Conference Agenda

Monitoring of agricultural crops is of vital importance for efficient food production and sustainable development of agricultural sector. The main objective of the present study was to evaluate the possibilities for crop recognition, using Sentinel-2 data in Parvomay test area in Bulgaria. For that purpose, the classification methods Support Vector Machines (SVM) and Random Forest (RF) were evaluated. These methods were applied to satellite multispectral data acquired by the Sentinel-2 satellites, for the growing season 2020-2021. Main crops grown in the research area are winter wheat, rapeseed, sunflower and maize. In accordance with their development cycles, we developed temporal image composites for the suitable moments of time when each crop is most accurately distinguished from other crops. Ground truth data was available from the integrated administration and control system (IACS) - a vector database containing information about crops sown in individual agricultural parcels for the territory of Bulgaria. The IACS data was used for both training the classifiers and accuracy assessments of the final maps.

The sentinel series satellite in Europe and GF series satellite in China are providing the data options for agricultural monitoring as well as enhancing the capability of agricultural monitoring in general. This project has made the great progress since the inception of this project. In Heilongjiang Sanjiang site, the 15 farms in the plain were selected as the study area. The sentinel 2 time series images during the growing season of 2020 to 2022 were downloaded from the official website https://scihub.copernicus.eu. The samples were collected from field by the queries with farmers and field survey. The Random Forest algorithm was applied to implement the classification. Finally, the field preparation, different rice planting methods and the harvested fields were identified from the images. In Jinzhong site, a heave flood happened in early Oct. 2021. In the middle Oct., the research team was sent to the field and collected the crop types in the field, crop types in the flooded water and others harvested already in the field. The 3 good images in September were available to retrieved the crop types before the flood. The image on Oct 17, 2021 was used to retrieved the crop types and flood crops. Then the results were layered and used to analysis the flood situation. The field situation in winter also was investigated. The bare fields, fruit tree fields, vegetated fields and others were classified with a Random Forest classifier. In Shanxi site, the apple fields and winter wheat/maize fields were also identified from the satellite image. The Venus satellite images are agreed by the team to provide this year. But it seems that there are some issues and the data are not made available right now. Through this joint project and the heavy involvement of young scientists from Europe and China, the satellite data finely processing and information retrieval algorithm is being exchanged and it is expected to bring a step forwards to support agricultural monitoring at fine scale.

Agriculture is the largest consumer of water worldwide, accounting for about 70% of the global fresh water withdrawals. Irrigation efficiency and crop water use efficiency are key concerns for agricultural water management. The objective of the project is to assess irrigation water needs and crop water productivity based on the integrated use of satellite data with high resolution, ground hydro-meteorological data and numerical modelling, which is particularly significant for large un gauged agricultural areas. In such studies, satellite observation-based products or information with high accuracy and continuously spatial and temporal coverage are essential to support monitoring and modelling of agricultural water use and efficiency at farm and basin scales.

The following progresses have been made in the two years of project implementation:

1) FEST-EWB model improvement in evapotranspiration estimates over crop trees areas for optimizing crop irrigation efficiency.

Remotely- sensed data at different temporal and spatial resolution of vegetation parameters (leaf area index (LAI), fractional coverage of vegetation, albedo) which are used as inputs to hydrological model are obtained at high spatial and temporal resolution merging Sentinel 2 data with Landsat 8 and 9. Satellite LST is further retrieved from Landsat 8 and 9 at 30 m spatial resolution to be used for the hydrological model calibration.

Indeed, the energy–water balance FEST-EWB model (flash flood event-based spatially distributed rainfall–runoff transformation—energy–water balance model) computes continuously in time and is distributed in space soil moisture (SM) and evapotranspiration (ET) fluxes solving for a land surface temperature that closes the energy–water balance equations.

The model can work both in the single-source version (Corbari et al., 2011) and in its double-source one (Paciolla et al., 2022). The former uses a single balance equation for the pixel, while the latter, although requiring the same amount of input data, distinguishes between the vegetated and non-vegetated areas in the pixel. This improvement may result crucial in analysing heterogeneous agricultural areas such as those of fruit tree crops, where arboreal canopy is interspersed with bare soil or low-cut grass land cover.

The FEST-EWB model uses as input data the meteorological information, soil type data and satellite vegetation information (as LAI, fv and albedo from Sentinel-2 data). The comparison between modelled and observed LST was used to calibrate the model soil parameters with a newly developed pixel to pixel calibration procedure. The effects of the calibration procedure were analysed against ground measures of soil moisture and evapotranspiration.

Preliminary results of the amount of precision irrigation water supply and the Evapotranspiration deficit at pixel scale will also be shown.

The FEST-EWB modelling approach has been applied, both in its single- and two-source structure, over field sites featuring walnut (Italy, 2019-21), vineyard (Spain, 2012 and Italy, 2008) and pear trees (Italy, 2022).

2) Estimation of Cropland Gross Primary Production by Integrating Water Availability Variable in Light-Use-Efficiency Model. A light-use-efficiency (LUE) model for cropland Gross Primary Production (GPP) estimation, named EF-LUE, driven by remote sensing data, was developed by integrating evaporative fraction (EF) as limiting factor accounting for soil water availability. Model parameters were optimized using CO2 flux measurements by eddy covariance system from flux tower sites, and the optimized parameters were spatially extrapolated according to climate zones for global cropland GPP estimation in 2001–2019. According to the site-level evaluation, the proposed EF-LUE model explained 82% of the temporal variations of GPP across crop sites, which is much better than original LUE model. The fraction of absorbed photosynthetically active radiation (FAPAR) data from the Copernicus Global Land Service System (CGLS) GEOV2 dataset, EF from the ETMonitor model, and meteorological forcing variables from ERA5 data were applied to EF-LUE model For the global cropland GPP estimation. The results showed overall better accuracy than other existing global GPP products, and it could capture the significant negative GPP anomalies during drought or heat-wave events, indicating its ability to express the impacts of the soil water stress on cropland GPP. This work was published in Remote Sensing.

3) Calibration and validation of SWAT model in ungauged basins. To meet the challenge of model calibration and validation in ungauged basins, we developed a new approach to calibrate SWAT hydrological model using remote sensing evapotranspiration data. This procedure is designed to deal with spatially variable parameters and estimates either multiplicative or additive corrections applicable to the entire model domain, which limits the number of unknowns while preserving spatial variability. Different remote sensing ET datasets were tested in model calibration, i.e., ETMonitor, GLEAM, SSEBop, and WaPOR, and the calibration results based on ETMonitor ET showed the best performance with R² > 0.9 and Nash–Sutcliffe Efficiency (NSE) > 0.8. The calibrated SWAT model simulation was validated against remote sensing data on total water storage change with acceptable performance (R² = 0.57, NSE = 0.55), and remote sensing soil moisture data from ESA CCI product with R² and NSE greater than 0.85. Based on the proposed procedure, a case study focused on Lake Chad Basin in Africa was carried out and the paper was published in Remote Sensing.

4) A multi-temporal and multi-angular approach for systematically retrieving soil moisture and vegetation optical depth from SMOS data. Microwave retrieval of soil moisture is an underdetermined issue, as microwave emission from the land surface is affected by various surface parameters. Increasing observation information is an effective means to make retrievals more robust. We developed a multi-temporal and multi-angular (MTMA) approach using SMOS (Soil Moisture and Ocean Salinity) satellite L-band data for systematically retrieving vegetation optical depth (VOD_p , p indicates the polarization ─ H: horizontal, V: vertical), effective scattering albedo (ω_{p_eff}), soil surface roughness (Z_{p_s}), and soil moisture (SM_p). The retrieved by MTMA shows generally good agreement with in-situ measurements with overall correlation coefficients of larger than 0.75, and the overall ubRMSE of (0.050 m³/m³) and (0.054 m³/m³) which are lower than that of SMOS-IC Version 2 (V2) (referred to as SMOS-IC) (0.058 m³/m³) and SMOS-L3 (SMOS Level 3) (0.066 m³/m³) products. This work was published in Remote Sensing of Environment.

Net primary productivity (NPP) and Net ecosystem productivity (NEP) plays an important role in understanding ecosystem function and the global carbon cycle. In this study, based on the long-time series NPP data products, the global NPP spatiotemporal dynamic change analysis is realized by using the methods of trend analysis, catastrophe analysis and wavelet analysis. And then, the change trend and law of NPP in different regions are analyzed, which can provide a reference for the study of global carbon cycle. In addition, on this basis, by coupling the improved CASA model, geoscience statistical model (GSMSR) and the soil respiration–soil heterotrophic respiration (R_s-R_h) relationship model, the NEP of terrestrial ecosystems at intercontinental, national and regional scales is estimated, which provides scientific basis and technical conditions for carbon balance assessment and carbon neutralization policy formulation at different scales.

The first two years of this project focused on retrieval of biophysical variables of vegetation, estimation of bare soil properties, and monitoring pest and diseases. The remote sensing data used include non-imaging hyperspectral data and image data from satellites such as Sentinel-2, GF-1 and PRISMA. The field experiments were carried out at Maccarese farm in Central Italy and farms in the Quzhou district in China.

For vegetation, field campaigns were primarily performed contemporary to the Sentinel-2 or PRISMA acquisitions on the Maccarese (IT) farm along the winter wheat and maize growing season. A Decision Trees Ensembles (DTE) model has been developed by training a database of plant canopy spectral and directional reflectance generated by using PROSAIL-PRO. A reduction of dimensionality performed by selecting a subset of PRISMA bands (i.e. excluding water vapor absorption bands and noisy bands) were applied prior to the training phase. DTE model has been selected among the tested algorithms as appears a robust method in dealing with noise, less prone to overfitting and less sensitive to the difference between simulated and real data. LAI, LCC, FCOVER and FAPAR were retrieved on the Maccarese test site with a RMSE value of 1.15 m² m⁻², 11.45 μg cm⁻², 0.16% and 0.16%, respectively. RMSE values are always lower than the ones derived by Sentinel-2 with SNAP. Further steps will consider the retrieval of additional biophysical variables as carotenoid, nitrogen and leaf mass area.

For topsoil characterization the test site is located in the Quzhou district and more than 100 soil samples were collected and analyzed in the lab to assess SOC. PRISMA data preprocessing was similar to the one for vegetation with the inclusion of data pre-treatment i.e. Savitzy–Golay filter (SG), Standard Normal Variate (SNV) and Continuum Removal (CR). PLSR and Cubist algorithms were applied only to bare soil pixel by using a 10-fold cross validation. Results show a significant accuracy for SOC showing a RPD higher than 2. The result for Quzhou reflects the small field size with respect to the PRISMA GSD [30m]. To minimize the field size effects, an optimized pan sharpening procedure by using Sentinel-2 bands and the 5m/pixel PRISMA PAN, is under development.

For what concerns crop threats among which pests & diseases, Italian and Chinese research groups are working on the development of a high resolution crop mapping and a crop early warning to be combined with the pest/diseases monitoring model to accurately estimate its possible effect in terms of production loss and food security. To this purpose, the activity carried out in the framework of Dragon 5 will take into account results coming from other initiatives in the context of the GEO work programme (CROP_PEST_MONITORING and AFRICULTURES). High resolution (10m) crop mapping based on Sentinel-2 image series methodology has been developed and it will be integrated with pest/diseases monitoring model in order to be able to forecast/estimate the potential affected area.

For desert locust presence risk forecasting, we have proposed a dynamic prediction method of desert locust presence risk at Somalia-Ethiopia-Kenya. First, we obtained six static environmental factors, including DEM, land cover, soil sand content, clay content, silt content, and coarse fragment content. Then the time lag variables of four dynamic environmental factors, namely rainfall, soil moisture, NDVI, and LST, were then extracted based on multi-source time-series data. Finally, we have constructed a dynamic remote sensing-based risk forecast model of desert locust presence combining a multivariate time lag sliding window technique and machine learning algorithms. Monthly prediction experiments from February to December 2020 were then conducted, extracting high, medium and low risk areas of desert locust occurrence in the study area. Results demonstrated that the overall accuracy was 77.46%, and the model enables daily dynamic forecasting of desert locust risk up to 16 days in advance, providing early warning and decision support for preventive ground control measures for the desert locust.

The project, therefore, at the Mid Term event, is well aligned to the timetable described at the time of the submission and most of the activities have been started. Moreover, the PhD candidate, selected within the YS program, is working on the project using common data.

The recognition of intra-urban heat islands depends on the capacity to identify, at varying temporal and spatial scales, of variations in urban form and urban fabric. In this paper Earth Observation is used to define Land Surface Temperature, recognize intra-urban heat islands and identify the impact of urban form/fabric characteristics [land cover (greenery), the impervious fraction, the building density and height and the type of construction materials) to surface thermal dynamics. Particular attention is given to the relation between high/medium/low density areas and LST on the one hand, and on the cooling potential of greenery (in the above areas) on the other, in view of defining urban patterns to be taken into consideration once adaptation plans to urban heat are developed. Results are examined against spatial resolution, time of day and season, whereas special attention is given to the impact of climate extremes, namely heat waves, to any recognized patterns as well as to the presence of intra-urban heat islands. Landscape metrics are also used in support of the extraction of urban patterns and the understanding of their links to surface thermal dynamics. Finally, results are rolled out as a climate service for Climate friendly and smart cities.

Urbanization is continuing at an unprecedented rate in many cities across globe. Rapid urbanization poses significant social and environmental challenges, including sprawling informal settlements, increased pollution, urban heat island, loss of biodiversity and ecosystem services, and making cities more vulnerable to disasters. Therefore, timely and accurate information on urban changing patterns is of crucial importance to support sustainable and resilient urban planning and monitoring of the UN 2030 Urban Sustainable Development Goal (SDG). The overall objective of this project is to develop innovative, robust and globally applicable methods, based on Earth observation (EO) big data and AI, for urban land cover mapping and urbanization monitoring.

Using ESA Sentinel-1 SAR, Sentinel-2 MSI and Chinese GaoFen-1 images, the EO-AI4Urban team has developed varous deep learning-based methods for urban mapping and change detection during the past two year. For urban mapping, a novel Domain Adaptation (DA) approach using semi-supervised learning has been developed for urban extraction. The DA approach jointly exploits Sentinel-1 SAR and Sentinel-2 MSI data to improve across-region generalization for built-up area mapping [1]. For urban change detection, several novel methods have been developed including a dual-stream U-Net [2] and a Siamese Difference Dual-Task network with Multi-Modal Consistency Regularization [3]. Further, a high-resolution feature difference attention network (HDANet) is proposed to detect changes using the Siamese network structure. To tackle the loss of the spatial features of buildings caused by the multiple successive down-sampling operations using fully convolutional networks (FCNs), a multi-resolution parallel structure is introduced in HDANet, and the image information with different resolutions is comprehensively employed Using Landsat time series, a machine learning method was developed to map built-up areas and to analyze changes during 2000 to 2020 [4]. In addition, techniques for the characterization of urban changes exploiting sequences of SAR and/or optical data, with focus on Chinese cities are further developed. The temporal fitting of the coherence sequence using Sentinel-1 SAR data has been explored. This method has shown potential to discriminate and label accordingly the changes in time for intra-urban areas [5]. Another approach has been developed to focus on an automatic labeling of these temporal sequences based on a statistical analysis of the changes in time, clustered into different behaviors (e. g., stable clusters, fast changes, etc.). This approach has been tested in Chinese cities and achieved significant results [6]. In consideration of the difficulties to find an object of interest in large-scale scenes, a deep learning-based visual grounding method in remote sensing images has been developed to automatically locate a target object from large-scale scenes by a language description [7]. A robust cloud detection approach has also been developed [8]. Finally, using Landsat time series, a machine learning method was developed to map built-up areas and to analyze changes during 2000 to 2020 [9].

Experiments conducted on a test set comprised of sixty representative sites across the world showed that the proposed DA approach achieves strong improvements upon fully supervised learning from Sentinel-1 SAR data, Sentinel-2 MSI data and their input-level fusion. The fusion DA offers great potential to be adapted to produce easily updateable human settlements maps at a global scale [1]. Using the Onera Satellite Change Detection (OSCD) dataset, the results showed that the dual-stream U-Net outperformed other U-Net-based approaches together with SAR or optical data and feature level fusion of SAR and optical data [2]. Using bi-temporal SAR and MSI image pairs as input, the Siamese Difference Dual-Task network with Multi-Modal Consistency Regularization have been tested in the 60 sites of the SpaceNet7 dataset. The method achieved higher F1 score than that of several supervised models when applied to the sites located outside of the source domain [3]. Using the LEVIR-CD dataset, the results confirmed that HDANet can improve the differential feature representation for building change detection [4].

Using Landsat time series, the results show that machine learning method could extract built-up areas effectively. To analyze urbanization in 13 cities in the Beijing–Tianjin–Hebei region, SDG indicator 11.3.1, the ratio of land consumption rate to population growth rate (LCRPGR) is calculated and the results show that the LCRPGR in Beijing–Tianjin–Hebei region fluctuated from 1.142 in 2000–2005 to 0.946 in 2005–2010, 2.232 in 2010–2015 and 1.538 in 2015–2020. Apart from the megacities of Beijing and Tianjin, the LCRPGR values were greater than 2 in all cities in the region after 2010, indicating inefficient urban land use [9].

References:

[1] Hafner, S., Y. Ban and A. Nascetti, 2022. Unsupervised Domain Adaptation for Global Urban Extraction Using Sentinel-1 and Sentinel-2 Data. Remote Sensing of Environment (in press).

[2] Hafner, S., A. Nascetti, H. Azizpour and Y. Ban, 2022. Sentinel-1 and Sentinel- 2 Data Fusion for Urban Change Detection Using a Dual Stream U-Net. IEEE Geoscience and Remote Sensing Letters, vol. 19, pp. 1-5.

[3] Hafner, S., Y. Ban and A. Nascetti, 2022. Multi-Modal Consistency Regular- ization Using Sentinel-1/2 Data for Urban Change Detection. Submitted to ISPRS Journal of Photogrammetry and Remote Sensing.

[4] Wang, X., J. Du, K. Tan et al., 2022. A High-Resolution Feature Difference Attention Network for Change Detection. To be submitted.

[5] Che, M. and P. Gamba, Temporal and Spatial Change Pattern Recognition by Means of Sentinel-1 SAR Time-Series. IGARSS 2020 - 2020 IEEE International Geoscience and Remote Sensing Symposium, 2020, pp. 160-163.

[6] Che, M., A. Vizziello, P. Gamba, "Spatio-temporal Change Mapping with Coherence Time-Series", submitted to IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing.

[7] Sun, Y., S. Feng, X. Li, Y. Ye, J. Kang, X. Huang. 2022. Visual Grounding in Remote Sensing Images. Proceedings of the ACM International Conference on Multimedia, 2022.

[8] Li, X., X. Yang, X. Li, S. Lu, Y. Ye, Y. Ban. 2022. GCDB-UNet: A novel robust cloud detection approach for remote sensing images. Knowledge-Based Systems, 238: 107890.

[9] Zhou, M., Lu, L., Guo, H., Weng, Q., Cao, S., Zhang, S., & Li, Q. (2021). Urban Sprawl and Changes in Land-Use Efficiency in the Beijing–Tianjin–Hebei Region, China from 2000 to 2020: A Spatiotemporal Analysis Using Earth Observation Data. Remote Sensing, 13(15).

In recent years, deep neural networks have become a hype tool for various optical and SAR remote sensing image interpretations including object detection, recognition, land use land cover classification, change detection, and multi-temporal image analysis. However, in real-world scenarios, there still are many challenges, such as lack of large annotation data, heterogeneous data transferring, open set recognition, etc. In this talk, we will present our latest efforts to address such problems with the support of the Dragon Project, including： 1) Self-supervised learning for remote sensing image classification and recognition; 2) Heterogeneous SAR domain adaptive object detection; 3) open set recognition and new class discovery for remote sensing images. Moreover, we will introduce an interactive deep learning remote sensing image annotation system and an explainable analysis system for the remote sensing imagery system we are developing.

As an important oceanic physical process, mesoscale eddies play a key role in the processes of ocean mass transport and energy exchange, such as biogeochemical cycles, marine ecosystems, and marine heat balance.

In recent years, these automated ocean eddy identification and tracking algorithms can be divided into two categories: Eulerian- and Lagrangian-based approaches. Based on instantaneous sea surface height (SSH) or velocity field, the major circular structures of mesoscale eddies can be detected, which are called as Eulerian eddies. Previous SSH-based and sea level anomaly-based (SLA-based) methods have shown the best performance due to their ability to avoid extra noise and excess eddy detections. Although SSH-/SLA-based methods worked well at the basin scale, the calculation efficiency decreases distinctly at the global scale, mainly due to the high-order computation complexity on contour iterations. With the improvement of data resolution, existing hardware fails to meet the requirement of long-time-scale global eddy recognition due to the increase of the number of SSH/SLA contours. Thus, based on satellite altimeter data, a highly effective orthogonal parallel algorithm for identifying and tracking global eddies is proposed. Surprisingly, this algorithm is ~100 times faster than the previous SSH-based method on global eddy detection without reducing the accuracy of mesoscale eddy recognition. According to this orthogonal parallel algorithm, the global mesoscale eddy dataset for the past 28 years (1993-2020) was generated, which provides a data foundation for the subsequent study of mesoscale eddies. In terms of the mesoscale eddy dataset, an automatic recognition method of global dipole eddy pairs that consist of two counter-rotating eddies moving together for a period of time within a limited space distance is proposed by using the K–D tree for cutting space. Simultaneously, the transmission modes and characteristics of dipoles are analyzed, including the characteristics of long life, high propagation speed, and entanglement trajectory. In addition, an algorithm named EddyGraph for tracking mesoscale eddy splitting and merging events is proposed and the corresponding data set of eddy trajectories in the Northwest Pacific is available, which would fill the gaps in data sets to support studies on eddy splitting and merging in the Northwest Pacific.

Different from other methods based on instantaneous flow fields, Lagrangian eddies are the cumulative results of the state of the fluid within a given time scale, which can maintain material coherence over the specified time intervals. Firstly, by using the elliptic Lagrangian Coherent Structures, the boundary of a black-hole eddy was extracted based on the data of the geostrophic flow velocity field in the Western Pacific Ocean. Combined with multi-source satellite remote sensing data and in-situ data, the consistency of material transport in the horizontal direction and the coherence of material in the vertical direction of the vortex are analyzed and verified. The results show that the black hole vortex boundary can describe material transport more objectively than the Euler vortex boundary on a longer time scale. Then, the Lagrangian eddies in the western Pacific Ocean are identified and analyzed. By introducing the Niño coefficient, the lag response of the Lagrangian eddy to El Niño is found. Through normalized chlorophyll data, it is observed that Lagrangian eddies can cause chlorophyll aggregation and hole effects. These findings demonstrate the important role of Lagrangian eddies in material transport. Nevertheless, although Lagrangian eddies work well at estimating material transport, the high calculation cost during the integration process has become a bottleneck, especially when the data resolution is improved or the study area is enlarged. Therefore, SLA-based orthogonal parallel detection of global rotationally coherent Lagrangian eddies is built, whose runtime is much faster than that of a previous nonparallel method. Finally, a data set of long-time-scale global Lagrangian eddies is established.

Furthermore, an integrated marine visualization system, named i4Ocean, has been presented. The system is designed and implemented to investigate and study physical marine processes by visualizing and analyzing spatiotemporal marine data. Notably, these actions are realized by providing various GPU-based interaction and visualization techniques for displaying multidimensional data. The system achieves three goals: high visibility, good performance and interactive capabilities. The techniques of z-coordinate calibration and sphere rendering, which restore the most authentic marine environment, provide excellent feedback for oceanographers. The efficient ray sampling technique including a preintegrated transfer function and adaptive sampling methods, increases the rendering efficiency of ocean data. By further introducing a transfer function, users can extract the region of interest in the system and analyze diverse marine phenomena. A data-centric approach was adopted to guide the design of the transfer function by analyzing the scalar field and its properties. The best parameters of the transfer function were obtained to maximize the visibility of important features, which helps to analyze mesoscale eddies of typical ocean phenomena.

Orbital sensors have the capability to provide global repeated observations that can cover the whole world, so they have the potential to observe the earth in a large range, and are widely used in monitoring key phenological events of forest vegetation, monitoring ecosystem environment and other fields. However, due to the limitation of time and space resolution, remote sensing technology faces many challenges in the application of Earth observation. For example, the current satellite image data cannot directly reflect the occurrence and development process of key phenological events. In addition, different methods are used to indirectly estimate ecological indicators to obtain estimation results that are far different, which brings great difficulties to correctly explain and understand the correlation between phenological change and ecosystem and climate change. In a word, the monitoring and research of ecological events based on satellite images need reliable ground reference data, and the lack of relevant ground phenological event observation reference information is the recognized main bottleneck in the field.
In the field of forest phenological monitoring, there are many technical challenges to assess the exact date of phenological events such as the germination and abscission of leaves based on satellite observations. This study aims to explore the feasibility of studying the change characteristics of light reflectivity of vegetation during the occurrence period of key forest phenological events based on the high spatial and temporal resolution long-term forest observation timing of laser radar, and optimizing the monitoring accuracy of satellite phenological events based on the relevant characteristics. The research relies on LiPhe (Lidar Phenological Monitoring) platform and relevant data established by Geospatial Information Research (FGI) of the Finnish National Surveying and Mapping Bureau. In February 2020, LiPhe platform was established in Finland's 111 year old national experimental forest area (Hyytiälä 61°51 ́N,24°17 ́E)。The Liphe platform mainly comprises a Riegl VZ-2000 ground-based lidar scanner (TLS), as well as relevant edge computing and remote control facilities. The scanner is installed on the SMEAR II meteorological observation tower established in 1995, 30 meters from the ground, and the scanning range covers about 10 hectares of forest around the tower. Since April 2020, the scanner will collect data every half an hour and continuously provide point cloud data with high spatial resolution (100 meter distance and 1 cm three-dimensional point distance).

As of October 2021, a year and a half of high spatial and temporal resolution 3D point cloud observation sequence has been obtained.
This study shows the important potential of relevant data as accurate phenological ground observation reference data to support the study of tree level phenological change mechanism and optimize satellite image phenological event monitoring results. With the help of machine learning method, the expected results of the occurrence time of key phenological events in the same forest area based on LiPhe data and satellite data are compared, and the key parameters affecting the prediction of the occurrence time of phenological events are analyzed. The correlation comparison reveals the characteristics of forest light reflectivity changes caused by the dynamic changes of forest structure and the changes of its related physical properties, and demonstrates the optimization role of the relevant characteristics in the interpretation of satellite spectral information phenological events.
In the field of ecological environment quality assessment, scholars have learned the complex and diverse characteristics of the ground ecosystem through in-depth research, and it has natural defects to simply use a single indicator to evaluate the ecological quality. Relying on the advantages of remote sensing data, such as easy access, large area and multi temporal observation, scholars have proposed a variety of comprehensive ecological remote sensing inversion models, coupling a variety of ecological indicators to reflect the quality of the ecological environment. Among them, the RSEI model proposed by Xu Hanqiu couples the four indicators of greenness, humidity, heat and dryness, and uses the principal component analysis method to adaptively determine the weight of the indicators, eliminating the risk of too strong subjective factors of AHP experts' experience to confirm the weight, which has been widely used in the field of ecological remote sensing. This paper studies the ecological environment quality of Shanxi Province in 2013-2019 based on the RSEI model. Specifically, through the Google Earth Engine (GEE) cloud platform coding, the winter RSEI results of 2013-2019 were obtained, and the spatiotemporal differentiation characteristics of the ecological environment in winter in Shanxi Province were studied. The experimental process was mostly based on the GEE cloud platform, which has important research significance for the ecological environment protection in Shanxi Province and even the whole country.

Forests play a critical role in the Earth's ecosystem and strongly impact the environment. Under the threat of global climate change, remote sensing techniques can provide information for a better understanding of the forest ecosystems, early detection of forest diseases, and both rapid and continuous monitoring of forest disasters. This project concerns the topic of ecosystems and spans the subtopics estimation of forest quality parameters and forest and grassland disaster monitoring. The aim is to study and explore the application of multi-source remote sensing technology in forest parameter extraction and forest disaster monitoring, using data fusion of satellite images, drone-based laser scanning and drone-based hyperspectral images. The research contents are tree species classification, forest parameter estimation, and forest insect damage detection.

1. Work performed

We applied for satellite images through ESA and Ministry of Science and Technology of the China. These included images acquired from RADARSAT-2, WorldView-3, Sentinel-1/2, and Gaofen series. These data cover several study areas in China and Sweden, including Gaofeng, Yantai, Fushun, Lu'an, Wangyedian, Genhe and Pu'er in China and Remningstorp in Sweden. Field investigations were carried out in Gaofeng, Fushun, Lu'an and Remningstorp. In these study ares, drone-based multispectral and hyperspectral images and laser scanning data were also acquired and studied.

(1) Satellite images

The satellite images acquired for the study areas in China are:

RADARSAT-2, one image covering Fushun and one image covering Qingyuan.

Sentinel-1, time-series images from 2019 to 2021, covering Gaofeng and Wangyedian.

Sentinel-2, time-series cloud-free images from 2019 to 2021, covering Gaofeng, Lu'an, Wangyedian, Genhe, and Pu'er.

Gaofen-1/2/6, 295 images from 2021 to 2022, covering Gaofen, Genhe and Pu'er.

The satellite images acquired in Remningstorp, Sweden are:

WorldView-3, one SWIR image (June 2021).

Sentinel-2, time-series cloud-free images from 2018 to 2021.

RADARSAT-2, one image each in 2020 and 2021.

Pleiades, one image (29 Apr 2021).

(2) Field investigations data

Field investigation of forest parameters was conducted in Gaofeng, China. The inventory recorded diameter at breast height, tree height, under branch height, and the coordinates of the plots.

Spectral information was collected from healthy and pine nematode-infested forests at different stages in the Fushun and Lu'an study areas.

The forest information of the sample plots in Remningstorp was updated. A controlled experiment was conducted for bark beetle infestation, and the infestation symptoms were recorded.

(3) Technical progress

Tree species classification. We proposed three deep learning models using drone-based hyperspectral data: an improved prototype network (IPrNet), a CBAM-P-Net model of the prototype network combined with an attention mechanism, and a Proto-MaxUp+CBAM-P-Net model of the CBAM-P-Net combined with a data enhancement strategy. We developed ACE R-CNN, an attention mechanism, edge detection and region-based instance segmentation algorithm, to accurately identify individual-tree species using UAV LiDAR and RGB images. The performance of these models was demonstrated in the Gaofeng study area. A tree species classification method based on multi-temporal Sentinel-2 data was developed and compared with the classification using mono-temporal data. The performance was verified at Remningstorp.

Forest parameters extraction. We proposed a mean-shift individual-tree crown segmentation algorithm based on canopy attributes using UAV oblique photography data, and developed an individual-tree biomass estimation model fusing multidimensional features, which has good performance in the Gaofeng study area. We proposed a method to automatically extract high-resolution tree height products by combining ZY-3 stereo images and DEM. We developed a forest aboveground biomass estimation model using Sentinel-2 data and tree height data, which obtained accurate forest aboveground biomass maps in the Wangyedian study area. The potential of using PolSAR data acquired at C- and X-band was investigated to estimate forest aboveground biomass at a test site in southern Sweden. The polarization decomposition method was used to RADARSAT-2 and TerraSAR-X data for estimating forest aboveground biomass.We also investigated the potential of time-series TanDEM-X for monitoring forest growth by using the phase height data. We proposed methods to quantify the effects of thinning and clear-cuts on the phase height and apply the methods on detecting silvicultural treatment.

Detection of forest biotic disturbance. About D. tabulaeformis, we proposed a spectral-spatial classification framework combining drone-based hyperspectral images and RGB images to identify damaged tree crowns. For Bursaphelenchus xylophilus, we analyzed the spectral characteristics of two tree species (Pinus tabulaeformis and Pinus koraiensis) in the study areas of Weihai and Fushun during different infection stages. Sensitive bands were selected and a detection model was constructed to identify the infection stages of Bursaphelenchus xylophilus. For European spruce bark beetles (Ips typographus [L.]) infestation, methods of early detecting infestations were proposed using drone-based multispectral images. We investigated how early the infestation can be detected after an attack. We also compared the sensitivity of Sentinel-2 and WorldView-3 SWIR images in detecting early-stage infestations.

(4) Collaborative Research

Co-supervising 1 PhD student.

One joint research paper (accepted for publication in Ecological Indicators), one manuscript, and one published conference paper (IGARSS 2022).

2. Future Plans

(1) Data acquisition

We would like to apply for TanDEM-X and WorldView-3 images covering the study areas of Gaofeng, Genhe, and Pu'er. We plan to obtain drone-based multispectral and hyperspectral imagery covering Remningstorp. We will continue the controlled experiment and field observation of bark beetle infestations.

(2) The research contents

For tree species classification, we will explore deep learning models for individual-tree and stand-scale tree species classification using WorldView-3 and Sentinel-2 imagery.

For tree forest parameters, we will explore crown extraction methods combining satellite imagery and LiDAR, and monitor regional biomass dynamics using Sentinel-1 SAR data. We will develop a method of detecting forest biomass change using RADARSAT-2 imagery.

For forest insect damage detection, we will explore the early identification method of Bursaphelenchus xylophilus using the acquired hyperspectral data. We will study early identification methods of Ips typographus [L.] based on multispectral and hyperspectral images from UAVs.

(3) Cooperation plan:

Co-supervising 1~2 PhD students.

Co-publishing 2~3 research papers.

Co-organizing an international summer school on forest parameters and deforestation mapping using remote sensing data.

Monitoring the status and dynamics of forest is a major issue in the frame of current climate change analysis, as carbon stock variations for the biosphere represent the major source of uncertainties within the global carbon cycle. Synthetic Aperture Radar (SAR) is an active remote sensing device, able to image the reflectivity of wide environments from space, in a systematic way, independently of weather or light conditions. The penetration of electromagnetic waves into vegetated media makes SAR a unique tool for forest 3-D remote sensing applications. Series of SAR acquisitions are used in this study according to two different processing modes.

The first one uses the coherent information contained in a set of SAR images, acquired with a Polarimetric diversity (PolSAR), Interferometric, i.e. spatial, diversity (InSAR), or Tomographic, i.e. multi-baseline InSAR, diversity (TomoSAR), 3-D imaging purposes. Some very significant results have been found regarding the characterization (forest height, underlying grund topography, and Above Ground Biomass) of tropical forests measured at P band, and participate to the preparation of the upcoming ESA BIOMASS mission. Estimated quantities were further analyzed by comparison with other sources of information, such as the GEDI spaceborne lidar acquisitions , or by evaluating the geophysical properties of the retrieved topographic indicators. At higher frequency bands, the difference between the correlation time of radar echoes measured over forests and the revisit time of a spaceborne SAR platform does not allow to apply classical repeat pass imaging techniques. This working group experienced different approaches able to cope with this serous limitation, and based on the model-based analysis of single-pass InSAR pairs on the one side, and on the reconstruction of classical repeat-pass information from a time series of InSAR pairs. Both approaches led to very promising results at L and X bands.

The second processing mode used in this project is related to the incoherent analysis of time series of SAR images, in order to detect changes an relate them to specific properties of forest. In particular the Tropisco service, providing operational deforestation maps derived from Sentinel 1

data, has been launched recently Details may be found at the following links

https://www.spaceclimateobservatory.org/tropisco-amazonia
https://www.spaceclimateobservatory.org/tropisco-south-east-asia

This joint project combines the use of field, airborne and drone remote sensing, and spectroradiometer data to validate and calibrate innovative new satellite sensors from CNAS, ESA and NASA for forest inventory, assessment and health monitoring applications in China and the UK.

1.Vegetation parameter distribution using airborne LiDAR data - UK

Work with airborne LiDAR has been applied for the estimation of mensuration variables, such as Yield Class, Top Height, Basal Area, Volume and mean DDB plus other variables like fractional cover and LAI, over 120K ha of woodlands across the North of England. Derived point cloud metrics were combined with current growth models in sub-areas of 30x30 m² inside each forest stand. As different parts of the country were surveyed in different years, the growth models helped us to reduce this temporal decorrelation and to bring all estimates to 2022. The comparison of results against field observations produced R² above 0.95 for all estimates. Afterwards, Sentinel-1 has been analysed to detect the perimeter of areas affected by wind damage after the Arwen storm in November 2021. Then, volume lost has been calculated as the intersection of polygons representing wind damage and pixels containing volume estimates.

2. Assessing the effects of drought and stress on productivity using satellite data, field experiments and biomass modelling - UK

Monitoring of drought stress processes in Sitka spruce plantations in Kielder Forest using time-series analysis of different vegetation indices derived from Sentinel-2 and Landsat. In parallel, polytunnel experiments with control and drought groups for this species tracked hyperspectral changes in reflectance to determine early onset drought indicators and identify differing drought tolerances between breeding population clonal types. This process helps us to isolate relevant pigment changes under mild-to-moderate drought stress and to identify drought sensitive indices. Additional work with NDVI, derived from MODIS and Sentinel-2 satellites, was used to drive a biophysical vegetation model to study carbon sequestration over commercial Sitka Spruce plantations. Future work will look to integrate more drought sensitive indices into existing models to improve model accuracy under drought conditions.

Other attempts to detect signs of plant stress in the time series of Vegetation Indices (VIs) computed from Sentinel-2, involved the use of Principal Component Analysis applied to the pixel time series of the VIs to learn features of the data common to most observations (e.g. be that phenological or instrumental) and the learned principal components are used as basic functions in a simple linear regression model to filter out these signals. Remaining features in the observations that cannot be well explained by these components may be indicative of stress.

3.Forest Cover Mapping using Sentinel-2 and GF-6 Data, Pu’er Study area, China

The research area focuses on the surroundings of Pu'er City in Yunnan Province, using Sentinel-2 images analysed through the Google Earth Engine (GEE) platform to extract the spectral features, texture features, and terrain features, combined with field survey data, airborne remote sensing data, and terrain data. A classification data set containing the optimal features was obtained by feature screening. The object-oriented and pixel-based classification methods were used to respectively carry out random forest classification. The results show that the classification accuracy of the object-oriented classification method is higher than that of the pixel-based classification method, with an overall classification accuracy of 88.21% and the Kappa coefficient of 0.865. This was followed by a comparison and verification of classification results. The classification results are compared at pixel level with other published land cover products (including the Dynamic World, ESA, ESRI) to analyze their area consistency and spatial consistency. Accuracy evaluation of all products was carried out combining Pu'er airborne hyperspectral data and LULC ground truth data. Finally, the product inconsistency factors were analysed to improve the quality of classified products. Furthermore, the GF-6 data will be used for Pu'er forest classification, and the classification results will be compared with the Sentinel-2 classification results to obtain the optimal forest covered map.

4. Multi-Scale Biomass Mapping Using Airborne LiDAR Data - China

The LiDAR biomass index (LBI) was extended to airborne LiDAR data for multi-scale forest biomass estimation. Through suitability compensating the laser point clouds of each tree and using a small number of trees measured in the field for model calibration, robust and highly accurate results were obtained. The method was verified by the field measurement data of 20 analytical trees, 133 sample plots and 39 subcompartments with larch plantations in three forest farms. Good performances were demonstrated (R2=0.98 and RMSD=11.85 kg at tree scale, R2=0.77 and RMSD=28.74 t/ha at plot scale and R2=0.86 and RMSD= 144.15 t at stand scale). Through comparing with the existing methods of predicting DBH based on tree height and crown width to calculate biomass, the proposed method shows higher accuracy and obvious versatility among different forest farms. In comparison with the existing LiDAR metrics method, it can obtain more detailed results, but only a small amount of measurement data was needed to calibrate the model.

5. Forest gap identification based on UAV LiDAR

The remote sensing of UAV LiDAR can quickly obtain the three-dimensional spatial information of the forest. The study site is located in the Puer Sun River Reserve in Yunnan, China. The canopy height model (CHM) was derived from the point cloud data of UAV LiDAR. The fixed threshold method was used to identify forest gaps in CHM. The reference data from visual interpretation of images was used for accuracy assessment of forest gap identification. The overall accuracy of the fixed threshold method was 92%, and the spatial distribution of the gap was aggregation. The forest gaps in the study site area were mainly small and medium gaps, showing that there were fewer disturbance events. The spatial distribution of forest gaps and its spatial characteristics in small area subtropical natural forests can be mapped by UAV LiDAR data. Forest gap information from UAV LiDAR can be used for the accuracy assessment and validation for the forest gap derived from GF-7 satellite imagery for large area.

Because of the sheer size of grasslands, grassland degradation makes a far more important contribution than cropland degradation to the global environmental change phenomenon of desertification, which is land degradation in dry areas. About 42% of China's national land area consists of grasslands, and if all the grasslands in the world were brought together they would cover 50% more than the total area of Europe. Yet since there are no reliable estimates of the scale and degree of grassland degradation in the world, the global extent of desertification is equally uncertain, and this also makes it difficult to monitor compliance with the Land Degradation Neutrality (LDN) target of the Sustainable Development Goals (SDGs).

Devising a reliable method for mapping grassland degradation is therefore imperative, and is the aim of this project. In the first part of this project we have critically evaluated all previous research into mapping desertification with satellite data in order to devise such a method. We report eight main findings. First, grassland degradation, like desertification, is a very complex phenomenon, comprising multiple forms of soil degradation and vegetation degradation. Second, grassland degradation is poorly monitored at national scale in most countries. Third, even though monitoring dryland degradation is essential to estimate compliance with the LDN target, little progress in using satellite data and other types of "big data" to compensate for the lack of national data has been reported in the literature since the SDGs were finalized. Fourth, matching the spatial resolution of a sensor to the areal variability of each form of vegetation degradation and soil degradation is necessary to minimize spatial systematic errors, while matching the temporal resolution of a sensor to the turnover time of each component is necessary to minimize temporal systematic errors. Fifth, consequently, a multiple sensor approach is needed to monitor grassland degradation, since while medium (20-80 m) resolution data can be used for initial land use mapping, key grassland biophysical parameters(e.g. biomass, productivity, soil carbon content) estimation and the mapping of some forms of land degradation (such as sandy area expansion), very high (≤ 1 m) resolution data is often needed for mapping most forms of vegetation degradation (e.g. tree density, shrub encroachment) and soil degradation (e.g. gully expansion) in dry areas. Sixth, a multilayer multisensor approach, with land use classification and key grassland parameters estimation preceding degradation measurement, will optimize image processing time and the overall accuracy of grassland degradation assessment. Seventh, reliable methods are still lacking for using satellite data to measure wind erosion and soil compaction. Eighth, there is great scope to experiment with radar and LIDAR sensors to improve the accuracy and coverage of grassland degradation monitoring.

The continuous monitoring of natural multi-hazards, namely large landslides, is of fundamental importance to minimize and prevent actual and future risks. In this Dragon-5 project, Earth Observation (OE) data are used and integrated into various methods for the identification and monitoring of landslide phenomena at different spatio-temporal scales. The upper-Jinsha River region, composed of steep terrain and broken structures, is known to cause frequent landslide disasters. Distributed scatterer SAR interferometry (DS-InSAR) was used to monitor and analyze the Woda landslide area. In the study, 106 Sentinel-1A ascending and 102 Sentinel-1A descending images, covering a time period of almost 5 years (from 2014/11/05 to 2019/09/04 and from 2014/10/31 to 2019/09/11, respectively). The two-dimensional deformation of the landslide revealed that the maximum surface deformation rate in the normal direction is about -80 mm/yr, and in the east-west direction is about 118 mm/yr. According to the rescaled range (R/S) analysis, the Hurst index the deformation trend will continue for some time.

The Gilgit region, Pakistan, was also analyzed using a time series of Sentinel-1 images, covering a time period of about a year (2019 – 2020). Several deformation areas were mapped in the region. A 3D prototype platform, enabling the visualization of the deformation results of the typical active landslides, has been developed, allowing the identification of 3 landslides with potential risk.

The use of Artificial Intelligence (AI) techniques in EO data, to develop/optimize data processing and analysis methods, is also one of the major goals of this project. In this regard, the multitemporal difference interferometric phase diagram of subsidence mining area is obtained by using synthetic aperture radar differential interferometry (DInSAR) techno machine learning (ML) FCN-8s, PSPNet Deeplabv3 and U-Net models were applied to DinSAR data to extract mining subsidence. The results show that the U-Net model presents high detection accuracy and takes short time to run. In order to improve the semantic segmentation and extraction accuracy of mining subsidence, the efficient channel attention (ECA) module is introduced into the traditional U-Net model for training. The detection of earthquake fringes with Vision Transformer (ViT) technic was also tested in the scope of this project. In our methodology we use a dataset composed of 5110 interferograms from LicSAR dataset, that were used to apply DL techniques to recognize patterns. The F1-Score and AUC clearly show that the ViT outperforms VGG19, respectively, with 0.88 against 0.69, and 0.97 against 0.86. However, the major disadvantage of this method consists in the detection of false negatives in overlapped patches. Sometimes the model identifies as “no deformation” patches with deformation fringes but that change in the overlapped patches for the same interferogram. For a better confirmation the method is evaluated applied to the whole interferogram. It can be concluded that the results are higher in both models, proving that some false negatives are adjusted in overlapped patches. ViT remains the model with the best performance. The precision is now equal for both (1.0) and with a significant rise of the Recall and F1-Scores. As main conclusions, we can say that the new and different model proposed in this work to detect fringes in SAR interferograms achieves better results, comparing to Convolutional Neural Networks. This way, we are even closer of obtaining results that can create practical products that may help providing a faster response in earthquakes scenarios.

Monitoring structures of great heritage and historical value, more quickly and effectively, is also one of the major goals of our project. However, only the use of Artificial Intelligence techniques will allow to deal with the huge amount of data that will be generated. The Vilariça Valley, located in the north of Portugal is crossed by an active fault and 3 classified buidings (churches of Torre de Moncorvo, Freixo de Espada à Cinta and Foz Côa, in Portugal) are being used as test sites. More than 1000 Sentinel-1, ERS and Envisat images are already processed providing thousands of points in the villages where these buildings are located. However, only a few points were identified directly in the buildings and in their proximity. To increase the density of points in the building and in its proximity, a time series of 25 PAZ images was commissioned, with the support of ESA. In the near future, these results will be integrated with data from field sensors and LiDAR models, using AI techniques.

Based on the big archived Sentinel-1 satellite radar data during the last 7 years, we developed parallel processing techniques for the detection of subtle ground motion on a phase-stable point basis, so that InSAR deformation could be used for accurately determining tectonic motion or strains induced by anthropogenic activities. In particular, we strengthen our analysis results by joint analysis with 3D GPS data within the same time span if available. In this way, we are able to separate ground motion with long-wavelength errors of InSAR results, which is the major source of errors when tectonic motion is considered. Joint analysis of GPS and InSAR data also provides us a way to separate InSAR surface motion in the Line-Of-Sight direction into vertical and horizontal deformation in particular orientations. But it is worthless to have 3D measurements with InSAR only as it cannot effectively discern the signals in a nearly north-south direction.

We apply our method in various tectonic regions for seismic deformation detections, including the population densified North China Plain, the fast shale-gas development Sichuan basin, and the tectonic active Eastern Tibetan Plateau, etc. In this study, we reduce various InSAR errors, in particular the unwrapping errors as before, while the most important long-wavelength errors are overcome by GPS-assistant joint analysis in this version of processing, so that tectonic motion can be well detected and compared with GPS measurements.

The project aims to develop and apply innovative data analytic methods underpinned with machine (deep) learning technology to analyze and detect seismic anomalies from electromagnetic data observed by the SWARM and CSES satellites along with CSELF network. In this report we will present the progress of anomaly detection algorithms underpinned with Deep Neural Networks (DNN), which focuses on predicting Swarm data as long as possible from the Swarm historic data. We report our investigation into several architectures of Recurrent Neural Networks (RNN), in particular we investigate the development of a Long-Short Term Memory (LSTM) based methods. We present our design and implementation of the LSTM algorithm and predicted results of applying the algorithm to learn from the Swarm historic data. We will also describe error metrics for measuring the accuracy of Swarm data reconstruction and the methods of detecting anomalies in the Swarm data, which are in relation to three earthquakes.

Landslides are destructive geohazards to people and infrastructure, resulting in hundreds of deaths and billions of dollars of damage every year. Therefore, mapping the rate of accumulation of such geohazards and understanding their mechanics is of paramount importance to mitigate the resulting impacts and properly manage the associated risks. In this mid-term project report, the main outcomes relevant to the joint European Space Agency (ESA) and the Chinese Ministry of Science and Technology (MOST) Dragon-5 initiative cooperation project ID 59339 “Earth observation for seismic hazard assessment and landslide early warning system” are reported. The primary goals of the project are to further develop advanced SAR and optical techniques to investigate seismic hazard and risk, detect potential landslides on wide regions, and demonstrate EO-based landslide early warning system over selected landslides. Regarding the landslide hazard, in order to achieve these objectives, next tasks were developed up to now: a) a procedure for phase unwrapping errors and tropospheric delay correction; b) improvement of a cross-platform SAR offset tracking method for the retrieval of ground displacements; c) InSAR and PolInSAR monitoring and semiautomatic mapping of active displacement areas on wide regions, identification of triggering factors and modelling; d) application of InSAR-based landslide early warning system on selected sites. The achieved results, which mainly focus on selected sensitive regions including the Tibet Plateau and the Three Gorges in China and the Alcoy valley in Spain, provide essential assets for planning present and future scientific activities devoted to monitoring landslides. These analyses are crucial for an optimal prevention and management of these geohazards, as well as for a rapid response after their occurrence.

In the framework of Dragon-5 project, Northeastern University (NEU) from China and the National Institute of Geophysics and Volcanology (INGV) from Italy analyzed the multiple geohazards over the heavy industrial base in Northeast China using time Series SAR images. Moreover, we have also considered a new study site, the Changbaishan active volcano (Jilin Province, ~300 km east from Shenyang). This volcano last erupted in 1903 and was responsible for the largest eruption of the last millennium in 946 CE. Changbaishan is affected by landslides, earthquakes, degassing, and ground deformation. Deformations occurred during the 2002-2006 unrest episode and in 2017, when a nuclear test in North Korea triggered landslides. The multi-hazard exposure of Changbaishan is relevant because a population of ~135000 in China and 31000 in North Korea lives within 50 km from the volcano. We analyze the Changbaishan 2018–2020 deformations by using remote sensing data and detect an up to 20 mm/yr, NW-SE elongated, Line of Sight movement on the southeastern flank and a −20 mm/yr Line of Sight movement on the southwestern flank.

These data reveal an unrest occurring during 2018–2020. Modeling results suggest that three active sources are responsible for the observed ground velocities: a deep tabular deflating source, a shallower inflating NW-SE elongated spheroid source, and a NW-SE striking dip-slip fault. The depth and geometry of the inferred sources are consistent with independent petrological and geophysical data. Our results reveal an upward magma migration from 14 to 7.7 km. The modeling of the leveling data of the 2002–2005 uplift and 2009–2011 subsidence depicts sources consistent with those responsible for the 2018–2020 unrest. The past uplift is interpreted as related to pressurization of the upper portion of a spheroid magma chamber, whereas the subsidence is consistent with the crystallization of its floor, this latter reactivated in 2018–2020. Therefore, Changbaishan is affected by an active magma recharge reactivating a NW-SE striking fault system.

The analysis and modelling of the Changbaishan volcano has been the topic of a joint published paper on Frontiers in Earth Sciences (doi: 10.3389/feart.2021.741287).

Concerning the Fushun open pit mine, in these first 2 years of the project, the 2 research teams have collaborated to following the MT-InSAR processing updating the results from DRAGON-4 project until the end of 2021. We have also performed new processing technique as the OT time series analysis during 2013 to 2016 for this area.

Fushun west Opencast coal mine (FWOCM), located in the southwest of Fushun city, China, is the largest opencast mine in Asia. Since the 1920s, more than 90 landslides have been reported in FWOCM, especially the huge landslide on the south slope, which named Qiantaishan landslide. The Qiantaishan landslide has experienced a fast moving period during 2013 to 2016, and has stabilized since 2017. During the fast moving period, the landslide mass has moved approximately 90 meters. However, since 2017, displacements of the Qiantaishan landslide is less than 150 millimeters per year. In order to analyze the spatial pattern and temporal evolution of different periods of the Qiantaishan landslide, both MT-InSAR and multi-temporal pixel offset tracking has been conducted. Multi-temporal pixel offset tracking is conducted based on 53 Cosmo SkyMed SAR images collected from 2013-07-03 to 2016-12-18, to monitor displacement of the fast moving period of Qiantaishan landslide. The results show that the landslide moves very fast during 2014, and slows down during 2015 to 2016. Besides, displacement of the Qiantaishan landslide shows very strong correlation with precipitation, which accelerates in rainy season and decelerates in dry season. Starting from the beginning of 2017, the Qiantianshan landslide gradually stabilized. The MT-InSAR analysis is conducted based on Sentinel-1 images collected during 2017-01-11 to 2021-12-16, to monitor the slow-moving period of Qiantaishan landslide. The MT-InSAR results show that the displacements rate of the Qiantaishan landslide is within 150 mm/year, which has basically stabilized. The area with the largest displacement is located near the former Liushan old river channel, and the maximum displacement rate is approximately 120mm/yr. This is due to the undercutting of the bedrock near the old river channel and the existence of river pebble layer, which has good permeability, allowing rainwater to penetrate through the cracks on the slope, reducing the tensile strength and increasing the mobility of the landslide body.

With the key goal to exploit satellite Synthetic Aperture Radar (SAR) imagery and advanced processing methods for archaeological prospection and heritage sites protection, the Dragon-5 SARchaeology international collaboration project is making step changes to demonstrate the capability of SAR to detect objects of archaeological significance, and monitor the status and stability of cultural and natural heritage sites and their assets. These are Earth observation applications of paramount importance in the field of land monitoring and Earth system science. The project focuses on a range of study sites in China, Russia, Mongolia, Italy, Norway and Bulgaria, including a wealth of heritage asset types, namely burial mounds, partly buried archaeological ruins, standing monuments within urban centres, natural reserves, paleo-channels and ice patches with organic remains.

This work reports on the key achievements from the first two years of the project, during which the research activities focused on: state-of-the-art review of heritage applications of imaging radar; multi-sensor SAR and optical data collection and tailored tasking of new acquisition campaigns over the study sites; SAR image processing with feature extraction, image classification, change detection and Interferometric SAR (InSAR) methods; analysis and interpretation; field data collection, ground truthing and validation of EO-based evidence and observations.

In the wider Province of Rome (Italy), a long-term InSAR ground deformation analysis was carried out with big data stacks of Sentinel-1 IW SAR imagery, and land subsidence hotspots that may represent a potential threat to heritage assets were identified. An initial investigation on the detectability of buried archaeological features was also performed across sub-urban and rural landscapes of the province by analysing multi-frequency SAR data collected in C-band in RADARSAT-2 Fine Beam and Sentinel-1 IW dual-pol. and in X-band by COSMO-SkyMed. The interpretation of SAR imagery has been aided by very high resolution optical data from DEIMOS-2, WorldView-3, Pléiades-1 and Google Earth, and validated by evidence collected in the field.

In Wuhan (China) long-term SAR and InSAR analyses were carried out to estimate risks for local cultural heritage sites due to urbanization and surface motion. Long time series of COSMO-SkyMed data, acquired via the Wuhan-CSK project – a cooperation between Wuhan University and the Italian Space Agency (ASI) – as well as TerraSAR-X data were used for long-term deformation estimation and to survey the urbanization development. Additionally, ERS-1/2 and ENVISAT ASAR data acquired via the Dragon-5 project were processed. The available historical data from Keyhole sensors allowed for manual mapping of the urban areas into the mid-1960s. The 3D development of the urban area was in the focus of the processing of high resolution SAR data, so that the detailed 2D and 3D urbanization analysis allowed for identification of the urban development and therefore a better risk assessment for cultural heritage sites in Wuhan.

For the research on burial mounds, the work focused on improving the methodologies and better monitoring the sites with respect to climatological factors. This is important as the most valuable burial mounds are to be found in or close to permafrost areas. Global warming and thawing of permafrost endanger the organic remains in some of the sites in question that are currently still frozen and therefore extremely valuable for archaeological analysis. Learning more about the current extent of permafrost, monitoring spatial changes and hopefully being able to predict the spatio-temporal patterns of future changes is of crucial importance for the planning and prioritization of future archaeological excavations.

The detection of looting activities is an important task for cultural heritage protection. SAR interferometric coherence is a very sensitive change detection tool and, in combination with the high temporal availability of SAR data, could make for a good approach of looting detection. The main challenge is the very high sensitivity of the coherence to change and other factors, such as soil moisture and spatial baseline differences. Successful change detection with low false alarm rates is therefore difficult, especially when looking for changes in sub-pixel dimensions. To test new approaches based on detecting statistical inhomogeneities within adaptive local and non-local neighbourhoods, a test is prepared at the satellite receiving station of LIESMARS, Wuhan University. On the test site, looting structures will be artificially created to conduct tests of detectability and support method development using high resolution TerraSAR-X data and medium resolution Sentinel-1 data.