MNDWI: The Water Index That Stops Mistaking Cities for Lakes
The original NDWI has a habit of painting rooftops and parking lots as water. Xu's modification fixes that with one swap — SWIR for NIR — and it has become the default water index for anywhere people live.
The physical basis
Built-up surfaces reflect strongly in SWIR while water absorbs it almost entirely, so replacing NIR with SWIR widens the water/built-up gap that plain NDWI collapses. Water stays firmly positive; concrete and asphalt drop clearly negative.
The math behind it
where Green is green reflectance, SWIR1 is shortwave-infrared 1 reflectance (~1.6 µm).
How to interpret the output
> 0 water · < 0 non-water. Separates water from built-up far better than NDWI, so it is the default urban water index.
The bands it needs
| Sentinel-2 | Green B3 · SWIR1 B11 |
| Landsat 8/9 | Green B3 · SWIR1 B6 |
Real-world use
MNDWI is the workhorse for urban water mapping, flood delineation in populated areas, and long-term shoreline and reservoir change detection.
Watch out for
SWIR is coarser than the visible bands (20 m on Sentinel-2, 30 m on Landsat), so fine channels and narrow streams can be missed, and terrain or cloud shadow can still fool it — pair with a shadow mask in mountainous scenes.
Compute MNDWI on your own study area
Skip the code. Draw or upload a boundary and Spatial Research Suite runs MNDWI on live Sentinel-2 or Landsat imagery — with cloud masking, exports and citations built in.
Run this analysis in GISforus →Frequently asked
MNDWI or NDWI — which is better for cities?
MNDWI, by a wide margin. Swapping NIR for SWIR suppresses the built-up false positives that plague the original NDWI.
Why can MNDWI miss small streams?
The SWIR band is coarser (20–30 m), so channels narrower than a pixel or two can fall below detection.