Conference Agenda

With an increasingly warming climate globally, hydro-climate extremes such as heatwaves and droughts have surged, challenging the stability and resilience of social-ecological systems. In particular, the frequency, intensity, and duration of compound heatwave and drought extremes (compound hot-dry extremes) largely increased in recent decades, which intensifies the spatio-temporal mismatch between water supply and demand in the agricultural water community. Furthermore, compound hot-dry extremes are foreseen to be more frequent and intense in the coming decades under a projected warming scenario. This underscores the imperative need for well-informed and effective adaptation strategies within the agricultural water system to address the disasters induced by compound hot-dry extremes. The AgriWATER project encompasses three key aspects: (1) detection and mechanism analysis of compound hot-dry extremes in both Europe and China for the past two decades; (2) quantifying impacts of compound hot-dry extremes on agriculture and water resources across the major crop areas of Europe and China; and (3) developing adaptative strategies to reduce the vulnerability of agricultural water system to compound hot-dry extremes. Using multi-source remote sensing retrievals, machine learning methods, and advanced hydrological and crop models, the AgriWATER project aims to improve the understanding of interactions between compound hot-dry extremes and agricultural water systems over the core food production areas in Europe and China. This provides an important reference into water resources management and food security for both Europe and China, and delivers valuable insights for other agricultural areas globally.

The objective of the project is to evaluate the role of satellite earth observation (EOs), and in particular of optical-thermal-microwave data, in support of water management at different spatial scales (field, regional, and global). Satellite data can play an important role in assessing water availability as well as crop water needs and stress conditions, either as a stand-alone source of information or in combination with hydro-meteorological modelling tools. In this context, satellite data represent an optimal compromise between the need to monitor large areas in a cost-effective manner, and the need for detailed local estimations akin to ground data.

A large suite of satellite-based products is available for water balance modelling at multiple spatial scales, which are relevant to climate, meteorology, basin-scale water management models and field-scale precision irrigation. Nevertheless, there exists a need to understand how the accuracy of such data products affects the reliability of hydrological model simulations when used for operational applications at multiple spatial scales. These needs are exacerbated by the increasing expansion of areas devoted to water-intensive crops and the concurrent impact of climate change on water availability.

Examples of Chinese and Italian / European case studies will be presented on several practical topics to support the discussion on the SAA4Water project.

Chinese case studies will be presented and discussed in relation to the following application:

• Evaluating the Irrigation Performance Using Remote Sensing Data and the Budyko Hypothesis at a regional scale. A novel metric of consumptive irrigation water use was developed after partitioning the total actual evapotranspiration (ET) into green ET and blue ET, and it was applied to the irrigated areas in northwest China to evaluate the performance of irrigation. The irrigation water use efficiency, estimated as the ratio of BET to net irrigation water, was evaluated in detail. The results show that the irrigation water use efficiency has improved in the last 10 years, although some regions still face severe net irrigation water deficits.
• Monitoring cropland water use efficiency (WUE) at global scale. By combining the improved EF-LUE model with the ETMonitor model, a 20-year (2001–2020) global cropland WUE dataset was developed with a spatial resolution of 1-km. It is driven by consistent forcing data and considers the coupling relationship between the carbon and water exchange. This dataset can help to assess and explore the changes in the global cropland WUE and support the monitoring and assessment of the SDG 6.4.
• ET and precipitation derive from remote sensing observations in supporting drought’s impact evaluation at continental scale. We investigated the impact of drought on water availability in the Belt and Road region using high-resolution remote sensing data from 2001 to 2020. An overall decreasing trend in water availability is observed with an expanding water deficit area and increasing of the water deficit intensity associated with drought frequency. The critical role of ET drived from remote sensing observations in assessing seasonal variability of water availability evaluation is also highlighted.

European / Italian case studies will be presented and discussed in relation to the following application:

• Precision irrigation supported by satellite information was analysed, providing a discussion on the required satellite spatial resolution and temporal frequency.
• Estimation of evapotranspiration using observations from several satellites, ground data and hydrological models at different spatial and temporal resolutions.
• Water balance analysis using a suite of satellites observations, ground data and hydrological modelling for operational water availability detection at river cross sections in basins of different sizes.

The results reinforced the synergistic use of satellite data in water and energy balance models is a robust approach for irrigation.

Effective and timely monitoring of crop water stress is critical for achieving agricultural sustainability. In our Dragon 6 project, we aim to synergistically utilise thermal infrared (TIR) and solar-induced fluorescence (SIF) observations from ESA and Chinese satellites to monitor crop water stress The key scientific objectives of this project are: 1) refining the land surface temperature (LST) retrieval method, 2) advancing evapotranspiration partitioning into soil evaporation and plant transpiration using an analytical model driven by LST, 3) developing models for estimating plant transpiration using SIF observations, 4) developing downscaling methods for LST and SIF data to fine spatial resolutions, and 5) enhancing crop water stress monitoring capability based on plant transpiration and SIF estimates. The progress in these five work packages (WP) for the last year is described as follows:

WP1. LST retrieval algorithm.

A new LST retrieval algorithm was developed by integrating the physical mechanisms of the split window (SW) and temperature emissivity separation (TES) algorithms into the deep learning (DL) model, constructing the DL-SW-TES framework. This new framework directly retrieves LST from easily accessible parameters without requiring any prior knowledge of LSE information and atmospheric profiles.

WP2. ET partitioning.

An advancement of the analytical surface energy balance (SEB) model STIC is envisaged. A “top-down” decomposition approach will be adopted based on the SW85 model. Such a method requires lumped estimates of energy fluxes (latent heat flux (ET), sensible heat flux, net radiation, and ground heat flux) in conjunction with air and dewpoint temperatures, LST, canopy-air vapor pressure, and aerodynamic conductance (g_A). The energy fluxes, canopy-air vapor pressure, and g_A are retrievable from STIC.

WP3. Modelling transpiration from SIF.

Two modelling approaches are planned. The first one is based on the connection of plant transpiration and photosynthesis through stomatal conductance. The second method is based on the FEST-2X2-EWB model. The FEST-2X2-EWB model computes continuously in time and distributes in space the total ET as well as the soil and vegetation components, and soil moisture at multiple soil layers. FEST-EWB computes LST as an internal variable and has physically-based non-residual formulations for the energy balance components (and can be used for model calibration or for updating the model state through data assimilation.

WP4. LST and SIF downscaling

The LST downscaling is planned to use a deep-learning based method. The estimated LST using the developed LST retrieval algorithm in WP1 will be input into the deep-learning framework. High spatial resolution LST will be obtained by combining ECOSTRESS and SEVIRI LST estimates.

To address systematic errors in SIF downscaling caused by canopy shadow, we implemented BRDF correction for vegetation indices and utilized SIF_n to minimize angular and PAR effects, generating a 500m SIF product. Subsequently, a Random Forest model was developed using VPD, LAI, and the generated SIF to estimate evapotranspiration, achieving high fitting accuracy across cropland, broadleaf forests, and needleleaf forests.

WP5. Crop water stress indicator development

we developed a novel crop water stress indicator, SIF_n, by minimizing the impact of angular effects and PAR variations on satellite-derived solar-induced fluorescence (SIF). During the 2019 North China Plain drought event, SIF_n exhibited an earlier decline and greater sensitivity in detecting the onset of water stress compared to traditional vegetation indices, raw SIF, and fluorescence quantum yield (ΦF). Furthermore, SIF_n anomalies showed a notable correlation with rainfall and meteorological factors, revealing the dynamic contributions of both canopy structure and vegetation physiology to SIF across different drought phases. These findings underscore SIF_n's potential to enhance the accuracy and timeliness of crop water stress monitoring.

Overall, these advancements in LST and plant transpiration retrieval, SIF downscaling will support the crop water stress monitoring with soil-independent information at high spatial resolutions.

In the framework of the ESA-MOST Dragon-6 project (id.95348), the Istituto Nazionale di Geofisica e Vulcanologia (INGV) from Italy, the Northeastern University (NEU), the Jilin Earthquake Agency and Jilin University from China are carrying out collaborative research using time-series radar and optical images on different geohazards in Northeast China.

Stemming from the experience of the previous Dragon-4 and Dragon-5 projects, the joint research team from Italy and China collaborated to monitor the deformation due to industrial activities and volcanic dynamics. In particular, we are continuing to measure and analyze the ground motion at the Fushun open pit mine and Changbaishan Tianchi volcanic complex to support local hazard assessment. In addition, we analyze the ground deformation at a new study site, the Songliao Basin. Songliao is a Mesozoic-Cenozoic continental basin characterized by significant oil and gas extraction activities and water exploitation. To investigate the surface motion affecting the basin, we mainly use InSAR data from Sentinel-1. The Songliao Basin appears to be affected by ground motion (subsidence and uplift) caused by natural phenomena and anthropogenic activities.

Our main goal is to identify and characterize the processes responsible for the deformation (fluid migration, slope mass movement, volcano dynamics, and human activities), hence quantifying the deformation caused by each source. To this aim, we model the degassing and magmatic activity at Changbaishan Tianchi by using the detected surface deformation and explore the correlation between deformation and the amount of oil/gas extraction or fluid pumping in deep wells in the Songliao Basin.

In detail, the Changbaishan Tianchi intraplate volcano is one of the most active and hazardous volcanoes of Northeast Asia, characterized by a summit caldera formed after the 946 CE ‘Millennium’ Plinian eruption. From December 2020 to June 2021, the frequency and magnitude of earthquakes were significantly higher than during background periods with hundreds of earthquakes (46 events per month on average) reaching a local maximum magnitude of ML 3.1. Our study reports a comprehensive deformation analysis and a geophysical inversion scheme aimed at unveiling the volcano dynamics in this period. Multi-temporal InSAR analysis results of 32 ALOS-2 images from 2018 to 2022 show that the surface deformation is a combination of seasonal fluctuations (± 25 mm in average, with a maximum ± 45 mm) and a long-term positive component. The least squares linear regression of the deformation time series and temperature data isolates the seasonal fluctuations, revealing a clear uplift and subsidence process from June 2020 to July 2021 in the caldera area. To constrain the Tianchi plumbing system dynamics, a combined inversion scheme consisting of three deformation sources is proposed. The inversion results and the seismic records indicate that the Tianchi volcano experienced a low-level unrest episode from December 2020 to June 2021. The shallower plumbing system, located at about 5–9 km depth and modeled by pressurized spheroids, underwent a cumulative volume increase of 26 × 10⁶ m³ from November 2018 to April 2021, followed by a volume decrease of 9 × 10⁶ m³ from April to July 2021. This suggests magma rising from a 14 km deep storage zone to the shallower plumbing system, followed by depressurization due to the escape of fluids. In the next years of the project, we will continue to monitor the Tianchi volcano complex using InSAR techniques and also considering images acquired by the ALOS-2 sensor operating in the L-Band, thus allowing a larger spatial coverage due to a lower sensitivity to vegetation.

Aiming at testing new monitoring procedures based on SAR and optical satellite data, the NEU and INGV teams have also investigated the Hunga Tonga-Hunga Ha’apai (HTHH) submarine volcano, providing an exhaustive analysis able to interpret the disastrous eruptive event that destroyed the island in January 2021. HTHH in the Tonga-Kermadec island arc initiated a new eruptive phase lasting approximately 1 month after 7 years of dormancy. This eruptive phase culminated in the January 15, 2022, Plinian eruption and the associated destructive tsunami. To analyze the complete eruption sequence from 2021 to 2022, satellite optical images and synthetic aperture radar images were analyzed, revealing the morphological changes of the volcanic island(s) during the time. The preparatory phase preceding the first eruption on December 19, 2021, was studied by considering descending Sentinel-1A images from 2020 to 2021. The obtained surface deformation before the December 2021 eruption shows an uplift of nearly 6.4 cm on the line of sight. Results reveal that a possible intrusion of magma started in May 2020 and gradually increased until December 2021, leading to a 19-month-long deformation phase before the eruption. The observed deformation is possibly explained by (a) a NNW-SSE striking, magma-filled dike upraising since the beginning of 2020 from a spheroid-like source located at 5000 m depth.

As far as the Songliao basin is concerned, we have analyzed Sentinel 1-A/B images belonging to different tracks to obtain the largest possible spatial coverage for such an area. To achieve this goal, we have selected SAR data along the descending orbit in the temporal interval between 2021-2024 for Sentinel-1A data for the northern part and Sentinel-1B data for 2017-2021. The latter was because the southern part of the basin was only covered by Sentinel-1B data, which was unfortunately discontinued in 2022. We processed more than 600 SAR images using the Small Baseline Subset technique and obtained the mean ground velocity and the displacement time series. Some very interesting phenomena were detected from these products, highlighting different behaviors, like subsidence and uplifting, for contiguous areas. In the Daqing area, where one of the larger oil fields of the Songliao basin is located, we have detected an uplift in the oil extraction area and a subsidence in the neighboring water network. Data show that for every ton of crude oil produced in the oil field area, 2-3 tons of water collected from the neighboring areas (the water basins and the city) need to be re-injected, and there are 36,879 m³/yr of oil-polluted wastewater. The detected ground deformation shows a subsidence in the water network area and uplift due to water injection in the oil field.

The resilience of vulnerable communities and agricultural systems against natural hazards—particularly landslides and land subsidence—needs to be strengthened. These events cause numerous fatalities and significant economic losses each year. Landslides primarily affect linear infrastructure and urban areas in mountainous regions, resulting in serious communication disruptions and loss of life. Meanwhile, land subsidence is emerging as an increasingly severe geohazard, expected to impact millions of people globally in the coming years by raising flood risks in urbanized areas and damaging critical infrastructure. In this contribution, we present the results obtained after one year of the ReSeLMAIN project (ID: 95355 – Remote Sensing for Landslide Monitoring and Impact Assessment on Infrastructure), developed within the framework of the Dragon-6 cooperation program—an initiative between the European Space Agency (ESA) and the Chinese Ministry of Science and Technology (MOST). The ReSeLMAIN project aims to employ remote sensing techniques, particularly Synthetic Aperture Radar Interferometry (InSAR), to identify, map, and monitor landslides and land subsidence, as well as assess their impacts on infrastructure.

We are working in various study areas across China and Spain. In Beijing, the capital of China, high-resolution TerraSAR-X data was used to analyze land subsidence induced by groundwater withdrawal, revealing a maximum mean displacement rate of over 110 mm/year between 2010 and 2019. A methodology based on slope gradient, vertical radius of curvature, and angular distortion calculated from InSAR-derived displacements has been developed to assess potential damage along the Beijing–Tianjin high-speed railway. In the eastern Tibetan Plateau, China, multi-platform SAR images from 2007 to 2022 were used to identify and characterize landslide groups, which pose grave peril to community members and critical construction along the Jinsha River. A novel InSAR-based method was developed to infer the slip surface of active landslides, along which the maximum displacement reached 1.5 m. We also found that in our study area, the 2018 Baige landslide failure had caused persistent acceleration to downstream wading landslides, highlighting the prolonged cascading impact of landslides on the watershed’s geological environment. In Monóvar, Spain, an Active Deformation Area (ADA) was identified using data from the European Ground Motion Service (EGMS). This area shows maximum vertical and east–west displacement rates of 64.6 mm/year and 23.8 mm/year, respectively, during the period from 2019 to 2024, affecting more than 6 hectares of partially urbanized land. While the deformation was initially attributed to a landslide, subsequent investigations have linked it to a subsurface dissolution process. Supporting evidence includes the concentration of structural damage in buildings located over gypsiferous substrata, the concentric pattern of surface cracking, and the identification of a characteristic subsidence bowl. The findings indicate that a significant number of buildings, roads, and other infrastructure in Monóvar are affected by these ground deformations, underscoring the need for improved monitoring and mitigation strategies.

The aim of this project focuses on the development and application of processing methodologies to address the 3D characterization of sub-surface targets using stacks of spaceborne SAR data acquired over natural scenarios. The investigated applications will include 3D imagery of forests, ice sheets, and desert areas, and are therefore mapped into Dragon topic Solid Earth - Subsurface target detection. The topics above are of fundamental importance in the context of present and future spaceborne SAR missions, which will allow increasingly more systematic use of multiple acquisitions thanks to improved hardware stability and more strict orbital control. Specifically, the proposed activities are intended to support use of multi-pass data stacks from:

• the upcoming P-Band mission BIOMASS.
• in-orbit L-Band missions, such as the Argentinian SAOCOM constellation, the Chinese missions LuTan-1 & LuTan-4, and potentially ALOS-4 and NISAR.

Research activities will consider SAR data stacks acquired by P- and L-band spaceborne SARs over dense tropical forests, ice sheets, and desert areas, as well as campaign data from ESA campaigns such as TomoSense, AfriSAR, AlpTomoSAR, IceSAR and the Second Tibetan Plateau Scientific Expedition and Research led by the Chinese Academy of Sciences (CAS). The activities will be concentrated on processing SAR image stacks to extract information about vertical vegetation structure and sub-surface terrain topography in forested areas, and also about the internal structure of sand dunes in desert area as well as snow-ice volume in glacier area. Estimation and compensation of ionospheric and tropospheric propagation effects will be investigated as well. Leveraging the unprecedented availability of P-Band spaceborne data from the BIOMASS mission, the research will as well be extend to investigating the 3D of the ionosphere. Whenever possible, validation activities will exploit the availability of reference data gathered at campaign sites, for which we plan to analyze spaceborne acquisitions at the same sites

The first year of the Dragon 6 project focused on the application of satellite remote sensing methods to monitor geohazards and support agricultural water use assessment. Activities were carried out in Brazil, Portugal China, and Pakistan, combining Synthetic Aperture Radar (SAR), optical data, ground measurements, and artificial intelligence models.

In Maceió, Brazil, ground deformation due to underground salt mining was analyzed using 145 Sentinel-1A images acquired between June 2019 and April 2024. Initial Persistent Scatterer Interferometry (PSI) identified general deformation trends but showed limitations in low-coherence areas. To address this, a Quasi-Persistent Scatterer (QPS-InSAR) approach was applied, increasing point density by over 400% (from 4,494 to 23,460 scatterers) and revealing cumulative vertical displacements ranging from −1,750 mm to +10 mm. Notably, 17 scatterers exhibited a loss of amplitude stability in 2023, with 11 of these changes occurring in December—correlating with the collapse of Mine 18. These results support the use of amplitude time series as an indicator of pre-failure instability.

In the Jamari National Forest, also in Brazil, Sentinel-1 SAR data were used to detect the impact of selective logging in tropical forest areas. Backscatter-based change detection methods were applied to areas logged in 2017 and 2020 and a control area. The detected canopy loss was 6.90% and 4.11% in the logged plots, respectively, while the undisturbed area showed only 0.16% change. Spatial overlap with mapped logging infrastructure reached 60.94% and 51.65%, confirming the method’s suitability for operational-scale forest monitoring.

In Portugal, a study in the Vilariça Valley addressed soil moisture estimation for irrigation management in olive orchards. Sentinel-1 SAR imagery (VV and VH polarizations) was combined with field measurements taken every 30 minutes from July 2020 to December 2021 at 10 cm depth. Two Artificial Neural Network (ANN) models were trained: one using 161 ascending images (D1), and another combining 246 ascending and descending images (D2). The D1 model achieved higher performance (RMSE: 2.78%, R²: 0.69, MAPE: 8.26%) compared to D2 (RMSE: 3.96%, R²: 0.59, MAPE: 12.41%), highlighting the importance of acquisition geometry consistency.

Collaborative work conducted by the Chinese team included surface deformation analysis in the Gilgit section of the China–Pakistan Highway and in the North China Plain (NCP). Multi-temporal InSAR, GNSS, and hydraulic head data from 2015–2019 were used to characterize aquifer behavior and land subsidence in NCP. Seasonal vertical displacements of up to 25 mm were identified, with long-term subsidence averaging 29 mm/year and localized peaks reaching 120 mm/year. Variability across regions was linked to differences in groundwater management and the influence of the South-to-North Water Diversion (SNWD) infrastructure.

These results demonstrate the effectiveness of combining EO data, ground observations, and AI models to monitor dynamic environmental processes. The activities developed during this first year contribute to building operational workflows for deformation risk detection, support tools for irrigation optimization, and future developments such as early warning systems and integration of heterogeneous data sources.