Interactive research vision
Research Vision
A research vision connecting multi-source Earth observations, scientific discovery, process understanding, and hybrid hydrological modeling through an iterative learning cycle.
Observation, Discovery, and Modeling
This future-facing framework connects complementary observations with scientific and process discovery, model refinement, and hybrid hydrological modeling. The goal is an iterative learning cycle in which observations improve understanding and models, while model behavior and uncertainty reveal what should be observed next.
Individual parts of this vision are grounded in completed and ongoing research, but the full closed-loop framework represents a direction for future research rather than an operational system that has already been implemented.
Scientific learning cycle
A closed scientific learning cycle
Each stage passes distinct scientific information to the next. Observations support discovery and model development, while diagnosis exposes unresolved processes, uncertainty, and observational gaps.
01 Observe
Multi-source measurements across spatial and temporal scales.
Evidence & constraints lead to Discover ↓
02 Discover
Patterns, process hypotheses, biases, scale dependencies, and uncertainty.
Process hypotheses lead to Refine ↓
03 Refine
Process understanding, structure, parameters, representations, human influences, and scale relationships.
Structure & parameters lead to Model ↓
04 Model
Process-based, statistical, differentiable, physics-informed, and hybrid approaches.
Simulations & estimates lead to Diagnose and Learn ↓
05 Diagnose & Learn
Evaluate simulations, reconstructed states, estimates, uncertainty, deficiencies, and information needs.
Observation priorities guide the next Observe stage ↻ Observe
Close the loop: Model–data mismatch, uncertainty, and missing processes identify the variables, locations, scales, and times where additional observations may be most informative.
Explore the Observe componentObserve / Earth Observation
Connecting multi-source observations across scales
Satellites, in-situ networks, UAV and low-altitude sensing, and hydrometeorological forcing provide complementary evidence at global, regional, and local scales. This concept explores how these observations could be integrated to characterize water-storage dynamics, constrain hydrological states and parameters, and support scale-aware model learning.
Interactive scientific pathway
From measurement to model-ready information
Highlight platforms, spatial footprints, and the hydrological variables they represent.
Multi-scale observation concept
The accessible concept follows four steps: measure with complementary platforms, integrate their spatially different evidence, estimate hydrological states and conceptual uncertainty, and pass model-ready information downstream.
- Basemap: simplified geographic land outlines over a distinct ocean surface
- Global: representative satellite coverage and four synthetic water-state regions
- Regional: a conceptual basin integrates satellite, ground-network, and UAV evidence
- Local: a process cross-section links surface water, soil moisture, groundwater, and sensors
- Estimate: deterministic annual states, observation samples, and conceptual uncertainty
- Inform: states, constraints, and uncertainty are passed toward downstream model learning
The interactive WebGL view will replace this diagram when supported.
Drag to rotate, use the wheel or pinch gesture to zoom, select an object for details, or use the controls to change scale, layers, water variable, concept region, simulated date, and playback speed. Regional and Local progressively focus on the same selected region; Reset view restores the current scale preset.
Static concept view shown.
Research grounding
Selected observation systems, published studies, and an open-source reconstruction project ground the measurement, state-estimation, scale, and model-learning elements of this concept. Accelerated playback makes the same deterministic seasonal trajectory easier to inspect; it does not alter the synthetic states or represent faster physical processes.
Observation systems
Representative satellite perspectives on soil moisture, surface-water elevation, and terrestrial water storage.
SMAP · SWOT · GRACE/GRACE-FO
State estimation through time
Data assimilation and complementary process-based and machine-learning pathways for interpreting evolving water states.
Li et al. (2020), Geoderma · Li et al. (2020), Journal of Hydrology
Regional and cross-scale dynamics
Water-storage reconstruction and process or parameter relationships across regional and spatial scales.
Model learning and reconstruction
Multi-source model learning and an open implementation for reconstructing terrestrial water-storage anomalies.
These references provide research grounding rather than evidence that the full framework is already implemented. The interactive water states are deterministic simulations, not measurements, forecasts, or reproductions of the cited studies. Playback speed changes only how quickly the annual concept is displayed; it does not change values, uncertainty, sampling support, or observation-platform motion.