Fact-checked by Grok 2 weeks ago

Normalized difference water index

The Normalized Difference Water Index (NDWI) is a remote sensing spectral index developed to delineate and enhance open water features, such as lakes, rivers, and wetlands, in satellite imagery by contrasting the reflectance properties of water in the green and near-infrared (NIR) bands. Introduced by Samuel K. McFeeters in 1996, it provides a normalized ratio that highlights water bodies against surrounding vegetation, soil, and built-up areas, with the index calculated using the formula NDWI = \frac{\text{Green} - \text{NIR}}{\text{Green} + \text{NIR}}, where Green represents reflectance in the green band (typically around 0.52–0.60 μm) and NIR in the near-infrared band (around 0.77–0.90 μm). For Landsat Thematic Mapper imagery, this corresponds to Band 2 (Green) and Band 4 (NIR). NDWI values range from -1 to +1, where positive values (typically > 0) indicate the presence of due to its high in the and strong , while negative values (≤ 0) correspond to non-water features like and , which reflect more in the . In practice, thresholds such as 0.3 may be applied in environments to refine detection and reduce false positives from shadows or built structures. The index's sensitivity to also allows for qualitative assessments of , though it can overestimate extent in areas with dense or confuse it with . A variant, the Modified Normalized Difference Water Index (MNDWI), proposed by Hanqiu Xu in 2006, improves upon the original by substituting the mid-infrared (SWIR) band for NIR in the formula MNDWI = \frac{\text{Green} - \text{SWIR}}{\text{Green} + \text{SWIR}}, enhancing open water detection while better suppressing noise from built-up land, soil, and vegetation. Note that a separate NDWI, developed by Bo-Cai Gao in 1996, uses NIR and SWIR bands to monitor vegetation liquid water content rather than surface water bodies, and the two should not be conflated despite the shared acronym. NDWI and its variants are applied across , including inundation mapping, where they delineate temporary water extent during events like the using Landsat data. In and , NDWI aids in identifying standing water for breeding sites, achieving high accuracy (e.g., 78.4% in detecting swimming pools). Agricultural and hydrological studies employ it for tracking water body changes, irrigation assessment, and drought monitoring via platforms like Landsat, , and MODIS. These applications leverage freely available satellite data, enabling cost-effective, large-scale analysis of and dynamics.

Introduction

Definition

The Normalized Difference Water Index (NDWI) is a remote sensing-derived that utilizes the difference between two specific bands in or aerial imagery to enhance and detect signals associated with liquid . It operates on the principle of contrasting properties where water exhibits strong in certain wavelengths and high reflectance in others, thereby isolating water-related features from surrounding . The primary purpose of NDWI is to quantify the presence and extent of liquid , either as moisture content within canopies or as open bodies such as lakes, rivers, and wetlands. This makes it a valuable tool for , including assessment, mapping, and ecosystem evaluation in various geospatial applications. Mathematically, NDWI follows the normalized difference structure, expressed as \frac{\text{Band A} - \text{Band B}}{\text{Band A} + \text{Band B}}, where Band A and Band B are bands selected for their sensitivity to characteristics. This formulation normalizes the ratio to produce values ranging from -1 to +1, with the division by the sum of the bands ensuring and improved contrast for detection across diverse imaging conditions. Variants of NDWI exist, tailored to emphasize either vegetation water content or open water surfaces, though both adhere to the core normalized difference framework.

Key Variants

The Normalized Difference Water Index (NDWI) encompasses two primary variants, each tailored to distinct applications through specific spectral band selections. Gao's NDWI focuses on quantifying liquid water content within canopies, leveraging the near-infrared () and short-wave infrared (SWIR) bands to capture subtle absorption features associated with moisture in leaves. This variant is particularly effective for detecting variations in water status, such as those indicative of stress. In contrast, McFeeters' NDWI is optimized for extracting open bodies from imagery, employing the and bands to exploit the contrast between water surfaces—which exhibit low and higher —and surrounding features like and . This approach enhances the delineation of standing , such as lakes and , by highlighting high-contrast boundaries. The core distinction between these variants lies in their targeted phenomena: Gao's addresses biophysical within for agricultural and ecological monitoring, whereas McFeeters' prioritizes mapping for hydrological applications. Both are routinely implemented on multispectral data from platforms like Landsat and , where the choice of bands dictates the suitability for versus body analysis.

History and Development

Gao's Contribution (1996)

In 1996, introduced the Normalized Difference Water Index (NDWI) in his seminal paper published in Remote Sensing of Environment, addressing a critical gap in capabilities for monitoring . The motivation stemmed from the limitations of established indices like the (NDVI), which primarily reflects density but saturates at high leaf area indices (typically ≥3) and shows insensitivity to variations in . aimed to develop a complementary index suitable for global-scale assessment of vegetation moisture, particularly in agricultural and forestry applications, by leveraging the near-infrared (NIR) and shortwave infrared (SWIR) spectral bands to detect liquid water absorption features. Gao's initial validation involved hyperspectral data from the Airborne Visible/Infrared Imaging Spectrometer (AVIRIS), testing the index across diverse ecosystems to demonstrate its sensitivity to vegetation water status. Key datasets included AVIRIS imagery from Jasper Ridge, (acquired on June 2, 1992), featuring mixed oak woodlands, grasslands, and , and from the High Plains region of (acquired on August 7, 1990), encompassing agricultural fields and rangelands. Laboratory measurements of leaf water content from these sites showed strong linear correlations with NDWI values, with the index yielding positive values for green, hydrated and negative values for dry or senescent areas, effectively highlighting spatial patterns such as drainage influences in Jasper Ridge. This contribution pioneered quantitative of canopy-level vegetation water content, establishing NDWI as a foundational tool that has influenced subsequent models for monitoring and risk assessment. For instance, it has been integrated into risk mapping frameworks, where higher NDWI values indicate reduced flammability due to elevated levels. The work's emphasis on hyperspectral data also spurred advancements in multi-band indices for detection.

McFeeters' Contribution (1996)

In 1996, S.K. McFeeters published the seminal paper introducing the Normalized Difference Water Index (NDWI) specifically for delineating open water features in remotely sensed imagery. The work appeared in the International Journal of Remote Sensing, volume 17, issue 7, pages 1425–1432, and proposed NDWI as an advancement over traditional methods for extracting water bodies from satellite data. McFeeters developed NDWI to address limitations in simple band ratioing techniques applied to Landsat Thematic Mapper (TM) imagery, which often struggled to isolate water pixels amid mixed . The index was designed to maximize the contrast between water surfaces and surrounding non-water elements, such as and , by leveraging the and near-infrared bands where water exhibits distinct properties. Initial testing of the NDWI involved applying it to multiple Landsat TM scenes from diverse environments, including urban and rural settings with varying water body sizes and turbidities. Results demonstrated that NDWI values for water pixels clustered positively, enabling straightforward thresholding to accurately map open water features while minimizing false positives from . This approach proved effective for both clear and turbid waters, highlighting its robustness for practical applications. McFeeters' NDWI has since become a foundational tool in , widely adopted as a standard for mapping, inventory, and extraction in . The 1996 paper underscores its enduring influence on water-related geospatial analysis.

Formulation and Calculation

Gao's NDWI Formula

Gao's Normalized Difference Water Index (NDWI) is formulated as the normalized difference between near-infrared () and short-wave infrared (SWIR) values, specifically designed to quantify in canopies from . The index leverages the contrasting spectral responses of these bands to : (typically in the 0.7–1.3 μm range) is strongly reflected by healthy but absorbed by bodies, while SWIR (1.3–2.5 μm range) is highly absorbed by liquid within leaves and , amplifying sensitivity to variations. The precise formula, as proposed by Gao, is: \text{NDWI} = \frac{\rho_{\text{NIR}} - \rho_{\text{SWIR}}}{\rho_{\text{NIR}} + \rho_{\text{SWIR}}} where \rho_{\text{NIR}} represents the reflectance in the NIR band (e.g., centered at 0.86 μm) and \rho_{\text{SWIR}} the reflectance in the SWIR band (e.g., centered at 1.24 μm). This structure normalizes the difference to a range of -1 to +1, where values approaching +1 indicate high vegetation water content (due to elevated NIR and reduced SWIR absorption) and values near -1 suggest low water content or non-vegetated surfaces. The choice of these wavelengths ensures both bands penetrate to similar depths within the canopy, minimizing confounding effects from soil background or atmospheric interference. To compute Gao's NDWI, surface reflectance values are first extracted from calibrated for the selected NIR and SWIR bands, ensuring atmospheric correction to derive accurate \rho values. These reflectances are then substituted directly into the , yielding a per-pixel map suitable for quantitative analysis of water stress or monitoring. For modern sensors like , Band 8 ( at 0.842 μm) and Band 11 (SWIR at 1.61 μm) are commonly used as proxies for the original MODIS channels specified by , maintaining the index's sensitivity to liquid water.

McFeeters' NDWI Formula

The McFeeters' Normalized Difference Water Index (NDWI) is computed using the formula: \text{NDWI} = \frac{X_{\text{[green](/page/Green)}} - X_{\text{[nir](/page/NIR)}}}{X_{\text{[green](/page/Green)}} + X_{\text{[nir](/page/NIR)}}} where X_{\text{[green](/page/Green)}} represents the in the band (typically 0.5–0.6 μm) and X_{\text{[nir](/page/NIR)}} is the in the near-infrared band (0.7–1.3 μm). This formulation leverages the spectral properties of , which exhibits high in the wavelengths due to low and very low in the near-infrared due to strong , resulting in positive NDWI values for open water features. In contrast, terrestrial and reflect highly in the near-infrared while showing lower , leading to negative or near-zero NDWI values that help suppress these non-water elements. Computation involves pixel-wise application of the formula to calibrated imagery, where reflectance values are derived from digital numbers after atmospheric correction to ensure comparability across scenes. The in the denominator mitigates effects from topographic variations, illumination differences, and noise, providing robustness for mixed pixels containing partial water coverage. For example, in imagery, McFeeters' NDWI utilizes Band 3 (, centered at 0.533 μm) for X_{\text{green}} and Band 5 (near-infrared, centered at 0.865 μm) for X_{\text{nir}}.

Required Spectral Bands

The computation of the Normalized Difference Water Index (NDWI) requires multispectral data from sensors capable of capturing in specific regions, depending on the variant: the green band (approximately 0.52–0.59 μm) and near-infrared () band (centered around 0.77–0.86 μm) for McFeeters' formulation, or the band and shortwave infrared (SWIR) band (typically 1.57–1.65 μm or 2.11–2.29 μm) for Gao's formulation. Atmospheric correction is essential to convert raw radiance data to top-of-atmosphere (TOA) or, preferably, surface values, as uncorrected data can introduce errors from and absorption by atmospheric constituents. Commonly used satellite sensors include the Landsat series, , and MODIS, each providing the requisite bands with varying spatial resolutions and revisit frequencies. For the Landsat Thematic Mapper (TM) and Enhanced Thematic Mapper Plus (ETM+), McFeeters' NDWI utilizes Band 2 (green: 0.52–0.60 μm) and Band 4 (: 0.76–0.90 μm), while Gao's NDWI employs Band 4 () and Band 5 (SWIR: 1.55–1.75 μm). Similarly, for Operational Land Imager (OLI), McFeeters' variant uses Band 3 (green: 0.533–0.590 μm) and Band 5 (: 0.851–0.879 μm), with Gao's using Band 5 () and Band 6 (SWIR: 1.566–1.651 μm).
Satellite/SensorVariantBands UsedWavelengths (μm)Spatial Resolution
Landsat 5 TM/ETM+McFeeters2 (Green), 4 (NIR)0.52–0.60, 0.76–0.9030 m
Landsat 5 TM/ETM+Gao4 (NIR), 5 (SWIR)0.76–0.90, 1.55–1.7530 m
Landsat 8 OLIMcFeeters3 (Green), 5 (NIR)0.533–0.590, 0.851–0.87930 m
Landsat 8 OLIGao5 (NIR), 6 (SWIR)0.851–0.879, 1.566–1.65130 m
Sentinel-2 MSIMcFeeters3 (Green), 8 (NIR)0.560 (central), 0.842 (central)10 m (B3), 10 m (B8)
Sentinel-2 MSIGao8 (NIR), 11 (SWIR)0.842 (central), 1.610 (central)10 m (B8), 20 m (B11)
MODIS (Terra/Aqua)McFeeters4 (Green), 2 (NIR)0.545–0.565, 0.841–0.876500 m
MODIS (Terra/Aqua)Gao2 (NIR), 5 (SWIR)0.841–0.876, 1.230–1.250500 m
For MultiSpectral Instrument (MSI), McFeeters' NDWI is calculated with Band 3 (green: central 0.560 μm) and Band 8 (: central 0.842 μm), whereas Gao's uses Band 8 () and Band 11 (SWIR: central 1.610 μm). MODIS, with its broader swath for global monitoring, applies Band 4 (green: 0.545–0.565 μm) and Band 2 (: 0.841–0.876 μm) for McFeeters' variant, and Band 2 () and Band 5 (SWIR: 1.230–1.250 μm) for Gao's, as the latter SWIR band aligns closely with the optimal sensitivity for water content. Preprocessing steps are critical for reliable NDWI computation, including conversion to surface reflectance via algorithms like Landsat's Collection 2 processing or Sentinel-2's Sen2Cor atmospheric correction, which account for aerosols, , and . Cloud and cloud shadow masking must also be applied using quality flags (e.g., Landsat pixel QA band or Sentinel-2 scene classification layer) to exclude contaminated that could bias water index values.

Interpretation

Value Ranges and Meanings

The Normalized Difference Water Index (NDWI) produces output values ranging from -1 to +1 due to its normalized difference formulation, which scales the difference in between two bands relative to their sum. For McFeeters' NDWI (open detection), negative values (typically ≤ 0) indicate non-water features, such as dry , built-up areas, or , where near-infrared () exceeds green band . Values approaching +1 signify open surfaces due to strong green and absorption by . A value of 0 represents balanced green and , often associated with sparse or transitional conditions. McFeeters' NDWI provides sharper delineation for open , where values exceeding +0.4 typically denote pure water bodies due to pronounced green- contrast. For Gao's NDWI (vegetation water content monitoring), negative values indicate low , such as in dry soil or water-stressed , where short-wave infrared (SWIR) exceeds . Values approaching +1 signify high vegetation water content, corresponding to saturated or well-hydrated canopies due to exceeding SWIR. A value of 0 represents balanced and SWIR . Gao's NDWI variant demonstrates higher to subtle variations in vegetation , with values in the range of +0.1 to +0.3 commonly indicating moderately hydrated . Negative values, in particular, arise from dominant SWIR reflection without significant water presence, as seen in bare or dry conditions. NDWI results are typically visualized in (GIS) software using grayscale or color ramps to enhance water-related features, facilitating clearer interpretation of moisture patterns.

Thresholding Techniques

Thresholding techniques for the Normalized Difference Water Index (NDWI) involve applying specific cutoff values to classify pixels as , , or non-water features, enabling the delineation of bodies or vegetation water status from imagery. For McFeeters' NDWI, which emphasizes open water detection, a standard of ≥0.3 is commonly used to identify water pixels, while values <0.3 classify as non-water, helping to minimize confusion with built-up areas in urban environments. In contrast, Gao's NDWI, designed for monitoring vegetation , typically employs thresholds such as <0 or <0.1 to detect water-stressed vegetation, where positive values indicate sufficient canopy water and negative or low values signal stress or dry conditions. To address variations due to seasonal changes, atmospheric conditions, or sensor differences, dynamic thresholding methods adapt thresholds automatically rather than relying on fixed values. One widely adopted approach is , an technique that determines the optimal threshold by maximizing inter-class variance in the NDWI , effectively separating water from non-water classes without manual intervention. Machine learning-based adaptive thresholding, such as support vector machines or random forests trained on local image statistics, further refines this by incorporating contextual factors like heterogeneity, improving robustness across diverse landscapes. Post-processing enhances the binary masks generated from thresholding by applying morphological operations to reduce and refine boundaries. Erosion removes small isolated pixels or thin features that may represent artifacts, while expands water bodies to reconnect fragmented areas, often using a structuring element like a 3x3 for optimal smoothing. Additionally, integrating digital models (DEMs) allows for terrain corrections, such as combining NDWI with topographic wetness indices to adjust thresholds in hilly or shadowed regions, preventing misclassification due to slope-induced illumination variations. Validation of these thresholding techniques typically involves accuracy assessments against data, such as manually delineated water maps or high-resolution imagery. In settings, where built-up noise is prevalent, studies report coefficients exceeding 0.8 for NDWI-based classifications, indicating substantial agreement beyond chance and confirming the reliability of refined thresholds and post-processing.

Applications

Vegetation Water Content Monitoring

Gao's NDWI, formulated as the normalized difference between near-infrared and shortwave infrared bands, has been widely applied to monitor water content, enabling the assessment of moisture status across various ecosystems. In , it supports monitoring by detecting reductions in canopy water, facilitating scheduling to mitigate crop stress and aiding in prediction through correlations with water availability during critical growth stages. In , NDWI-derived estimates of live moisture content inform fire danger ratings by quantifying the levels of foliage, which influence ignition probability and fire spread rates. Case studies demonstrate NDWI's utility in tracking phenological changes and drought impacts in rangelands. For instance, MODIS-derived NDWI time series over the U.S. revealed rapid declines in vegetation water content during severe events following 2010, such as the 2012 episode, allowing for the monitoring of grassland responses and recovery patterns over multiple seasons. Similarly, in , NDWI time-series analysis has been used to detect anomalies in forest water stress during extreme events, including the 2023 heatwave, where it highlighted spatially variable declines in tree moisture levels across temperate regions. More recently, as of 2025, NDWI from has been applied to assess vegetation responses to the 2024 European droughts, aiding in early detection of water stress in agricultural areas. NDWI is often integrated with the (NDVI) to enhance biophysical modeling of vegetation dynamics, combining structural and hydrological information for more robust simulations of productivity and water use efficiency. Time-series applications of this combined approach enable in moisture patterns, supporting early warnings for propagation in diverse landscapes. The primary benefits of NDWI for vegetation water content monitoring include its non-invasive nature, allowing large-scale assessments over remote or expansive areas without ground-based measurements, and its sensitivity to changes in canopy water thickness equivalent to up to 0.5 cm of liquid water, which captures subtle variations in plant hydration before visible wilting occurs.

Surface Water Body Delineation

McFeeters' Normalized Difference Water Index (NDWI) has been widely applied in and environmental management for delineating surface water bodies, including extent mapping, lake and monitoring, and . In extent mapping, NDWI leverages optical to identify inundated areas rapidly, enabling real-time assessment of impacts. For instance, during the , NDWI derived from data was used to verify extents, delineating an inundation area exceeding 30,000 km² across affected regions. Similarly, in lake and monitoring, NDWI facilitates the tracking of water surface area changes over time, supporting by quantifying seasonal or long-term variations in storage volumes. efforts employ NDWI to map open water components within complexes, aiding in conservation planning and assessment by distinguishing water from surrounding and . Integration of McFeeters' NDWI with () data enhances its utility, particularly for all-weather surface water delineation in cloudy or vegetated environments. Optical NDWI provides high spectral contrast for clear conditions, while complements it by penetrating clouds and vegetation, allowing combined approaches to achieve more robust and water body mapping. In , NDWI assists in separating water bodies from impervious surfaces, supporting land-use classification and management by highlighting small water features amid built environments. Key benefits of McFeeters' NDWI include its ability to provide high contrast for detecting small water bodies, such as larger than one , which is crucial for detailed inventories in heterogeneous landscapes. Additionally, its sensitivity to temporal changes enables monitoring of dynamic processes like or through multi-date , offering insights into environmental shifts without extensive ground surveys.

Limitations and Improvements

Identification of Common Limitations

The Normalized Difference Water Index (NDWI) is susceptible to atmospheric interference, including and by aerosols and , which can distort values in the relevant spectral bands and reduce mapping accuracy, particularly under hazy or turbid conditions. These effects are more pronounced for variants relying on visible and near-infrared bands, as atmospheric path radiance adds noise to the signal, often necessitating preprocessing with correction algorithms to mitigate biases in water detection. Mixed pixels pose another inherent challenge, where sub-pixel fractions of water mixed with land cover in heterogeneous landscapes, such as river edges or urban fringes, lead to underestimation or overestimation of water extent due to spectral blending. This issue is exacerbated in moderate- to coarse-resolution imagery, where narrow water bodies like streams are often unresolved, resulting in diluted NDWI values that complicate delineation without sub-pixel unmixing techniques. Variant-specific limitations further constrain NDWI applications; McFeeters' formulation, using green and near-infrared bands, is prone to false positives from built-up areas and shadows, as urban reflectance and shadowed regions exhibit low near-infrared signals similar to open water, leading to misclassification in or mountainous terrains. In contrast, Gao's NDWI, designed for vegetation liquid water content via near-infrared and short-wave infrared bands, is affected by leaf variations, such as angle dynamics, which alter canopy reflectance and confound water content estimates in layered forests. Temporal factors also limit NDWI reliability, as seasonal vegetation changes, including phenological shifts in water content, alter baseline index values and require site-specific to avoid misinterpretation of or signals. Additionally, the index depends on daytime, cloud-free optical acquisitions, as obscures surface and nighttime data is unavailable, restricting consistent monitoring in frequently overcast regions.

Modified NDWI (MNDWI) and Other Variants

The Modified Normalized Difference Water Index (MNDWI), introduced by Hanqiu Xu in , addresses limitations of the original NDWI by replacing the near- () band with the mid- (MIR) or shortwave (SWIR) band to better suppress noise from built-up areas, , and while enhancing open water features. The formula is given by: \text{MNDWI} = \frac{X_{\text{green}} - X_{\text{MIR}}}{X_{\text{green}} + X_{\text{MIR}}} where X_{\text{green}} is the green band reflectance and X_{\text{MIR}} is the MIR band reflectance. This modification leverages the higher absorption of water in the MIR spectrum compared to built-up surfaces, improving delineation in complex urban environments. Other variants include the Automated Water Extraction Index (AWEI), developed by Feyisa et al. in 2014, which incorporates multiple spectral bands—including green, NIR, SWIR, and thermal—to automate water mapping and reduce errors from shadows and terrain. AWEI exists in two forms: one accounting for shadows (AWEIsh) and another without (AWEInsh), enhancing separability between water and non-water pixels through band differencing and weighting. Additionally, variants integrating thermal bands with NDWI have been explored to link water content with land surface temperature, as thermal data reveal cooler water bodies relative to surrounding land, aiding in applications like urban heat island analysis. Recent advancements hybridize NDWI variants with machine learning, such as convolutional neural network (CNN)-based thresholding on Sentinel-2 data for refined water extraction, as demonstrated in 2023-2024 studies that combine spectral indices with deep learning to handle variability in illumination and shadows. Hyperspectral extensions, like the Hyperspectral Difference Water Index (HDWI) proposed in 2014, adapt the NDWI framework to narrow-band hyperspectral data for higher spectral resolution, enabling finer discrimination of water types in urban and vegetated landscapes. Emerging approaches as of 2025 include fusing NDWI with synthetic aperture radar (SAR) data to mitigate cloud cover and atmospheric limitations. MNDWI has seen widespread adoption for urban water monitoring, with studies up to noting improved accuracy over the original NDWI in built-up areas by minimizing false positives from impervious surfaces. For instance, in mixed urban-rural settings, MNDWI achieves overall accuracies exceeding 95%, particularly when integrated with multi-temporal imagery.

Comparison with NDVI

The (NDVI) is calculated as \frac{X_{NIR} - X_{RED}}{X_{NIR} + X_{RED}}, where X_{NIR} is the near-infrared reflectance and X_{RED} is the red reflectance, primarily indicating content and density. In contrast, the Normalized Difference Water Index (NDWI) developed by uses \frac{X_{NIR} - X_{SWIR}}{X_{NIR} + X_{SWIR}}, with X_{SWIR} being the shortwave infrared reflectance, making it particularly sensitive to in leaves rather than overall greenness. This distinction allows NDWI to detect physiological earlier than NDVI, as water absorption in the SWIR band provides a direct measure of canopy moisture independent of pigment levels. While NDVI excels at mapping vegetation vigor and cover in sparse to moderate canopies, it saturates in dense , where values plateau around 0.8–0.9 regardless of further increases, limiting its utility in forests or advanced stages. studies in agricultural settings, such as croplands, typically show moderate to strong positive relationships between the two indices, with Pearson coefficients ranging from 0.6 to 0.8, reflecting their shared sensitivity to but divergent responses to . The complementary nature of NDVI and NDWI enables their combined use to identify stress more effectively; for instance, high NDVI paired with low NDWI signals water-deficient but photosynthetically active , improving early detection over NDVI alone. In , multi-index models integrating both have been applied in projects like those using ESA data since the early 2020s to optimize and yield forecasting in croplands. The Modified Normalized Difference Water Index (MNDWI), proposed by in 2006, enhances the original NDWI developed by McFeeters by replacing the near-infrared () band with the shortwave infrared (SWIR) band, which reduces from built-up areas and shadows in environments. This modification results in fewer false positives for detection in complex landscapes, where NDWI often confuses impervious surfaces with due to similar reflectance. For instance, evaluations using Landsat imagery have shown MNDWI achieving overall accuracies around 98% and coefficients around 0.7 in tasks, outperforming NDWI by minimizing commission errors from non-water features. The Normalized Difference Water Index (NDWI) developed by in 1996 (also known as the Normalized Difference Moisture Index or NDMI) employs and SWIR bands to assess vegetation and content rather than open bodies. While this NDWI is effective for monitoring liquid in vegetated canopies, it is less sensitive to surface delineation compared to McFeeters' NDWI, as its focus on moisture absorption in SWIR makes it prone to overestimation in dry soils or underestimation in turbid waters. Similarly, the Land Surface Water Index (LSWI), developed by et al. in 2002 for MODIS data, uses and SWIR bands akin to 's NDWI to provide complementary insights into inundation under canopy cover, where McFeeters' NDWI may fail due to saturation, but it requires multi-sensor integration, increasing complexity over NDWI's standalone multispectral approach. In performance comparisons, the Automated Water Extraction Index (AWEI), formulated by Feyisa et al. in 2014, demonstrates advantages over MNDWI in handling shadows and turbid waters, reducing both commission and omission errors by approximately 50% in Landsat-based mappings of heterogeneous landscapes. Specifically for small water bodies, such as ponds and narrow rivers, AWEI variants like AWEIsh exhibit lower omission errors than McFeeters' NDWI, capturing features that NDWI misses due to its sensitivity to variability in shallow or vegetated edges. Hyperspectral indices, such as the Hyperspectral Difference Water Index (HDWI) proposed by et al. in 2014, further outperform NDWI in for fine-scale water extraction, achieving significantly higher accuracies in urban and coastal settings, though they demand specialized hyperspectral sensors and greater computational resources. Selection of water indices depends on scene complexity and data availability: NDWI remains ideal for straightforward multispectral applications like basic surface water mapping in clear environments, while advanced indices such as MNDWI or AWEI are preferred for or turbid scenes with high interference, and hyperspectral options suit detailed studies in coastal mangroves where fine spectral discrimination is essential.

References

  1. [1]
    The use of the Normalized Difference Water Index (NDWI) in the ...
    The Normalized Difference Water Index (NDWI) is a new method that has been developed to delineate open water features and enhance their presence in remotely- ...
  2. [2]
    Using the Normalized Difference Water Index (NDWI) within ... - MDPI
    The NDWI was first proposed by McFeeters in 1996 to identify non-urban surface water associated with wetlands [19]. Kim et al. used eCognition® Developer to ...
  3. [3]
    (PDF) Modification of Normalized Difference Water Index (NDWI) to ...
    Aug 5, 2025 · McFeeters , S. K. 1996. The use of normalized difference water index (NDWI) in the delineation of open water features.. International Journal of ...
  4. [4]
    NDWI - MODIS Web - NASA
    In this article, another index, namely, the normalized difference water index (NDWI), is proposed for remote sensing of vegetation liquid water from space. NDWI ...Missing: applications | Show results with:applications
  5. [5]
    [PDF] Using Satellite Images to Determine Surface-Water Cover during the ...
    Sep 13, 2013 · Normalized Difference Water Index (NDWI)=(Green–near infrared)/(green+near infrared). The. NDWI is useful for delineating clear water cover ...<|control11|><|separator|>
  6. [6]
    Assessment of Landsat-8 and Sentinel-2 Water Indices: A Case ...
    Sep 18, 2023 · Generally, the normalized difference water Index (NDWI) is used to rate changes in the water area [3]. In this regard, several studies have ...
  7. [7]
    USGS EROS Archive - Vegetation Monitoring - eVIIRS Global NDWI
    Jun 21, 2023 · Composite Calculation for the Normalized Difference Water Index (NDWI). The software used to create 10-day dekadal NDWI composites was developed ...
  8. [8]
    NDWI—A normalized difference water index for remote sensing of ...
    NDWI is defined as (ϱ(0.86 μm) − ϱ(1.24 μ m )) (ϱ(0.86 μ m ) + ϱ(1.24 μ m )) , where ϱ represents the radiance in reflectance units.
  9. [9]
    NDWI—A normalized difference water index for remote sensing of ...
    In this article, another index, namely, the normalized difference water index (NDWI), is proposed for remote sensing of vegetation liquid water from space.
  10. [10]
    A Global Multi-Sensor Dataset of Surface Water Indices ... - Nature
    Jul 17, 2025 · This study aims to provide monthly global surface water datasets utilizing three water indices, NDWI, MNDWI, and WI2015, derived from Landsat-8 ...
  11. [11]
    [PDF] Harmonized Landsat Sentinel-2 (HLS) Vegetation Indices (VI ...
    NDWI is particularly useful for monitoring changes in surface water bodies, such as lakes, rivers, and wetlands, and for assessing water-related environmental ...
  12. [12]
    Risk of Climate Change Impacts on Drought and Forest Fire Based ...
    Jan 18, 2018 · Normalized Difference Water Index. Bo-Cai Gao (1996) proposed a normalized difference water index (NDWI), which uses the wavelengths 0.86 and ...
  13. [13]
    A critical assessment of pre-fire wildfire risk mapping - ScienceDirect
    Sep 20, 2025 · NDWI is essentially a negative (inverse) indicator of fire danger: higher NDWI values denote wetter vegetation and thus lower fire risk, whereas ...
  14. [14]
    Estimation of leaf water status to monitor the risk of forest fires by ...
    In this paper, laboratory measurement confirmed a relationship between leaf water status and the normalized difference water index (NDWI).
  15. [15]
  16. [16]
    NDWI: Index Formula, Value Range, And Uses In Agriculture
    Sep 29, 2021 · The Normalized Difference Water Index (NDWI) is used to highlight open water features in a satellite image, allowing a water body to “stand out” against the ...
  17. [17]
    Landsat Collection 2 Surface Reflectance | U.S. Geological Survey
    Landsat 8-9 OLI Band 8 and Landsat 7 ETM+ Band 8 (panchromatic band) are not processed to Top of Atmosphere or Surface Reflectance; Landsat 8-9 OLI Band 9 ...
  18. [18]
    [PDF] Landsat 8 (L8) Data Users Handbook - Amazon AWS
    Nov 26, 2019 · The OLI sensor collects image data for 9 shortwave spectral bands over a 190 km swath with a 30 meter (m) spatial resolution for all bands ...
  19. [19]
    Landsat 8 OLI and TIRS and their uses - USGS.gov
    By Frequently Asked Questions 2016 (approx.) Chart listing Landsat 8 bands, corresponding wavelengths, and how the bands can be used. Original Thumbnail Medium ...
  20. [20]
    404 Not Found
    - **Status**: Not Found
  21. [21]
    NDWI - Normalized Difference Water Index Calculation - GeoAI
    NDWI is a satellite-derived remote sensing index that enhances water features, distinguishing open water from land, vegetation, and built environments.
  22. [22]
    Normalized Difference Water Index (NDWI)
    Jul 25, 2023 · NDWI evaluates vegetation liquid water content and water-inundated areas, sensitive to changes in liquid water content. Its range is -0.1 to 0. ...Missing: paper | Show results with:paper
  23. [23]
    [PDF] Analysis of Dynamic Thresholds for the Normalized Difference Water ...
    Abstract. The normalized difference water index (NDWI) has been successfully used to delineate surface water features. However,.
  24. [24]
    Improved surface water mapping using satellite remote sensing ...
    This study designed an automated surface water detection framework that optimizes the Otsu thresholds for water indices and picks the optimal water index.
  25. [25]
    Augmented normalized difference water index for improved ...
    Further, we implement a dynamic thresholding method, the Otsu algorithm, to enhance ANDWI's performance. Applied to a variety of environmental conditions, ANDWI ...
  26. [26]
    Sub-Pixel Surface Water Mapping for Heterogeneous Areas ... - MDPI
    Apr 7, 2023 · This study proposes a sub-pixel surface water mapping method based on morphological dilation and erosion operations and the Markov random field (DE_MRF)
  27. [27]
    Combining topography and reflectance indices for better surface ...
    The greater the NDWI or the MNDWI, the higher the likelihood of waterbody. The MNDWI is identical to another index for detecting ponds, namely the Normalized ...
  28. [28]
    High-Resolution Mapping of Urban Surface Water Using ZY-3 Multi ...
    ... NDWI achieved the highest accuracy with kappa coefficient of 0.89. However, it was still 0.07 lower than that of the UWEM. Compared with NDWI, the total ...
  29. [29]
    Normalized difference water index for remote sensing of vegetation ...
    In this paper, another index, namely the normalized difference water index (NDWI), is proposed for remote sensing of vegetation liquid water from space. NDWI ...
  30. [30]
    Evaluating spectral indices for water extraction - ScienceDirect.com
    McFeeters (1996) proposed the normalized difference water index (NDWI) based on the strong water reflectance in the green band and strong absorption in the ...
  31. [31]
    Mapping of Urban Surface Water Bodies from Sentinel-2 MSI ... - MDPI
    3.2.​​ Nevertheless, McFeeters' NDWI is limited by its inability to suppress noise from the land cover features of built-up areas, the low spatial resolution ...<|separator|>
  32. [32]
    Evaluation of Water Indices for Surface Water Extraction in a ...
    Aug 7, 2018 · Upon selecting optimum thresholds, the overall accuracy (OA) and kappa coefficient (kappa) was improved, but not satisfactory. NDVI and NDWI ...
  33. [33]
    A satellite‐based method for monitoring seasonality in the overstory ...
    Jan 8, 2010 · The dNDWI was more sensitive to LAIo at lower values of LAIu. At higher values of LAIu such as LAIu = 2.0, dNDWI became saturated with ...
  34. [34]
    Temporal dynamics in vertical leaf angles can confound vegetation ...
    Oct 2, 2024 · We show that leaf angle dynamics systematically confound widely applied vegetation indices. Moreover, we demonstrate that these effects are not random.
  35. [35]
    Evaluation of Normalized Difference Water Index as a Tool ... - MDPI
    Jan 1, 2019 · The NDWI is sensitive to changes in leaf water content because green vegetation spectra in the SWIR region is dominated by water absorption ...2.2. Climate · 2.3. Pasture And Soil... · 3. Results And Discussion
  36. [36]
  37. [37]
    Modification of normalised difference water index (NDWI) to ...
    The normalized difference water index (NDWI) of McFeeters (Citation1996) was modified by substitution of a middle infrared band such as Landsat TM band 5 for ...<|control11|><|separator|>
  38. [38]
    A new technique for surface water mapping using Landsat imagery
    McFeeters (1996) introduced the Normalized Difference Water Index (NDWI) to delineate open water features using the green (band 2) and near-infrared (band 4) of ...
  39. [39]
    Examining the relationship between land surface temperature and ...
    A greater correlation (0.98) between NDVI and NDWI but a negative (-0.98) correlation with NDBI was found in this investigation. Abstract. Land surface ...2. Materials And Methods · 2.5. From Landsat 8 Oli/tirs · 4. Discussion
  40. [40]
    Water Body Extraction from Sentinel-2 Imagery with Deep ... - MDPI
    Feb 24, 2023 · A common task in land-cover classification is water body extraction, wherein each pixel in an image is labelled as either water or background.
  41. [41]
    Water body extraction from high spatial resolution remote sensing ...
    Jul 12, 2024 · This paper presents a novel network model named EU-Net, specifically designed to extract water bodies from high-resolution remote sensing images.
  42. [42]
    New hyperspectral difference water index for the extraction of urban ...
    Jul 24, 2014 · In this paper, we introduce a multiple-band hyperspectral water index, called the hyperspectral difference water index (HDWI), with the ...
  43. [43]
    Extracting Urban Water Bodies from Landsat Imagery Based ... - MDPI
    Nov 16, 2022 · We propose a novel urban water extraction algorithm (UWEA) that is efficient in distinguishing water and other low-reflective objects.
  44. [44]
    NDVI, the Foundation for Remote Sensing Phenology - USGS.gov
    High NDVI values (approximately 0.6 to 0.9) correspond to dense vegetation such as that found in temperate and tropical forests or crops at their peak growth ...<|control11|><|separator|>
  45. [45]
    A five‐year analysis of MODIS NDVI and NDWI for grassland ...
    Mar 27, 2007 · NDWI calculated from the 500-m SWIR band of MODIS has recently been used to detect and monitor the moisture condition of vegetation canopies ...
  46. [46]
    Status for: Vegetation Indices (MOD13) - MODIS Land - NASA
    MODIS NDVI exhibits saturation in dense vegetation canopies, including forests, while MODIS EVI maintains good sensitivity and correlation with tower flux ...
  47. [47]
    Mitigating NDVI saturation in imagery of dense and healthy vegetation
    This study investigates the underlying causes of NDVI saturation and proposes a two-stage saturation mechanism.
  48. [48]
    Methodological evaluation of vegetation indexes in land use and ...
    It was also observed that for the native forest, the relationship between NDVI and NDWI expressed a correlation coefficient of 0.68, representing one of the ...
  49. [49]
    Exploration and advancement of NDDI leveraging NDVI and NDWI ...
    After comparing NDVI and NDWI, it was discovered that NDWI values respond more quickly to drought conditions than NDVI values, and the correlation between NDVI ...
  50. [50]
    [PDF] Satellite Data Applications in Crop Management
    Sentinel (ESA). • Landsat (NASA / USGS). • Planet Labs. These services enable a variety of agricultural applications for our customers, including crop.
  51. [51]
    Comparison of surface water extraction performances of different ...
    For example, the NDVI value of the city site is easy to misclassify water bodies with shadow and built-up area because they overlap with each other (Figure 6 ).Missing: limitations | Show results with:limitations