There are a large number of functions which can be further performed on the digital data in order to meet specific requirements. Enhancements are designed to aid the analysis of images and pictures so that the maximum use can be made of them, e.g. the use of colour rather than grey tones means that more discrimination can be applied. We cannot examine all display and enhancement techniques, but some are shown in Figure 4.19 and the major ones are described:
Contrast Stretching. Sensors are designed to record a full range of values from a very strong signal (usually displayed as white) to no signal (black), with in-between some 254 grey tone steps. Few scenes require such a dynamic range and data values for any scene tend to occupy less than half of the available range. The values can be changed to occupy the full scale by using a stretch programme that can be applied to give a linear, uniform distribution, gaussian or other arbitrary stretch.
Edge Enhancement. This technique can be visualized as using the computer to recognize subtle but sudden steps in the grey scale, where the sudden step in local grey scale is continuous over a regional variation in background digital values. Its effect is to increase the grey scale of the step so that its edge is more clearly displayed. It is particularly useful for the detection of linear elements, either oceanic (warm or cold water fronts) or man-made (canals, waterways, etc.).
Density Slicing. This involves delineating the different ground features by colour coding, i.e. areas having similar reflectance values are assigned to one colour. Up to eight “slices” might be taken through the range of pixel values which would produce an image showing eight key features in eight colour ranges. The computer can calculate when to best define a threshold to separate individual colours and it can also calculate the area covered by the different features (colours or colour tones). It can be particularly useful in water depth analyses. A simpler way of using colour is to select three spectral bands and feed the data from each through a different colour gun on the image processors monitor. This results in either a natural-colour or false-colour composite. False-colour indicates that colours have been assigned to information sensed in bands which are beyond the visible range, i.e. they simulate radiation which cannot be seen.
Figure 4.19 Some Important Image Processing Techniques
Spatial Filtering. This term includes a family of operations which use values for groups of pixels to spatially filter areas to produce images with different effects such as smoothing, or line enhancement. For example, “low pass” filters assign to each pixel of the radiance values to be suppressed. On the other hand “high pass” filters can enhance differences of reflectance values in the image.
Principle Component Analysis (Data Compression). This analysis compresses multi-spectral data sets as an aid to image classification and pattern recognition. The normally highly correlated spectral bands are transformed into a new set of variables or components that are uncorrelated. The new components represent more efficiently the variance of the data set. Using Landsat MSS data the first two principle components generally account for over 95% of the original variance, although lower order components may include useful image detail.
Image Classification. Spectral classification techniques can be grouped into supervised and unsupervised classifications. We can only briefly describe classification here-details are given in Lantieri (1988). If a RS user wishes to classify all the tones on an image which have a certain range of, for instance, crop categories, then by finding out the typical pixel values for each crop (using field investigations), the findings can be applied to the rest of the image, i.e. the computer can be “trained” to seek out various spectral signatures for one small area and then apply it across the image. Several different statistical methods are available for determining how to allocate pixel values to categories. Image classification is usually successful if the targets are spectrally very different; unfortunately in most land use analyses many crops register similar pixel values.
Multi-Temporal Analysis (Change Detection). Data sets collected at different dates (multi-date) or at different times (multi-temporal) of the same day can be registered and compared pixel by pixel. The pixel value of each pair can then be ratioed or subtracted. Other procedures can also be used, such as the statistical procedures of Principle Component Analyses, to compose the multiple data sets. The resulting digital numbers will highlight differences between the images. This is of particular use in studies of crop state, seasonal vegetation anomalies, rivers, thermal changes, coastal processes, etc.
Digital Mosaics. A display application of interest is the “patchworking” of images from the same sensor to form mosaics covering large areas. This poses radiometric and geometric problems since the selected scenes may have been imaged at different seasons of the year, i.e. when colour balances are different. Mosaics are of value in studying regional features of land cover, geology, natural resources and in cartography.
In this section we will look at how RS might be applied to the development and management of aquacultural and inland fisheries. The accent will be on potential applications rather than actual applications since examples of the latter will be given in Chapter 7, and because one of our basic aims is to exemplify ways in which RS can be of use in areas where location optimizing suffers from a lack of existing data. Our approach will be from the viewpoint of examining the individual production functions and to see how far RS methods might be applied. In doing this there will clearly be instances where remotely sensed data can be used directly, but there will be many more instances where various “proxy” remote data, or as Kapetsky and Caddy (1985) call it, “inferred” data, can be used. For most usable images a degree of “ground truthing” would be desirable and the extent to which this is necessary will vary greatly from function to function.
Again, we will concentrate here upon images obtained from satellite-based sensors since these are the most widely available, they are able to give wide area coverage, they may be available for many time periods and they are potentially cost-effective. However, given the restricted resolution available, we will be discussing location optimization on the macro rather than micro scale. For micro scale analysis airborne sensors are particularly useful and they would be especially important for specific site selection and for other purposes such as monitoring existing fishing levels or for any micro environmental analyses.
By the end of this section the reader should be able to assess the extent to which RS can help in providing information for the management and development of fisheries and aquaculture. Clearly, there are many functions which can never be aided by RS, and for these land-based detection methods will always be applicable. Cheney and Rabanal (1984) note that the decision to use RS for aquaculture siting should be based on a number of factors:
Though we will exemplify many potential RS applications, Cheney and Rabanal (1984) also provide a comprehensive list, as does Pettersson (1989), and Cordell and Nolte (1988a) provide a discussion of the problems of applying RS to aquaculture and inland fisheries location.
In section 2.3.1 we saw that a large number of variables many influence water quality, and thus potential fish productivity. Though it is not yet possible to use RS techniques to detect variations in most of these qualitative factors, it is possible to gain a number of inferences as to water quality, e.g. by looking at:
Because some of these variables exhibit similar radiometric values, it has proved necessary to establish various empirical models to show the relationship between water radiance values and different environmental parameters. Even when these models have been verified they are not always globally valid. Further problems with assessing water quality via RS exist, such as the fact that the spatial resolution is insufficiently detailed for most river assessment and the fact that most radiance values will only refer to near surface water. Additionally, since many water qualitative parameters undergo frequent changes, the monitoring frequency from satellite-derived data is insufficient, and indeed even “real time” images may not be real enough in the case of some pollutants.
Acid Rain. It is recognized that this is an increasing problem in several major world regions, and that areas where vegetation is showing signs of damage will also exhibit water having adverse acidic conditions. Using Landsat TM imagery obtained over southern Germany, in the wavelengths of 0.75 to 1.3 um, damage assessment could be determined since affected trees gave significant spectral changes. Information derived from reference data on tree species, age, degree of damage, etc., is being used to train and test image classification and for the interpretation of results.
Land Use and Crop Yields. Trolier and Philipson (1986) have shown the feasibility of again using TM scenes, only this time by visual analysis, to determine classes of land use (from 22 different classes) which could have a major effect on the quality and quantity of watershed run-off. This proved quite successful using TM Bands 3, 4 and 5 or a composite of them. Clearly, different land use types will give clues as to the likely degree of pollution to be expected, though it would be wise to seek verifiable data to back up that gained from remotely sensed imagery. Larger urban and industrial areas can readily prove to aggravate water quality as can areas having intensive farming with high crop yields. Also areas which have exploitive activities such as mining, quarrying or forestry are likely to be qualitatively affected. Most land uses which lack human intervention are easily detected and would exhibit better quality water. Figure 4.20 gives examples of land use reflectance values from TM data.
Figure 4.20 Mean Reflectance Values for Several Categories of Land Use
Effluent Discharges and Other Pollutants. RS techniques are practicable for the detection of high volume effluents whose spectral characteristics are markedly different from the surrounding water, e.g. effluents which are highly coloured or at elevated temperatures. Examples of pollutants monitored from satellites include sewage sludge dumped at sea, sewage outfalls, titanium dioxide and oil discharges, spills or blowouts. Figure 4.21 illustrates both high suspended sediment loads and a titanium dioxide effluent plume. Scanners used successfully to detect pollutants include MSS, CZCS, microwave detection by SMMR and SAR, fluorescence detection by LIDAR (laser-radar) and thermal detection by IR scanner.
Suspended and Cohesive Sediment Concentrations. Radiation in the visible part of the spectrum penetrates significantly below the water surface where it is absorbed and scattered. The presence of suspended particles causes much “backscatter”, giving the water a cloudy appearance. The Landsat MSS, the SPOT HRV and the CZCS have all allowed false-colour composite images to be prepared which clearly highlight sediment concentration and plumes (see Figure 4.21). A colour slice enhancement from a single band (of the MSS) has been used to quantitatively map sediment concentrations (Figure 4.22), though reliable calibration requires the collection of water samples at the time of the satellite overpass. Sediment concentrations can be used as natural tracers to identify water movement and frontal boundaries. Cheney and Rabanal (1984) report much of the work done on the measurement of suspended sediment concentrations from satellite data, warning that there is an identity overlap with information on chlorophyll content. Ritchie and Cooper (1988) report the success of the satellite monitoring and estimation, using Landsat MSS data, of sediment concentrations in small inland water bodies.
Salinity. Both Butler et al (1988) and Cordell and Nolte (1988b) report that salinity level measurements from satellite data are not yet recorded in the literature, but that they are certainly feasible with the use of microwave sensors, i.e. changes in salinity cause significant changes in the emissive brightness temperature of water for frequencies less than 5 GHz. The measurement of this brightness should be adequate for mapping the near-surface spread of freshwater at a river mouth or estuary. Although salt wedges at some depth could not be plotted, brightness measurements could be a great help in site selection for many forms of aquaculture, both in saline conditions and at the mariculture/fresh water interface.
Phytoplankton Concentrations. This water qualitative factor is certainly the best documented, possibly because detection is both very important and it provides dramatic large-scale imagery which is not usually visible, e.g. occurrences of significant algal blooms in the northern Adriatic and the Scandinavian Skagerrak area (Figure 4.23). The presence of phytoplankton is detected by their green pigment (chlorophyll) - increasing quantities cause water colour to change from blue to green. Phytoplankton presence can be considered as an index of biological productivity and it can be related to fish production - in moderate amounts it is ideal, but when so-called algal blooms occur the planktonic concentrations are sufficient to reduce dissolved oxygen levels dramatically, often resulting in massive fish kills. The spectral reflectance of algae is quite distinctive in the near infrared range. MSS measurements along with field data sampling, allow for the calibration of algal concentrations, and estimates of algal front movements can be reliably made (Petterson, 1989). Phytoplankton quantitative estimates have been successfully made in lakes as well as in the sea, and several sensors have been used for detection, e.g. MSS, CZCS and an Ocean Colour Experiment (OCE) instrument.
Figure 4.21 Effluent and Sediment Yields in the Humber Estuary in Northern England
© Aquatonics
Ltd
Landsat-1 image of the Humber estuary, UK 27 June 1976
| 1 | Effluent plume from factory producing titanium dioxide | 4 | Quarry |
| 5 | Clay pits | ||
| 2 | Ships at anchor in the outer estuary | 6 | Convenham reservoir |
| 3 | Oil refineries | 7 | A15 (to Humber bridge) under construction |
Figure 4.22 Colour Slice Enhancement Showing Suspended Sediments Which Allows the Quantitative Map to be Compiled
A colour slice enhancement of a single Landsat MSS band (fig 3), not only shows up the suspended sediment distribution patterns, but is capable of quantitative calibration as a map of suspended sediment concentration (fig 4).
Fig 3 Colour-slice enhancement of Landsat 2 MSS band 5–19 Aug 1978
Fig 4 Map of suspended sediment at 10.00 GMT 19 Aug 1978 (2 hrs before high water, Portsmouth) based on fig 3. Arrows show tidal stream.
Figure 4.23 CZCS Composite of Bands 1, 2 and 3 of the Northern Adriatric Showing Chlorophyll Concentrations in Orange
CZCS composite of bands 1,2 and 3 of the Venice lagoon showing chlorophyll concentrations - high amounts in red, low in blue.
A number of sensors, working in the thermal spectral range of 2 to 18 um, are capable of detecting surface water temperatures. The main sensors used to data have been the MSS, CZCS, AVHRR, SMMR and the TM. Sea surface temperature (SST) sensors operate from both geostationary and orbiting satellites. Since data obtained from the GOES geostationary satellite necessarily has a large spatial resolution (8 kms), it is only useful for sea, large lake or oceanic (macro-scale) areas, though on the plus side data can be obtained at 30 minute intervals. For smaller water bodies, the AVHRR instruments aboard the NOAA satellite series is most frequently used. It offers a spatial resolution of about 1.1 kms and a thermal resolution of 0.2 degrees C. Many SST algorithms have been developed by NOAA and tested in a wide variety of situations, and the validity of satellite-derived SST data is now widely recognized for bodies of water of various sizes (Robinson and Ward, 1989). For smaller bodies of water Band 6 of the TM on-board Landsats 4 and 5, has provided satellite thermal imagery at a resolution of 0.12 kms. An example of TM data (Figure 4.24) clearly shows a well developed thermal plume emanating from a power station in the N.E. corner of a lake. The actual or relative temperature gradients can be calculated and the movement of the plume correlates strongly with wind direction. Since SST data has been collected over a long time, NOAA now publish comprehensive real time and long-term water temperature maps covering many areas of the world.
We have reported that there are many countries or regions which have not yet documented or created an inventory of their existing waterways or inland lakes, e.g. Shimang (1989) reports the problem in Nigeria and Vincke (1990) in Malawi. The latter author describes in some detail how the use of Landsat MSS and TM data, helped by extensive ground truthing, can facilitate an inventory of all lakes of over about 1 ha. Adeniyi (1988) reports similar plans for N.W. Nigeria. As well as the actual existence of surface water supplies, RS can give clues to variable water flow rates, e.g. Cheney and Rabanal (1984) report several methods of relevant hydrologic modelling and Blair Rains (1985) reminds us that areas having extensive forest cover removal are liable to have accentuated hydrological characteristics with a greater incidence of flash flooding.
Estimates of the extent of snow cover can help with predicting river water levels following Spring thaws, and marked seasonal variations in both lake and river levels have been monitored, even with the use of the comparatively poor resolution Meteosat instruments, e.g. in Figure 4.25 the purple indicates open water and the green is vegetation. (Finch 1990) shows recent advances in the ability to assess sub-surface water resources using TM images, especially in semi-arid environments where these supplies might be critical to the long-term sustainability of fish production ventures. Since a huge potential exists for fish production in very small ponds, Kapetsky and Caddy (1985) recommend the use of airborne sensors for detection of same, though skilled interpretation of SPOT HRV data will enable many small water bodies to be located.
A large number of production function are concerned with spatial distributions of a “man-made” social and economic environment, e.g. such factors as distance to the various markets, size of markets and the availability of various economic inputs. Since oblique remotely sensed images are analogous to maps, i.e. showing 2D distributions of spatially related objects at different scales, then they can be used in virtually the same way as maps. Once the scale has been determined, RS images can give direct information concerning distances, but they can only give inferred information on market size, input availability and land costs. SPOT HRV data is ideal for mapping, and updating maps, at scales of >1:25 000. Most human developments of importance, e.g. roads and urban areas, can be discriminated and Iisaka and Hegedus (1982) have shown how, using appropriate models, reliable population estimates (inferring market size) can be made from Landsat MSS images. Once the digitizing of existing base maps has been accomplished for any given area, then updating can be continuously undertaken as more recent geometrically corrected imagery is acquired, and as automatic feature recognition techniques are applied and improved.
Figure 4.24 Development of a Thermal Plume in a Lake in North Wales
Figure 4.25 Density-Sliced Meteosat Images of Lake Chad in November and April
Many factors relating to weather and climate can, and are, being obtained and compiled from satellite RS imagery, and this data has been of help in compiling both short and long-term meteorological information. However, this data is of little direct use to prospective fish producers since, for the optimizing of production locations, the enterepreneur must be concerned with not only average long-term conditions but also known climatic extremes. This information would already be in existence for virtually all areas, i.e. at the only realistic scales which are very small. Short-term forecasting could be of value in indicating that a potentially hazardous event was imminent - one for which precautionary measures might be advisable, e.g. exceptionally strong winds or very heavy rain. Models are now being developed where by future runs of particular weather conditions can be reliably inferred from present weather forecasts, e.g. of temperatures, wind directions or air pressure, though for the immediate future these are only likely to be successful for oceanic areas where there is spatially extensive uniformity of physical conditions.
Soil types can be inferred from RS images both directly and indirectly. Since natural vegetation types correlate strongly with soils, then the various vegetation categories will give a clue to the suitability of soils for pond construction. Soil type may also correlate with altitude, various climatic factors and with slope, and certain soil types will relate to areas of intensive crop production. Any indirect RS determination of soils must be carefully calibrated and substantiated by ground truthing. Several studies have shown that Landsat TM imagery is the most suitable for recording the very varied spectral reflectance characteristics of different soils, i.e. spectral signatures will vary with the amount of organic matter, mineral content, iron oxides present, moisture content, surface roughness as well as the percentage balance of sand, silt and clay. Post-soil classification evaluations of the predicted results show that soils can be categorized with almost total accuracy, and the area of each soil type can also be accurately estimated. Some confusion may be apparent between bare soils and areas of very low vegetation cover.
The ability of the SPOT HRV instrument to perform off-nadir viewing means that it is possible to obtain stereo-pairs of images, which in turn can be used for the plotting of contours, i.e. with the use of an Ortoster instrument. 50-metre contours can be plotted at a scale of 1:50 000. This is generally sufficient to determine whether slopes in a particular area would prove to be disadvantageous to aquaculture site location.
The existence of shelter for sea cage culture will be a function of several factors, i.e. strength and direction of prevailing wind, length of fetch and the height of surrounding land. We have seen that relief is now able to be determined - this will give an indication as to potential shelter. The Seasat satellite, which was operational for a short time in 1978, proved the possibility of gaining very accurate measurements of wave height, i.e. with the RA instrument an accuracy to within 10 cms could be obtained, and another instrument on board, the SASS, measured surface wind speeds and direction in all weather conditions. From these instrument readings fairly accurate charts could be drawn up, which gave long-term average or extreme wave or wind conditions. The prospective ERS-1 satellite, will soon be able to continue to provide this data.
Again knowledge of this function is important to cage culture. There are several methods of determining water depth in areas where it is relatively shallow (<20 m). Where water visibility is clear, the Landsat MSS instrument (using Band 4) can detect reflectance from the sea bed. At different depths, a different degree of brightness will be measured and bathymetric contours can then be drawn based on lines of equal brightness. In more turbid areas, then images using Band 4 reflectance captured on the Landsat TM sensor, can be preprocessed and then density-sliced to discriminate between water and sandbanks. The actual height of sub-surface sandbanks can be determined by using multiple images at different tidal states, and, by using images captured over a period of time, sandbank dynamics can be calculated.
There are some other production functions which can be inferred from satellite imagery, e.g. monitoring of sea ice; monitoring of vegetation levels or biomass estimates in inland waters and the existence of coastal flora or habitat types such as mangroves, but space precludes an analysis of these. During the next decade it is anticipated that sensors with higher resolutions will become available, and not only will this improve upon the imagery available but it will also allow satellite platforms to be used to gain data on production functions which presently rely on airborne sensing.
RS is a field of study which has not only generated a vast amount of new and usable data at an exponential rate, but it has spawned information about itself at a similar rate. Cordell and Nolte (1988b), in a study which was aimed at testing the feasibility of using RS to help in aquacultural location, noted - “The quantity of information (RS) to choose from is so large that it will tend to confuse and discourage customers who are unfamiliar with remotely sensed information.” (p3.10). It has been our experience too, that those who are interested in RS will start to delve and very soon they will suffer from a surfeit of information. This being the case we need only briefly suggest where to start.
This section will endeavour to show firstly the range of products which are generally available from commercial companies and other organizations. These products will include both different forms of direct sensor output and various “spin-off” products. We will describe the sources of products, their availability, cost, etc., and finally we will look at the guidance and support facilities which will inevitably be required in any field as complex and internationalized as is RS.
The wide range of RS products will have been created via the enactment of a large number of highly technological processes. Though we have discussed some of the data capture and data imaging processes, space precludes a discussion of those processes necessary to create the many different forms of products available from any one set of digital RS data. A number of satellite systems are producing different forms of output which spawn different products, but since the vast majority of output is presently, and easily, available from two main satellite series, landsat and SPOT, we will concentrate mainly on these. The quality of data output is quite variable and therefore customers may stipulate minimum quality acceptance levels, maximum cloud cover acceptance and preferred times of year for an image. It is possible to obtain “quick-look” microfiche or photographic images before purchase to assess image quality. To order products the user should be familiar with the spatial referencing system appertaining to the satellite series. Referencing can be given by latitude and longitude coordinates, by SPOT reference grid nodes (see Figure 4.26), or for Landsat by a “World Reference System”. Information about products from other sensor platforms can be obtained from some of the sources mentioned in section 4.8.2.
Figure 4.26 Part of the Referencing System Used in the SPOT Series
For SPOT images CCTs can be recorded at either 6250 or 1600 bits per inch (bpi). Given that the total volume of a SPOT scene (which is normally 60 × 60 kms but 60 × 80 kms for the maximum oblique viewing angle) is 27 to 100 Mbytes, then a single 6250 bpi tape can accommodate the entire scene. Two or three 1600 bpi tapes would be needed per scene depending on the scene data volume. Alphanumeric characters are coded by default in ASCII or optionally in EBCDIC, and files are organized according to “Band Interleaved by Line” mode or “Band Sequential” mode. Imagery on CCTs will have been pre-processed to one of six levels depending on customer choice.
CCTs showing Landsat imagery contain similar characteristics to SPOT for data volume and format. However, the levels of pre-processing will be different and more varied since data is derived from three sensors (MSS, RBV and TM) and there are three basic scene sizes available - full scene, quarter scene or a user defined map scene. A single Landsat full scene will cover about 9 times the area of a SPOT scene, though at a reduced resolution.
Both SPOT and Landsat images are available in black and white or colour formats, in film or as prints, and they may be at varying sizes which correspond to the scale of the image (see Tables 4.10 and 4.11). For both Landsat and SPOT special photographic products are available, e.g. for colour composites various specific bands from each Landsat sensor are used by default, but any combination of bands can be requested. SPOT special photographic products include various mosaics, colour composites, linear combinations of spectral bands, panchromatic photographs at 1:25 000 scale, stereoscopic images, “shift along track” or extended frame scenes, digital elevation models and orthophotos. Landsat also provide a similar range of special products. The range of photographic products available from the Kosmos series (via PRIRODA) is shown in Table 4.12.
| Sensor | Nominal Image Size | Film | Paper | Scale |
|---|---|---|---|---|
| RBV Landsat 1,2 and | 18.5cm (7.3ins) | X | X | 1:1 000 000 |
| MSS Landsat 1 to 5. | 37.1cm (14.6ins) | X | 1:500 000 | |
| 74.2cm (29.2ins) | X | 1:250 000 | ||
| RBV Landsat 3. | 18.5cm (7.3ins) | X | X | 1:500 000 |
| 37.1cm (14.6ins) | X | 1:250 000 | ||
| 74.2cm (29.2ins) | X | 1:125 000 | ||
| TM Landsat 4 and 5. | 18.5cm (7.3ins) | X | X | 1:1 000 000 |
| 37.1cm (14.6ins) | X | 1:500 000 | ||
| 74.2cm (29.2ins) | X | 1:250 000 |
| Scene | Nominal Image Size | Film | Paper | Scale | Processing Level |
|---|---|---|---|---|---|
| Whole | 241cm×241cm | X | 1:400 000 | 1A,1B,S1,2,S2 | |
| 300cm×350cm | X | 1:400 000 | 2,S2 | ||
| 482cm×482cm | X | X | 1:200 000 | 1A,1B,S1 | |
| 964cm×964cm | X | 1:100 000 | 1A,1B,S1 | ||
| Quarter | 241cm×241cm | X | 1:200 000 | 1A,1B,S1 | |
| 482cm×482cm | X | X | 1:100 000 | 1A,1B,S1 | |
| 964cm×964cm | X | 1:50 000 | 1A,1B,S1 |
RS products are available via a number of companies and organizations whose status varies from government level through to private commercial companies. With incrasing volumes of data available, with rapidly rising costs in the space industry and with changing government priorities, a distinct trend towards a commercial outlook in the provision of data is obvious. Three major commercial companies are the primary source of all SPOT, Landsat and Kosmos imagery:
SPOT Image. This company, based in France, was formed in 1982 to handle all commercial distribution of SPOT imagery, to promote the system and to generate and process data. They have established a network of distribution points in over 20 countries. All SPOT scenes are archived and then referenced in a computerized data base called the “SPOT Image Catalogue” which can be accessed via telephone, telex and switched-packet international networks at any time. They also offer extensive guidance and support facilities. They have increased their business considerably since 1986, though they are never likely to cover satellite development costs.
EOSAT. The Earth Observation Satellite Corporation was formed in 1985 as a joint venture by the Hughes Aircraft and the R.C.A. Corporations. It aims to promote future satellites in the Landsat and Omnistar series, develop new sensor systems, to promote and market Landsat imagery and to decrease government costs in RS. It presently receives a government subsidy of $19 million annually to cover Landsat 4 and 5 costs. It operates a network of receiving stations around the world and has its base at Lanham in Maryland. It holds over two million Landsat Images in a worldwide accessible digital archive and it also provides extensive user support facilities.
Soyuzkarta. This company was set up by PRIRODA in 1987, both to help achieve economic accountability of the Russian RS industry and to accelerate efforts to increase sales of photographic products on a world-wide basis (Soyuzkarta-Kartex, 1987; Morrison and Bond, 1989). A vast geographic range of imagery is available, though it is all in photographic form and it would therefore need to be digitized before using in a GIS. Over 100 long-term contracts have already been made with “western” countries to supply imagery which uis supplied through a company called Central Trading Systems. There is a desire by PRIRODA to increase contacts with foreign firms to help economically and to acquire better technology and expertise.
These three companies have various agreements with other companies and organizations to supply RS imagery. These include:
| Survey camera and image type | Film positive | Film negative | Film negative in set with film positive | Contact print on paper | Paper print, 2X enlargement | Paper print, 4–5X enlargement |
|---|---|---|---|---|---|---|
| KFA-1000 | ||||||
| Scale and image format | 1:270,000 | 1:270,000 | 1:270,000 | 1:270,000 | 1:130,000 | 1:50,000 |
| 30×30 cm | 30×30 cm | 30×30 cm | 30×30cm | 4 parts, 30×30 cm | 25 parts, 40×40 cm | |
| Black and white | $1,180 | $1,260 | $1,360 | $440 | $660 | $1,150 |
| Color | $1,560 | $1,650 | $1,770 | $550 | $820 | $1,400 |
| MK-4 | ||||||
| Average scalea and image format | 1:900,000 | 1:900,000 | 1:900,000 | 1:900,000 | 1:500,000 | 1:250,000 |
| 18×18cm | 18×18cm | 18×18cm | 18×18cm | 36×36 cm | 4 parts, 40×40 cm | |
| Black and white, 1 band | $1,050 | $1,110 | $1,190 | $210 | $300 | $440 |
| Black and white, 3 bands | $2,520 | $2,660 | $2,860 | $420 | $600 | $880 |
| Color | $1,950 | $2,070 | $2,250 | $700 | $1,020 | $1,600 |
| Color synthesizedb | $3,220 | $3,470 | $3,800 | $750 | $1,070 | $1,600 |
| KATE-200 | ||||||
| Scale and image format | 1:1,350,000 | 1:1,350,000 | 1:1,350,000 | 1:1,350,000 | 1:700,000 | 1:350,000 |
| 18×18 cm | 18×18 cm | 18×18 cm | 18×18 cm | 36×36 cm | 4 parts, 40×40 cm | |
| Black and white, 1 band | $230 | $250 | $280 | $100 | $140 | $220 |
| Black and white, 3 bands | $550 | $580 | $630 | $210 | $310 | $470 |
| Color | $450 | $480 | $510 | $200 | $280 | $420 |
| Color synthesizedb | $690 | $750 | $790 | $235 | $300 | $450 |
aThe image scale varies from 1:650,000 to 1:1,200,000 depending on the altitude of the orbit.
EURIMAGE. This is a consortium of companies, formed in 1986, which have the right to market Landsat data on a commercial basis throughout Europe. They are based in Rome and work under the auspices of the European Space Agency (ESA).
EARTHNET. This is the ESA distribution network, based in Rome, which handles the European data from SEASAT, the Nimbus 7 CZCS, the HCMM, and from the NOAA and Meteosat satellites.
EROS. The Earth Resources Observation Systems Data Center is part of the U.S. Geological Survey, being located at Sioux Falls, South Dakota. It is responsible for archiving, reproducing and distributing the Landsat data held by the U.S.G.S., through current Landsat data is distributed through EOSAT. It therefore has a large selection of older images.
NESDIS. The National Environmental Satellite, Data and Information Service is part of NOAA in the U.S., and it provides a large range of RS products associated with meteorology, the ocean and the environment.
NCIC. The National Cartographic Information Center, in the U.S., also provides a range of map, photographic and satellite imagery.
NRSC. The National Remote Sensing Centre is situated at Farnborough in the U.K., where it is the main source for all satellite imagery. It provides an extensive user support system (see section 4.8.3).
All these distributors have various forms of on-line, microfiche or microfilm data retrieval systems, many of which are accessible at other points or via computer telecommunication links. Computer listings of the data held are usually available. Although some of these companies and organizations purport to supply data obtained from Russian, Japanese or Indian satellite systems, in reality this may be hard to procure.
Spatial coverage of satellite imagery is now very comprehensive. Landsat TM and MSS data covers the whole world except a small exclusion zone over western China and the central Soviet Union. Having said this, there would be wide variations in the number of scenes available for any particular location. Figure 4.27 illustrates the amount of this variation over the comparatively small area of England, Wales and Scotland. The variations are mostly due to cloud cover problems, but they may also be a function of the satellite orbit return frequency and the degree of swath overlap, plus the systems performance of the spacecraft, e.g. systems performance of Landsat was very poor during 1983 and 1984 before Landsat 5 was in full operation.
The time taken to acquire imagery is usually two to three weeks, though by paying a high premium turn-around time can be as short as 48 hours. Prices for satellite imagery vary greatly according to products; an indication of 1989 Landsat prices is given in Table 4.13. SPOT prices are generally higher per km2 of scene, especially for high resolution products and for those having undergone maximum pre-processing. Because of the desire to earn foreign exchange, prices for high resolution Soviet imagery are cheaper than either Landsat or SPOT, even after essential digitizing costs have been met.
It must be remembered that the costs shown in Table 4.13 are just for the data. Hardware and software costs for data manipulation are high and the costs of analyzing the data can also be high, especially if consultancy fees are necessary or if training costs are considered. It is likely that the commercial companies are charging prices which are too high for the research/scientific community, e.g. sufficient Landsat CCT images to cover the entire land area of the earth would be >$50 million. Having said this, if the costs of using RS imagery are calculated per km2, then they may be considered to be very reasonable. In a recent study to test the cost-effectiveness of RS techniques for aquaculture site assessment (Chacon-Torres et al, 1988), for the Lake Patzcuaro area in Mexico, it was found that costs for digital MSS imagery plus necessary ground truthing were about 33% lower than for a conventional ground-based survey, and 38% lower than for a survey which combined aerial photography with ground-based methods. The authors did mention that these figures could be highly variable depending upon numerous extraneous factors.
Figure 4.27 Variations in Landsat Data Acquisition Over the U.K. for 1976–1986.
In the last section we inferred that the major companies and organizations involved in RS offered a degree of guidance and support. Here we will look at this support in more detail and we will outline other avenues where help may be obtained. The range of guidance and support facilities will vary as a reflection of the primary aims of the company or organization. To exemplify this we will examine support given at three levels:
Commercial companies. Since their primary aim must be to maximize financial returns, their range of support will be limited. It will mostly fall under the broad notion of “marketing”. Apart from selling the main CCT and photographic products they produce helpful material which maximizes the likelihood of selling more primary RS data. Examples of this would be:
| TM DIGITAL DECODED: | ||
| Full scene | 185×170km | $4900 |
| Quarter scene | 100×100km | 3600 |
| Map sheet | 55.5×91.2km | 2500 |
| TM DIGITAL MOVABLE SCENE (User defined): | ||
| 100 × 100km | $3200 | |
| 50 × 100km(N-S × E-W) | 2400 | |
| 100 × 50km(N-S × E-W) | 2400 | |
| TM COLOUR PRODUCTS (185 × 170 km): | ||
| Positive transparency | 1:1 000 000 | $1800 |
| Paper | 1:1 000 000 | $1300 |
| Paper | 1:500 000 | 1400 |
| Paper | 1:250 000 | $1500 |
| TM BLACK AND WHITE PRODUCTS (185 × 170 km): | ||
| Positive transparency | 1:1 000 000 | $500 |
| Paper | 1:1 000 000 | 550 |
| Paper | 1:500 000 | 600 |
| Paper | 1:250 000 | 700 |
| MSS DIGIOTAL PRODUCTS: | ||
| 185 × 170km | $1000 | |
| MSS COLOUR COMPOSITES (185 × 170km): | ||
| Positive transparency | 1:1 000 000 | $600 |
| Paper | 1:1 000 000 | 550 |
| Paper | 1:500 000 | 700 |
| Paper | 1:250 000 | 1000 |
| MSS BLACK AND WHITE PRODUCTS (185 × 170km): | ||
| Positive transparency | 1:1 000 000 | $155 |
| Paper | 1:1 000 000 | 95 |
| Paper | 1:500 000 | 190 |
| Paper | 1:250 000 | 290 |
National Remote Sensing Organizations. These will be government organizations whose raison d'être is to promote an “industry” in which much government investment may have been made. A much broader approach to promotion can be made because financial considerations are not paramount. As well as providing the support that commercial companies do, and acting as agents to distribute the products from the commercial companies, they may also:
The FAO Remote Sensing Centre (FAORSC). Operating on a global scale, the organization's objective is to use space technology to help bring about self-reliance of developing countries. They offer many of the support facilities listed for commercial companies and for national organizations, but dealings are usually at a governmental level and a list of the type of support offered reflects this:
World-wide there are now a large number of learned and professional societies which may be related to RS in various ways. These societies propagate information and support by holding conferences and symposia, by issuing special publications, technical papers and conference proceedings and by organizing courses, workshops and regular meetings. World-wide there are also numerous university courses which are either completely devoted to RS or which RS forms a significant part of. For a select bibliography, which covers RS at a general or introductory level, we would recommend: Curran (1985), Lo (1986), Mather (1987), Hyatt (1988) and Drury (1990). Cordell and Nolte (1988a) provides a detailed practical summary, directed towards aquaculture, of how to acquire RS information. Finally, Table 4.14 offers a list of some available journals whose content is mostly devoted to RS.
Inherent in the introduction of any comparatively new, large-scale and complex technology there will be problems-indeed the very fact of making progress is achieved by overcoming problems. So in some senses, it could be argued, we should not be concerned with them.
However, since some authors have argued that RS has failed to live up to its proclaimed potential (Young and Green, 1987), and within the RS community it is freely acknowledged that its capabilities may have been oversold, then it is only reasonable to highlight a few of the major problems. Space again precludes a detailed discussion and some problems have previously been alluded to.
| a) | Arianespace. |
| b) | Atlantic Canada Remote Sensing Newsletter. |
| c) | Aviation Week and Space Technology. |
| d) | EARSEL News. |
| e) | Earth Observation Quarterly. |
| f) | E.S.A. Journal. |
| g) | European Space Report |
| h) | Geocarto International. |
| i) | Geoscience and Remote Sensing. |
| j) | ISPRS Journal of Photogrammetry and Remote Sensing. |
| k) | ITC Journal. |
| l) | International Archives of Photogrammetry and Remote Sensing. |
| m) | Int. Journal of Aerial and Space Imaging, Remote Sensing and Integrated Geographical Systems. |
| n) | Int. Journal of Remote Sensing. |
| o) | Journal of Geophysical Research. |
| p) | Journal of IEEE Geosciences and Remote Sensing. |
| q) | La Lettre du CNES. |
| r) | Mapping Sciences and Remote Sensing. |
| s) | Photogrammetric Engineering and Remote Sensing. |
| t) | Remote Sensing in Canada. |
| u) | Remote Sensing Letters. |
| v) | Remote Sensing of Environment. |
| w) | Remote Sensing Society Newsletter. |
| x) | Space News. |
| y) | Space Policy. |
We have briefly explained that data accessibility is variable spatially, i.e. within a comparatively short distance it may vary by a factor of five or more. This variability is caused mostly by cloud cover and temporal variations in the functioning of sensors. We should add here that certain climatic zones pose particular problems to RS, i.e. especially the humid tropics where cloud cover is frequently high, plus west-facing coasts in the mid to high latitudes where onshore prevailing winds cause high cloud cover, and those areas which have seasonal rainfall due to movements in the inter-tropical convergence zone. Some areas may lack data for other reasons, e.g. politically sensitive areas, areas beyond the receiving range of some satellite transmitters and high latitude areas.
We have seen that it is necessary to verify the imagery, i.e. by matching up various known classes of land cover with pixel values displayed on images for the given area. This is necessary since pixel values for any specific point will vary from day to day according to a number of extraneous factors. This means that RS use has to constantly consider this inconvenience in terms of cost, problems of accessibility and the necessity of ground truthing at approximately the same time as the satellite overpass.
It is very clear that RS produces vast amounts of digital data and that whatever form this is used in expensive hardware will be required to process it. The more that the use of the data can be maximized, the more expensive will be the necessary hardware. The spatial areas and the types of programmes where RS techniques are likely to prove most valuable are also those which may be at a physical distance from any suitable hardware. Although data is beginning to be available in cheaper forms, e.g. floppy disks, and hardware will undoubtedly proliferate and fall in price, cost and accessibility will continue to make many potential RS using programmes non-viable.
For many purposes airborne sensing still offers the highest resolution and the lowest price, and until resolutions get down to about the five meter size then RS imagery will continue to have its limitations. Although the technology exists to secure better resolution levels, there are political, cost and legal barriers to be overcome before it is generally available. Though the present SPOT resolution of 10 meters is valuable, there are many facets of the human landscape, especially in developing countries, which are not easily determined or cannot be detected at all. Though Wright (1988) has noted, with a degree of truth, that the present crude pixel size could be an advantage given that the data volume problem would be greatly exacerbated with a smaller pixel size, RS is likely to have far greater user acceptance and project applicability with an improvement in resolution.
The increasing trend to commercialize RS output throws up a number of quite significant problems. Clearly products sold will need to have very stringent copyright regulations which themselves will need careful interpretation, e.g. who has the right to purchase individual scenes and can anyone else then purchase exactly the same scene? Agreements which must be carefully set up on a world-wide basis to operate product outlets or agencies, and this has led to a new and complex field of “space law”. Allied to these commercial problems are those of where to fix prices-there is the divergence in the need to recover large costs yet the need to sell data, often to organizations in the less developed world who can ill afford expensive imagery.
Most potential users of RS require assured access to imagery over a long temporal period, even if it is simply to justify the necessary hardware and software expenses for one project. A major problem that the industry faces is, what has been called elsewhere, “planning blight”. Virtually all satellite launches have suffered from uncertainty in scheduling and payloads, with most programmes being significantly behind schedule. For example, ERS-1 will have suffered at least a three-year launch delay, and Landsat 6 was supposed to have carried a SEA-WIFS sensor but this has been abandoned. Delays have been caused by a lack of funding, technological problems, major disasters such as Challenger, etc. It is difficult to sell a technology which has such an unreliable track record, and any short or long-term planning is at the best vague.
Whilst there are many other minor problems, e.g. delays in data acquisition, lack of trained personnel, etc., the application of RS to aquaculture and inland fisheries does not in fact suffer problems as much as the application would to many other enterprises or fields. This is because, except for some of the water qualitative parameters, production functions controlling fish production do not require real-time or regular repeats of imagery, and given that RS can only really help location optimizing at the macro scale, then resolution is also not a problem.
We have no intention here of cataloguing future RS satellite missions - for those interested we recommend Voute (1986) and Petterson (1989). Instead we wish to mention a few observations which might be particularly pertinent to readers of this study.
Though there are many problems and uncertainties in the RS domain, we are convinced there is a very healthy future for the “industry”. This belief is based on a number of factors:
Something of this growth scenario is encapsulated in Figure 4.28.
Figure 4.28 Future Developements in Computer Power, Data Acquisition and Earth Observation Sensors (from NASA, 1988)
Some of the important trends which we see occurring are:
New types of sensor platforms are being planned, e.g. Omnistar is being developed by EOSAT. This is a flexible platform which is designed to be serviceable from the space shuttle. It therefore allows for the “in-orbiting” update of new sensors, for a variety of sensor systems to be carried and it will ensure the RS user a continuity of data flow.
Whereas previous experimental work was carried out using normal polar orbiting satellites, this is now more likely to take place aboard the earth shuttle, aboard manned space stations or even in ground-based facilities.
There are a number of new countries entering the satellite RS field in the immediate future, e.g. Canada, Brazil, China, Indonesia/Netherlands. This is likely to lead to a greater access to data since these countries will wish to make a return on their expensive investment. Much of this data is likely to concentrate on areas where aquaculture and inland fisheries has a huge growth potential.
As an increasing amount of data becomes accessible it creates the ability to monitor temporal changes via time-series modelling. This might be important to production functions which relate to water quality and land use.
By the mid-1990s the cloud cover problem will be substantially overcome because three forthcoming satellite series are planned which carry SAR sensors, i.e. operating in the microwave portion of the spectrum, thus allowing for weather independent high resolution imagery. This will greatly benefit the humid tropical areas where aquaculture and inland fisheries has the greatest potential future.
There is likely to be closer cooperation between countries to launch joint satellite ventures. This will be done for cost cutting and political motives.
There are likely to be more user-oriented imagery capture systems, especially operated by the commercial companies such as SPOT Image or EOSAT. This allows the satellites to be programmed to individual user requirements.
There is certain to be a lot of attention given to the problem of: “When is a satellite commercially viable?” Unless this problem can be satisfactorily solved then future pricing, and the actual supply of products, will remain uncertain.
To considerably broaden the potential use of RS data, there will be an urgent need to make image processing software more available and affordable for microcomputers.
