10 Indices and Band Ratioing

Kumari Anamika Kumari Anamika

Objectives

• Student will get to know why band rationing is required.
• Student will acquire skill how to study statistics of band rationing
• Student will be equipped with knowledge to study further
• about the background functioning of band rationing

   Outline

1. Introduction

     a. Image Enhancement

     b. Band Ratioing

2. Selecting a proper band combination for an RGB image using Landsat Imagery

3. Spectral Indices

4. Reflectance Characteristics of earth’s surface features

     a. Vegetation

     b. Water

     c. Soil

5. Vegetation Indices

      a. Normalized Difference Vegetation Index (NDVI)

      b. Transformed Vegetation Index (TVI)

      c. Vegetation Ratio Index (VRI)

6. Effects of soil background on Vegetation Index

      a. Soil Adjusted Vegetation Index (SAVI)

     b. Perpendicular Vegetation Index (PVI)

7. Atmospheric Effects on Vegetation Index

8. Aerosol Free Vegetation Index (AFRI)

9. Reducing both Soil and Atmospheric Effects

10. Normalized Difference Water Index (NDWI)

11. Questions

12. References

Introduction

Pictures are the most common and convenient means of transmitting information. They portray spatial information that we recognize as objects. They convey information about positions, sizes, and interrelationships between objects.

A digital remotely sensed image is typically composed of picture elements (pixels) located at the intersection of each row i and column j in each k bands of imagery. Associated with each pixel is a number is known as Digital Number (DN) or Brightness value (BV) that depicts the average radiance of a relatively small area within a scene. A smaller number indicated low average radiance from the area and the high number indicates high radiant properties of the area.

Figure 1: Structure of multispectral image

Image Enhancement:

Image enhancement techniques improve the quality of an image perceived by a human. These are applied separately to each band of a multispectral image in order to display or visually interpret the data more effectively. The objective is to create “new” images from original raw data in order to increase the amount of information that can be visually interpreted from the data. Image enhancement is attempted after the image is corrected for geometric and radiometric distortions.

Band Ratioing:

Sometimes differences in brightness values from identical surface materials are caused by various environmental factors such as topographic slope and aspect, shadows, or seasonal changes in sunlight illumination angle and intensity. These factors may limit the ability of an interpreter to correctly identify the surface materials or landuse in a remotely sensed image. This ratio transformations can be applied to reduce the effects of such environmental conditions.

Ratioing is the enhancement that provides unique information and differences between the spectral reflectance of surface features that are difficult to detect in a standard image. It is also useful to discriminate the soils from vegetation.

A band ratio is a result of dividing each pixel of one band by the corresponding pixels of another band. Many researchers have developed many different band ratios over the years that exploit the specific reflectance properties of the surface materials. For example, vegetation has an extremely high reflectance in near infrared wavelength region, relative to red wavelengths.

The number of possible ratio combinations for a multispectral sensor with K bands is n=K (K-1). Thus for Thematic Mapper (for six reflectance bands), the number of ratio combinations is 30 out of which 15 are original and another 15 are reciprocal.

Ratio function can be mathematically expressed as:

Where BVi,j,r = output ratio value of the pixel at ith row and jth column

BVi,j,k = brightness value at the same location for band K.

BVi,j,l = brightness value at the same location for band l

The output ratio value ranges from 0 to 255 (since the value of BVi, j can be zero).These new numbers can be stretched or expanded to produce images with considerable contrast variation in a black and white rendition.

Certain features or materials can produce distinctive gray tones in certain ratios; TM band 3 (red) divided by 1 tends to emphasize on red or orange colored features or materials, such as natural hydrated iron oxide as light tones. Three band ratio images can be combined as color composites which highlight certain features in distinctive colors. Ratio images also reduce or eliminate the shadow effects.

                    Figure 2: Landsat TM image of Bisalpur Dam near Tonk, Rajasthan

Ratio images can be directly related to spectral properties of materials. Ratioing can be thought of as a method of enhancing minor differences between materials by defining the slope of the spectral curve between two bands. But we must understand that dissimilar materials having similar slopes but different Albedo values can be easily separable in the standard image but become inseparable in ratio images.

Figure 3: Reduction of Scene Illumination effect through spectral ratioing

(source Lillesand and Kiefer,1993)

Table 1: Spectral Ratioing output

Selecting a proper band combination for an RGB image using Landsat Imagery:

Selecting the proper bands to use in the imagery has a huge impact to interpret the features in a particular image. The Below mentioned list describes some of the features of seven bands of Landsat Thematic Mapper Imagery for detecting different features.

Band1 (0.45-0.52µm):

This band with short wavelength penetrates more than the other bands hence it is preferred more in the aquatic ecosystems. It is used to monitor sediments in the water, mapping coral reefs, and water depth. Since the blue light is scattered more than other bands, it has the maximum noise of all the Landsat bands.

Band2 (0.52-0.60µm):

This has similar qualities to band1 but not extreme. It is sensitive to water turbidity differences.

It matches the wavelength for the green features such as vegetation.

Band3 (0.63-0.69 µm):

It senses in the strong chlorophyll absorption region and strong reflectance region for most soils and thus it is sometimes called as chlorophyll absorption band. This band can be useful to discriminate soil and vegetation in monitoring vegetation health.

Band4 (0.76-0.90 µm):

It operates in the best spectral region in order to distinguish vegetation varieties and conditions.

Since water absorbs nearly all light at this wavelength water bodies appear very dark.

              Figure 4: Band 4 of Landsat TM (Bisalpur Dam near Tonk, Rajasthan)

Band 5 (1.55-1.75 µm):

This band is very sensitive to moisture and hence is widely used to monitor vegetation and soil moisture. It has separated water body from barren and croplands (light gray tones). Since urban area and vegetation have the same spectral response it is difficult to distinguish between them.

Figure 5: Band5 of Landsat TM not able to distinguish between settlement and vegetation since it has same gray tone for both the features.

Band6 (10.4-12.5 µm thermal infrared):

It is a thermal band and hence is used to measure surface temperature. It is primarily used for geologic applications but is sometimes used to measure plant heat stress. It is used to distinguish clouds from soils since clouds tend to be very cold. It has half the resolution of other multispectral TM bands (i.e. 60m than 30m).

Band7 (2.08-2.35 µm):

It has the ability to sharply separate land and water features and hence is also used for vegetation moisture. Generally, band 5 is preferred for this as well as soil and geological mapping.

Band Ratios:

Let us understand how different band ratio combinations are used for the interpretation of particular feature. Below is the sample image of Landsat TM of the area near Kota, Rajasthan and the natural drainage is the Chambal river flowing between Kota district.

                                                                  Figure 6: Kota district, Rajasthan

Band4/Band3 (NIR/Red):

It enhances the presence of vegetation, the brighter the tones the denser the vegetation. The below-given ratio image shows healthy vegetation by bright tone while water body is seen in dark tones. Along with these features line features such as roads are also clearly visible.

Figure 7: Band4 / Band3 of Landsat TM showing the same area as in the previous figure.

Band5/Band2 (MIR/Green):

This ratio can be used to detect the moisture level contained in the surface features. Darker the tone the greater the moisture content in the surface features. Hence water and soil moisture can be easily detected by this ratio.

Figure 8: Dark to bright tone represents high to low level of moisture content.

Image shows the part of Bhilwara and Kota districts from left to right.

Band3/Band4 (Red/NIR):

Urban, Water and line features like roads and drainage can be easily enhanced using this band ratio. Waterbodies are seen in brighter tone while vegetation is visible in dark black tone. The level of water content is easily detectable in the imagery.

     Apart from this, the ratio of band 2 to band 5 i.e. green by MIR is used to detect water bodies, ratio 3/1 to detect ferric iron rock, ratio 2/3 to sharply detect croplands & barren lands while ratio 4/5 enhances the waterbody, vegetation, and presence of moisture contents in the vegetation. In this combination, waterbody appears as dark tone and vegetation as light tone. Also, the ratio 7/2 separates forests from croplands but fails to discriminate between forests and water bodies as both appear as dark. It also enhances the highways, urban & built-up areas, and croplands and all of them appear in light tones.

Aside from the above-listed ratios, many other combinations are also possible with the Thematic Mapper sensor (total of 30 ratios are possible). Landsat TM uses a different number of bands from those of Landsat MSS, Spot, and IRS (Indian Remote Sensing) sensors.

Spectral Indices:

For decades, various spectral indices have been used by scientists to predict, model or infer surface processes to assess and monitor several different land change processes such as

1. Vegetation health and status

2. Burned Area

3. Fire Severity, etc

This section describes the spectral response of different earth surface features (vegetation, soil, and water) and various indices used in the field remote sensing.

Various spectral indices that are used to aid the interpretation of the imagery is broadly advanced in the following four stages:

Initially intrinsic indices were developed from simple band ratios, which highlighted the properties of vegetation at different stages of plant growth.

Secondly, indices were produced to compensate for background effects such as that caused in areas in which the soil response dominates the vegetation.

Thirdly indices were used to compensate the effects of atmospheric distortion.

Finally, new spectral indices were computed for the applications other than vegetation health which includes burn area assessment and fire severity.

Reflectance Characteristics of earth’s surface features:

Spectral index criterion was used to

1. Maximize the sensitivity of certain surface features properties( biophysical properties of the plant).
2. Normalize or reduce the effects of sun angle, viewing angle, atmosphere, topography, instrumental noise, etc to allow consistent spatial and temporal comparisons
3. Link the specific and measurable surface processes ( such as leaf area index, biomass, APAR, etc) (Jensen (RSE book,2000))

The spectral characteristics of three main features of the earth viz. vegetation, water, and soil are discussed below:

                                         Figure 10: Spectral response of vegetation, soil, and water

Vegetation:

The spectral characteristics of vegetation vary with the wavelength. Spectral response of healthy vegetation is minimum significant in the visible portion of the electromagnetic spectrum resulting from the pigments of the plant leaves. Chlorophyll strongly absorbs radiation in the red and blue wavelengths but reflects green wavelengths. The internal structure of healthy leaves acts as a diffuse reflector of near infrared wavelengths and thus reflectance dramatically increases in this region. Many plant stresses alter the reflectance in this region (IR region), and sensors operating in this region are often used for detection of stress vegetation.

Water:

Water (in the soil, vegetation, or water bodies) absorbs radiation at near-infrared (NIR) wavelengths and beyond. Reflectance from a waterbody can be due to the interaction of EM wave with a) the water’s surface (specular reflection), b) with the material suspended in the water and c) with the bottom of the water body. Because of this, water bodies, as well as all the features containing water, can easily be detected, located and delineated in the remotely sensed data. Clear water absorbs relatively little energy with the wavelengths less than 0.6µm, resulting in the high transmittance in the blue –green portion of the spectrum. Turbid water has a high reflectance in the visible region than clear water. This is also true for the waters containing high chlorophyll concentrations. The reflectance patterns are used to detect algae colonies as well as contaminations such as oil spills or industrial wastes, etc.

Soil:

The majority of the radiation incident on a soil surface is either reflected or absorbed and a little is transmitted. Factors such as moisture content, soil texture (concentrations of sand, silt, and clay), surface roughness, an iron oxide present under the soil, and organic matter content affect the soil reflectance curve and act over less specified spectral bands. These factors are less dominant than absorbance features in the vegetation reflectance spectra.

The moisture contained in the soil decreases the reflectance which is observed maximum in the water absorption bands at about 1.4, 1.9, 2.2 and 2.7 µm. soil moisture highly correlated with the soil texture.

Figure 11: Variation in spectral reflectance characteristics of soil as per the soil moisture content.

Spectral signatures can be obtained by measuring the energy that is reflected by targets on earth’s surface over a variety of different wavelengths. And different features can be identified by comparing their respective spectral responses, which may not be possible or may be difficult if we compare them only at one wavelength. For example, water and vegetation have a similar response in the visible wavelength but is different in infrared wavelength.

                  Figure 12: Variations in spectral reflectance of soil as per soil texture.

Vegetation Indices:

The ratio of near infrared band to red band is very high for healthy vegetation and is comparatively lower for not healthy vegetation. The resulting vegetation indices have been commonly used for the quantification of greenness in the vegetation and also to quantify the biomass values. Following are some of the popular vegetation indices commonly used in the field of remote sensing.

Normalized Difference Vegetation Index (NDVI):

The most common intrinsic index is the Normalized Difference Vegetation Index (NDVI) which represents the vegetation graphically. It is used to assess the healthy green vegetation under the target being observed. The NDVI identifies the photosynthetic affinity or “greenness” through the reflective properties of chlorophyll and mesophyll layers within the plants. It compares the normalized difference of brightness values of near infrared and red band. It might get affected by varying atmospheric conditions, illumination, viewing angles and soil reflectance etc.

Mathematically, NDVI can be represented as

Where NIR is the near infrared band and R denotes the red band. The resultant NDVI values lie between -1 and 1 which can be classified from very sparse vegetation to very dense vegetation as shown in the table2.

Table 2: NDVI parameters and their ranges

                                                     Source: Manual on Key indicators of

Desertification and Mapping Environmentally

Sensitive Areas to Desertification

Vegetation indices based on NDVI are extensively used to find the amount of vegetation all over the world. As proposed by Groten in 1963 maximum of 10 days NDVI images can be used for crop forecasts.

Transformed Vegetation Index (TVI):

Deering et al, 1975 produced a variant to NDVI by adding 0.5 to NDVI and taking the square root of the expression which resulted in the transformed vegetation index (TVI). It is mathematically expressed as

1 or close to 1. VRI values for bare soils generally are near to 1(1.21 for bare dry soils and 1.33 for bare wet soils). VRI increases with the increase in the amount of vegetation. While VRI values are not bounded it can be beyond 1.

Effects of soil background on Vegetation Index

Until the soil is fully covered with the vegetation the soil background will influence the vegetation index. For incomplete canopies, the wetting of previously dry soils (and vice versa) can cause a change in vegetation index. The change is further complicated by the fact that transmission of light through vegetation is considerably greater in NIR than in R band.

Both the ratio and the linear classes of vegetation index rely on the existence of the soil baseline in the red and NIR wavelength space for soil normalization.(Huete,1988). The intercept of this line is close to but does not pass through the origin and usually, there is some scatter of soil points away from the principal soil line. Such secondary soil influences are most noticeable with red and yellow colored soils (Kauth and Thomas, 1976).these two factors affect the discrimination of low amounts of vegetation from bare soil, and are significant in the arid regions and in the early stages of the vegetation growth. (Huete et al.1984).

Soil Adjusted Vegetation Index (SAVI):

Soil Adjusted Vegetation Index was developed by Huete (1988). He transformed the NIR and R reflectance axes in order to minimize the error obtained from the variation in the soil brightness values. For this, he proposed two parameters viz. L1 and L2 which when added to NIR and R reflectance bands reduces the variation caused by soil brightness. Mathematically, SAVI is:

Where L is a soil adjustment factor (it is 0 for low vegetation and 1 for high vegetation). Huete (1988) showed that optimal L values were obtained for different amounts of vegetations, but concluded that L=0.5 was optimal for the wide range of conditions. Although SAVI was developed using ground based data, Huete and Warrick (1990) demonstrated that it successfully minimized soil background effects using satellite data.

Several modifications have also been proposed over the years. For example, Steven in 1998 proposed the optimized soil vegetation index (OSAVI) which is given by

The development of improved vegetation index is an ongoing process. Major et al. (1990) used modeled data (in contrast to Huete’s experimental data) to adjust the RVI for soil background. Baret and Guyot (1991) discussed the potentials and limits of several VI for estimating leaf area index.

Perpendicular Vegetation Index (PVI):

Another index that can be used for the reduction of soil effects is the perpendicular Vegetation Index which is developed by Richarson and Wiegand in 1977. This index indicates the plant development which is based on the plot of the radiance values in the near infrared band to those in the visible red band.

To better understand the PVI index, let us first understand the concept of the soil line. The soil has a specific spectral signature that differentiates it from other Landcover features. The soil reflectance value increases with the increase in wavelength in the visible and NIR region; however, it is affected by many different parameters. Soil moisture and organic matter affect the soil absorbing capacity while soil texture and soil structure reflect the behavior of the soil i.e. whether the soil is a diffuse or Lambertian reflector. However, the reflectance of the particular soil type remains constant for NIR-R relationship.

Following is the plot of red against the near infrared reflectance of same soil under different moisture content conditions.

                                                      Figure 13: example of soil line between Red

Against NIR reflectance scatterplot

   It is clear from the above plot that reflectance values of soil for different moisture content conditions fluctuate proportionally and are correlated and have the linear relationship between them. The line showing this linear relationship is the soil line.

The perpendicular Vegetation Index used the red and NIR bands to calculate the perpendicular distance between the vegetation spot and the soil line in the R-NIR plot. The vegetation has high NIR and low Red reflectance values than the underlying soil and thus vegetation spot will lie in the top left corner of the scatterplot. The increase in the vegetation density will move the vegetation spot towards the top left corner away from the soil line. This indicates that the pixel whose reflectance value or the corresponding spot that lie near or on the soil line corresponds to soil and not the vegetation.

The limitation of PVI index is that it assumes that there will only be one type of soil beneath the vegetation, however, this is not true in most of the cases as is a mixture of soils available in the environment ( for example mixture of soil and rocks). To fix this problem Soil Adjusted Vegetation Index (SAVI) was proposed by Huete (1988).

Other indices such as Transformed Soil Vegetation Index (TSAVI; Baret and Guyot, 1991) and Modified Soil Vegetation Index (MSAVI; Qi et al. 1994), proposed different adjustment factors that better performed than SAVI in certain cases. Also, specific vegetation indices have been proposed in the past, using high spectral resolution data and observing specifically at chlorophyll and plant vigor.

Atmospheric Effects on Vegetation Index

Obtaining quantitative information from satellite data requires an accurate accounting of atmospheric effects. According to Switzer et al(1981), atmospheric path radiance contributes to falsely low values of NIR/R ratio in Landsat data. Jackson et al. (1983) examined the atmospheric effect on a number of vegetation indices by simulating Landsat MSS data for a data set comprising season long reflectance measurements over wheat. In general, both ratio and linear combination indices decrease with increasing atmospheric turbidity.

In order to remove the effect of the atmosphere on NDVI, Kaufman and Tanre (1992) replaced the red band with the linear combination of red and blue bands. The index so obtained was entitled as Atmospherically Resistant Vegetation Index (ARVI) and is mathematically expressed as:

Where NIR, B, R is the reflectance in near infrared, blue and red band respectively and RB = R-ɣ(B-R) where ɣ(gamma value) is the weighting function whose value depends on the aerosol type.

    Due to shorter wavelength blue band is much more easily scattered by the atmospheric particles than the red band. Thus the ARVI index takes the advantage of different scattering responses from the blue and red band to retrieve information regarding the atmospheric opacity.

The difference between the reflectance of the highly sensitive blue band and the less sensitive red band acts as an indicator of different atmospheric conditions. Here ɣ serves as a weighting function for the difference of reflectance values for these two bands. There can be various values that can be chosen for this function but according to Kaufman and Tanre’s statement in 1992, it is best to select gamma value as ‘1’ “when information on the aerosol type is not available”. Consequently, the main purpose of RB is to reduce the influence of the atmosphere, where a more accurate assessment of the red reflectance value can be obtained. The resultant value of this index ranges between -1 to 1.

Aerosol Free Vegetation Index (AFRI):

The AFRI model primarily uses the SWIR wavelength for the development of vegetation index. The SWIR band has both the advantage of being sensitive to vegetation while at the same time being susceptible to the influence of aerosols. This is because its wavelength is much larger than the radius of most aerosols, except for the very large dust particles, which allows it to be more capable of circumventing the suspended particles. This index simulates the reflectance of the red band with the SWIR wavelength i.e. the value of actual red reflectance is better calculated with the SWIR under the influence of aerosols. Even under the absence of any aerosol, this index could be “a match for the NDVI index” (Karnieli et al., 2001). For various landcover types, these ARVI index values are highly correlated with NDVI index (correlation value may reach up to 0.98). The mathematical expression of AFRI is:

     Where the subscript 2.1 and 1.6 refers to the reflectance of different wavelengths in µm situated within the SWIR region and ρNIR denotes the reflectance in the near infrared region. Both 2.1 and 1.6 are less affected by the absorbing atmospheric gases since they are situated within atmospheric windows. The index produces dynamical range similar to that of NDVI and AVRI which is between -1 and +1.

Reducing both Soil and Atmospheric Effects:

Pintyand Verstraete in 1992 defined the Global Environment Monitoring Index (GEMI) to correct both soil and Atmospheric effects.

Where ρ1 and ρ2 are the top of the atmospheric reflectance of the red and NIR region respectively.

GEMI was designed to be more sensitive to changes in vegetation than in NDVI, but less sensitive to atmosphere.

Normalized Difference Water Index (NDWI):

Remotely sensed imagery has long been used in water resources assessment and coastal management. These applications have involved the delineation of open water using thematic information extraction techniques. There are various methods for the extraction of water information from remote sensing imagery. One of the methods is through analyzing signature features of each ground target among different spectral bands, finding out the signature differences between water and other targets based on the analysis, and then using an if-then-else logic tree to delineate land from open water (Yu et al.1998, Xu 2002). The other one is a band-ratio approach using two multispectral bands. One is taken from visible wavelengths and is divided by the other usually from near infrared (NIR) wavelengths. As a result, vegetation and land presences are suppressed while water features are enhanced. However, the method can suppress non-water features but not remove them, and therefore the normalized difference water index (NDWI) was proposed by McFeeters (1996) to achieve this goal.

The NDWI is expressed as

Where Green and NIR are the green and near infrared bands respectively.

This index maximizes the water reflectance by using green wavelengths, minimizes the low reflectance of NIR by water features and uses the high reflectance of NIR by vegetation and soil features. As a result, waters have positive values and are enhanced, while vegetation and soil have zero or negative values and therefore are suppressed (McFeeters, 1996). Many built up land features also have positive values and as a result, they are often mixed with the extracted water information.

Wilson et al. in 2002 proposed a Normalized Difference Moisture Index (NDMI) which had an identical band composite with Gao’s NDWI. Both Gao’s NDWI and Wilson’s NDMI detects the vegetation water liquid and thus are different from McFeeters’ NDWI.