Information and Advsiory Note Number 63, November 1996
1.1 For many centuries people have exploited the advantage of a “bird’s eye view” of the landscape, where often high vantage points have been used in order to spy on or interpret areas of the earth’s surface. The collection of such information about the landscape from a distance is termed ‘remote sensing’ and the data collected in this way are called ‘remotely sensed data’.
1.2 Remote sensing has contributed to:
1.3 Within SNH remote sensing now assists in the following key areas:
1.4 This Note describes some remote sensing principles, data used for mapping the environment, and techniques used by SNH in some of the key areas above.
2.1 Remote sensing began formally with aerial photography, when in 1862 balloons were used to get a “bird’s eye view” of the landscape, by the American military during the American civil war. Progress was rapid at the turn of the 20th century, with the development of photographic technology and use of cameras fixed to aircraft. Following the Second World War, airborne technology was developed further by the Americans in order to map and survey Eastern Block countries. With the evolution of earth-orbiting satellites and digital electronic imaging systems in the 1960s, remote sensing moved into a new revolutionary age. Today, many different types of remote sensing technology are available resulting in a plethora of remotely sensed data.
2.2 Environmental remote sensing is now firmly established within many disciplines such as geography, geology, botany, zoology, civil engineering, forestry, meteorology, agriculture and oceanography. Within these, remotely sensed data usually refers to the recording and measuring of electromagnetic radiation by a variety of sensors. Further reading on this topic can be found at the end of this Note.
3.1 Environmental remote sensing systems record electromagnetic radiation. These have four main components (as shown in the diagram below).
(1) A source of electromagnetic radiation, which is natural e.g. the sun’s reflected light, the earth’s emitted heat, or man-made such as micro-wave radar.
(2) There is an interaction between electromagnetic radiation and the ground (soil, water, vegetation), where some of this radiation is often absorbed. This determines the makeup and amount of radiation emitted by, and reflected from, the earth’s surface.
(3) This radiation is then modified by different degrees of interaction with atmospheric components (e.g. water vapour, dust and gasses).
(4) Electromagnetic radiation is finally recorded by a sensor (e.g. camera) which is mounted in an aircraft or satellite.
3.2 Electromagnetic radiation (EMR) is the energy transmitted through space in the form of electric and magnetic waves. EMR forms as a continuous range of wavelengths, and the frequencies of these together group to produce the electromagnetic spectrum. Those which are of use in remote sensing lie mainly in the visible, near infrared and microwave ranges.
3.3 All types of land cover, different rock types, water bodies, vegetation, etc., absorb a portion of the electromagnetic spectrum within a range of possible values described as a “signature” of electromagnetic radiation. This signature is recorded by a sensor. Thus it is possible to classify a remotely sensed image on the basis of an understanding of which wavelengths are associated with which features, and how much is reflected from these.
3.4 The radiation wavelength values and signature are translated into discrete digital numbers within the sensors system. These numbers are referred to as greyscale values or brightness values, and usually range from 0-255 depending on the characteristics of the sensor. This is the base information for making an image or picture of the scene recorded from the ground. With airborne sensors and satellite sensors each sensor is designed to record a specific portion of the electromagnetic spectrum.
3.5 Many types of sensors are currently involved in recording features of the earth for land mapping purposes. Those most commonly used are listed in Table 1. The spatial resolutions referred to are defined here as a measure of the smallest object which can be resolved by the sensor on the earth’s surface. This unit is referred to as a ‘pixel’. This pixel provides the ‘building block’ of an “image” or “scene”.
4.1 One of the most common types of remote sensing data utilised, and often considered as a “tool of the trade” by many habitat specialists, is aerial photography. Although this was one of the earliest methods of remote sensing it is still the most popular and widely used. The wide use of benefits include:
4.2 As with many types of survey (e.g. ground based survey), air photo remote sensing survey has limitations:
4.3 Air photography is still the most popular basis for describing landscape features, habitat variations, human impacts and landcover types.
4.4 For large-scale surveillance using air photography it is necessary to devise techniques in Air Photo Interpretation (API). This is the identification of different tones of colour on photos and labelling these with a feature type (e.g. young plantation, heather moorland, blanket bog etc.). It is important in API that the interpretation of these colour tones is standardised between interpreters. It is also important that subsequent computations of these types/labels are fully documented so that the process is repeatable. API, like all remote sensing interpretation techniques must be tested and fully validated in the field, so that any confusion between interpreters and cover type labelling can be quantified. An example of an air photo interpretation dataset is the MLURI Land Cover of Scotland (MLURI 1993).
5.1 Digital imaging systems (scanners) can be mounted on both satellite and aircraft platforms. Interpretation and processing of these data require moderately powerful computer equipment, specialist computer software and a skilled image processing specialist/interpreter.
5.2 The lack of familiarity with such data has lead to a limited use for mapping landcover in the UK. However these data are more routinely utilised in continental Europe, Australia and America. The data produced from these systems should be viewed as a complementary data source to aerial photography and ground survey. Scanners offer a wide variety of benefits:
5.3 The use, however, of satellite imagery does have certain drawbacks:
6.1 Aerial photography is commissioned from air survey companies. Typically the survey company supplies 9” square prints with a 60% overlap between successive photographs along the flight line. The overlap enables the interpreter to view areas in stereo. Aerial photographs can be converted into a digital format using a flat-bed scanner.
6.2 Alternate photos from a flight line can be used for most digital analyses. Airborne scanner images are also obtained from air survey companies. Imagery may be supplied ready- georeferenced, typically on CD-ROM. Satellite images can be acquired from nationally-appointed distributors, e.g. National Remote Sensing Centre Ltd. Images can be requested from the archives or as new acquisitions. New acquisitions for an area of interest can be ordered with specific cloud cover content (typically less than 10% of the scene) and within a specific time frame. Images are supplied on CD-ROM or magnetic tape.
7.1 Image processing equipment: Digital remotely-sensed images usually require additional processing in order to be used as a mapping tool or to extract information from the data. This is achieved within an image processing computer software system. These are available for UNIX workstations and increasingly for PCs, due to the recent improvements in computer hardware.
7.2 How to process an image: A typical image processing sequence for the production of a land-use map from a satellite image is: (elements in bold are discussed in detail below):
1. Georeference image to Ordnance Survey National Grid.
2. Select the appropriate image wavelength band(s).
3. Improve contrast and brightness within the image to increase the visual discrimination of features.
4. If necessary, apply a ‘sharpening’ filter to enhance the detailed lines within an image or a ‘smoothing’ filter to remove noise. An image can contain noise due to minor variations in the response of the detectors in the scanner.
5. Transform the data using a statistical technique such as Principal Components Analysis to increase the likelihood of discriminating the different land use types within the scene.
6. Create masks to ensure that the analysis is focused upon the area of interest.
7. Apply a statistical classification procedure, i.e. automatically assign a class number or class name to each pixel using the ‘signature’ of that pixel.
8. Assess the robustness and accuracy of the land cover classification.
9. Produce a thematic map at a specified scale showing the land cover details and ancillary information derived from a GIS.
10. Export the land cover map to a format suitable for use within a GIS.
7.3 Georeferencing: Satellite images supplied from the data vendor are distributed in a Space Oblique Mercator projection (known as Level 1B format). The vendor has removed distortions due to the orbital distortion (satellite wobble) and the tilt of the sensor (for the SPOT and IRS satellites). If the satellite data are to be integrated with other geographically-referenced information, such as that held within GIS systems, further processing is required. Here several easily-identifiable points need to be distributed around the satellite image. The map co-ordinates for each point are entered by measuring these points on a map.
For Ordnance Survey maps these points lie on a grid, specified by a Transverse Mercator projection. From these points the image processing software computes a transformation in which all points on the satellite image are relocated to the same projection as the map. A transformation is then applied to the satellite image to produce a georeferenced image.
A refinement of the above process involves taking into account the distortions in the image due to varying topography (altitude). With the aid of a grid of heights known as a Digital Elevation Model (DEM), a more accurate georeferencing procedure called orthorectification can be performed.
7.4 Understanding wavelength combinations: Each wavelength band in the satellite or airborne scanner image has been chosen for a particular purpose (e.g. Table 2). Some of the bands have been selected by the instrument designer for the purpose of vegetation mapping, or geological mapping or shallow water bathymmetry etc.
These wavelength bands can be viewed either singly on a computer screen as a monochrome or greyscale image; or as three bands combined to produce a colour composite image. A colour composite is constructed by assigning a different wavelength band to the red, green and blue components which make up a colour image on a colour display. If band 3 is assigned to the red component; band 2 to the green component; and band 1 to the blue component a ‘true colour’ composite (view/image) similar to a high altitude air photograph is produced. For vegetation mapping a typical combination is to assign:
This type of combination is known as a ‘false colour’ composite. Typically, grassland appears bright green; woodland dark green; and moorland areas as a mixture of browns and dark pinks. By using the appropriate wavelength bands the false colour composite enables the interpreter to identify the different land cover types more easily than the true colour composite.
7.5 Typically a simple enhancement is applied to the image to improve the range of colours and the contrast in the image. Both of the colour composites shown in Figure 1 have been enhanced in this way. The colour composite can be printed as a map for field reconnaissance or used as a reference for interpreting the results of an image classification.
7.6 Thematic mapping: A thematic map shows specific observations within a theme. One depicting vegetation cover can be produced using remote sensing either by manual photo-interpretation techniques using the information from colour composite images, or by computer-aided analysis.
Computer-aided techniques use a statistical classification procedure. Two approaches can be employed:
7.7 Both of the above computer aided classifications can use wavelength bands selected by the user or derived from Principal Component or vegetation index images.
8.1 Satellite image processing: The Scottish Blanket Bog Inventory (SBBI) is currently involved in a programme of work which aims to map and characterise the blanket bogs of Scotland using current techniques in image processing and peatland ecology. The map shown in Figure 2 was produced using an unsupervised classification with specific techniques developed and employed for mapping peatlands.
For details on the methods and applications within the SBBI please refer to the further reading at the end of this document.
8.2 Monitoring: The National Countryside Monitoring Scheme (NCMS): This is a project which has assessed the degree of landcover change in Scotland using three sets of temporal air photographs: 1940s 1970s and 1980s. This sample based project (covering 7.5% of Scotland’s land area) identifies landcover types (feature types) through the interpretation of different ‘grey-scale tones’ on black and white air photos. Where different tones are identified these are given a feature label (e.g. blanket mire). These labels are arranged within a GIS along with line or polygon vectors which show the extent of the cover type identified. In this way it has been possible for NCMS to determine changes in landcover feature types through time.
(Further detailed information is available in Information and Advisory Note No. 38).
8.3 The Landcover of Scotland (MLURI 1993). The Landcover of Scotland (LCS) survey was proposed by the Under-Secretary of State for Scotland in 1987, with the aim of producing a detailed census of landcover throughout Scotland, using API. The LCS is hierarchical and recognises the principal, major and main land cover features identified by API methods. e.g. For one given principal feature we have:
There are 5 other principal features, (farms and developed rural land, bare ground, and miscellaneous features, woodland, and agricultural land), all of which contain major features, main features and sub-categories (see further reading for details of these). LCS can allow evaluation of the distribution of categories deemed to be of conservation importance, for example, heather moorland (Thompson et al 1995). This dataset is the only near complete census of the landcover of Scotland available to SNH. It is currently held within the SNH GIS and at MLURI.
This Note has provided an introduction to the general principles and use of environmental remote sensing. Remote sensing often provides much of the core information in a GIS. The two technologies therefore provide a complementary capability for a cost-effective means of assessing ecological characteristics over landscape-scale areas and to periodically update basic data. As the variety and volume of ecological data increases, the combined use of GIS with other computerised applications, including image processing from environmental remote sensing, will prove increasingly valuable for allowing the collation, storage, integrated analysis and dissemination of spatial data. This will in turn improve the quality of geographic information and its accessibility to future ecologists and conservationists.
Alaric, V. (ed). (1994). Remote Sensing and GIS in Ecosystem Management. Island Press, Washington, U.S.
Gardner, S. (1996). An Introduction to Geographic Information Systems. Information and Advisory Note Number 34. Scottish Natural Heritage, Perth.
Macaulay Land Use Research Institute. (1993). The Landcover of Scotland 1988 Final Report. The Macaulay Land Use Research Institute, Aberdeen.
Pooley, M.R & Jones, M.M (1995). Application of Remote Sensing to Habitat Mapping and Monitoring. Scottish Natural Heritage Research and Review Series, Perth.
Quarmby, NA., Everingham, F, & Reid, E (in press, 1996). Determining the composition of the blanket bogs of Lewis and Harris using Landsat TM. Scottish Natural Heritage Research, Survey and Monitoring Series, Perth.
Reid, E., Mortimer, G.N., Thompson, D.B.A and Lindsay, R.A. (1994). Blanket Bogs in Great Britain: An Assessment of large-scale pattern and distribution using Remote Sensing and GIS. In Large Scale Ecology and Conservation Biology (edited by PG. Edwards, R.M. May and N.R. Webb) pp. 229-246. 35th Symposium of the British Ecological Society, Blackwell Scientific Publications, Oxford.
Reid, E., Quarmby, NA., and Thompson D.B.A., (1996). Characterisation of blanket bogs using satellite imagery. Information and Advisory Note. Scottish Natural Heritage, Perth.
Reid, E., Ross, S.Y., Thompson, D.B.A. and Lindsay, R. A. (in press, 1996) From Sphagnum to a satellite: towards a comprehensive inventory of the blanket mires of Scotland. In: Conserving Peatlands (Ed. by L. Parkyn, R.E. Stoneman and H.A.P. Ingram), pp. 204-216. Centre for Agriculture and Bioscience Publications, Wallingford.
Thompson, D.B.A., Hester, A.J. & Usher M.B. (eds). (1995). Heaths and Moorland: Cultural Landscapes. HMSO, Edinburgh.
Tudor, G.J. (1996). Landcover change in Scotland: National Countryside Monitoring Scheme results for the 1940s and the 1970s. Information and Advisory Note Number 38. Scottish Natural Heritage, Perth.
Tudor, G.J., Mackey, E.C. & Underwood, F.M. (1994). The National Countryside Monitoring Scheme: the changing face of Scotland, 1940s to 1970s Main Report. Scottish Natural Heritage. Perth.
We would like to express special thanks to Des Thompson for advising on this note, and Gavin Tudor for providing specialist advice and editorial comments.