🪩 Data Exploration

The dataset is the core of QUANTUM AQUA . In this section you can understand how the system:

- 🧮 Examine the dataset used to train the model.
- 📊 View descriptive statistics.
- 🖥️ Generate quick visualizations.
- 📈 Explore variable distributions and relationships.

🛰️ Satellite Optical and SAR:

  • NDVI: indicates vegetation health → vegetation stress often signals groundwater decline.

  • NDWI / NDMI: measure surface and vegetation moisture → correlated with shallow groundwater availability.

  • Spectral bands (blue, NIR, SWIR): detect soil dryness, crop conditions, water bodies, and surface reflectance tied to infiltration.

  • Soil indices (BSI, SAVI): indicate exposed soil, erosion, and moisture retention capacity.

  • VV & VH: respond strongly to soil moisture and vegetation structure.

  • VV/VH ratios: differentiate vegetation, waterlogged soils, and bare dry areas.

  • Dry–Wet contrast: reveals seasonal moisture patterns.

  • Coherence: identifies stable vs. changing surfaces — important for detecting land shifts due to groundwater extraction.

  • SAR is crucial because it detects moisture variability and structural changes, even under cloud cover.

🏔️  Topographic and Geological:

Topography explains natural recharge pathways.

  • Elevation: influences climate, recharge zones, and hydraulic gradients.

  • Slope: steep slopes → runoff; gentle slopes → infiltration.

  • Aspect: controls solar radiation and evaporation.

  • Curvature: concave areas accumulate water; convex areas shed it.

  • TWI (Topographic Wetness Index): estimates potential groundwater recharge zones.

  • Distance to rivers/valleys: proximity to drainage networks often correlates with groundwater storage.

🌦️  Climate and Hydrological:

Climate determines long-term groundwater sustainability. Hydrology explains surface–subsurface connectivity.

  • Rainfall: primary source of natural recharge.

  • Evapotranspiration (ET): water lost to the atmosphere.

  • Aridity Index: ratio of precipitation to evaporative demand → drought severity.

  • Drainage density: low density → high infiltration; high density → high runoff.

  • River order: indicates stability and groundwater–surface water exchange.

  • Upstream area: defines catchment dynamics and recharge inputs.

🌍 GRACE and Georeferenced real wells :

GRACE provides the deepest signal of subsurface conditions. Wells are the gold standard for verifying subsurface predictions.

  • TWS (Total Water Storage): total change in water mass.

  • Anomalies: detect droughts and extraction impacts.

  • Trends: reveal long-term aquifer decline or recovery.

  • Real Wells validate predictions against measured groundwater levels

  • Real Wells improve accuracy through calibration

  • Real Wells anchor the model’s outputs to real-world observations

  • Real Wells enable continuous refinement of the predictive engine

🧮 The training dataset integrates a comprehensive set of variables, including:

– 🛰️ Satellite Optical: Optical satellites reveal what happens on the land surface, which directly influences what occurs below the surface.
–📡 Satellite SAR: SAR penetrates clouds, works day/night, and captures information about surface roughness, structure, and moisture.
–🏔️ Topographic: Topography controls where water flows, accumulates, or infiltrates.
–🪨 Geological: Geology defines the subsurface architecture where groundwater is stored and transmitted.
–🌦️ Climate: Climate governs recharge, evaporation, and water balance.
–🌍 GRACE: GRACE measures mass changes below the surface — the only satellite system that can directly sense groundwater depletion at large scales.
–💧 Hydrological: These capture how surface water interacts with subsurface systems.
- 📍Georeferenced real well locations Real wells are essential for ground-truthing the model.

📊 Descriptive statistics:

This graphic summarizes the key variables used to analyze and predict groundwater conditions, including satellite-derived optical and SAR indicators, topographic and geological characteristics, GRACE water storage signals, and real geolocated wells used for calibration and validation.

🖥️ Quick Visualizations:

This visualization combines satellite data (GRACE gravity signals and optical imagery), hydrological variables, and real geolocated wells to generate a predictive groundwater map. The colored surface represents the model’s estimated groundwater conditions, the contour lines highlight spatial gradients, the colored dots indicate pixel-level predictions, and the red markers show actual well locations used for calibration and validation.

📈 Variable distributions and relationships:

This figure illustrates how key variables interact within the groundwater system: the network graph shows relationships among satellite, climatic, geological, and hydrological indicators, while the 3D scatter plot displays their multidimensional distribution across the study region.