NDWISentinel-2wateranalysistutorial

Mapping Water Bodies from Space: NDWI and MNDWI Explained

Kazushi MotomuraJune 9, 20255 min read
Mapping Water Bodies from Space: NDWI and MNDWI Explained

Quick Answer: NDWI uses green and NIR bands to detect water based on strong NIR absorption. MNDWI replaces NIR with SWIR, which better suppresses built-up area false positives. Use NDWI for natural water bodies in rural areas. Use MNDWI when urban areas or bare soil might be confused with water. Both produce values from -1 to +1, where positive values indicate water.

Why Water Detection Isn't as Simple as It Looks

Water appears dark in most satellite imagery — so why do we need specialized indices to detect it? Because several other surfaces also appear dark:

  • Shadows from clouds, mountains, and buildings
  • Asphalt and dark rooftops in urban areas
  • Dark volcanic soils
  • Burned areas after wildfires

A simple brightness threshold will misclassify all of these as water. Spectral indices exploit the specific way water absorbs and reflects different wavelengths to separate it from these dark-but-not-water surfaces.

The Two Main Water Indices

NDWI (Normalized Difference Water Index)

Proposed by McFeeters (1996), NDWI uses the contrast between green light (which water reflects somewhat) and near-infrared (which water absorbs strongly):

NDWI = (Green - NIR) / (Green + NIR)

For Sentinel-2: (B3 - B8) / (B3 + B8)

  • Water: positive values (0.1 to 0.8)
  • Vegetation: negative values (-0.8 to -0.1)
  • Bare soil: negative values, close to zero

MNDWI (Modified Normalized Difference Water Index)

Proposed by Xu (2006), MNDWI replaces NIR with SWIR to reduce confusion with built-up areas:

MNDWI = (Green - SWIR) / (Green + SWIR)

For Sentinel-2: (B3 - B11) / (B3 + B11)

  • Water: positive values (0.1 to 0.8)
  • Built-up areas: strongly negative (unlike NDWI where they can be near-zero or positive)
  • Bare soil: negative values

Why MNDWI Outperforms NDWI in Urban Areas

The key problem with NDWI is that built-up surfaces (concrete, asphalt) can produce near-zero or slightly positive NDWI values — dangerously close to actual water signatures. This happens because urban materials have relatively low NIR reflectance.

MNDWI solves this by using SWIR instead of NIR. Built-up surfaces have high SWIR reflectance, pushing them firmly into negative MNDWI territory. The separation between water (positive) and urban (strongly negative) becomes much cleaner.

Surface TypeNDWIMNDWI
Deep water0.4 to 0.70.5 to 0.8
Shallow/turbid water0.1 to 0.30.2 to 0.4
Vegetation-0.5 to -0.1-0.4 to 0.0
Bare soil-0.3 to 0.1-0.5 to -0.1
Built-up areas-0.2 to 0.2 ⚠️-0.6 to -0.2 ✅

Practical Applications

Flood Extent Mapping

After a flood event, comparing pre-flood and post-flood NDWI images reveals inundated areas. Areas where NDWI changed from negative to positive indicate new water presence. This complements SAR-based flood mapping — SAR works through clouds during the event, while optical-based NDWI provides cleaner water boundaries once skies clear.

Reservoir and Lake Monitoring

Track seasonal water level changes by computing NDWI time series. The water boundary (NDWI = 0 threshold) shifts inward during dry seasons and expands during wet seasons. This is particularly useful for monitoring drought impacts on reservoirs.

Wetland Delineation

Wetlands present a challenge because they mix water and vegetation signals. In these transitional zones, both NDWI and NDVI are useful — high NDVI with near-zero NDWI suggests vegetated wetland, while positive NDWI with low NDVI indicates open water.

Urban Water Features

Parks, ponds, rivers running through cities — MNDWI is essential here. NDWI alone will produce too many false positives from dark rooftops and parking lots.

Tips for Better Water Mapping

Choose the Right Threshold

The default threshold of zero (water = positive, land = negative) works as a starting point, but optimal thresholds vary by scene:

  • Clear, deep water: threshold of 0.1-0.2 reduces noise
  • Turbid water: you may need to lower the threshold to 0.0 or even slightly negative
  • Mixed pixels at water edges: a lower threshold captures more of the water boundary but may include wet soil

Watch for Cloud Shadows

Cloud shadows over land can produce positive NDWI values that look like water. Always inspect your results against the true-color image. If a "water body" appears next to a cloud and has an irregular shape matching the cloud outline, it's a shadow.

Consider Seasonal Timing

The best time to map permanent water bodies is during the dry season, when temporary water features have receded. For flood mapping, compare against a dry-season baseline.

Combine with SAR

For the most robust water mapping, combine optical water indices with SAR backscatter analysis. Water appears dark in both SAR and NIR, but the failure modes are different — SAR isn't affected by cloud shadows, and optical isn't affected by wind-roughened water surfaces (which can appear bright in SAR).

Try It in Off-Nadir Delta

Off-Nadir Delta supports both NDWI and MNDWI as built-in visualization options for Sentinel-2 data:

  1. Search for Sentinel-2 imagery over a lake, river, or coastal area
  2. Select NDWI or MNDWI from the visualization options
  3. Compare results — especially in areas with mixed urban and water features
  4. Toggle between NDWI and MNDWI to see how built-up area false positives disappear with MNDWI

For flood monitoring, load imagery from before and after a flood event and switch between NDWI views to visualize the change.

Kazushi Motomura

Kazushi Motomura

Remote sensing specialist with 10+ years in satellite data processing. Founder of Off-Nadir Lab. Master's in Satellite Oceanography (Kyushu University).