Dragon 4 cooperation executive summary


Stimulating Scientific Exchange in Earth Observation Science and Applications



"The Dragon Programme, a flagship of China-Europe Science & Technology cooperation in Earth observation, has achieved impressive results by bringing together top scientists, training young talents, and sharing satellite data from both sides."

Dr. Huang Wei, Vice Minister MOST-China

"The Dragon programme is an example of international cooperation bringing out the best science from European and Chinese researchers, using both European and Chinese Earth observation data sources."

Prof. Jan Woerner, ESA DG


The Dragon programme started in 2004 as a joint undertaking between ESA and the National Remote Sensing Center of China (NRSCC), under the Ministry of Science and Technology (MOST) of China. This cooperation promotes the use of ESA, Copernicus Sentinels, ESA Third Party Missions and Chinese Earth observation satellite data within China for science and applications.

Dragon 4

Dragon 4 projects across 8 themes
Study areas covering China, Europe, South East Asia and globally


Dragon provides a unique platform for the joint exploitation of Earth observation data from optical, infrared, thermal and microwave sensors for science and application development. These multi-instrument data are used for monitoring the ocean, land and atmosphere and the provision of archive data allows for time series analysis and change detection from the 1990’s to the present day.

Use of High Bit Rate EO Data for Dragon Cooperation (Cumulative)

Dragon has delivered > 200 K high bit rate scenes for science and application development

The scientific community are leading users of Earth observation products, and the scientific research that employs that data is the foundation of most operational satellite applications. With up-and-coming new launches, the volume of satellite data will increase substantially in the near future, providing even more research opportunities.

Dragon exploits data from more than 30 European and Chinese satellites.

Training the next generation of scientists

As well as the formation of joint Sino-European teams and academic exchanges, Dragon organise thematic training courses for Young European and Chinese scientists. Courses provide hands-on training in EO data processing, algorithm and product development for geo-science applications, using ESA open source toolboxes.

Ocean training course poster session, November 2018, Shenzhen University China.


In addition to regular Dragon Symposia, meetings, and workshops, dedicated Dragon sessions are organised at leading international conferences and Symposia, for example ISRSE and ESA Living Planet.
Co-authored results of the research and applications development are delivered at the midterm and final stages of each programme.
Many of the innovative projects emerging from Dragon have resulted in high-level publications in quality peer-reviewed journals and spin-off activities (e.g. long term and global datasets).

Dragon-4 Projects

Through Dragon, not only have applications of remote-sensing expanded, but also high level results in scientific research have been achieved.

Dragon-4 has teamed up scientists to work on 28 projects and 77 sub-projects across a wide range of themes, which address societal issues facing Europe and China today. A selection of these projects are included in this summary document.

Monitoring Greenhouse Gases from Space

Carbon dioxide (CO2) is a major anthropogenic greenhouse gas, but substantial uncertainties remain about the magnitude, location, and durability of natural fluxes.

In particular, regional retrievals of carbon dioxide over China are especially challenging because of persistent cloud cover and heavy aerosol loading.

Through Dragon-4, a joint Chinese and European team of scientists and young researchers are calibrating TanSAT data using a test site in Finland. The aim is to characterize and improve the TanSat XCO2 observations to reduce the uncertainty of flux estimation and support studies on climate change.

TanSat XCO2 measurements. Initial results are presented for CO2 retrievals for 6 months in 2017. Comparison with other sensors show encouraging results.
Credit: Dragon 4 ID. 32301

Earth Observation Data to support agricultural resource monitoring and management

Sustainable food production remains a pressing challenge. Early warning of disease can lead to a more efficient crop production and improved food security. This project uses optical Earth observation data to help detect and monitor the outbreak of Xylella Fastidiosa, a disease that has affected Italian olive oil production in the Province of Lecce, (Apulia) in Italy.

Olive trees affected by Xyella. Photo taken from the Italian test site in Apulia 2016.

Copernicus Sentinel-2 data has been used to monitor affected olive groves and assess the density of trees affected (trees/ha). The technique relies on the fact that for affected trees the background reflectance dominates the signal as the leaves of the affected trees change colour and die off reducing the tree canopy reflectance.

Spectral reflectance from healthy olive tree compared with spectrum of an infected plant.

A temporal series of the standard deviation (STD) of annual variation of the NDVI (Normalized Difference Vegetation Index) is able to track this phenology change in the pixels.

Temporal behaviour of the NDVI-SD (Standard Deviation) for the pixels corresponding to two of the test sites where Xylella presence has been detected.
Credit: Dragon 4 ID.32275

The higher spatial resolution and improved temporal revisit of Sentinel-2 A and B imagery can be used to assess the olive trees density (trees/ha) and growth cycles.

Red=ch4, Green=ch3, Blue=ch2

Red=ch4, Green=NDVI, Blue=ch2

Portion of a Copernicus Sentinel-2A image from 26/12/2016 (Lecce province, tile 34TBK), covering a Corine Land Cover polygon (yellow polyline) corresponding to class 223 (olive groves). The red box corresponds to the main area, in the polygon, affected by Xylella.
Credit: G. Laneve, unpublished

Subsidence Monitoring Using InSAR

Beijing is one of the most populous cities in the world, with over 21 million people. Land subsidence induced by excessive groundwater withdrawal has caused serious social, geological, and environmental problems in the city. Dragon teams used long-term InSAR observations to monitor and measure this phenomenon.

Copernicus Sentinel-2 image of Beijing, the capital of China.
Credit: contains modified Copernicus Sentinel data (2016), ESA

Spatio-temporal dynamics of land subsidence over Beijing plain area was derived - based on 53 TerraSAR-X images acquired from April 2010 to February 2016 - using Interferometric Point Target Analysis (IPTA) technique.

Map showing two subsidence zones at the east of Chaoyang district. Maximum displacement rate (LOS) reached 140mm/year during 2010-2016, compared to 100mm/year during 2003-2010.
Copyright: TerraSAR-X data, processed by Capital Normal University

Characterisation of time series of land deformation using Earth observation data contributes to a better understanding of the spatio-temporal development of land subsidence.

PS points time-series classification
Credit: Dragon 4 ID. 32248

Land Degradation Surveillance of Drylands in China

Between 1978–2015 China experienced a 347% increase in urban population and a 9.2% increase in annual per capita income of urban households. This explosive growth, along with climate change has left a negative environmental footprint on China’s drylands.

This project used Earth observation and climate data to assess how vegetation is influenced by anthropogenic and climate factors.

Monitoring and predicting land condition trends presents a highly complex challenge.

Novel approaches to trend-mapping include the two-dimensional implementation of Rain Use Efficiency (2dRUE) for land degradation assessment.

The results show that 11.5% of drylands are degrading with over 20% in a state of flux.

Land condition trends for China Drylands after 2dRUE implementation.
Credit: Dragon 4 ID. 32396

Significant Time-coefficients frequency after random-stratified sampling for land-uses in China drylands.
Credit: Dragon 4 ID. 32396

A more recent study aims at developing a useful method for vegetation parameters retrieval in dryland based on both European and Chinese EO data and compares the application ability of both types of EO data in the extraction of sparse vegetation above-ground biomass in dryland.

Distribution of AGB in Otingdag sandyland based on on Sentinel-2 MSI data (10m).
Credit: Sun.Bin et al. 2019 ESA Living Planet Symposium

Distribution of AGB in Otingdag sandyland based on GF-1 WFV data (16m).
Credit: Sun.Bin et al. 2019 ESA Living Planet Symposium

Forest mapping using multi-temporal and multi-resolution EO sensors

Forests play a crucial role in Earth’s carbon cycle. Mapping key parameters as indicators of forest condition - such as Forest height, Crown closure, Leaf Area Index, and Above Ground Biomass - is essential for both forest management and ecological assessment.

This project demonstrated forest parameter retrievals - using different EO techniques - to support forest ecosystems mapping within China and neighbouring regions.

Forest parameter retrievals were performed using EO data from optical, SAR, 3D radiometry, as well as from simulated results based on airborne and space borne data.

Height map masking areas with simulated bias greater than 20%
Credit: Dragon 4 ID. 31470

Multi-track Tandem-X InSAR data was combined for forest height mapping in regions of steep topography.

Simulated forest height inversion bias (%)
Credit: Dragon 4 ID. 31470

Forest height map by combing two scenes of Tandem-X InSAR data
Credit: Dragon 4 ID. 31470

Validation data confirms that techniques appear to have been effective for retrievals over difficult sloping mountainous terrain.

Synergistic Monitoring of Ocean Winds, Waves and Storm Surges from Multi-sensors

The monitoring, mapping and forecasting of extreme events such as hurricanes or storm surges, has been a focus of research in both China and Europe for marine meteorology and climate and environment studies.

Dragon projects have provided an excellent opportunity for Chinese and European ocean research communities to utilize satellite remote sensing data from China, ESA and Third Party Missions (TPM) to actively monitor ocean swell, wind and other relevant parameters.

The animation shows the global distribution, propagation and direction of ocean swell across the world’s oceans. This is derived from S1A and S1B wave mode data over a 10 day period. The initial research was started with Envisat ASAR and has been developed further with the improved temporal resolution of the Sentinel-1A and B satellites operating 180 degrees apart.
Credit: Collecte Localisation Satellites (CLS), 2017

Similar to S1 A/B SARs, Chinese researchers have been investigating GF-3 SAR for ocean swell tracking. The early results are promising but are limited in temporal and spatial coverage as only GF-3 SAR is operating.

Significant Wave Heigh (SWH) algorithm development from Wave Mode Data provided by the Chinese Satellite Gaofen-3
Credit: Dragon 4 ID. 32249, Wang, H., et al., Remote Sens. 2018, 10(3), 363

Monitoring from Space for Ocean and Coast Sustainability

The project team have developed a 10 year dataset that consists of ocean wave integral parameters of significant wave height (SWH) (Figure 1) and mean wave period (MWP) data derived from the Advanced Synthetic Aperture Radar (ASAR) onboard the ENVISAT satellite over its full life cycle (2002-2012) covering the global ocean. Both parameters are calibrated and validated against buoy data. Cross-validation between the ASAR SWH and radar altimeter (RA) data is also performed to ensure that the SAR-derived wave height data are of the same quality as the RA data (see cross calibration Figure 2). The data set can be down loaded on-line.

Figure 1. Global sea state product based on the ten-year ASAR wave mode data
Credit: Li, X.; Huang, B. A Global Sea State Dataset from Spaceborne Synthetic Aperture Radar Wave Mode Data. Preprints 2020, 2020010200 (doi: 10.20944/preprints202001.0200.v1).

Figure 2. In the figure the calibrated ASAR SWH also displays good agreement with the calibrated JASON-1 SWH, with bias, RMSE, correlation coefficient and S.I. values of 0.18 m, 0.53 m, 0.93 and 16.64%, respectively.
Credit: As the above figure

Earth Observations for Geohazard Monitoring and Risk Assessment

Landslides are a widespread hazard worldwide and present an increasing socio-economic risk as vulnerable populations grow in areas of seismic hazard, especially across India-Asia.

Under this Dragon project, joint teams mapped the pre-landslide motion of the landslide, which struck the Xinmo Village, Maoxian County, Sichuan Province in 2017.

Regular over-passes from Copernicus Sentinel-1A and B radar satellites allowed the detection of pre-landslide movements using InSAR data. Optical images were used to detect and map landslide source areas and boundaries.

Although the processing was done after the landslide occurred, it demonstrates the potential of the methodology for landslide risk assessment and warning in the future.

Pre-event movements exhibited in the source area during the period from 14 May to 19 June 2017 for the Xinmo event.
Credit: Professor Zhenhong Li, Newcastle University Dr Keren Dai, Chengdu University of Technology Dr Tengteng Qu, Tongji University.

Earth Observation for Water Resource and Quality Monitoring

The Poyang Lake in Jiangxi province is the largest freshwater lake in the country and is globally important as an international conservation area for bird species. The basin is also one of China's most important rice-producing regions, although local inhabitants must contend with massive seasonal changes in water level.

As a recent example of the collaboration between Europe and China, scientists have identified an overall drop in water level in the Poyang Lake, using multi-scale, multi-temporal optical and SAR data. This information is useful for flood mitigation, habitat mapping, ecological characterisation and measuring the water cycle’s impact on human health.

In this animation, radar images from Europe’s Sentinel-1 satellite mission show the evolution of the lake from July 2015 through to May 2016.

Contains modified Copernicus Sentinel data (2015- 2016), processed by ESA with SNAP

Annual wetland inundation conditions
Blue areas show maximum inundation during summer monsoon season.(GF-1 data).
Credit: Dragon 4 ID. 3244


High-mountain Asia stretches from the Tien Shan and Hindu Kush in the northwest, to the eastern Himalayas in the southeast. The area is also part of what is known as ‘the third pole’. These high-altitude ice fields contain the largest reserve of freshwater outside the polar regions, providing freshwater for over 1.3 billion people in Asia – nearly 20% of the world’s population. In recent years, rising temperatures have caused rapid melting.

Scientists need to understand what regulates glacial flow speed to predict how meltwater will affect the region’s supply of freshwater in the future, and how meltwater adds to sea-level rise.

This long term study investigates glacier mass balance in the Third Pole Environment (TPE). Results reveal a complex distribution of changes across the region, but generally in the 5 year period 2012 to 2017, the rate of loss was higher compared to the previous 12 years.

Longer term monitoring is required to confirm the trends and if there is indeed acceleration in glacier loss as a result of regional / global warming.

Glacier height changing rates for 2000 to 2012
Credit: Dragon 4 ID. 32388

Glacier height changing rates for 2012 to 2017
Credit: Dragon 4 ID. 32388


Within the Dragon 4 cooperation, the Third Pole Environment (TPE) has been the subject of a long term study of climate change and land / atmosphere exchanges at different temporal and spatial scales. As part of this long term observation, the ESA L-band microwave radiometer (ELBARA-III ) has been deployed in situ by a joint Chinese and European team at the Maqu site on the eastern Tibetan Plateau. After several years of continuous operation, a unique multi-year dataset has been collected. Fig. 1 shows the radiometer in position and the various angular configurations deployed.

Fig. 1 A schematic overview of the ELBARA-III tower setup (top panel) and the footprints (bottom panel). The footprints vary from 3.31 m2 to 43.64 m2 for incidence angle from 40° to 70°. The half-axes of the elliptic footprint are indicated as a and b for a given incidence angle θi (top panel) and the projected ground distances from the radiometer to the closest- and the farthest-side of the elliptic footprints at -3 dB sensitivity of the antenna are indicated as dmin and dmax (bottom panel). The locations of the installed in situ soil moisture and soil temperature sensors are indicated as SMST_Z and SMST_LC. The fence (25m × 45m) is not drawn to scale.

The dataset contains measurements of L-band brightness temperature in horizontal and vertical polarization, profiles of soil moisture and soil temperature, turbulent heat fluxes, and meteorological data from the beginning of 2016 till August 2019. Auxiliary vegetation and soil texture information have also been collected in dedicated campaigns. The system was installed by the team of Dragon 4 principal investigator Professor Jun Wen (CAREERI/CAS).
This dataset can be used to:

  1. validate the Soil Moisture and Ocean Salinity (SMOS) and Soil Moisture Active Passive (SMAP) satellite based observations and retrievals;
  2. verify radiative transfer model assumptions and validate land surface model and reanalysis outputs;
  3. retrieve soil properties, as well as to quantify land and atmosphere exchanges of energy, water and carbon;
  4. help to reduce discrepancies and uncertainties in current Earth System Models (ESM) parameterizations.

The measurement cases include winter, pre-monsoon (Fig. 2), monsoon (Fig. 3) and post-monsoon periods.

Fig. 2 Seasonal variations of the Maqu ELBARA-III radiometry dataset for pre-monsoon season (late March to late June) in 2018. Plotted are soil moisture at 2.5 cm depth (SM_2.5 cm), albedo, ground surface temperature (TG), air temperature (Tair), soil temperature at 2.5 cm depth (ST_2.5 cm), the nominal freezing point as a reference (273.15 K), and the brightness temperature in horizontal and vertical polarization (,) at 40º incidence angle and precipitation (Pre). Trend lines (dashed lines) are added to SM_2.5 cm and (,) time series to assist interpretation.

Fig. 3 Same as Fig 2 but for monsoon period (late June to mid-August) in 2018.

The text in this article and the 3 figures have been taken and summarised from: Su, Z., Wen, J., Zeng, Y. et al. Multiyear in-situ L-band microwave radiometry of land surface processes on the Tibetan Plateau. Sci Data 7, 317 (2020).
For the print version of the paper and the data set access:

Discover more