Remote Sensing Indices

Regen Network
Published in
12 min readFeb 19, 2018


The following post has been prepared to walk the reader through a range of different methods utilized to derive information about vegetation in landscapes from satellite data. The term “Vegetative Index” describes an algorithm that processes spectral data for the purpose of determining information about plant health. The term “Remote Sensing,” in this instance, describes the use of satellite imagery to make discernment’s about landscape phenomena.


Vegetative Indices (VI) enable the acquisition of ecological information from satellite and drone data through the analysis of multi- or hyperspectral imagery bands. The reflectance of light changes with chlorophyll content, plant type, sugar content, water content within tissues and other factors. Due to this fact, the spectral reflectance responses captured by satellite imagery can reflect the interaction and coupling of carbon, nitrogen, and water cycles (Chang et al., 2016; Xue et al., 2017). A wide range of plant characteristics can be inferred through various indices. These indices are also used to improve the accuracy of classification algorithms. Indices enhance the spectral information and increase the separability of the classes of interest. All of these factors result in an increase in quality of the Land Use Land Cover (LULC) mapping produced (Ustuner et al., 2014). Indices have a dual function: providing information about plant growth and health as well as helping categorize different land classifications (mining, forest, bare soil, pasture, water surfaces, industrial etc.).

Figure 1 - Spectral curves for various natural features (NASA, 2018)

The different combinations of vegetation indices also enhance spectral characteristics of some crops while suppressing others (Kuzucu et al., 2017).

In this blogpost I will offer an introductory explanation of the difference between a variety of algorithms:

  • Normalized Difference Vegetation Index (NDVI)
  • Enhanced Vegetation Index (EVI)
  • Normalized Difference Water Index (NDWI)
  • Modified Soil-Adjusted Vegetation Index (MSAVI2)
  • Soil Adjusted Total Vegetation Index (SATVI)
  • Soil Adjusted Vegetation Index (SAVI)
  • Soil-Adjusted Vegetation Index optimized for Agricultural Monitoring (OSAVI)

Forest Canopy Density (FCD) methodology

  • Advanced Vegetation Index (AVI)
  • Shadow Index (SI)
  • Bare Soil Index (BI)
  • Temperature Index (TI)

This blogpost will not identify which eco-indicators correspond to specific ecological outcomes. This will be discussed in further blog posts.

Multispectral satellite imagery

The main spectral bands that are relevant to the application of VI’s are the following:

Visible light

  • Blue: 450–495 nm
  • Green: 495–570 nm
  • Red: 620–750 nm


  • Near Infrared (NIR): 750–900 nm
  • Short Wave Infrared (SWIR): 900–3000 nm
  • Thermal Infrared (TIR): 3000–14000 nm
Figure 2 - Spectral chart (NASA, 2018)

The wavelength domains vary from satellite to satellite. For example, the red band of the Worldview-3 satellite ranges from 630 to 690 nm whereas the PlanetScope satellites use 590 to 670 nm wavelength. Different satellites have different sensors that generate images with various spectral bands.

Figure 3 - Landsat 8 Operational Land Imager (OLI) and Thermal Infrared Sensor (TIRS) (Barsi, 2014)

NASA’s Landsat 8 satellite has eleven bands in total. Of those eleven, nine spectral bands have a spatial resolution of 30 meters. The resolution for Band 8 (panchromatic) is 15 meters. Thermal bands 10 and 11 are useful in providing more accurate surface temperatures and are acquired at 100 meter resolution, but are resampled to 30 meter. The Landsat 8 satellite images the entire Earth every 16 days in an 8-day offset from Landsat 7. (NASA, 2018)

ESA’s Sentinel-2 has thirteen bands in total. Of those thirteen bands, four spectral bands have a resolution of 10 meter. Six bands have a resolution of 20 meter and the remaining three have a resolution of 60 meters. The temporal resolution of the sentinel satellites (2A and 2B) is 10 days, combined this gives a 5-day refresh rate.

Figure 4 - Comparison of Landsat 7 and 8 bands with Sentinel-2 bands (NASA, 2018)


Photosynthesis requires water, carbon dioxide and light in order to produce sugars and oxygen. Chlorophyll, which gives plants their green color, absorbs visible light. Leaves reflect near-infrared light (NIR); this makes sense evolutionarily-speaking because plants use only visible light for photosynthesis. This means that a healthy plant with good photosynthesis activity can be analyzed by comparing NIR with visible red light. Unhealthy vegetation will reflect more visible light and less NIR. Healthy vegetation will absorb most of the visible light falling on it. Originally developed by NASA, the NDVI index is widely used in this application. NDVI values range between 0 and 1 (due to the normalization procedure). Very low values of NDVI (<0.1) correspond to barren areas of rock, sand or snow. Free standing water tend to be in the very low positive to negative values. Soils tend to generate rather small NDVI values (0.1–0.2). Sparse vegetation such as shrubs and grasslands may result in moderate NDVI values (0.2–0.5).

Table 1 - Typical NDVI values for different ecosystems (Pettorelli, 2013)

High NDVI values correspond to dense vegetation such as that found in temperate and tropical forests or crops at their peak growth stage (Pettorelli, 2006). As an index, it is often used in large-scale monitoring of forest disturbances and global vegetation assessments. More specifically, NDVI has been used to map ecosystem distribution, predict disturbances and asses their impact, monitor changes in the functional attributes of ecosystems, monitor habitat loss and degradation, carbon assimilation and evapotranspiration. On the farm scale, NDVI is used as a predictor of plant attributes, plant physiological status, yield predictions and crop distribution, and can also be used to detect and monitor aquatic vegetation (Pettorelli, 2013).

The main drawback of NDVI is that it is sensitive to the effects of soil (brightness and color), atmosphere (cloud cover and cloud shadow) and leaf canopy shadow (Xue et al., 2017). Another problem with NDVI is that in dense vegetation, it quickly reaches saturation. This is due to the fact that the NDVI index is non-linear. In conclusion, NDVI is good to study large areas and get a rough sense of the photosynthetic activity. Its sensitivity to soil and aerosols means it has limitations. For more qualitative analysis other indices should be used that have built in feedback mechanisms.


The Enhanced Vegetation Index (EVI) is the most common alternative vegetation index which addresses some of the issues with NDVI (soil and atmosphere limitations). In a 1995 study, Liu and Huete analyzed multiple soil types and atmospheric-enhanced vegetative indices; they concluded that, as a result of the interaction between the soil and the atmosphere, reducing one may increase the other. They introduced a feedback mechanism by building a parameter to simultaneously correct soil and atmospheric effects. This algorithm reduces the influence of atmospheric conditions and corrects for canopy background signals making it more sensitive to changes in high biomass areas (Liu & Huete, 1995; Xue et al., 2017). The EVI is described by the following formula:

L corrects for soil background, C1 and C2 are coefficients that correct for aerosol scattering in the atmosphere. The standard EVI used by NASA in MODIS sensor: L=1, C1=6 and C2=7.5.

Note: as proposed in the 2008 paper by Jiang, the EVI2 index can be used for sensors without a blue band, such as the AVHRR and ASTER imagery, to produce an EVI-like vegetation index, complementary to NDVI but without the problem of saturation. Especially when red reflectance is low and NDVI becomes saturated, the EVI2 index may reveal different vegetation dynamics (Jiang et al., 2008).

Figure 5 -Cross-plot of EVI2 and EVI using QA-accepted MODIS 1 km, 16-day composite VI data over the 13 EOS Land Validation core sites from 2000 to 2006 (Jiang et al., 2008)


MSAVI2 is the simplified version of the MSAVI algorithm. It was created to deal with the soil brightness problem, which is one of NDVI’s largest limitations. Whenever MSAVI is used, it is almost always the MSAVI2 version, which does not require a soil line (slope). It is mainly used in the analysis of plant growth, desertification research, grassland yield estimation, LAI assessment, analysis of soil organic matter, drought monitoring, and the analysis of soil erosion (Xue et al., 2017). MSAVI2 a good index for areas that are not completely covered with vegetation and have exposed soil surface.

Some sensitivity to photosynthesis is lost, due to soil brightness correction, . For that reason, MSAVI2 is best used in cases where there is a lot of exposed soil, such as in desert-like climates. It also is quite susceptible to atmospheric conditions which makes it difficult to use for bi-temporal change detection.


When soil brightness plays an important role — such as in areas where vegetative cover is low and the soil is exposed — the NDVI can be influenced by the reflectance of the soil. The Soil Adjusted Vegetation Index (SAVI) is a modification of the NDVI with a correction factor for soil brightness. The value of L is adjusted based on the amount of vegetation. L=0.5 is the default value and works well in most situations. With L=0, NDVI=SAVI.

The OSAVI is the soil-adjusted vegetation index optimized for agricultural monitoring. In a 1996 study by Rondeaux, it was found that, for agricultural use, the OSAVI outperformed all other indices (Rondeaux et al., 1996). In general, OSAVI is more sensitive to vegetation and shows differences in vegetation better than SAVI. Just like EVI, OSAVI works well in areas where the vegetation density is high (Rondeaux et al., 1996).


Soil-Adjusted Total Vegetation Index (SATVI) measures both photosynthesizing and dead vegetation. It requires short-wave infrared band (SWIR), which results in less-frequent use of this index compared to those reliant on just NIR and visible light, as many satellites don’t collect SWIR band. The SATVI is useful in mapping photosynthesizing biomass, ground residue, plant litter, surface conditions, and in calculations of aboveground biomass while compensating for varying soil brightness and background artifacts. SATVI can be used on Landsat, ASTER, Sentinel and MODIS imagery. The bands needed are RED, SWIR 1 and SWIR 2. In the case of Sentinel, this is band 4, 11 and 12. L is usually set at 0.5 and corresponds to the soil-line slope (Marsett et al., 2006; Qi et al., 2002; Qi, 2000; Torbick, 2016; Hagen et al., 2012).

Forest Canopy Density model (FCD)

The Forest Canopy Density model (FCD) was originally developed as a tool to assess the regrowth of a forest canopy in logged-over tropical forests. It only uses the bands of the landsat sensor and models canopy density using data derived from four indices: Advanced Vegetation Index (AVI), Bare Soil Index (BI), Shadow Index or Scaled Shadow Index (SI, SSI) and Thermal Index (TI).

The Advanced Vegetation Index (AVI):

The Bare soil Index (BI):

The Shadow Index (SI):

Figure 6 -AVI, BI, SI and TI indices in the FCD model (Rikimaru, 2002)

Compared to NDVI, the AVI reacts more sensitively to vegetation quantity and is able to highlight subtle differences in canopy density. The Shadow Index (SI) increases as the forest density increases and this shadow pattern affects the spectral response. For example, young and evenly spaced trees have a low canopy shadow index compared to mature natural forest stands. The Bare Soil index (BI) enhances the identification of bare soil areas and fallow lands. The Thermal Index (TI) is the calibrated values of the thermal band and increases as the vegetation quantity increases. It is lower inside the canopy of a forest due to blocking and absorption of the sun’s rays and because of the cooling effect of evaporation from leaves. The TI is therefore used to further differentiate bare soil from grassland and forest. The below figure demonstrates the interaction of the four different indices in the FCD model (Azizia et al., 2008; Rikimaru et al., 2002; Baynes, 2004).

In their 2016 article, Akike & Samanta recommend the FCD methodology for use in forest management activities with little need for ground truth, which makes the FCD method efficient and effective. They recommended that this model be used to estimate forest canopy density and possible forest loss in deforestation areas (Akike & Samanta, 2016). Accordingly to Baynes in his 2004 article, the FCD methodology successfully discriminates between forest types over a wide range of heights as long as canopy closure was maintained. From his observation, the FCD method appeared to be useful for analyzing deforestation or conversion from native forest to plantations (Baynes, 2004).


The Normalized Difference Water Index (NDWI) measures the change in the water content of leaves by using the NIR and SWIR bands. Because NDWI is sensitive to the water content of plants as well as bodies of water, it is often used for drought monitoring, recording yield reductions, reservoir discharge, lowering of groundwater levels etc.

Values for water bodies are larger than 0.5. Vegetation has much smaller values, which makes distinguishing between vegetation and water bodies easier. Built-up features have positive values between zero and 0.2 (Sentinel-hub, 2018).

Works Cited

Spatial — Resolutions — Sentinel-2 MSI — User Guides — Sentinel Online,


Akike, Slady, and Sailesh Samanta. “Land Use/Land Cover and Forest Canopy Density Monitoring of Wafi-Golpu Project Area, Papua New Guinea.” Journal of Geoscience and Environment Protection, vol. 04, no. 08, 2016, pp. 1–14., doi:10.4236/gep.2016.48001.

Azizia, Najafia , Sohrabia. “Forest canopy density estimating, using satellite images”, Natural Resources and Marine Sciences Faculty of Tarbiat Modares University — The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B8. Beijing 2008

Barsi, Julia, et al. “The Spectral Response of the Landsat-8 Operational Land Imager.” Remote Sensing, vol. 6, no. 12, 2014, pp. 10232–10251., doi:10.3390/rs61010232.

Baynes, Jack. “Assessing Forest Canopy Density in a Highly Variable Landscape Using Landsat Data and FCD Mapper Software.” Australian Forestry, vol. 67, no. 4, 2004, pp. 247–253., doi:10.1080/00049158.2004.10674942.

Chang, Liu, et al. “A Review of Plant Spectral Reflectance Response to Water Physiological Changes.” Chinese Journal of Plant Ecology, vol. 40, no. 1, 2016, pp. 80–91., doi:10.17521/cjpe.2015.0267.

Hagen, Stephen C., et al. “Mapping Total Vegetation Cover Across Western Rangelands With Moderate-Resolution Imaging Spectroradiometer Data.” Rangeland Ecology &amp; Management, vol. 65, no. 5, 2012, pp. 456–467., doi:10.2111/rem-d-11–00188.1.

Kuzucu, A. Karakacan, and F. Bektas Balcik. “Testing The Potential Of Vegetation Indices For Land Use/Cover Classification Using High Resolution Data.” ISPRS Annals of Photogrammetry, Remote Sensing and Spatial Information Sciences, IV-4/W4, 2017, pp. 279–283., doi:10.5194/isprs-annals-iv-4-w4–279–2017.

Liu, Hui Qing, and A. Huete. “A Feedback Based Modification of the NDVI to Minimize Canopy Background and Atmospheric Noise.” IEEE Transactions on Geoscience and Remote Sensing, vol. 33, no. 2, 1995, pp. 457–465., doi:10.1109/36.377946.

Marsett, Robert, et al. “Remote Sensing for Grassland Management in the AridSouthwest.” Rangelands, vol. 59, no. 5, 2006, doi:10.2458/azu_jrm_v59i5_marsett.

Pettorelli, Nathalie. Normalized Difference Vegetation Index. Oxford University Press, 2013.

Qi, J. “Spatial and Temporal Dynamics of Vegetation in the San Pedro River Basin Area.” Agricultural and Forest Meteorology, vol. 105, no. 1–3, 2000, pp. 55–68., doi:10.1016/s0168–1923(00)00195–7.

Rikimaru, P.S. Roy and S. Miyatake, 2002. Tropical forest cover density mapping. Tropical Ecology Vol. 43, №1, pp 39–47.

Rondeaux, Geneviève, et al. “Optimization of Soil-Adjusted Vegetation Indices.” Remote Sensing of Environment, vol. 55, no. 2, 1996, pp. 95–107., doi:10.1016/0034–4257(95)00186–7.

Torbick, Nathan, et al. “Regional Mapping of Plantation Extent Using Multisensor Imagery.” Remote Sensing, vol. 8, no. 12, 2016, p. 236., doi:10.3390/rs8030236.

Ustuner, M., et al. “Crop Type Classification Using Vegetation Indices of RapidEye Imagery.” ISPRS — International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, XL-7, 2014, pp. 195–198., doi:10.5194/isprsarchives-xl-7–195–2014.

Xue, Jinru, and Baofeng Su. “Significant Remote Sensing Vegetation Indices: A Review of Developments and Applications.” Journal of Sensors, vol. 2017, 2017, pp. 1–17., doi:10.1155/2017/1353691.

Xue, Jinru, and Baofeng Su. “Significant Remote Sensing Vegetation Indices: A Review of Developments and Applications.” Journal of Sensors, vol. 2017, 2017, pp. 1–17., doi:10.1155/2017/1353691.