1 The interaction of visible and near infrared EMR with soil

PROPERTIES OF EARTH SURFACES AND THEIR INTERACTIONS WITH ELECTROMAGNETIC RADIATION AT VISIBLE AND NEAR INFRARED WAVELENGTHS For a fixed distribution of incident radiation, the particular properties of a surface will determine how much EM energy is radiated in the direction of a sensor. The reflectance varies with wavelength. In the early 1970s it was thought that every surface would have a unique spectral signature and that once this had been established from laboratory experiments it would be possible to map any surface type from multispectral imagery. As you will learn in this section, however, although there are general differences between surfaces there exists considerable variation in the reflectance of a single surface type. These variations and the reasons for them are at the heart of remote sensing and I advise you to read the following very carefully. You should note the general differences in reflectance between soil, rocks, water and vegetation and consider the effects of other factors on the inherent variability of each of these surfaces. In addition, you should read the following: Price, J. C., 1994. How unique are spectral signatures?, Remote Sensing of Environment, 49, 181-186. 1 The interaction of visible and near infrared EMR with soil In general, soil surfaces are brown to the human eye. Brown colouring is a product of green and red EM radiation such that brown surfaces absorb more blue EMR than either green or red. Furthermore, very little energy is transmitted through soils, the majority of the incident flux is absorbed or reflected. The technical term for these types of surface is single scatterer. In the case of soil surfaces the level of reflectance gradually increases with wavelength in the visible and near infrared spectral regions. You can see from Figure 1 that maximum soil reflectance occurs at near infrared wavelengths. Figure 1: Reflectance of bare soil surfaces

The five most important characteristics of a soil that determine its reflectance properties are: Most important: Least important: moisture content organic matter content texture structure iron oxide content Source: Stoner and Baumgardner (1981) 1.1 Soil moisture content You should have found that the presence of soil moisture reduces the surface reflectance of soil at all visible wavelengths (Jensen, 1983). This occurs until the soil is saturated, at which point further additions of moisture have no effect on reflectance. Reflectance at near infrared wavelengths is also negatively related to soil moisture; an increase in soil moisture will result in a particularly rapid decrease in reflectance due to water (H20) and hydroxyl (HO) absorption features at 0.9 μm, 1.4 μm, 1.9 μm, 2.2 μm and 2.7 μm. The effect of water and hydroxyl absorption is more noticeable in clay soils because they have much bound water and very strong hydroxyl absorption properties. 1.2 Organic matter content Soil organic matter is dark and its presence will decrease the reflectance from the soil up to an organic matter content of around 4-5%. When the organic

matter content of the soil is greater than 5%, the soil is black and any further increases in organic matter will have little effect on reflectance (Curran, 1985). 1.3 Texture and structure Texture (the proportion of sand, silt and clay particles) is related to structure (the arrangement of sand, silt and clay particles into aggregates) and, therefore, they will be discussed together. The relationship between texture and structure can best be described by reference to two contrasting soil types. A clay soil tends to have a strong structure which leads to a rough surface on ploughing, causing small shadows and lowering reflectance values. In contrast a sandy soil exhibits weak structure which leads to a fairly smooth surface on ploughing with few shadows. It should be noted that the effects of soil structure complement other properties such as low moisture and organic matter content to increase the level of sandy soil reflectance. 1.4 Iron oxide content Iron oxide gives many soils their rusty red colouration by coating or staining individual soil particles. Iron oxide selectively reflects red light (0.6-0.7 μm) and absorbs green light (0.5-0.6 μm). This effect is so pronounced that Vincent (1973) used a ratio of red to green reflectance to locate iron ore deposits (Curran, 1985). Look again at Figure 1 - can you identify the increase in contrast between red and green reflectance of the iron dominated soil? 2 The interaction of visible and near infrared wavelengths of EMR with water Unlike soil, the majority of the radiant flux incident upon water is not reflected but is either absorbed or transmitted. At visible wavelengths of EM radiation little energy is absorbed, a small amount, usually under 5%, is reflected and the majority is transmitted. If you happen to be sitting in a deep bath whilst reading this unit then this is the reason why you can see your toes through the water. Water absorbs strongly at near infrared wavelengths, leaving little radiation to be either reflected or transmitted (Figure 2). You would not see through water as clearly at these wavelengths. Figure 2: Reflectance of water with different sediment loadings

Although water surfaces may be more homogeneous than soil surfaces we can still expect some variability in the reflectance of a body of water. The two most important factors are the depth of the water and the materials within the water. In shallow water some of the radiation is reflected not by the water itself but from the bottom of the water body. Therefore, in shallow pools and streams it is often the underlying material that determines the water body's reflectance properties. Anderson and Robbins (1998) used this observation to spectrally discriminate between acid and neutral streams based on the reflectance characteristics of acid and neutral iron precipitates. Field spectral measurements were conducted in situ on iron-oxide precipitates coating bottom substrates of an acidified stream (ph < 4) in the Virginia Gold-Pyrite Belt. Spectra of two neutral (ph 6 and 7) streams, one with iron precipitates present, were also measured. Acid precipitates were found to have higher spectral reflectance than neutral precipitates in the 0.65-0.75 μm region. With regard to water quality, the acid stream possessed low ph and high specific conductance. Guided by the field spectral data, a remote sensing technique using a digital multispectral video system was used to detect mine drainage in two Virginia streams. Where acidic discharge was present, falsecolour composites showed precipitates having high reflectance in both the 0.65 μm and 0.75 μm wavebands. In contrast, neutral iron-rich streams were low in reflectance, and streams unimpaired by mine drainage had virtually no detectable reflectance. Notwithstanding this example, the three most common materials suspended in water are non-organic sediments, tannin and chlorophyll: Non-organic silts and clays increase the reflectance at visible wavelengths due to interaction with and scattering by soil-like particles (Figure 2). If you are lucky enough to have visited a glacier then you may recall the milky white colouration of a proglacial stream. This is caused by the high concentration of fine rock flour sediment suspended in the water.

In agricultural scenes the main colouring agent is tannin produced by decomposing humus. This is yellowish to brown in colour and results in decreased blue and increased red reflectance. A good example of the effects of tannin can be found in streams that drain peat moorlands. Chlorophyll content must be very high before changes in reflectance can be detected (Piech et al., 1978). Water bodies that contain excessive levels of chlorophyll have reflectance properties that resemble, at least in part, those of vegetation with increased green and decreased blue and red reflectance. 3 The interaction of visible and near infrared wavelengths of EMR with vegetation You are beginning to observe a familiar pattern to this subsection. For soil and water surfaces you have described their general spectral signatures and then explained variations to those reflectance properties. The reflectance of vegetation may be a little more complex but we can still follow a similar approach. For you to understand variations in the multispectral reflectance of a vegetation canopy it helps if you first consider the reflectance properties of an individual leaf. 3.1 Reflectance of a leaf A leaf is built of layers of structural fibrous organic matter, within which are pigmented, water-filled cells and air spaces (Figure 3). Subsequently, three features - pigmentation, physiological structure and water content - have an effect on the reflectance, absorption and transmittance properties of a green leaf (Curran, 1985). Each feature dominates the reflectance in particular spectral regions - pigment absorption at visible wavelengths, physiological structure at near infrared wavelengths and absorption by water molecules at specific wavelengths in the near infrared region. Figure 3: Physiological structure of a leaf (cross-sectional view)

3.1.1 Pigment absorption at visible wavelengths What colour is healthy vegetation? Unless you re living in a particularly arid region then I hope that we can agree that vegetation is green i.e. it reflects more green EMR than either blue or red EMR. This is because higher plants contain four primary pigments, chlorophyll a, chlorophyll b, ß carotene and xanthophyll, all of which absorb visible light for photosynthesis. Chlorophyll a and b, which are the more important pigments, absorb portions of blue and red light; chlorophyll a absorbs at wavelengths of 0.43 μm and 0.66 μm and chlorophyll b at wavelengths of 0.45 μm and 0.65 μm. These absorption features occur in the blue (0.4-0.5 μm) and red (0.6-0.7μm) spectral regions. Subsequently, healthy vegetation has relatively less absorption (and higher reflectance) at green wavelengths (Figure 5). The less important carotenoid pigments, carotene and xanthophyll, both absorb blue to green light at a number of wavelengths (Curran, 1985). Pigment absorption only occurs at visible wavelengths and does not affect reflectance at near infrared wavelengths. Unfortunately, that is not the end of the story. Later in this section you will consider the effects of seasonal variations in pigment concentrations within a leaf.

3.1.2 Physiological structure and reflectance at near infrared wavelengths In the UK the recently privatised rail companies have produced some wonderful explanations for delays to trains. Delays have been caused by the wrong type of snow and more recently because leaves falling onto the track were too juicy! Leaves are juicy because they consist of water-filled cells and air spaces. You may recall that light bends when it passes across the boundary of two media with different refractive indices - the process of refraction. Air and water have different refractive indices and therefore as rays of light are transmitted through a leaf they are refracted at every cell boundary in the mesophyll and at other discontinuities, such as those between membranes and cytoplasm, in the upper part of the leaf (Figure 4). Unlike visible EMR, near-infrared EMR is not absorbed by the various pigments present in green leaves. The amount of near-infrared radiation reflected by a leaf at these wavelengths is the sum of specular reflection from the cuticle and upper epidermis and diffuse reflection produced by the scattering effect of multiple internal refractions (Figure 4). We should note that the diffuse reflection is by far the greater component. Figure 4: EMR pathways through a leaf

The combined effects of leaf pigments and physiological structure give all healthy green leaves their characteristic reflectance properties: low reflectance of red and blue light, medium reflectance of green light and high reflectance of near infrared radiation (Figure 5). The sharp contrast between red and near infrared reflectance is commonly known as the red edge. Figure 5: Reflectance, transmittance and absorption characteristics of a leaf at visible and near infrared wavelengths The major differences in leaf reflectance between species are dependent upon leaf thickness which affects both pigment content and physiological structure. For example, a thick wheat flag leaf will tend to transmit little and absorb much radiation whereas a flimsy lettuce leaf will transmit much and absorb little radiation (Curran, 1980). Reconsider the pathways of different rays of near infrared light through a leaf (Figure 4). Imagine many more (thousands, millions...) rays incident upon the surface of a leaf. If we assume that internal refraction occurs so frequently so as to have the effect of random scattering then we might suppose that approximately 50% of incident near infrared radiation is reflected towards a sensor, a figure confirmed by field observations of healthy green vegetation.

3.1.3 Water absorption at near infrared wavelengths The previous section suggests that a leaf should reflect approximately 50% of incident radiation. You have already found that the reflectance at visible wavelengths is much lower than this because of absorption by pigments. The near infrared reflectance of a leaf may be less than 50% due to absorption by water. If you have any doubts about the spectral properties of water then look back at Section 2. Leaf reflectance at all visible and near infrared wavelengths is reduced as a result of absorption by water. In particular, three major diagnostic absorption features occur in the near infrared region near wavelengths of 1.4 μm, 1.9 um and 2.7 μm and two minor features occur near wavelengths of 0.96 μm and 1.1 μm. The reflectance of the leaf at these wavelengths is negatively related to both the amount of water in the leaf and the thickness of the leaf. Water vapour in the atmosphere, however, also absorbs radiation at these wavelengths and precludes their use for satellite remote sensing. To avoid these wavelengths measurements are made at three atmospheric windows within the visible/near infrared region that are relatively free of water absorption (0.3-1.3 μm, 1.5-1.8 μm and 2.0 2.6 μm). You should note that within these broad wavebands the level of reflected EMR is still sensitive to leaf moisture. You are now in a position to explain the spectral reflectance of a leaf but the reflectance of an individual leaf is insufficient to describe the reflectance of a vegetation canopy. This is because a vegetation canopy is not a large leaf but is composed of a mosaic of leaves and other canopy components. 3.2 Reflectance of a vegetation canopy We have just said that vegetation canopy reflectance is a product of the reflectances of several different components. We should add that the 3- dimensional architecture of the canopy causes intra-canopy shadowing, that neighbouring plants may cause inter-canopy shadows and that, unless the canopy is complete, the background will also contribute to canopy reflectance. The net effect is that the reflectance recorded by a sensor is primarily related to the area of leaves within the canopy, rather than to the pure reflectance of the component leaves (Colwell, 1974). 3.2.1 Leaf area index The area of leaves within the canopy can be recorded by the leaf area index or LAI - the unit leaf area per unit area of ground. For canopies with a high

LAI, as is often the case for grasslands, then the reflectance properties of the canopy are similar to the properties of the leaf. In these cases the canopy will have low blue and low red reflectance, medium green reflectance and high near infrared reflectance. While these reflectance properties hold good for a healthy grass crop on a medium toned soil they show considerable variation with the environment. With no change in the reflectance of the individual leaves the canopy reflectance could vary appreciably due to the effect of the soil background, the presence of senescent vegetation and occurrence of phenological canopy changes, the angular elevation of Sun and sensor and the canopy form. 3.2.2 Background soil effects The reflectance of the soil has a considerable effect on reflectance of the vegetation canopy, and in particular for sparse vegetation canopies. If you simulate green, red and near infrared reflectance for vegetation with different LAI against light and dark soil backgrounds, you will find some soil/waveband combinations that are unsuitable for the remote sensing of vegetation. For example, on dark toned soils with low red reflectance there is little change in the red reflectance of the canopy with an increase in the canopy LAI as the leaves have similar reflectance to the soil. On a light toned soil with a high reflectance, the relationship between near infrared reflectance and LAI is weaker than on a dark soil, because the contrast between leaves and soil is lower at near infrared wavelengths. The angular elevation of the sensor determines the amount of soil that is visible; for as the angle of elevation moves from the vertical, less soil and more vegetation is seen. Although the vast majority of remotely sensed measurements are vertical there has been an increased use of sensors which sample data off-nadir. For example, the SPOT HRV sensor views the Earth surface at zenith angles up to 27o and ATSR onboard ERS has a forward view zenith angle of 55o (Figure 6). Figure 6: View zenith angle of SPOT HRV and ERS ATSR sensors

3.2.3 The effect of vegetation phenology Phenology is the study of the relationship between vegetation growth and environment. The phenology of a plant defines its seasonal pattern of growth, flowering, senescence and dormancy. Each season, plants experience chemical, physical, and biological changes, known as senescence, that result in progressive deterioration of leaves, stems, fruit and flowers. In contrast to the situation with green leaves there is negligible transmission through dead leaves at all wavelengths and the reflectance spectra of dead leaves are similar for all species. Vegetation senescence results in a 5-10 % rise in reflectance at blue and red wavelengths due primarily to the breakdown of chlorophyll into anthocyanin compounds with lower absorption. In contrast, the near infrared reflectance of the leaf remains relatively high, possibly slightly higher than for a green leaf. The high reflectance of brown leaves is most likely due to both the hardening of the cell walls, reducing transmission, and a reduction in cell water content thereby reducing absorption in the moisture absorption bands. As described in Section 3.2.1, it is the LAI that controls the near infrared reflectance of a vegetation canopy, although obviously senescence is also associated with changes in LAI due to leaf fall. In consequence, dead vegetation tends to have relatively high near infrared reflectance soon after it has started to brown off and to become darker with age. In the mid-latitudes, annual plants typically lose most or all of their roots, leaves and stems each year. Woody perennials typically retain some or all of their roots, woody stems and branches but shed leaves. Evergreens, including tropical plants, experience much more elaborate phenological cycles; individual leaves may senesce separately - trees do not necessarily shed all leaves simultaneously - and individual trees or branches of trees may shed leaves on cycles quite distinct from others in the same forest.

Figure 7: Field experiment to investigate the effects of biophysical and biochemical changes on the multispectral reflectance of a vegetation canopy Source: University of Kassel (2007) A study of savannah vegetation in the eastern Kalahari has shown that areas suffering from bush encroachment green up before grass dominant areas (Trodd et al., 1997). This is because bushes have deeper roots and access to subsurface water such that they are able to respond immediately once growing conditions are favourable. In contrast, grasses rely on rainfall to stimulate growth and therefore delay their response until a critical level of rain has fallen. Other phenological effects include changes to canopy form and leaf orientation. For example, a change from vertical to horizontal leaf inclination results in decreased red and increased near infrared reflectance without any change in the LAI of the canopy (Suits, 1972). If leaf angle changes are due to the canopy wilting then the situation may be further complicated by associated changes in the leaf water content. Recent experiments on legume/grass swards have confirmed the interaction of biophysical and biochemical factors on diurnal variations in reflectance (Figure 7). 3.2.4 The effect of canopy form The form of a vegetation canopy will determine the amount of shadow seenby the sensor. For example, an orchard of trees with a distinct 3-dimensional canopy architecture will have a lower reflectance than a field of uniform (2- dimensional) grass due to more inter-canopy shadows. Shadowing effects are more pronounced at low solar elevation angles.

4 The interaction of visible and near infrared wavelengths of EMR with rocks Rocks, like soils, are single scatterers and exhibit relatively simple spectral properties. Unlike soils rock reflectance is less dependent on water content and completely independent of organic matter content, texture or structure. Rock spectral reflectance primarily depends on their mineral composition. Mineral reflectances measured in the laboratory (Figure 9) show 2 sources of variation the overall level of reflectance and specific absorption features. Lower reflecting minerals, such as geothite, have similar spectral properties to soil surfaces low to moderate reflectance at visible wavelengths increasing into the near-infrared - whereas high reflecting minerals, such as quartz and calcite, exhibit almost uniformly high reflectance throughout the visible, near- and shortwave infrared spectrum. Differences between high reflectance minerals occur at specific wavelengths and these absorption features are known as 'diagnostics' or diagnostic absorption features. See, for example, the effect of carbonate on the reflectance of calcite at 2.35 μm. Some of the most pronounced absorption features are associated with clay minerals. As clays such as kaolinite may be associated with the presence of precious metals it was not surprising that the mineral exploration community were the drivers behind the decision to include a SWIR waveband on the Landsat TM sensor that measures reflectance at 2.08 μm 2.35 μm. They are now leading the way in hyperspectral remote sensing at these wavelengths. Unlike other minerals the strength of bonding between clay particles and water produces some absorption features at 1.4 μm and 1.9 μm. Figure 9: Reflectance of selected minerals

To find out more about the measurement and analysis of high resolution spectra and hyperspectral remote sensing then visit the USGS Spectroscopy Laboratory web site for detailed explanations of reflection and absorption processes and an impressive library of spectra. A link to the site is available in the Resources for this section. 5 A few final comments Before concluding this section I wish to draw your attention to the fact that the descriptions of the spectral properties of soil and water surfaces and vegetation canopies are typical of mid-latitude temperate climates. Those of you living or working in other parts of the world may experience a different set of surfaces or different canopy types. For example, in arid and semi-arid regions you find that low vegetation cover exposes rocks at the surface, rather than soil, which may exhibit particular spectral characteristics due to its mineral content; at high latitudes you may observe snow or ice surfaces with exceptionally high reflectance that saturates conventional sensors. You must also be aware that solar elevation is generally lower and therefore objects cast longer shadows at high latitudes. You have now completed this section and I hope that you understand the principles of multispectral remote sensing so that you are able to explain the bidirectional reflectance of any surface you wish to measure. 6 What you have learnt in this section Every target has a spectral signature but that these signatures are rarely unique The different properties - chemical composition and physical form

of soil, water and vegetation determine their reflectance characteristics It may be possible to discriminate between these surfaces at specific wavelengths or at a combination of wave bands It may be possible to infer the properties of a known surface from variations in the measured reflectance