The Insight Creation window allows you to create new analytical insights customized to your specific analysis requirements. This window is structured into two main sections: Master Data and Insight Types.

Master Data (common for all insights) #

This initial section captures fundamental details about the insight:

• Name: Enter a clear and descriptive name for the insight.
• Category: Select from the dropdown the appropriate category to organize your insight effectively.
• Description (Optional): Provide additional details or notes about the insight.

Note: If no description is provided, the platform will automatically generate one, highlighting key attributes of your newly created insight.

Instrument Compatibility #

The Insight Creation window will showcase the bands from the currently selected instrument. However, if all the bands used for the created insight are present in other instruments, the insight will automatically be available to be used with those additional instruments as well.

Types of Insights #

Insights can be created from five distinct types, each suited for specific analytical purposes:

1. Expression
   • Create insights based on mathematical expressions involving spectral bands.
2. Principal Component
   • Performs Principal Component Analysis (PCA), identifying major data variations. This includes setting specific bands for reflectance and absorption to highlight particular material characteristics.
3. Mixture Analysis
   • Compares spectral signatures (either from the platform library or uploaded by the user) to identify material mixtures.
4. Distance Analysis
   • Uses spectral signatures to measure spectral similarity or distance, helping to pinpoint material matches or differences.
5. Wavelength Analysis
   • Utilizes specific wavelength ranges from instruments to analyze reflectance, absorption, or abundance of particular materials.

Each insight type has unique characteristics and configuration methods, detailed in their respective sections.

1. Expression #

This type allows the creation of insights based on custom mathematical expressions involving selected spectral bands.

• Selecting Bands: Choose the bands needed from the available list. You can insert these into your expression by clicking or manually typing the band codes.
• Mathematical Operators: Operators can be clicked from the provided interface or manually typed.
• Composite Insights (RGB Channels):
  • Create composites insights using three separate expressions for the Red, Green, and Blue channels.
  • Palettes are not required when using RGB channels.
• Set the absolute minimum and maximum values for visualization. Typically, values between 0 and 1 suffice, as relative stretching adjustments can be applied later.

2. Principal Component #

Performs an Acosta Variant of the Principal Component Analysis (PCA) to identify major variations within your selected spectral bands.

• Selecting Bands: Choose all relevant bands from the provided selector.
• Band Designation:
  • Specify which selected band should represents maximum reflectance.
  • Specify which selected band should represents minimum absorption.
• PCA returns as many components as the number of selected bands. The platform calculates eigenvalues for each component:
  • The component selected is the one with the greatest eigenvalue difference between reflectance and absorption bands.
  • If the eigenvalue for reflectance is smaller than that for absorption, the resulting component is inverted.
• Palette: Select an appropriate color palette to represent your PCA results clearly.

3. Mixture Analysis #

Performs a Spectral Unmixing by comparing spectral signatures from uploaded materials with your data to identify material mixtures.

• Selecting Spectral Signatures: Choose three spectral signatures from the library (public library or user-uploaded).
• RGB Channel Assignment:
  • First signature → Red channel
  • Second signature → Green channel
  • Third signature → Blue channel
• The analysis is normalized (sum-to-one), producing results scaled between 0 and 1.
• No palette selection required, as results are displayed using RGB channels.

4. Distance Analysis #

Uses spectral signatures to measure spectral similarity or distance, helping to pinpoint material matches or differences.

• Selecting Spectral Signature: Choose one spectral signature from the library.
• Metrics: Select the distance metric to apply:
  • Spectral Angle Mapper:
    • Measures the angle between the spectral signature and each pixel’s spectral vector. It evaluates how similar their shapes are, regardless of intensity.
    • Ideal when you’re looking for spectral shape similarity, even under different lighting conditions.
  • Spectral Information Divergence: 
    • Compares the information content between the signature and pixel spectrum by treating them as probability distributions. Sensitive to relative differences across bands.
    • Best when fine spectral structure matters—especially in hyperspectral data.
  • Squared Euclidian Distance: 
    • Calculates the sum of squared differences between each band of the signature and the pixel spectrum. Emphasizes absolute intensity differences.
    • Straightforward, but affected by lighting variations and scale.
  • Earth Movers Distance:
    • Estimates how much “effort” is needed to reshape the pixel’s spectrum into the spectral signature, accounting for both band value and position.
    • Great for identifying shifted absorption features, especially in minerals.

• All metrics are normalized from 0 (least similar) to 1 (most similar).
• Palette: Choose a color palette to visualize distance results effectively.

5. Wavelength Analysis #

Utilizes specific wavelength ranges from instruments to analyze reflectance, absorption, or abundance of particular materials.

• Selecting Wavelength Range: Choose a minimum and maximum wavelength.
  • Wavelengths must have at least one band between the selected minimum and maximum wavelengths.
  • Maximum separation allowed between wavelengths is 17 bands.
• Analysis Type: Choose the analysis method:
  • Band Depth Analysis: Measures absorption intensity within the specified wavelength range (higher values indicate stronger absorption).
  • Absorption Feature Position: Identifies the exact position (wavelength values) of the absorption peak within the selected range for each pixel.
• Palette: Select a palette for visualizing results.

Best Practices #

• Be aware that, even though all insight types are compatible with any instrument that includes optical bands (whether public or private) the effectiveness of spectral distance or mixture analysis may vary depending on spectral resolution:
  • Instruments like EMIT or EnMAP (225–285 bands) allow for more precise analysis.
  • Instruments like Landsat-9, Sentinel-2, or ASTER (12–15 bands) provide lower spectral granularity, which may limit the ability to detect certain spectral signatures.
• Similarly, Wavelength Analysis is available for any optical instrument, but the selected wavelengths must exist within the instrument’s available range.
• Clearly name insights to facilitate easy identification.
• Select appropriate palettes to enhance visualization and interpretation.
• Carefully select spectral signatures and bands based on analytical goals.
• Verify band and wavelength selections for compatibility across instruments to maximize insight applicability.