Introduction: From Pixels to the Real World
Every satellite image tells a story, but that story only makes sense when you can trace each pixel back to the ground. In remote sensing, this connection between imagery and reality is called ground truthing. It ensures that what we interpret from orbit reflects what truly exists on Earth.
As Campbell and Wynne (2011) emphasize, field spectroscopy is the cornerstone of reliable remote sensing. Without it, even the most detailed imagery remains unverified, colorful, yes, but speculative.
For this lab at Delta State University, students in the Geospatial Information Technologies (GIT) department conducted field measurements across campus using a Handheld FieldSpec spectrometer. My role was to process, analyze, and interpret the resulting data, transforming raw light readings into visual and numerical insights on vegetation health, spectral behavior, and image analysis.
From Raw Data to Reflectance
The dataset provided by my classmates included wavelength, reference panel readings, and reflected energy from each target surface, including grass, leaves, soil, concrete, and asphalt.
I converted all text files to CSV, cleaned the data, and calculated reflectance using the standard equation:
where Ltarget is reflected radiance from the surface, and Lreference is radiance from the white reference panel (Jensen, 2015). The resulting reflectance ratios were then plotted and compared across several vegetation samples.
The Signature of Life (and Decay)
The data revealed unique reflectance behavior for each vegetation type, a visual fingerprint of their biological condition.
Vegetation 1 – Broad Green Leaf
| Wavelength (nm) | Reference (W·m⁻²) | Target (W·m⁻²) | Reflectance (Target/Reference) |
|---|---|---|---|
| 325 | 997.1704 | 103.9585 | 0.104253 |
| 326 | 1049 | 116.7593 | 0.111305 |
| 327 | 1103.24 | 123.6953 | 0.112120 |
| 328 | 1169.317 | 117.176 | 0.100209 |
| 329 | 1236.683 | 124.8806 | 0.100980 |
| 330 | 1305.012 | 139.1611 | 0.106636 |
| 331 | 1376.507 | 139.5254 | 0.101362 |
| 332 | 1444.672 | 144.7024 | 0.100163 |
| 333 | 1510.975 | 158.2781 | 0.104752 |
| 334 | 1591.487 | 169.582 | 0.106556 |
The spectral curve showed minimal reflectance in blue and red regions, a slight rise in green, and a steep climb beyond 700 nm, the textbook pattern for healthy, chlorophyll-rich vegetation.
Vegetation 2 – Dense Ground Canopy
| Wavelength (nm) | Reference (W·m⁻²) | Target (W·m⁻²) | Reflectance (Target/Reference) |
|---|---|---|---|
| 325 | 983.4182 | 29.5896 | 0.030089 |
| 326 | 1040.699 | 28.04834 | 0.026951 |
| 327 | 1098.388 | 29.60889 | 0.026957 |
| 328 | 1159.736 | 32.14502 | 0.027718 |
| 329 | 1227.456 | 34.03638 | 0.027729 |
| 330 | 1296.439 | 36.34985 | 0.028038 |
| 331 | 1356.305 | 42.31494 | 0.031199 |
| 332 | 1423.762 | 40.64331 | 0.028546 |
| 333 | 1502.155 | 31.40747 | 0.020908 |
| 334 | 1578.85 | 33.5564 | 0.021254 |
This sample mirrored the broad-leaf pattern but with smoother transitions, suggesting dense, evenly hydrated vegetation.
Vegetation 3 – Dry Grass
| Wavelength (nm) | Reference (W·m⁻²) | Target (W·m⁻²) | Reflectance (Target/Reference) |
|---|---|---|---|
| 325 | 1083.566 | 50.99707 | 0.047064 |
| 326 | 1155.612 | 38.19531 | 0.033052 |
| 327 | 1226.659 | 21.21509 | 0.017295 |
| 328 | 1288.702 | 11.92773 | 0.009256 |
| 329 | 1360.558 | 27.0542 | 0.019885 |
| 330 | 1436.758 | 50.28784 | 0.035001 |
| 331 | 1502.411 | 49.18066 | 0.032735 |
| 332 | 1577.69 | 46.27222 | 0.029329 |
| 333 | 1666.812 | 47.70361 | 0.028620 |
| 334 | 1757.937 | 47.93896 | 0.027270 |
This dataset showed elevated visible reflectance and weaker NIR returns, confirming senescence or moisture loss.
Vegetation 4 – Waxy Leaf
| Wavelength (nm) | Reference (W·m⁻²) | Target (W·m⁻²) | Reflectance (Target/Reference) |
|---|---|---|---|
| 325 | 1027.857 | 615.4612 | 0.598781 |
| 326 | 1099.464 | 652.3022 | 0.593291 |
| 327 | 1169.632 | 680.7329 | 0.582006 |
| 328 | 1235.632 | 704.1807 | 0.569895 |
| 329 | 1308.058 | 752.803 | 0.575512 |
| 330 | 1382.805 | 809.8494 | 0.585657 |
| 331 | 1449.407 | 837.3228 | 0.577700 |
| 332 | 1523.088 | 877.0813 | 0.575857 |
| 333 | 1607.657 | 936.4009 | 0.582463 |
| 334 | 1696.7 | 984.0625 | 0.579986 |
Reflectance here was consistent across NIR, slightly lower than the broad-leaf sample, characteristic of thicker, wax-coated leaves that scatter less light (Lillesand, Kiefer, & Chipman, 2015).
Comparing Spectral Curves
Across 400–700 nm, reflectance remained low except for a minor bump near 550 nm (green). The sharp NIR rise beyond 700 nm in Vegetations 1 and 2 confirmed healthy foliage, while the dry grass (Vegetation 3) flattened, and the waxy leaf (Vegetation 4) reflected moderately less infrared energy.
This graphical behavior effectively links plant physiology to spectral geometry, moisture, pigment, and cell structure expressed as rising or leveling lines.
From Field Data to Spectral Library
Once I cleaned and processed all datasets, I imported them into ENVI’s Spectral Library Builder. The software generated professional reflectance plots that closely matched the Excel graphs but displayed smoother interpolation.
The similarity across both tools validated the dataset integrity and confirmed that ENVI’s endmember collection accurately represented field reflectance signatures (Campbell & Wynne, 2011).
Seeing Beyond the Visible
To extend the analysis, I used Landsat 8 imagery in ENVI to compare satellite reflectance behavior with field spectra. Two band combinations were tested:
- 4-3-2 (Natural Color): Earth as seen by the eye, vegetation in green, soil brown, water dark.
- 5-4-3 (False Color): Vegetation appears bright red, making it easier to assess health and moisture.
The 5-4-3 composite amplified the vegetation signal, showing stark red areas where chlorophyll activity was strongest — perfectly aligning with NIR-driven reflectance patterns (Jensen, 2015).
Spectral Signature Validation
Using ENVI’s Spectral Profile Tool, I extracted a sample pixel from the bright red area in the false-color image. Its curve displayed the same pattern seen in the field data, deep blue/red absorption and a steep NIR climb.
Minor differences occurred due to pixel mixing (soil and canopy within a 30 × 30 m footprint), but overall, the satellite and field data confirmed each other’s accuracy.
Deriving Vegetation Indices
To translate reflectance into measurable vegetation health, I used ENVI’s Vegetation Index Calculator to derive:
- Simple Ratio (SR) = NIR / Red
- Normalized Difference Vegetation Index (NDVI) = (NIR – Red) / (NIR + Red)
NDVI values near 0.8 represented healthy vegetation, around 0.2 for sparse vegetation, and negative values for water.
After classification, I exported the Regions of Interest (ROIs) into ArcGIS Pro, producing final vector overlays of water bodies, bare soil, and dense vegetation.
The Ground Truth
Although satellites provide the view from above, the real understanding begins below — in the data measured by instruments, processed by analysts, and verified by comparison.
This lab demonstrated that remote sensing is not remote at all. It’s a dialogue between sunlight, surface, and software. My contribution, the data processing, visualization, and interpretation, turned raw spectral values into a complete narrative of how the Earth reflects its light.
To see the planet clearly, you must process its numbers first.
References
Campbell, J. B., & Wynne, R. H. (2011). Introduction to Remote Sensing (5th ed.). Guilford Press.
Jensen, J. R. (2015). Introductory Digital Image Processing: A Remote Sensing Perspective (4th ed.). Pearson Education.
Lillesand, T. M., Kiefer, R. W., & Chipman, J. W. (2015). Remote Sensing and Image Interpretation (7th ed.). John Wiley & Sons.