Referring back to Figure 1.3 (p.6) it can be seen that we are now at the stage where the data streams, in their various forms, converge into the “box” labelled GIS. Entering this box are “flows” of data - both directly relevant and proxy, as well as the three map “flows” - it is these streams which collectively combine to form the inputs to GIS. In this chapter we intend to show how GISs have evolved, what the major processes are that make up the system, and then to examine something of the technology necessary for making the system function. We also intend to look at benefits and problems of GIS and then to give some guidance on system selection, including the support facilities available. We will conclude by attempting, in this complex and fast changing field, to examine some likely future trends in GIS. A varied selection of functional GISs, as applied to aquaculture, inland fisheries or related topics, are exemplified in Chapter 7.
GIS is a branch of science, or a field of learning, which has evolved, and is very much still evolving, at such a rapid pace that a definition of what it is or what it does, has changed and expanded so much that the only thing that we can be certain of is that any definition we give now would not describe what is being done in perhaps 5 or 10 years time! This rapid evolution, which is described in more detail in section 6.3, has meant that there is much controversy over not only a definition of GIS, but also where GIS lies in a hierarchy of similar fields and on what basis a typology of GIS should be determined.
Though it would appear that the nomenclature of “Geographic(al) Information System(s)” is coming to the fore and is becoming universally accepted as pivotal or central to those processes which we describe in this chapter (Clarke, 1986), there is still a body of opinion which considers GIS to be a “narrow term”, only one strand of several systems which, although similar, should retain their separate identities (Shand and Moore, 1989). Other names synonymous with GIS include:
“Geog-based information systems”,
“Natural resource information systems”,
“Spatial information systems”,
“Geographic data systems”,
“Digital mapping systems”,
“Land information systems”(LIS).
It is likely that most of these names will give way in favour of GIS, though LIS is likely to hold its ground for a while, along with other associated or specific application areas such as “Computer Aided Drawing” (CAD) and “Automated Mapping/Facilities Management” (AM/FM).
Actual definitions of GIS will be variable and range from the very simple: “A computer system capable of holding and using data describing places on the Earth's surface”, through the rather limited: “A GIS then is a software package,.....” (Butler, 1988, p.31) and through the novel: “GIS are simultaneously the telescope, the microscope, the computer and the xerox machine of regional analysis and synthesis.” (Abler, 1988, p.137), eventually to extremely “wordy” definitions. We would suggest that an actual definition is not as important as the basic ideas which GISs convey, e.g. the essence of GIS must involve:
That being “geographical” it contains data and concepts which are concerned with spatial distributions.
That “information” implies some notion of conveying data, ideas or analyses, usually as an aid to decision-making.
That being a “system” it involves the sequence of inputs, processes and outputs.
That the three strands mentioned above are given functionality within a recent technological scenario based on “Hi-tech” capabilities.
In very practical terms GIS comprise a collection of integrated computer hardware and software which together are used for inputting, storing, manipulating and presenting geographical data (Figure 6.1). The data may be in any textual, map or numeric form which are capable of being integrated within a single system. GIS exist in a variety of forms and embody the potential for an enormous range of applications. No single typology for GISs has yet emerged and clearly a number of categorizations are possible. For those interested we recommend Clarke (1986) and Bracken and Webster (1989).
Figure 6.1 Systems Diagram to IIIustrate GIS
The rapid evolution of GIS, especially over the last decade, has been caused by a complex amalgam of major factors, plus a number of minor ones. Here we identify the major factors before briefly examining the historical sequence of GIS development. For those interested further in these areas, details are given in Burrough (1986), Jackson and Mason (1986), Dept. of Environment (1987), Smith et al (1987), Crosswell and Clark (1988), Goodchild (1988), Tomlinson (1989) and Star and Estes (1990).
Over the last two decades there has been a surge in data volume, much of which has been available in digital format, e.g. from RS sources, from censuses and from the major mapping agencies. This surge was in response to the perceived need to have banks of information, in an easily manipulated form, so as to maximize the use of expensively procured data. Much of this data has been accessible using various on-line facilities associated with computer networking and communications.
Technological advances in computer hardware have enabled a dramatic increase in the performance/cost ratio of computer processors. Dept. of Environment (1987) estimates that processing costs have fallen by a factor of 100 in the past decade, and that this is likely to continue. Figure 6.2 illustrates how processor performance has also increased in terms of speed obtained relative to investment made. Performance increases are now tending to blur the traditional distinction between the hierarchy of “mainframe”, “minicomputer” and “microcomputer” - a minicomputer does today what only a mainframe could do five years ago. Actual hardware size reduction has allowed for significant office space saving costs.
Figure 6.2 Improvements in Processor Performance Over Time (from Croswell and Clark, 1988)
Advancing on the tide of the explosion in computing power and capability have been a number of parallel developments. These include: computer aided design (CAD), remote sensing (RS), spatial analysis, digital cartography, surveying and geodesy, etc. All these fields have a spatial perspective and can be inter-related, though other fields such as IT, image processing, computer graphics and photogrammetry have also contributed. GIS has emerged as a core methodology allowing for integration to occur if desirable, or allowing for each of the separate fields to greatly enhance their own efficiency. Thus GIS is “…the result of linking parallel developments in many separate spatial data processing disciplines.” (Burrough, 1986. p.6), and it allows for some considerable developmental effort and costs to cease being duplicated.
Paper maps have traditionally formed the basis of spatial enquiry and these were needed at a large range of scales. Paper maps occupy much space, are easily damaged, they date quickly, they are expensive to produce and data cannot be rapidly extracted from them. The inception of GIS has changed much of this. Both private and governmental organizations have quickly realized the tremendous social, environmental and commercial value of GIS for a range of applications - the main fields are in market location analysis, property management, social resource allocation, resource exploitation, inventory location and environmental analyses.
A consequence of this application and commercialization of GIS has been a proliferation in research and development. GIS has allowed decision makers, in all organizations, to explore a range of possibilities or scenarios before large investments are made or before plans and actions are implemented.
Though there have been claims for very early GISs, e.g. the British Domesday Book of the late 11th century, GIS as we recognize it had its origins in the Canadian GIS of 1964. This embodied the early recognition of what might be possible in terms of using computers for handling numerical data and with out-putting useful and timely information. GIS development was limited in the 1960s and early 1970s because of cost and technical limitations, though during this period the development of the minicomputer was important as was the creation of some original mapping packages, e.g. SYMAP, GRID, IMERID and GEOMAP.
During the 1970s there was a rapid rise in the related, parallel fields (section 22.214.171.124). Advantages were seen in linking data sets, utilizing spatial data in more ways and GIS associated equipment was beginning to be acquired by universities, research organizations and small private companies. By the late 1970s computer mapping had made rapid advances. There were hundreds of computer systems for many applications. Interactive capability was achieved and there were great advances in output devices capable of generating high resolution displays and hard copy graphic products. There were also many computer literate students emerging from universities.
In the 1980s GIS had really taken off, especially during the latter part of the decade, and it is now a growth industry of major proportions. We list some of the developments which have occurred recently - further details will occur throughout this chapter:
Advances in interactive query and graphics processing.
Proliferation of low-cost software, much of it suitable for personal computers (PCs). Improved instructions, menus, manuals, etc. have made GIS accessible to non-GIS specialists.
Distributed computing via networks for the sharing of resources and data.
The availability of “off-the-shelf” or “turn-key” GIS systems, sometimes comprising complete hardware and software packages.
Significant microprocessor developments have allowed for cost reductions and for huge amounts of memory concentration into very small “chips”.
A trend from the use of, or digitizing of, specific maps towards having archives of digitized data in a cartographic data bank which can be manipulated, analyzed and displayed in any desirable form.
A proliferation in the support side of GIS - journals, courses, education, symposia, etc.
Governments, utilities and other enterprises seeking increased efficiency in data handling.
During the whole recent developmental period there has been a “leap-frogging” of developments within specific areas of GIS in terms of them being applications-driven or technology-driven. Most of the developments have been occurring in North America though some have come from Europe. In most countries the government has played a large part in GIS progress since it has been the generator of large volumes of data, since it created needs in departments such as forestry, land use planning and natural resource development and since it is being increasingly called upon to take a leading role with environmental concerns. The global market for GIS systems and data is currently (1990) estimated at $4 billion, and is growing at 20% per annum (Tomlinson, 1989), and Figure 6.3 exemplifies a breakdown of the likely U.K. GIS market till 1999 (Rowley and Gilbert, 1989).
Figure 6.3 The Breakdown of the U.K. GIS Market Till 1999 (from Shand and Moore, 1989)
Figure 6.1 showed the overall functioning of GISs in a simplified form. In this section we describe the elements displayed within the GIS “box”, and show how they are integrated for the successful functioning of the system. There are a great number of functions which a GIS might be required to perform and the most important of these are set out in Table 6.1. The list is compiled from a variety of sources and the interested reader should consult the following: Knapp and Rider (1979), Rhind (1981), Dangermond (1983), Burrough (1986), Smith et al (1987) and Rhind and Green (1988). Throughout this section we will briefly discuss those peripheral hardware items which are directly related to GIS. Space prohibits a review of general computing hardware, e.g. processors, disk drives, alphanumeric terminals, tape drives, V.D.Us and other monitors, even though they may be essential to GIS. In section 6.7 we do look at ways of optimizing hardware systems configurations. Further details on GIS hardware can be obtained from Letcher (1985), Megarry (1985), Walsh (1985), Croswell and Clark (1988), Kadmon (1988) and Dangermond and Morehouse (1989).
|A)||DATA INPUT AND ENCODING|
|i)||Data capture, e.g. digitizing and integration of external data.|
|ii)||Data validation and editing, e.g. checking and correction.|
|iii)||Data structuring and storage, e.g. construction of different kinds of surfaces and data coding.|
|i)||Structure conversion, e.g. conversion from vector to raster.|
|ii)||Geometric conversion, e.g. map registration, scale changes, various transformations, map projection change.|
|iii)||Generalization and classification, e.g. reclassifying data, aggregating or disaggregating data.|
|iv)||Integration, e.g. combining layers of different surfaces.|
|v)||Enhancement, e.g. image edge enhancement.|
|vi)||Abstraction, e.g. calculations of area centroids and Thiessen polygons.|
|i)||Selective retrieval of information based on user-defined themes or criteria, including “browse” facilities.|
|i)||Spatial analysis, e.g. route allocation, slope and aspect calculations.|
|ii)||Statistical analysis, e.g. histograms, frequency analysis, measures of dispersion.|
|iii)||Measurement, e.g. line length, area and volume calculations, distance and directions.|
|i)||Graphical display, e.g. maps, graphs.|
|ii)||Textual display, e.g. report writing, production of tables.|
|F)||DATA BASE MANAGEMENT|
|i)||Support and monitoring of multi-user access to the data base.|
|ii)||Coping with system failures.|
|iii)||Communication linkages with other systems.|
|iv)||Up-dating of data bases.|
|v)||Organization of the data base for efficient storage and retrieval.|
|vi)||Maintenance of data base security and integrity.|
|vii)||Providing a “data-independent” view of the data base.|
All data being input to a GIS must be in digital format, in either numeric or alphanumeric form. Data may be input via a variety of mediums including computer compatible tapes (CCTs), floppy disks, Compact Disc-Read Only Memory optical discs (CR-ROMs), etc. It is obvious that if the data is originating from multifarious sources, then each data set may differ structurally. Given that “The creation of a clean, digital database is a most important and complex task upon which the usefulness of the GIS depends.” (Burrough, 1986. p.57), then, from a GIS viewpoint, it would be ideal and far more efficient if data could be:
In the absence of this ideal, GIS systems software must either ignore these factors, compensate for them or make do with sub-optimal input levels.
The input sources from where the data for GIS may originate have been discussed in Chapters 2 to 5 and are shown in Figure 6.1. In this section we need not discuss inputs from digital archives or inputs from other GISs, since these will both be already available (captured), frequently in a format suitable for immediate use. The choice of other capture methods will be governed largely by the available budget and the type of data being input. The methods of capture for data sources are:
A) For Tabular and Field Surveys
Capture may be by using manual methods, aided by a keyboard and VDU, to interactively create data bases or files, i.e. for entering the results of field work or questionnaire surveys. Data might be filed in either a standard spreadsheet package or in any of the many specialist data entry modules associated with particular computer packages or programs, e.g. statistical packages such as SPSS or Minitab, computer cartography packages such as MICROMAP or survey analysis packages such as SNAP. These programs or packages require that data is entered in a structured format - the data can then be edited and corrected and any numerical manipulations can be performed.
Data loggers can be used. These are specialist devices that automate the process of collecting and recording data in the field. They may be automatic or semiautomatic. They carry out a limited range of functions recording data on variables such as soil moisture content, water flow, sediment particle size, climatic variables, etc. For analogue data loggers, the data will need to be digitized. Specialist data entry devices are available which semi-automate the field collection of live questionnaire data, i.e. answers are fed directly into a pre-programmed memory within a battery operated, hand held portable terminal. There are a variety of other microcomputers which are now becoming available as a result of microprocessor advances and subsequent price reductions. Maguire (1989) provides further useful information on many field data capture methods.
B) For Map Data
Maps may be captured by the use of digitizers or various types of scanners, e.g.:
Electromechanical digitizing involves using a tiltable table, or tablet, on which the map is positioned (Figure 6.4), with an in-built Cartesian surface (grid), having energized intersections typically resolving to 0.01mm. Attached to the table is a pen or tracking cross (cursor or puck) which can be moved along lines or to points, and can detect the signal at any intersection of the grid. Cursors can be equipped with up to 16 buttons which are used for additional program control, e.g. to move from point to line or for adding identifier labels. The analogue signal detected is coded by the computer into a digital x,y co-ordinate, measured from a user-defined origin. These digitizers may work in either point or line/stream mode. In the former points are recorded at a signal from the operator; in the latter mode the digitizer records co-ordinates at fixed time or distance intervals. Some digitizers can operate in all three modes. Digitizers are increasingly linked directly to VDUs for monitoring purposes, and/or linked to the host computer for direct input of data and to allow the computer to set up the digitizing operating parameters.
Figure 6.4 An Electro-Mechanical Digitizer plus Various Types of Line-Following Cursors (from Kadmon, 1988)
In flat bed optical scanning the scanning head moves along both the x and y axis (Figure 6.5 (a)). Drum scanning involves using a scanning densitometer. A map is fixed to a rotating drum and a scanning photosensitive cell moves backwards and forwards along the x axis capturing, in digitized form, the image of the map as a matrix of pixels (Figure 6.5 (b)). Pixel values vary with differences in light intensity. The step size between each row controls the cell or pixel size in both scanners - this is very small in modern scanners. Scanners are now available for vector (line) scanning - here a laser beam automatically follows lines until junctions are reached or until the user selects alternative lines. With this scanner great accuracy can be achieved but it is user/time intensive. The problem of automatically scanning colour maps is overcome by scanning the colour print masters which are usually black and white thematic lithographs. Scanners have a typical resolution of up to 0.012mm and can collect data at the rate of 30 000 pixels per second.
Figure 6.5 Flat Bed and Drum Optical Scanners (from Burrough, 1986)
In electronic videodigitizing a video camera captures images as collections of up to 512 x 512 pixels each containing a level of light intensity. They are useful for capturing line boundaries from RS or aerial images and entering them into image analysis systems. This is a cheaper form of map capture than scanning, but the resolution is poor.
Data capture from maps is usually by layers, with each layer representing a different theme (Figure 6.6). Capture in this form permits easy updating and specific spatial analyses. Digitizers can input grid-based or vector-based data (section 126.96.36.199) plus textual map labels or special symbols. Most digitizing is carried out off-line using a separate microcomputer or PC for job control or data storage. Attribute data is usually entered from a keyboard. Common problems associated with digitizing are:
It can be very time consuming if done accurately, e.g. it takes about 30 person hours to digitize the boundaries of a typical 60 x 40cm soil map at 1:50 000. Digitizing costs can form a very high proportion of GIS costings, and scanners are an expensive alternative.
Huge data files can be created from single map sheets.
Automatic scanners are frequently unreliable where several lines are in close proximity, e.g. contours in mountainous areas.
Manual digitizer operators can only work for approximately four hours daily if accuracy is to be maintained.
Since most new maps series are now produced using digitizing methods, some of the problems associated with digitizing will ease once the pressures to digitize existing sheets have been dealt with.
Figure 6.6 Representation of the Layers Captured by RS or Map Digitizing (from Goodenough, 1988)
C) For the Integration of Remote Sensing into GIS
Space precludes much discussion of the procedures necessary for the integration of RS imagery into GIS - Jensen (1986) and Goodenough (1988) provide more detail on this. Here we will mostly concentrate on some precautionary advice.
We have shown, in section 4.6, that RS imagery is preprocessed to a variety of user-defined levels, and then filed on CCTs. This data can then be integrated into the GIS, perhaps via external packages such as ERDAS or GEMS, which have been used to perform additional processing. Some GIS software contains its own RS processing programs. The most important task in integration is ensuring that RS-derived data is referenced to exact ground co-ordinates so that registration with other GIS data is possible.
Before integrating the RS data (being held on a CCT) it is important to know the levels of pre-processing which may have been performed. If only crude radiometric and geometric corrections have been done then any of the further pre-processing levels described in section 4.6 might be necessary. Actual integration may give rise to a number of problems. The RS data may only be classified into 256 class levels, whereas GIS are capable of handling far larger arrays. This may make it difficult to assign detectable RS image features to classes in the GIS. There is inevitably difficulty in matching RS images to other thematic data which has been derived from topographic map sources or elsewhere - this is especially true in areas having varied relief. Goodenough (1988) found that it was not uncommon to have displacements of 200 meters at the 1:50 000 scale. Other problems include differences in land area shapes and sizes as well as the image interpretation problems described in Chapter 4.
Some authors argue that there has been very little success in integrating RS into GIS. Young and Green (1987) say generally that this is because of differences between the potential and the operational realization of this potential, and more specifically Wilkinson and Fisher (1987) note that too much RS data is available at a resolution which is not reliable for realistic GIS. Robinson Barker (1988) puts the lack of success down to a cooling off from an initial period of great interest in the mid-1970s and to government indecision and inaction. There was also “too much technology and data” and very sophisticated techniques were needed so that even now low cost interactive RS data does not exist. Finally, RS data is in gridded (raster) format whilst the majority of GIS work in vector format. It is important to bring attention to these limitations, not only to warn the potential user, but also to show that there is still a huge amount of research necessary to ensure reliable integration, and it could well be that the future of RS depends upon its ability to integrate successfully with GIS.
A bird's eye view of the world, as depicted on a mapped surface, reveals that the surface consists of either points, lines or 2D areas which are cartographically called polygons. Thus in Figure 6.7 (a) roads would be lines, houses are usually points and gardens or fields are polygons. All information captured by any method shown in section 188.8.131.52 must be capable of being displayed, and therefore must be appropriately encoded to show any of these three forms. There are two basic organizational modes which the computer may work in to display spatial forms, i.e. vector or raster mode (respectively (b) and (c) in Figure 6.7).
Vector Mode. Here a map will consist of points whose positions are defined by geographic x,y co-ordinates. A point may be defined in isolation to represent a single relatively small mapped object such as a telephone box, a building or a settlement, i.e. depending on the scale of the map. A series of points can be defined and joined to show a line - this might represent a field boundary, road, river, etc. Lines too can be defined and joined so as to enclose an area (a polygon) which might represent any 2D feature such as a field, lake or a country or any thematic area such as an individual soil zone.
Digitizing in the vector mode can be extremely accurate, e.g. in representing non-straight lines additional accuracy is obtained by registering (digitizing) a larger number of points around a curve. The vector mode is usually employed where it is necessary to integrate manual and computer graphics techniques and where annotations are frequently required. Because vector modes use quite complex data structures, the technology is expensive as is display and plotting, particularly for high quality colour products.
Raster Mode. Here the whole mapped surface is composed of a grid of cells which form a matrix of rows and columns. The size of each cell determines the resolution (or detail) of the mapped surface. Very small cells are referred to as pixels (as in RS imaging), and each cell or pixel is a data element showing, by digital encoding or by colour coding of the final map, the occurrence of different features at different cell locations.
Raster graphics are usually used where it is necessary to integrate topographical and thematic map data, either together or with RS data. The main problem of this mode is that the use of cells means that recognizable structures can be lost and there can be a serious loss of information. However, each of the two modes will have several advantages over the other (Table 6.2) and this means that they are best seen as complementary rather than competitive. Though in some ways the issue of structure mode is critical, because once established it is difficult to change it, GIS are increasingly able to handle data in both vector and raster structure since conversion programmes are now available.
Figure 6.7 Modes of Organizing Mapped Data (from Maguire, 1989)
|i)||Features can be accurately located.|
|ii)||Very small features can be shown.|
|iii)||The correct shape of features can be shown.|
|iv)||Total data storage requirements can be far less.|
|v)||Data on individual resources is easily retrieved, updated or annotated.|
|i)||Comparisons of different themes (layers) are easy to make.|
|ii)||Quantification and aerial measurements are easy.|
|iii)||Integration of RS, or other grid based imagery, can be efficient.|
|iv)||The data structure is very simple.|
|v)||The technology is cheap and it therefore gains a larger market.|
|vi)||Processing algorithms are simple and easy to write.|
|vii)||Grided data is more compatible with raster based output devices.|
For complete data capture it will be insufficient to simply digitize, in vector or raster format, the graphical data shown on a map or a similar type of data recorded from another source. All spatial data should also have specified:
Data on each of these need to be digitized in such a format that they can be efficiently attached (linked) to graphical mapped entities.
The actual location reference will be automatically registered via the Cartesian wire grid embedded in the digitizing table or tablet. Data about attributes such as the width of a river, its water quality, the variability of flow, etc., needs to be stored apart from the spatial data, usually in a tabular database. This is done by giving each type of data a common identifier which can be efficiently linked as required, i.e. by using a special computer programme into which the x,y co-ordinates have also been stored. Figure 6.8 shows how topological referencing is rather similarly achieved. Thus a typical polygon map is abstracted into seven nodes (points), eleven links (lines) and five polygons. By the use of the unique identifiers, the spatial relationship between these three possible forms used on maps can readily be tabulated as shown, and by adding x,y co-ordinates at each of the points a dual system of spatially identifying map elements is achieved. Topology is implicit in raster modes because all locations are simply defined by rows and columns.
Figure 6.8 The Creation of a Topological Data Structure (after Dangermond, 1983)
Figure 6.9 summarizes the steps needed to digitize a set of boundaries, and their non-spatial attributes, and link them together to form a topologically linked data base of individual polygons. The particular steps might vary according to the GIS being used, volume of data, hardware available, etc. Once the linking of spatial and non-spatial data has occurred, then verification and editing is easily performed.
Figure 6.9 Steps in Creating a Topologically Correct Vector Polygon Data base (from Burrough, 1986)
It has been intimated that data capture processes are likely to generate huge quantities of digital data. If the computer memory is large enough this might not provide handling problems, but for any smaller microprocessor then greater efficiency can be achieved by selectively storing (compressing) only relevant bits of the data. Space prohibits a detailed discussion of the many types of storage structures - those interested should consult Samet (1984), Jackson and Mason (1986) or Green and Rhind (1986). We briefly illustrate two simple methods:
Run length coding. This relies on the fact that for many kinds of thematic map, adjacent cells could have the same value. Consequently, it is only necessary to enter the data for each row (run), specifying a cell value and the column numbers where that value begins and ends. For example, Figure 6.10 shows how a simple vector map (a) is converted into a raster cell matrix (b), having encoded digital scores, and (c) shows its final conversion to a run length code. Each “run” is shown by a triplet of numbers indicating starting column, end column and value. In this example a digital saving of nearly 40% has been achieved; further savings could have been made if all “empty” areas were assumed to be zero.
Figure 6.10 Example of a Run Length Code Structure (from Burrough, 1986)
Quadtrees. This is a data structure based on the regular decomposition of a region into quadrants and sub-quadrants. Eventually the whole of a specified functional category within a gridded area will be delineated by cells of varying sizes, i.e. down to the smallest pixel for which data has been gathered (Figure 6.11). The relationship between the raster map, with its sub-divided cells, can be shown as links and nodes in a “directed tree”, with each link being directed from the “parent” node towards its “children” (sub-quadrants). From each node in the tree there are four “edges” representing the four quadrants of N.W., N.E., S.W. and S.E. Sub-quadrants only emanate from nodes which are shown as being sub-divided in the raster map. From the example shown it can be seen that nearly a 50% saving in data capture has been made.
Figure 6.11 A Raster Mode Map and its Quadtree Representation (from Jackson and Mason, 1986)
Once data has been captured, the data base files reside in CPU memory (RAM) and are thus immediately accessible for computation and manipulation. Because of the large amount of data in the files, they are usually stored on fast hard disk drives. Sometimes data may be stored off-line on magnetic tape or other media, e.g. floppy disks. To extract meaningful information from the data base, it will need to be queried, organized and perhaps changed in any number of ways to suit individual requirements, i.e. manipulated so that it can be retrieved in a useful format for later analysis. The distinctions between data capture, manipulation and data analysis are not hard and fast - they are all functions performed within the GIS and there is a considerable degree of overlap. The main data manipulative functions will be briefly examined, i.e. those that most GIS software will be programmed to perform - further information can be obtained from Berry and Tomlin (1981), Dangermond (1983) and Burrough (1986) and selected case studies will be provided in Chapter 7.
These functions are usually performed on the initial data soon after capture, through they might be necessary at any stage. All captured data must be interactively reviewed for errors, since these are common, especially in digitizing. Typical digitizing errors are:
“Double digitizing” - lines are digitized twice which introduces “slivers” into the data if the lines do not exactly correspond.
“Undershoots” or “overshoots”, where lines are either too long or too short.
Lines, line segments (arcs) or points may be entirely missing.
All errors require correction or editing. Some GIS software contains programmes for verifying the correctness of all geometric, topological and attribute data. This might involve making certain that all graphical data is suitably defined; that spatial attributes do not exceed expected ranges or values; that nonsensical combinations of attributes do not occur, etc. The importance of error correction cannot be overstated. Burrough (1986) devotes 30 pages to a detailed analysis of error propagation and Goodchild and Gopal (1989) write exclusively about the needs for spatial accuracy. If data is not scrupulously verified, and errors remain captured, then further manipulation of the data will lead to error magnification and the subsequent invalidity of the data.
We have previously mentioned (section 184.108.40.206) that data can be structurally converted so as to require less storage. The other main structural conversion which may be necessary is raster to vector conversion or vice versa. Figure 6.12 gives a simple illustration of what is involved. Note that in the vector to raster conversion a loss of accuracy is incurred, which can be magnified as a function of both cell size and “wiggliness” of the lines. In the raster to vector conversion, the computer performs a vectorizing process which consists of “threading” a line through a swarm of pixels, using a “thinning” algorithm (because the swarms of pixels are thinned to a line).
This heading covers several forms of manipulation all aimed at geometrically changing the mapped surface so that maps can be adjusted to one another, or to a reference system (Figure 6.13). Different algorithms will need adopting to facilitate the necessary changes.
Transformations from one co-ordinate system to another, or from one map projection to another.
Scale changes can be adopted by selecting a simple multiplier function.
Maps can be rotated to particular orientations.
Spatial data may suffer from various distortions, e.g. aerial photographs are not always scale correct because of aircraft tilt, relief differences and differences caused by the angle of view, and paper maps will suffer from paper stretch. These distortions (or “image warps”) can be manipulated by “rubber sheeting”, i.e. treating the distorted image as an elastic sheet which may be stretched or compressed until it exactly fits an accurate base map.
Figure 6.12 Summary of a Vector to Raster - Raster to Vector Conversion (from Robinson Barker, 1988)
Figure 6.13 Some Fundamental Geometric Manipulations of GIS Database Files (after Dangermond, 1983)
Map merging involves overlaying different data sets (layers), and then performing arithmetical calculations, usually on values assigned to polygons, in order to automatically create new mapped surfaces. Any number of layers can be progressively added, or subtracted, and both vector and raster based surfaces can be manipulated. Existing data on a single theme can be merged, e.g. to combine two classes of water quality in order to create a third class, or varied themes can be merged in order to build up a “variable scored surface” for any topic over any area.
Integration of tabular or statistical (numeric) data within the GIS, either with other such data or with maps, will depend on finding objects within the existing database that can be tied to geographic locations (x,y co-ordinates) and are uniquely identified. Integration can be complex since considerations must be made about data structures and formats, variable length records, plus ways to identify, access and manage identified objects.
These manipulations require that data is advantageously changed in some way. A large number of functions can be performed including:
Adding to, deleting or changing any data in files, i.e. similar to editing.
Aggregating or disaggregating numerical or attribute data.
Sorting or reclassifying data into user defined classifications.
Generalization and smoothing, i.e. these processes involve data reduction algorithms to change data structures or to remove excess co-ordinates in digitized lines.
Dissolving, i.e. where a new polygon map has been generated by overlaying, then adjacent polygons may retain the same value or characteristics. Lines can be deleted or “dissolved” to simplify the mapped surface.
Geo-coding, i.e. assigning attributes to spatial points, lines or polygons.
Addressing, i.e. annotations of maps by labels, text, legends or cartographic symbols.
These include a variety of functions such as:
Image enhancement, i.e. ensuring that the final mapped products are cartographically refined and optimally presented.
Abstraction, i.e. calculations of a more theoretical and statistical nature, e.g. area centroids, proximal features, Thiessen polygons.
Clipping, i.e. separating an area for particular attention.
Zooming or windowing, i.e. selecting a particular area for enlargement so that operations can be performed on just a window of a data set.
Buffer generation, i.e. the creation of buffer zones around any point, line or polygon features. These are zones at preset distances from any feature, e.g. they might be zones of exclusion or inclusion.
This is a crucial component of GIS since it greatly affects the users ability to interact with the data, and the way the user is able to restructure the data to solve specific problems. Many variations of the data will be retrievable, e.g. by specified area or region, theme or class, etc., and these may be retrievable in different ways using two separate retrieval systems - one for map data (lines, points, polygons), and one for non-map data (attributes, etc.).
As well as simple searching and retrieval by theme or area, retrieval using the rules of Boolean Logic is commonly employed. Here the simple operators of “AND”, “OR”, “XOR”, “NOT” are stated to see whether a particular condition is true or false. This can be portrayed in the form of Venn diagrams (Figure 6.14). Commands can be input to the GIS such that any combinations of attributes for a theme or an area can be retrieved or any combination of spatial areas. For example, a command could read “find all water bodies with a mean average temperature of >20C, in combination with a pH of >7.0, that exceed 3ha in area”. Complex retrieval commands could involve aerial shapes, lengths of waterway, intensity of current fishing levels, adjacency to urban areas, etc.
Figure 6.14 Venn Diagrams to Show the Results of Applying Boolean Logic to Two or More Sets (from Burrough, 1986)
Retrieval for analytical purposes would normally be done interactively. As well as the previously mentioned zooming and windowing facilities, retrieval functions usually have browse or search facilities.
GIS provide a large range of analysis capabilities that can operate on selected spatial aspects of the data, the non-spatial attributes of those data or on a combination of these. The user will work interactively to perform the analysis. In Table 6.1 we classified data analysis under three categories - spatial, statistical and measurement. These headings are arbitrary and other classifications could be used, e.g. Dangermond (1983) and Jensen (1986) differentiate between polygon-based and grid cell-based analyses noting that, although grid cell analyses are more generalized, they are much more efficient in terms of data storage and in the operation of analytical tasks. Figure 6.15 exemplifies some grid-based analytical techniques. Because there are so many analysis types, we will simply list some of the more common ones.
Figure 6.15 Some Grid-Based Analytical Techniques (after Dangermond, 1983)
Spatial Analyses. As well as the integration and merging of overlays to produce new mapped surfaces for analysis, other spatial analyses include optimizing route allocations (network analysis), calculating intervisibility, slope and aspect, plus digital terrain modelling, location-allocation optimization, trend surface analysis, etc. For most of these spatial analyses raster-based format is optimal, though rapid advances in software programming is likely to soon mean that data structure differences are irrelevant.
Statistical Analyses. These may range from summarizing and describing simple numerical data, e.g. calculating means, modes, medians, etc., to more complex frequency analysis, measures of dispersion, e.g. auto-spatial correlations, rank correlations and nearest neighbour analysis plus multivariate analyses.
Measurement. This includes a number of operations which may be carried out on one or more data layers. Examples include the simple enumerating of features (either in total or per polygon), measuring linear or curvilinear distances between or along objects, calculating areas, perimeters or volumes, calculating angles and recording direction measurements. From fairly simple measurements rather more sophisticated data can be derived, e.g. frequency/accessibility surfaces, various cost surfaces and any user-defined matrix analysis.
Here we will be concerned both with the types of display and the main devices used for capturing displays. If GIS is thought of as a true processing system then display represents the output from the system. However, unlike most systems where output represents the final processing stage, in GIS output (or display) can be achieved during any of the functional stages listed in Table 6.1. This facility is most important since it allows for user control, review, experimentation, etc. at any stage. Most good GIS software has a range of graphic display features which allows control of label size and fonts, shading ranges, line widths, graphic symbolism, map feature position or composition, etc.
Output can be in three main forms:
Temporary Display. This is that which is captured on the visual display unit (VDU). It is the functional user interface in that a visual display is shown of any action that the user takes, and it allows for interactive experimentation or manipulation at no material cost, at great speed and in an almost infinite variety of ways. Before final displays to other output devices, it is important that the prospective output is reviewed and is graphically refined in order to produce meaningful and presentable displays.
Hard-Copy Display. This is the output which is printed, via a variety of printers and plotters, onto paper or film. Hard copy displays can be in black and white or multi-coloured. The vast qualitative range achievable is dependent upon either cost factors, associated with achievable graphic quality of the hardware used, or with purpose factors related to the intended use of the output. The best quality achievable is now superior to that achieved by manual methods - this qualitative ability has been achieved because GIS functioning has been able to combine and absorb skills from the computer graphics and automated cartography fields.
Disk Files. These are useful if new data structures resulting from transformation and manipulation processes need to be permanently saved, i.e. for rapid later retrieval or for transfer to another data base, another package or to a separate GIS.
There is a variety of visual formats that the display can take.
Thus graphical displays will consist of an infinite variety of maps or graphs, whilst textual output will include both tables and automatic report writing on data base contents to standard summary sheets.
For the permanent capture of GIS output, there are a huge range of plotters and printers plus associated devices, each of which have their own uses and advantages.
Lineprinters. Using varying combinations of alphanumeric characters to produce different shadings, lineprinters were an early output device, though they are still frequently used because they are convenient for researchers or cartographers in giving inexpensive previews before subsequently, using high quality graphic output. Since lineprinters can only print sequentially, data must be organized by using a suitable software program. The quality of output can be greatly improved by photographic size reduction.
A more recent advance on conventional lineprinting is by the use of the dot-matrix printer/plotter. Here displays are produced as series of dots, with shading varying with dot density, and lines are represented by strings of dots (e.g. Figure 6.19). Alphanumeric or special characters can also be printed.
Wide Format and Ink Jet Printers. These are becoming state-of-the-art for hard-copy printing. Both types allow for printing high quality text and graphics at high speeds, and they are capable of producing hard-copy output in colour. They are both expensive but require little attention and can be set to run overnight.
Drum Plotters. These consist of a drum which can be rotated in both directions whilst driving a continuous roll of paper. A pen on a carriage can move along the drum to plot lines (Figure 6.16). All movement is controlled by the computer. Drum plotters work relatively quickly at typical speeds of 40 cms per second. They vary in width from 25cms to 150cms and they can produce multi-colour, high quality output.
Figure 6.16 The Operation and Rotation of a Drum Plotter (from Kadmon, 1988)
Flatbed Plotters. A drawing pen(s) can be moved over a flat drawing surface in both x and y directions, or in conjunction to produce curved lines (Figure 6.17). The movements will correspond to grid co-ordinates held in a computer file. Flatbed plotters typically have A3 or A4 plotting areas, facilities for multi-colour plotting and work at speeds of about 25cms per second. They have a number of advantages over other plotters:
Some can be used as digitizers.
They can use various forms of writing, scribing or laser implements, on drawing paper, acetate and lithographic film.
They are heavy and stable giving greater precision.
A large variety are available, i.e. by size, functional ability, qualitative output and cost.
A number of colour pens can be fitted which can each be operated by pre-set commands.
Pen plotter costs vary (for large plotters) from $5 000 to $12 000.
Figure 6.17 Flatbed Plotter Showing Pen Holder Variations (from Kadmon, 1988)
Electrostatic Plotters. These operate in a similar way to photocopiers. An array of nibs (electrodes) selectively deposit charges on chemically treated paper in the form of an image. Toner is then made to adhere to the charged areas. Once the data has been processed into a suitable format they work very quickly, producing up to 60 × A0 drawings per hour in monochrome, or 10 in colour. They produce medium to high quality hard-copy on high quality paper in up to four colour shades. The cost of these plotters decreased by 50% between 1986 and 1989 - though they are still about $20 000.
Direct Thermal Plotters. These can deliver monochrome A0 plots at a speed of 20 per hour. They use a paper heating process which affects the chemical properties of the special heat-sensitive paper, causing it to turn black where desired. The paper is expensive and being light sensitive, it is unsuited to archiving. By 1991 film media will be available to give long life impressions. A high resolution 400 dpi model costs about $30 000.
Screen Copy Devices. These devices are connected to the port of a graphics workstation and produce hard-copy of the graphic (or text) image appearing on the monitor. They can produce proofs for editing or final plots for use in documents. There are various different devices working on impact, electrostatic, thermal transfer and ink jet printing technologies. Trends in these devices include higher resolution, increased use of colour and falling costs. Present costs range from $2 500 to $5 000.
Other output devices include various cameras, optical film writers outputting to colour micro-film and colour slides and light spot projectors.
Again space prohibits a detailed discussion of data base management systems (DBMS), but there are many basic texts available, e.g. Laurie (1983), Walsh (1985), Date (1986), Austin (1989) and Maguire (1989) and Dale (1990). We must start off by defining a data base. This is a large organized collection of data which should be independent of any particular application and therefore usable in any desired way. It represents the top of a structured hierarchy which consists of fields, records, files and data bases. To exemplify this - if a data base were to be established on river water quality within a country or region, it would be necessary to take a sequence of steps:
At each river sampling point measurements of different water quality parameters would be made.
Each water qualitative parameter measured (observed) would constitute a field.
A collection of all the qualitative measurements (fields) taken at one location would be a record.
A collection of all records made along one river, or stretch, would make up a file.
A collection of all files for the country or region would make up a data base.
Individual files within the data base may be organized in various ways, and the whole data base can be structured in several ways, e.g. commonly using hierarchical, network or relational models.
A DBMS is a computer programme for creating, maintaining and accessing a data base. They can mostly handle numeric and alphanumeric data and they can provide the essential linkage between the user, graphic and other data, plus a range of external computer packages for mapping and statistical analyses. Relational DBMS are now acknowledged as offering the most flexible access to data since they allow independent sources of attribute information to be “joined” in order to create new relationships. Robinson et al (1989) show how a RDBMS fits centrally into a specifically designed GIS, made up by interfacing a number of independent software packages (Figure 6.18).
The reasons for having a DBMS are that all large collections of data, which are available to more than one user, require rules to maintain and manage them so that they remain usable. Since the volume of data is increasing exponentially, the importance of management is being increasingly magnified. It is also most important that large data bases have efficient storage structures, both to minimize storage space and so that they can be efficiently searched or analyzed, and users need to give careful consideration to the exact requirements of the data structure and contents before establishing files or data bases. Maguire (1989) points out that in reality a DBMS isolates end users from the complexities of computer systems, providing them with an easy to use management and analysis system, the functioning of which they will be completely ignorant of.!
The most important facility of a DBMS is their query language. This should provide for easy access to the data and allow for grouping, ordering, selecting and basic statistical procedures as part of the query, i.e. the query language offers opportunities for examination of specific relationships within the data and flexibility in manipulating data into the correct form for passing on to other packages within the GIS (Figure 6.18). DBMS should also allow security or protection to be maintained, if necessary by restricting access to parts of the data base. This may be done by making certain data unintelligible, perhaps by storing them in coded formats, by limiting levels of disaggregation or requiring that certain passwords are known in order to gain access to data or to the GIS.
Figure 6.18 The Integration of External Software into a GIS via a RDBMS
ORACLE: This RDBMS is used as the core of the system. ORACLE is one of the most widely used RDBMS currently available. It provides a full range of database facilities, making use of the SQL query language, an acknowledged worldwide standard.
GIMMS: This is graphics software developed within Edinburgh University. It offers high-quality graphics for the mapping of distributions and located point symbols.
MINITAB: A statistical package used to perform statistical operations such as correlation and regression and statistical hypothesis testing. Results of such operations can be relayed to ORACLE for storage or to GIMMS for mapping.
ARC/INFO: This is a commercially available GIS developed by ESRI Inc. of Redlands, California. The system consists of a graphics package (ARC) and a relational-type file handling system (INFO). In the GIS of Belizean agriculture INFO has been used purely as a store for data passed from ORCACLE. The software necessary to make this connection has been purpose designed in Edinburgh.
GEOLINK: Interfacing software, providing user-friendly menus and enabling linkages to be made between the other software packages.
In this section we will briefly consider three aspects of GIS software. After an introduction, we discuss the varied characteristics of the GIS software milieu. Unlike other sub-sections of GIS, any discussion of software is difficult to structure in a logical way. It seems important to show why this is so. Our intention will not be to confuse the reader - it is to make him or her aware of the extraordinary complexity of this aspect of GIS! We conclude with a review of some varied GIS software packages. Further sources of information on GIS software will be given in Section 6.8.
Any computer software are basically coded programs which are input to the computer to make it function. Programs may be written in various computer languages and maybe written to several kinds of input mediasuch as CCTs and disks. Software programs can be generally classified as:
Operating systems (systems software) whichcontrol the individual tasks of particular computers, e.g. MS-DOS or UNIX.
Assembler, compiler and interpreter programs whichare used to translate computer languages into machine codes.
Individually written programs (applications software) which anyone, having the necessary skills, can write.
Pre-written commercial applications programs - which allow any of a huge range of tasks or games to be performed.
Large multi-purpose programs - called packages, which usually perform a range of related tasks.
True GIS software would come under heading (v), though for the functioning of GIS, operating systems software is essential, and software classified under headings (ii), (iii) and(iv) might be used, especially the latter. User interface with software programs may be by typing commands, by selecting from “menus” which increasingly are of the “pull-down” or “pop-up” type.
GIS software is mostly designed to carry out some or all of the functions listed in Table 6.1. The first specifically GIS programs and packages were produced in the mid-1970s in North America, by a variety of private and public companies ordepartments. Their evolution and categorization has been complex but there are now over 50 genuine GIS software packages on the world market (Montgomery, 1989). This market is certain to grow rapidly as prices drop and as the trend towards producing GIS software for the microcomputer continues, i.e. over half of all GIS software now runs on microcomputers, and prices for some quite sophisticated packages are below $100. Table 6.3, compiled recently by Dennison Parker (1989), shows details on most available GIS software. Maguire (1989) argues that “Such is the importance of software at the present time that many computers are bought and sold, noton thespecifications of hardware, but on the ability to run popularsoftware.” (p.208).
In some ways it is impractical even to consider GIS software and hardware separately since in many cases they are integrally linked. Thus much software is specifically written for hardware systems and many GIS distributors are offering complete “turn-key” packages, having not only integrated hardware and software, but also complete user support facilities. Having said this, we can look at some ways in which GIS software is so variable.
Much of what might now be called fully functional GIS software has either evolved from other related fields or it has evolved from the needs to perform individual GIS functions. The major evolutionary force was perhaps that of Computer Aided Drawing (CAD), though more recently the field ofcomputer graphics has spawned much software. The RS industry has given incentives to image processing software, much of which is now integrated into full GIS. Some of the manipulative andstatistical software components have borrowed heavily from existing statistical programs and externally-derived DBMS programs are increasingly being integrated to GIS. Most recently there is an emphasis on devising software based on the needs of geographical (spatial) analysis. Software tends to retain its raster or vector data storage basis, depending on the evolutionary line it followed.
|System Name||Computing Environment||System Type||First Installed||Number of Users||Pricing||Data Structure(s)||DMBS Interfaces|
|ARC/INFO||DEC,PRIME, DG, IBM, etc.||GIS||1981||nr||nr||Vector||Info, Oracle, Ingres|
|ATLAS* Graphics||PCs/DOS||DM||1984||1000's||$450–1,200||Vector||DIF, Dbase, Lotus, etc.|
|Axis Mapping Info.||PCs/DOS, Sun Apollo, VAX IBM/UNIX||GIS||1978||25+||45£7500–15000||Vector, Raster||na|
|Deltamap||HP9000, SUN, APOLLO,SGI/UNIX||GIS||1986||100+||$8000-80,000||Vector, raster, TIN||Oracle, Ingres, Informix|
|Earth One||PCs/DOS||GIS||1986||40||$12,000–28,000||Vector & raster||na|
|EPPL7||PCs, PS-2/DOS||GIS||1987||335||$500–1,000||Raster||Rbase, Dbase III|
|ERDAS||PCs/DOS; SUN/UNIX; VAX/VMS||GIS, IP||1979||900+||$2,000||Raster||Infor|
|FMS/AC||PCs/DOS; SUN/UNIX; Macintosh||GIS, FM||1987||500||$2,500–7,500||Vector||Dbase, etc.|
|Gas, Electric, Water & Municipal FM||IBM 370/MVS, VM||GIS, AM,FM||1984–89||22||nr||Vector||IMS, DB-2|
|Geo Sight||PCs/DOS||GIS, AM||1987||65+||$4,450||Vector, quadtree||Dbase|
|Geo/SQL, MumMap||PCs/DOS; Sun/UNIX||GIS||1987||240||$9,500+||Vector||Rbase, Oracle, Ingres|
|Geo Vision||VAX/VMS, ULTRIX; SUN, IBM-RT/AIX||GIS,FM||1976||47||nr||Raster, Vector quadtree||Oracle|
|Geovision “GeoPro”||PCs/DOS Macintosh||AM||1988||2||$1,995–4,995||Vector||SQL & DBF supported|
|GFIS||IBM S/370 architecture systems||GIS||1977||180+||var.||Vector||IMS/DLI, SQL/ DB2|
|Gimms||Mainframes, Minis (inclu. UNIX); PCs/DOS, Macintosh||DM,GIS||1970||300||$1500–$3000||Vector, raster||Oracle, SAS, SPSS|
|GDS||VAX/VMS, DEC station/Ultrix||GIS, AM||1980||800+||$10,000+||Object (vector)||Oracle, etc.|
|IDRISI||PCs/DOS||GIS||1987||700||$50–300||Raster||Lotus, Quattro, etc.|
|IGDS/DMRS||DEC VAX/VMS||CAD-CAE FM-GIS||1973||1371||$7,500–110,000||Vector, raster||Informix|
|IMAGE||PCs/DOS||GIS, IP||1989||100+||$995+||Vector||Lotus, Dbase, etc.|
|Laser-Scan||DEC VAX/VMS||GIS||1985||150||£10,000–100,000||Vector, raster||RDB|
|Mac GIS(Cornell U.)||Macintosh||GIS, AM, FM||1988||nr||150||Vector, raster||na|
|MacAtlas, PCAtlas||Macintosh, PCs/DOS||GIS||1985||5000+||$79–199||Vector, raster||na|
|MacGIS (U. Oregon)||Macintosh||GIS||1987||30||$100–300||Raster||Hypercard, etc.|
|Manatron GIS||Unisys/DOS, UNIX||GIS||1983||60+||nr||Vector,raster||Oracle, Fas-port, Adept, Request, etc.|
|Map Grafix||Macintosh||GIS,AM||1987||nr||$8,500||Vector||4th Dimension, Oracle, Double Helix, Omnis, etc.|
|MicroStation GIS||Intergraph/UNIX||GIS||1989||11||$8,300||Vector,raster||Oracle, Ingres, Informix|
|MOSS||DG, Prime||GIS, IP||1977||>100||(public)||Vector,raster||DG/SQL, Oracle|
|Nucor GIS||PCs/DOS||GIS||1988||10||$500–4,500||Vector, raster||ZIM|
|Pamap GIS||Var/VMS, DOS, UNIX,ACS,CS/2||GIS||1983||200||$7,500–$60,000||Vector, raster||RDB, Oracle, Dbase|
|PC ARC/INFO||PCs, PS-2/DOS||GIS||1987||nr||nr||Vector||Info|
|SICAD||Siemens/UNIX||GIS||1978||250||£20,000+||Raster, vector, quadtree||DB2, Informix, etc.|
|SPANS||PCs/DOS, OS2||GIS||1985||400||$8,000+||Raster, vector, quadtree||nr|
|STRINGS||PCs/DOS||GIS/FM||1979||150||$3,500–5,000||Vector||Ingres, Sybase Britton Lee|
|Territory Mgt. Sys.||PCs/DOS||GIS||1988||25||$2,950–3,950||Vector, quadtree||Dbase|
|Topologic||PCs/DOS; OS-2; VAX/ VMS||GIS||1987||18||$2–7,000||Raster, Vector, quadtree||Dbase, RDB|
|UltiMap||Apollo, AEGIS Operating System||GIS,AM||1974||40||$19,000–50,000||Vector, raster||Oracle, Informix, Ingress, IMS, etc.|
|USEMAP||PCs/DOS||GIS, AM, FM, CADD||1973||3||$1,500–5,000||Vector, raster||Dbase III|
In the U.S. especially, from where most software originates, there are two main classes of GIS software in terms of legal ownership, i.e. that in the public domain (government funded and researched) and private (owned by the author or their employer), plus some which falls in the grey area where ownership is uncertain. This causes some problems since public domain software can be sold for prices which cover little more than distribution costs, and since they are so cheap they are often copied or “dubiously acquired” for integration in to expensive commercial packages. Most public domain packages provide little or no support facilities, though some, such as GRASS (developed by the U.S. Army Corps of Engineers), is well supported.
Because much of the software has its roots in various fields, i.e. in government departments, in universities or in large and small private companies, each of whom may be producing a different software commodity, a number of complex linkages are occuring as the industry is achieving both an optimum production equilibrium (in terms of systems variety), and economies of scale. Alternatively, some universities or government departments are marketing their own products or they are setting up private companies to do this.
Software packages are extremely difficult to price and prices show huge differentials. This results from:
It will be clear from the above that software ranges from single or modular programs, which are capable of being linked to a GIS to perform specific functions, to large integrated packages which may perform all GIS functions. Some packages are specifically designed for mainframes or minicomputers, others for microcomputers, though increasingly the larger GIS companies are designing their programs to suit any size of computer. Many packages are “customized” to suit specific end-user requirements. Because of this functional range, no two programs or packages will be the same, and so it is impossible to compare one with another or to instigate any meaningful software classification system.