NDMI: Technical Guide for Environmental Consultants - Calculation, Applications, Limitations

The NDMI (Normalized Difference Moisture Index) is a spectral index derived from Sentinel-2 images that measures the water content in vegetation canopies. Values range from −1 to +1: high values indicate greater leaf moisture and favourable water conditions, while low or negative values indicate water stress or absence of vegetation.

The ndmi formula — NDMI = (NIR − SWIR1) / (NIR + SWIR1) — exploits the sensitivity of SWIR1 (~1610 nm) to water in leaf tissues: at 1610 nm, water absorbs, and the NIR (~842 nm) depends on cell structure. The ratio isolates the water signal (Gao, 1996). The ndmi vegetation index complements the NDVI: reading them together is useful in ESG and EIA.

Those who do not build pipelines from scratch can use tools like the iLoveNatura NDMI task: it starts from Sentinel-2 L2A, applies the formula pixel by pixel, and provides monthly time series, 10 m maps, and a comparison between the site (ROI) and control area (CA) on an A–E scale. Useful for EIA/SEA and ESRS reporting.

To obtain the NDMI index, two spectral bands are required: near-infrared (NIR, around 860 nm) and SWIR1 (around 1610 nm). Work is conducted on surface reflectance that has already been atmospherically corrected, using Level-2A products and masking clouds and shadows with scene quality bands. Without these precautions, the returned values are difficult to compare over time.

The most commonly used satellite sources are two. ESA's Sentinel-2 offers a resolution of 10–20 m and a revisit time of about 5 days, utilising bands B8 and B11 and is free of charge; NASA/USGS's Landsat-8/9 operates at 30 m with a 16-day interval and uses B5 and B6. The choice depends on the scale of the analysis and the historical depth required.

On the software front, the most popular environments are Google Earth Engine, QGIS with the Semi-Automatic Classification plugin, and ESA SNAP. For those who do not wish to write code, the NDMI task available on iLoveNatura can be used, which produces maps ready for EIA/SEA and ESG.

The NDMI is a powerful tool, but not infallible. The most common mistake is confusing it with the NDVI: although both are normalised indices, they measure different things — moisture in the first case, photosynthetic vigour in the second — and are not interchangeable in an ESRS E4 report.

Another pitfall concerns atmospheric noise. Clouds, shadows, and fog degrade the signal: systematic masking is required, for example with the Scene Classification Layer of Sentinel-2 or the QA_PIXEL of Landsat. Skipping this step leads to false positives on water stress.

The SWIR band must be chosen carefully: for the NDMI, SWIR1 (~1.6 µm) is used; SWIR2 (~2.2 µm) is instead the basis of the NBR, typical for fires. Finally, the NDMI depends on species, phenology, and soil: it is better interpreted comparatively, on multi-year local baselines, rather than applying universal thresholds. Filtering pixels with low vegetation cover is almost always necessary.

In environmental consultancy, the NDMI is not just an academic indicator: it is applied in practical contexts. In EIA/SEA procedures, it is used to capture the pre-operation vegetative state, size mitigations, and monitor post-implementation progress, especially on linear infrastructures such as roads, power lines, or agro-photovoltaic plants.

On the ESG and CSRD/ESRS E4 front, the NDMI provides quantitative and traceable KPIs on impacts to vegetative ecosystems: a historical series allows for objective documentation of whether a mitigation is working or if a site is losing ecological functionality.

At a territorial scale, the NDMI helps identify degraded areas to prioritise for Nature-based Solutions, forest restoration, and renaturation projects. It also becomes a prevention tool: abnormal variations indicate early water stress, useful for fire prevention plans, forest management, and climate adaptation for businesses and administrations.