This dataset focuses on mine mining, urban expansion, cultivated land development, hydropower construction, and tourism development in the key areas of Qilian Mountain.Through high-resolution remote sensing images, compare the changes before and after the statistics. For the maps of the landforms in the Qilian Mountains, check and verify them one by one; re-interpret the plots that are suspicious of the map; collect the relevant data in the field that cannot be reflected by the images, check and correct the location. At the same time, unified input and editing of map attribute information. Generating a data set of human activities in the key areas of the Qilian Mountains in 2018.
QI Yuan，ZHANG Jinlong，JIA Yongjuan，ZHOU Shengming，WANG Hongwei
This dataset is blended by two other sets of data, snow cover dataset based on optical instrument remote sensing with 1km spatial resolution on the Qinghai-Tibet Plateau (1989-2018) produced by National Satellite Meteorological Center, and near-real-time SSM/I-SSMIS 25km EASE-grid daily global ice concentration and snow extent (NISE, 1995-2018) provided by National Snow and Ice Data Center (NSIDC, U.S.A). It covers the time from 1995 to 2018 (two periods, from January to April and from October to December) and the region of Qinghai-Tibet Plateau (17°N-41°N, 65°E-106°E) with daily product, which takes equal latitude and longitude projection with 0.01°×0.01° spatial resolution, and characterizes whether the ground is covered by snow. The input data sources include daily snow cover products generated by NOAA/AVHRR, MetOp/AVHRR, and alternative to AVHRR taken from TERRA/MODIS corresponding observation, and snow extent information of NISE derived from observation by SSM/I or SSMIS of DMSP satellites. The processing method of data collection is as following: first, taking 1km snow cover product from optical instruments as initial value, and fully trusting its snow and clear sky without snow information; then, under the aid of sea-land template with relatively high resolution, replacing the pixels or grids where is cloud coverage, no decision, or lack of satellite observation, by NISE's effective terrestrial identification results. For some water and land boundaries, there still may be a small amount of cloud coverage or no observation data area that can’t be replaced due to the low spatial resolution of NISE product. Blended daily snow cover product achieves about 91% average coincidence rate of snow and non-snow identification compared to ground-based snow depth observation in years. The dataset is stored in the standard HDF4 files each having two SDSs of snow cover and quality code with the dimensions of 4100-column and 2400-line. Complete attribute descriptions is written in them.
ZHENG Zhaojun, CAO Guangzhen
Snow cover dataset is produced by snow and cloud identification method based on optical instrument observation data, covering the time from 1989 to 2018 (two periods, from January to April and from October to December) and the region of Qinghai-Tibet Plateau (17°N-41°N, 65°E-106°E) with daily product, which takes equal latitude and longitude projection with 0.01°×0.01° spatial resolution, and characterizes whether the ground under clear sky or transparent thin cloud is covered by snow. The input data sources include AVHRR L1 data of NOAA and MetOp serials of satellites, and L1 data corresponding to AVHRR channels taken from TERRA/MODIS. Decision Tree algorithm (DT) with dynamic thresholds is employed independent of cloud mask and its cloud detection emphasizes on reserving snow, particularly under transparency cirrus. It considers a variety of methods for different situations, such as ice-cloud over the water-cloud, snow in forest and sand, thin snow or melting snow, etc. Besides those, setting dynamic threshold based on land-surface type, DEM and season variation, deleting false snow in low latitude forest covered by heavy aerosol or soot, referring to maximum monthly snowlines and minimum snow surface brightness temperature, and optimizing discrimination program, these techniques all contribute to DT. DT discriminates most snow and cloud under normal circumstances, but underestimates snow on the Qinghai-Tibet Plateau in October. Daily product achieves about 95% average coincidence rate of snow and non-snow identification compared to ground-based snow depth observation in years. The dataset is stored in the standard HDF4 files each having two SDSs of snow cover and quality code with the dimensions of 4100-column and 2400-line. Complete attribute descriptions is written in them.
ZHENG Zhaojun, CHU Duo