Conference Agenda

Crop water stress monitoring is critical for effective irrigation. In our Dragon 6 project, we target at utilising thermal infrared (TIR) and solar-induced fluorescence (SIF) observations from ESA and Chinese satellites synergistically to achieve efficient crop water stress monitoring. Five scientific objectives are envisaged for the respective work packages (WP) in this project: 1) refining the land surface temperature (LST) retrieval method, 2) advancing evapotranspiration (ET) partitioning into soil evaporation and plant transpiration using an analytical model driven by LST, 3) 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 that has been made for these five WPs in the first two years is listed as below:

WP1. LST retrieval algorithm.

A new LST retrieval algorithm, the deep learning coupled with split window and temperature-emissivity separation (DL-SW-TES) method, was developed by integrating the physical mechanisms of the split window (SW) and temperature emissivity separation (TES) algorithms into the deep learning (DL) model. The DL-SW-TES framework directly retrieves LST from easily accessible parameters without requiring any prior knowledge of LSE information and atmospheric profiles. The DL-SW-TES framework was evaluated using both the simulation dataset and high-resolution ECOsystem Spaceborne Thermal Radiometer Experiment on Space Station (ECOSTRESS) observations. The simulation analysis showed that the DL-SW-TES method achieved a root mean squared error (RMSE) of 1.05 K in LST retrieval and appeared robust across various uncertainty conditions. The evaluation of the ECOSTRESS LST estimates at the six radiometer sites revealed that the DL-SW-TES method achieved a better performance with an overall RMSE of 1.56 K and a bias of -0.06 K compared to the official ECO2LTSE product (with an RMSE of 1.94 K and a bias of -0.25 K).

WP2. ET partitioning.

A two-source surface energy balance (SEB) model based on the analytical framework of the Surface Temperature Initialized Closure (STIC) model, named two-source Surface Temperature Initialized Closure (TSSTIC), was developed based on the Priestley–Taylor equation after refining the radiation partitioning between soil and plant and the estimation of ground heat flux. The TSSTIC model was evaluated by comparing with the prognostic model Soil Canopy Observation of Photochemistry and Energy fluxes (SCOPE) when driven by the same meteorological data and vegetation traits. The results indicated good consistency between TSSTIC and SCOPE in the retrieval of canopy ET and plant transpiration, with RMSEs around 40 W m^-2 and biases around 15 W m^-2.

WP3. Estimating transpiration from SIF.

Based on plant physiology, the connection of plant transpiration and photosynthesis through stomatal conductance will be utilised. The canopy stomatal conductance will be inferred based on the SIF signal. The retrieved conductance will then be utilised to estimate plant transpiration.

WP4. LST and SIF downscaling

A novel LST downscaling model, the physically guided diffusion model (PGDM), was developed. In this model, the downscaling task was formulated as an inference problem, aiming to sample from the posterior distribution of high-spatial-resolution (HR) LST conditioned on low-spatial-resolution (LR) LST observations and a suite of HR geophysical priors. The performance of PGDM was evaluated on the three compiled datasets using the widely adopted “upscaling-downscaling” strategy. The results demonstrated that RMSEs of 0.91 and 1.126 K were achieved for the 10x and 20x downscaling tasks, respectively.

To address the systematic inconsistency between vegetation indices (VIs) and TROPOMI SIF arising from canopy shadow and angular effects, SIF was first normalized to a unified observation geometry and photosynthetically active radiation (PAR) condition to eliminate systematic biases. Building on this, a land cover-informed downscaling framework for TROPOMI SIF was developed, in which a ControlNet-based deep learning model incorporates land-cover information as structural constraints during spatial reconstruction, alongside multi-source predictors derived from Sentinel-2, Sentinel-3, and ERA5, including NIRv, EVI2, shortwave infrared reflectance, and LST. The framework downscaled TROPOMI SIF to 300 m spatial resolution. Evaluation focused the spatial continuity, land-cover boundary preservation, and cross-cover variability across the North China Plain. Results demonstrated that the land cover-informed framework better represented spatial gradients and boundary transitions in mixed-pixel regions compared to other deep learning baselines. The downscaled SIF product enables more reliable characterization of canopy structural and physiological responses under water stress conditions, providing a finer-scale input for crop water stress assessment.

WP5. Crop water stress indicator development

Building on the downscaled SIF product from WP4, a novel crop water stress indicator, SIFn, was further developed to enhance the sensitivity and timeliness of stress detection. A spatiotemporally continuous SIFn dataset at 300 m spatial resolution and 4-day temporal resolution was subsequently generated to support field-scale water stress monitoring. During the 2019 North China Plain drought event, SIFn 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), capturing physiological changes up to two weeks earlier than NIRv during the early drought phase. Furthermore, SIFn 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 SIFn's potential to enhance the accuracy and timeliness of crop water stress monitoring at field scale.

Overall, the advancements achieved during the first two years demonstrate the strong outcomes of the collaboration between the European and Chinese partners under the Dragon 6 framework, laying a solid foundation for the research to be carried out in the coming two years.