Conference Agenda

The MSS-1 (Macau Science Satellites-1) twin satellites were successfully launched on 21 May 2023 into a nearly circular low-inclination (41 degree) orbit at an altitude of about 450 km: Satellite A carries a payload of high-precision magnetometers to accurately measure Earth's magnetic field while Satellite B is equipped with other sensors such as a solar X-ray instrument to obtain space environmental data.

The MSS-2 twin satellites, marked by strongly eccentric polar orbits forming the MSS constellation of geomagnetic survey satellites, are scheduled to launch in the year of 2028.

The geomagnetic field measured by the satellites is closely associated with the whole Earth's system because the measured field consists of contribution from various sources: the core convection-driven dynamo, the induced field from the electrically conducting ocean and mantle, the crustal magnetization, and the ionosphere/magnetosphere currents.

I will report progress in the MSS constellation of geomagnetic survey satellites and its initial scientific results.

The main goal of the recently kicked-off Dragon-6 STAI4CH project is to demonstrate the significant capabilities of optical and radar EO data to provide essential information for archaeological and cultural heritage mapping and monitoring applications, by building upon first achievements by Dragon-5 research collaborations. In particular, STAI4CH aims to develop novel Artificial Intelligence (AI) and Deep learning (DL) methods to process medium and high resolution EO data to efficiently map waterways and waterscapes of ancient Empires in China, Italy and Iraq, and detect physical transformations and land cover changes due to modern development and urbanisation that might impact the heritage assets.

During its first months of activity, STAI4CH’s team has started to design and develop the heritage asset mapping and change detection approaches, which exploit semantic segmentation Neural Network (NN) based on EO self-supervised learning and change detection NN based on multi-source EO data, respectively. Medium resolution EO datasets acquired by Copernicus Sentinel-1 SAR and Sentinel-2 optical missions, and the Chinese sustainable development science satellite SDGSAT-1, have started to be collected for a transect of the UNESCO World Heritage Site (WHS) of the Grand Canal of China, which has been selected as the testing ground to train and validate the newly developed methodologies. In parallel, preparatory work on EO data interpretation approaches to map heritage features in archaeological landscapes of Southern Iraq has been carried out, with the aim to start to generate a knowledgebase to validate the methods when exported to the waterscapes of the UNESCO WHS of the Ahwar marshes. Pivotal has been the role of the Chinese and European Young Scientists in both the algorithmic development and initial experiments in China, and the archaeological mapping in Iraq.

The first results of the AI/DL-based experiments on the Grand Canal transect will be showcased during this presentation, including a discussion on how the methods performed to detect the heritage features and land cover changes, their limitations and pathway to technical improvement. Future work will focus on the enhancement of the performance of the developed methods, their implementation across the whole site of the Grand Canal, and then the testing of replicability through geographical exporting of the AI/DL-based methods on the Ahwar marshes in Iraq and the waterways of the Tiber river in Italy, part of the Historic Centre of Rome UNESCO WHS. Further algorithmic development will include tailoring of the methods to the processing of high resolution imagery from SkySat optical and ICEYE SAR missions, including fine tuning to the local scale for precise mapping and change detection, and also validation of the results based on medium resolution imagery.

Objective and methods:

With the increasing availability of different medium and high-resolution remote sensing data, the fusion of multiple data sources to generate high-temporal-spatial- resolution data for grassland monitoring has been widely used, among which the fusion of Landsat 8 with Sentinel-2 and MODIS is the most common. Data fusion processors have been developed in the EO community for a longer time already. Now, Harmonized Landsat Sentinel-2(HLS) could be download from the Earthdata Search and Land Processes Distributed Active Archive Center. However, there is currently limited research that considers fusing Chinese high-resolution data(such as GF-1 and GF-6 WFV) with foreign medium and high-resolution data(Such as Sentinel-2 and Landsat) and does not introduce low spatial resolution data information.

Sandy land is one of the results of land desertification, which refers to land mostly covered by sand or sandy soil (sand-covered land), including desert. Knowledge of the spatial distribution and variation in sandy land is important to better understand desertification processes, land resource management, and environmental research. The emissivity of sandy soil in the TIR bands has remarkable characteristics. In addition, the green transition increasing attention is paid to the thermal insulation efficiency of building. Several thermal mission can support related information needs such as Landsat TIRS, Ecostress and at coarser resolution also Sentinel-3. The upcoming Copernicus LSTM mission will provide systematic global TIR observations at high spatial resolution. Certainly, synergies between such mission need to be explored. Here the Chinese SDG-Sat mission is highly relevant. It provides 30m TIR observation yet over a small spatial footprint. Therefore, complementary use need to be explored and potential for fusion these sources need be assessed.

Based on the background above, the objectives of this proposal are to focus on the following aspects:

• Evaluate complementary EO data sources from Chinese and European missions in terms of their characteristics for integrated or fused applications. Special focus will be paid to Sentinel2 and Geofen6. Also thermal complementarities between SDG Sat and other LST data sources will be explored.
• Explore the value of different harmonization and fusion processors such as STARFM, sen2like and others.
• Perform a prototype implementation of a fusion processor in a cloud environment
• Evaluate fused data stream in different application contexts

The deliverables include：

(1) Image Processing Algorithms: A prototype of a fusion processor will be developed and transformed into an cloud based on demand service prototype.

(2) Geospatial Products: Fused China-EU EO data of study area. Spatial distribution of heat emission of study area. Spatial distribution of sandy land in Xilin Gol League, China

(3) Technical Reports: Comprehensive technical reports detailing the methods, processes, and techniques used in the project.

(4) Scientific Publications: 3-5 research papers or scientific publications presenting the project's findings, methodologies, and significant results.

(5) Young scientist training: 3-5 young scientist exchange and training

Three-dimensional ocean dynamic prediction holds significant implications for maritime activities, marine economy, and meteorological-climatic systems. Traditional numerical forecast models for ocean dynamics require substantial computational resources and time consumption. With advancements in artificial intelligence (AI) methodologies, data-driven AI-based ocean models can now achieve superior prediction accuracy compared to numerical models while demanding quite limited computational time and resources. However, current AI ocean models remain constrained by numerical forecasting systems as they continue to rely on numerical analysis fields for forecast initials. This study proposes an end-to-end AI-based ocean forecasting framework that directly utilizes satellite remote sensing observations for three-dimensional ocean dynamic prediction. The model architecture connects satellite observations (input) with three-dimensional oceanic parameter forecasts (output). Input parameters include sea surface temperature(SST), sea level anomalies (SLA), and sea surface current derived from satellite remote sensing. The model generates 10-day forecasts of three-dimensional parameters across 0-200 meters depth with 20 vertical layers. By eliminating the dependence on numerical model outputs, this AI framework enables near real-time prediction updates through instantaneous computation (seconds-level processing) upon acquisition of new satellite observations. This innovation facilitates rolling updates of three-dimensional forecasts, thereby fundamentally transforming the conventional role of satellite remote sensing from merely providing observational references to enabling predictive capabilities for future oceanic states. This advancement establishes a new operational framework that satellite remote sensing can transcend its traditional function as a diagnostic tool, emerging as a proactive predictive foundation for marine environmental forecasting.

Since the project launch last year, our project focuses on two critical challenges in current degradation monitoring, including large-scale quantitative assessment of shrub encroachment, and grassland degradation assessment with very high spatial resolution image. Through sensitivity analysis of optical and SAR features against shrub coverage, optimal monitoring periods and feature combinations were identified. By developing a feature fusion algorithm to amplify shrub signals, the team successfully constructed a shrub coverage index. The team conducted visual interpretation of Google Earth imagery in Hexigten Banner, Inner Mongolia. This enabled high-precision mapping of land cover and spatial patterns of vegetation/soil degradation, establishing a grassland degradation classification framework. The findings demonstrate that VHR image analysis coupled with cross-scale multidisciplinary frameworks serves as an effective tool for policy-making and land management decisions.

Aiming at information extraction of wetland environment factors, a signal photon extraction method based on convolutional neural network and a hyperspectral vegetation feature band selection method based on quantum genetic spectral angle classification algorithm were proposed based on ICESat-2 and measured hyperspectral data respectively. At the basin scale, statistical analysis and model simulation method were used to analyze the characteristics and driving factors of water and sediment change, and separate the contribution of natural and human factors based on the long-term hydro-meteorological, natural and socio-economic factor and remote sensing data. The digital twin technology was used to construct more than 8 types of migratory bird models and 12 types of wetland vegetation models in Poyang Lake wetland, dynamically display the habitat activities of migratory birds and simulate the vegetation growth process synchronously, so as to realize the accurate restoration of the seasonal growth and decline of the virtual Poyang Lake water area, the inter-annual change of vegetation growth and the migratory activities of migratory birds. Above works are benefit for evaluation of wetlands resilience and reveal the human-water relationship in watersheds.