Vegetation indices and GIS

Why this matters

Five bands of reflectance for every pixel is a mass of numbers no farmer can act on. A vegetation index combines the plant-related bands into one number per pixel, and zoning turns those numbers into areas you can act on, such as “this 5-rai zone needs a visit”. But without understanding what an index does and does not measure, you might fertilise a spot whose real problem is waterlogging.

Three vegetation indices

• NDVI (Rouse and colleagues, 1974; Tucker, 1979) uses the contrast between NIR, which leaves reflect strongly, and red, which chlorophyll absorbs. It ranges from −1 to 1: dense healthy vegetation is high, bare soil low and water usually negative
• NDRE (Barnes and colleagues, 2000) uses the red-edge band instead of red. The red edge is sensitive to chlorophyll content in dense canopies, and Barnes and colleagues used it with other indices to monitor crop nitrogen status
• GNDVI (Gitelson and colleagues, 1996) uses the green band instead of red

Because indices are ratios they partly cancel differences in brightness, but they do not replace the calibration of module 2.

pixels = {  # reflectance: green, red, red edge, NIR
    "healthy": {"green": 0.08, "red": 0.04, "rededge": 0.20, "nir": 0.50},
    "stressed": {"green": 0.10, "red": 0.09, "rededge": 0.22, "nir": 0.35},
}
nd = lambda a, b: (a - b) / (a + b)
for name, p in pixels.items():
    print(f"{name:<9} NDVI {nd(p['nir'], p['red']):.3f}  NDRE {nd(p['nir'], p['rededge']):.3f}  "
          f"GNDVI {nd(p['nir'], p['green']):.3f}")

healthy   NDVI 0.852  NDRE 0.429  GNDVI 0.724
stressed  NDVI 0.591  NDRE 0.228  GNDVI 0.556

Zoning the field

Our rice field is divided into a 3 × 6 grid of cells, each 50 × 53.3 m (about 1.67 rai). In practice each cell would hold the mean of its pixels, and cells are assigned to zones with criteria agreed before seeing the results.

NIR = [[0.48, 0.50, 0.47, 0.44, 0.36, 0.30], [0.49, 0.51, 0.48, 0.42, 0.34, 0.28], [0.47, 0.49, 0.46, 0.43, 0.37, 0.33]]
RED = [[0.05, 0.05, 0.06, 0.07, 0.10, 0.13], [0.05, 0.04, 0.05, 0.08, 0.11, 0.14], [0.06, 0.05, 0.06, 0.07, 0.10, 0.12]]
CELL_RAI = (300 / 6) * (160 / 3) / 1600

def zone(v): return "high" if v > 0.70 else ("medium" if v >= 0.50 else "low")

counts = {"high": 0, "medium": 0, "low": 0}
for n_row, r_row in zip(NIR, RED):
    values = [(n - r) / (n + r) for n, r in zip(n_row, r_row)]
    print(" ".join(f"{v:.2f}" for v in values))
    for v in values:
        counts[zone(v)] += 1
for z, c in counts.items():
    print(f"{z:<6} {c:>2} cells = {c * CELL_RAI:.2f} rai")

0.81 0.82 0.77 0.73 0.57 0.40
0.81 0.85 0.81 0.68 0.51 0.33
0.77 0.81 0.77 0.72 0.57 0.47
high   11 cells = 18.33 rai
medium  4 cells = 6.67 rai
low     3 cells = 5.00 rai

Doing it in QGIS

QGIS computes indices from reflectance rasters with the Raster Calculator, for example ("nir@1" - "red@1") / ("nir@1" + "red@1"). Zones are then made by reclassifying value ranges and converting to polygons to compute areas. Use a coordinate system in metres (such as UTM 47N) and divide by 1,600 to get rai. The map layout must show the image date, the index and criteria used, a scale and the ground-truth status (see map layout in UAT 361).

Module lab

Common mistakes

Summary

• NDVI uses NIR and red, NDRE NIR and red edge, GNDVI NIR and green; each has its own range
• Zone with thresholds set in advance, then compute areas in rai in a metre-based coordinate system
• Indices show where something is abnormal; ground truth shows why
• Delivered maps must state the date, index, thresholds and verification status

Check your understanding

1. NIR = 0.45 and Red = 0.05. What is NDVI?
2. NIR = 0.45 and RedEdge = 0.25. What is NDRE?
3. A field of 18 cells, each 1.67 rai, has 3 cells in the low zone. How many rai is the low zone?
4. Why should a low zone running along one edge of the field be checked for water first?
5. Does a low NDVI tell you the cause of the problem?