NDPI (Normalized Difference Pond Index)
NDPI (Normalized Difference Pond Index) is a computational water body index focused on detecting small water bodies that have vegetation or are surrounded by denser plant cover. The false-color image is calculated based on reflected shortwave infrared radiation (SWIR) and green light (G) values using the following formula:
The NDPI algorithm was developed by Lacaux et al. (2007) to distinguish vegetation and water bodies surrounding it. This index uses remote sensing data to separate ponds and smaller water bodies from surrounding vegetation, enabling more accurate mapping and monitoring in wetland areas. NDPI is particularly useful in densely vegetated regions where traditional methods may be ineffective at distinguishing water from land.
This map layer is most useful and informative during the period of active vegetation. Figure 2 visualizes the color tone changes of this index throughout different seasons. Generally, we observe that in early spring the map layer appears in yellow tones, becoming bluer and grayer as summer and greenery increase, and in autumn and cold seasons yellow tones dominate again. When the ground is bare in winter, a yellow background persists, but after snow melts, the map layer turns uniformly blue (Figure 2).
Figure 1. NDPI false-color image through different seasons
If more green light than shortwave infrared is reflected from an area, the index is negative. On the map, such areas appear from blue to dark blue. The darker the blue, the more likely it is a water body or water-saturated vegetation. Detecting and classifying ponds in densely vegetated areas can be a challenging task. NDPI can be a good tool to help better distinguish water bodies from surrounding soil and vegetation, refining mapping and analysis.
Figure 2. Highlighting small water bodies with NDPI false-color image
Exceptions include some narrower rivers, which due to edge effects can appear yellow on satellite images (Figure 3). In some cases, a pixel may contain data both from the water surface and the adjacent land (SWIR resolution 20m). Additionally, shoreline vegetation may influence the depiction of small or narrow water bodies on satellite images. Also, while NDPI is quite effective for detecting surface water like small water bodies and water-saturated vegetation, its accuracy is limited in densely vegetated areas where optical satellites cannot see through, and on open areas it can only detect thin surface water layers.
Figure 3. Võhandu River in NDPI false-color image shortly before the 2023 Võhandu marathon
It is also important to note that in this false-color image, clouds, snow, and ice appear blue like water bodies, which can lead to inaccurate analyses if unnoticed. Cloud haze (Figure 4) can distort data by making areas appear more water-saturated than they really are. To mitigate this, we recommend using the false-color image alongside the RGB image to confirm no clouds or shadows are present, and to remove cloud effects during data processing when possible.
Figure 4. Cloud haze near Vändra shown in RGB image (left) and cloud effect in NDPI image (right, haze in dark blue)
If more shortwave infrared radiation than green light is reflected from an area, the index is positive. On the map, such areas appear from grayish yellow to bright yellow. For example, plowed and mowed fields appear yellow because soil generally reflects more shortwave infrared than green light. Drained peat fields also appear yellow.
Figure 5. Peat fields and well-visible blue gravel roads in the NDPI false-color image (see more in Satiladu)
The negative NDPI values of mines and gravel roads are an interesting phenomenon resulting from their optical properties. This is not necessarily caused by strong green light reflection, but rather by low shortwave infrared reflection (SWIR), influenced by soil moisture content, mineral composition, or materials absorbing shortwave infrared radiation, such as clay or organic particles. On false-color images, this can create a visual similarity to water even though these are completely different surfaces. This highlights how important it is to understand the properties of analyzed areas when using spectral indices.
Figure 6. Mines in NDPI false-color image (reference to Satiladu)