Satellite Data Workflow

1. Overview

IDCubeCloud integrates free, open satellite imagery directly into the hyperspectral analysis workspace. Users can search for Sentinel-2 or Landsat 8/9 scenes by drawing a region of interest (ROI) on an interactive map, filter by date and cloud cover, and load the selected scene into the same analysis pipeline used for lab-acquired hyperspectral data.

Key points

• No API key required — all data comes from Microsoft Planetary Computer (free public STAC catalog).
• No file downloads — bands are read as Cloud-Optimized GeoTIFFs (COGs) streamed directly into server memory.
• The loaded data becomes a standard [H × W × B] hyperspectral cube, compatible with every IDCubeCloud tool (spectral plotting, vegetation indices, PCA, spectral library matching, etc.).

2. Data Source and Access

Microsoft Planetary Computer

How We Get the Data

1. Draw an ROI and set scene filters in the satellite browser.

2. The backend searches the Planetary Computer STAC catalog and returns matching scene metadata.

3. Choose a scene. The backend reads signed COG band assets within the ROI and resamples them to a common grid.

4. The bands form a cube with wavelengths in a workspace session.

3. Available Satellite Options

Option A: Sentinel-2 Level-2A (Surface Reflectance)

12 spectral bands loaded

Option B: Landsat Collection 2 Level-2 (Surface Reflectance)

Each Landsat platform is its own entry in the source tree, because the instruments differ: OLI (Landsat 8/9) carries a coastal/aerosol band that ETM+ (Landsat 7) and TM (Landsat 4/5) do not.

7 spectral bands loaded

The catch-all landsat-c2-l2 source picks its band table from each scene's own platform, so a Landsat 5 scene is read with the six TM bands rather than the OLI asset names.

Option C: Other Sources

When to Use Which

4. End to End Workflow

Step 1: Navigate to Satellite Browser

Click the green Satellite button in the toolbar → opens the Satellite Data Browser page.

Step 2: Define Region of Interest (ROI)

• Click "Draw ROI" (orange button).
• Click & drag on the map to draw a rectangle over your area of interest.
• The green rectangle stays visible showing your selected ROI.
• Max ROI size: 2° × 2° (~220 km per side).

Step 3: Set Filters

Step 4: Search

Click "Search Scenes" → the backend queries the STAC catalog once per underlying collection (sources that share one, such as Landsat 8 and 9, are merged into a single query) → results come back interleaved newest-first, each card showing its thumbnail, date, platform, source and cloud cover %.

Step 5: Load into Workspace

Click "Load into Workspace" on any scene card → backend downloads all bands for your ROI → you are redirected to the main analysis workspace with the data loaded in RGB true-color view.

5. Data Conversion into the IDCubeCloud Pipeline

Raw Satellite Data → Hyperspectral Cube

The satellite data goes through several conversions to become compatible with IDCubeCloud's pipeline:

1. Reproject the WGS-84 ROI to each raster’s coordinate reference system.

2. Read only the raster window within the ROI.

3. Resample band images to a common grid, with a maximum dimension of 2048 pixels.

4. Stack bands in wavelength order and store the cube and wavelengths in the workspace session.

After Loading What You Can Do

Once loaded, satellite data is indistinguishable from any other hyperspectral dataset in IDCubeCloud:

6. Data Values, Units and Accuracy

What the Values Represent

Important: These Level-2 products represent atmospherically corrected surface reflectance, but the stored pixel values use different encodings. For Landsat Collection 2, calculate reflectance as DN × 0.0000275 − 0.2. For Sentinel-2 L2A, use the quantification value and any band-specific BOA offset in the product metadata. The values displayed by IDCubeCloud depend on whether its loader applies that scaling; confirm the loader behavior before treating plotted values as physical reflectance.

Provider references: USGS Landsat Collection 2 Surface Reflectance; Copernicus Sentinel-2 L2A Data Quality Report, Processing Baseline 04.00.

For illustration, a Sentinel-2 encoded value of 2500 corresponds to 0.25 only when the quantification value is 10,000 and the BOA offset is zero. A Landsat Collection 2 encoded value of 2500 corresponds to −0.13125 after its scale and offset.

Is It True Value?

Yes, with caveats

Potential Errors and Artifacts

7. Known Limitations and Potential Errors

Technical Limitations

Comparison with Lab Hyperspectral Data

8. Application Use Cases

Agriculture and Precision Farming

Forestry and Ecosystem Monitoring

Water Resources and Wetlands

Urban and Land Use Planning

Geology and Mining

Disaster and Environmental Monitoring

Snow and Ice

Summary

IDCubeCloud's satellite integration provides free, immediate access to multispectral satellite imagery from Sentinel-2 (12 bands, 10 m) and Landsat 8/9 (7 bands, 30 m) via Microsoft Planetary Computer. The data is Level-2 surface reflectance — atmospherically corrected and radiometrically calibrated. It is loaded into the same analysis pipeline as any hyperspectral file, supporting all tools including spectral plotting, vegetation indices, PCA, and spectral library matching.

While not true hyperspectral data (the spectral resolution is coarser — 7–12 broad bands vs. hundreds of narrow bands), satellite imagery adds massive spatial coverage and temporal repeatability that lab instruments cannot provide. Together, they offer a powerful combination for remote sensing analysis at both local and regional scales.

Document generated for IDCubeCloud v1.0 — September 2026

Data providers: ESA/Copernicus (Sentinel-2), USGS/NASA (Landsat), Microsoft Planetary Computer (catalog & hosting)