Band Depth Analysis (BDA)

The Band Depth Analysis (BDA) index estimates the intensity of a spectral absorption feature by measuring how deep the reflectance curve dips relative to a modeled continuum. It is used to infer concentration or abundance of specific materials (e.g., minerals, moisture, pigments) based on how strongly a substance absorbs light in a given spectral region.

How it works #

Instead of assuming that the absorption minimum occurs at the center of the band range, this improved method performs a pixel-wise analysis to find the true minimum reflectance across the full spectral window. It then computes the band depth by comparing this value against the expected reflectance (continuum) at the same wavelength.

Steps involved: #

1. Continuum Modeling A linear continuum is constructed between the first and last bands in the selected wavelength range. This defines the expected reflectance if no absorption were present.
2. Continuum Removed Reflectance (CR): For each band in the range, the reflectance is normalized to the continuum to highlight the absorption: ${\mathrm{CR}}_{\lambda }=\frac{R_{\lambda }}{R_{\text{continuum}}\left(\lambda \right)}$
3. Minimum Detection The algorithm finds the band where CR is minimum, indicating maximum absorption.
4. Depth Calculation Using the original reflectance and the interpolated continuum at that specific wavelength, the band depth is computed as: $\text{BDA}=1-\frac{R_{\text{min}}}{R_{\text{continuum}}}$ Where:
   • $R_{\text{min}}$ is the reflectance at the point of deepest absorption
   • $R_{\text{continuum}}$ is the estimated baseline reflectance at the same wavelength
5. Clamping to Valid Range To avoid artifacts or overcorrections, all BDA values are clamped between 0 and 1:
   • 0 → no absorption
   • 1 → maximum possible absorption

Why it’s better #

This approach improves upon traditional methods that assume the absorption happens exactly at the midpoint, which can be inaccurate. Instead, it dynamically adjusts to the real spectral behavior, resulting in more accurate and geophysically meaningful maps.

Applications #

• Mineral identification and quantification in remote sensing for mining and geology.
• Monitoring of water stress and pigment concentration in vegetation.
• Environmental detection of chemical residues with known spectral absorption features.