Urban Land Use Mapping Using Register Data and standard GIS Martin Hvidberg Danish Forest and Landscape Research Institute Danish Ministry of Environment and Energy Hørsholm Kongevej 11 - DK2970 Hørsholm - Denmark
[email protected] http://www.fsl.dk/eng/index.htm
Abstract. This paper will show one way to generate land use maps from register data. The register data used here is the Danish Building and Dwelling Register (BBR). The thematic result map showing urban land use will be based on a regular grid net with cell size of 100 m by 100 m. Each cell will be classified, based on the information on the buildings falling inside that cell. This paper will also present a hierarchical class system called Hierarchical Urban Land Use Class System (HULUCS) for describing urban land use. HULUCS operates on two levels, i.e. a single building level and an area level. The later allows mix classes. Additionally the implementation of the classification is discussed in context of computer performance, and finally the resulting map of Denmark is presented.
1
Introduction
The demand for spatial data is input for the planning process, risk assessments, various spatial computer modes, etc have existed for as long as these tasks have been performed. [2],[6],[13],[21]. Working in an urban context a land use map is often preferred over a land cover map. While urban land use is mainly governed by the activities inside buildings it can not be monitored by user of satellite or aerial photographs though these techniques can provide detailed information on surface cover, e.g. [5]. With modern GIS it is tempting to produce land use maps based on publicly available register data. This paper is a description of the process and some of the lessons learned along the way. 1.1 Background The work presented here is made in association with two projects, namely: "Planning systems for Sustainable Development" (PSSD) and the "Area Information System" (AIS). PSSD are financed by the European Regional Development Fund, INTERREG II C and have partners in Denmark, Finland and Germany [8],[15]. AIS is financed and implemented by the Danish Ministry of Environment and Energy, with the purpose of generating a total land use data set for Denmark with special relevance for environmental considerations and analysis [16], [17].
Demands for an urban land use classification had been expressed on beforehand in the Danish Ministry of Environment and Energy. This demand were expected to exist in other public organisations as well, later this has been confirmed by a high frequency of response when publicising the existence of such a data set for the entire Denmark. It has not been investigated systematically what the intended uses of the data are amongst those who request a copy. Though it is the intention to collect feed back from the users. Mapping areas based on register data have been used for a long time: demographic distribution, rainfall distribution, transport flows, etc. [7], environmental resources and land use [19] real estate in Denmark [14],[9]. To make urban land use maps on a national basis requires both reliable homogeneous register data and a computer with a certain capacity. We are getting used to computer power increasing and are still more concerned with what we want to do, rather than what is possible. One of the things we want, is to use the computer analysing spatial problems using GIS. At present this is even possible with desktop equipment. The present project was made using a standard computer1 and the standard GIS ArcView™ 3.1. The dominant problem proved to be the size of the data set. The Danish “Building and Dwelling Register” (BBR2) holds 2.4 mill buildings and were to be used to classify approximately ½ mill. 1 ha squares. The entire task was originally intended to be implemented in ESRI’s programming language Avenue™. But frequent breakdowns of the windows system3 provoked a switch to ArcView™ 3.1 for UNIX™4, but though this was stable the estimated run times was still approximately 1000 hours. Finally an urban land use map was successfully generated using a combination of ArcView on different platforms and with the core classification routines ported to Borland Delphi™.
2
Method
The task of converting The Danish Building and Dwelling Register’s (BBR), point based information to area based thematic maps, is here referred to as aggregating point information into polygon information. The building points initially hold the input information, such as the usage of the building, its ground- and floor-area, etc. The point-based entity is therefore the individual building and the desired output is a map of regions, each associated with one urban land use class. This challenge contains three main sub-problems. a) Defining the limits of homogeneous areas. b) Definition of which output classes are acceptable an, c) Defining the rules for the classification process, i.e. the rules that describe which combinations of buildings are required to satisfy each of the previously defined classes, and to make sure they are mutual exclusive and exhausting. All three 1
Pentium II, 400 MHz, 128 Mb Danish abbreviation: ”Bygnings og Bolig Registeret” 3 NT 4.0 sp. 5 4 IRIX 6.5 on a Silicon Graphics Indigo2 RISC 4000
2
questions were sought answered mainly through experimenting with different solutions, to gain methods that provide optimal results, given the conditions. Probably adjustments are needed to apply these methods in other contexts. 2.1
Defining homogeneous areas
Several approaches have been suggested and tested [18]. They can be grouped in two main categories: Those who utilise additional data layers, and those who do not. The first group has the potential advantage of apriory knowledge. By utilising an existing map of the city as the frame, and then applying a land use class to each pre-defined sub area have the advantage that such sub-areas are likely to be relatively homogeneous. The second group are solutions using e.g. grid nets. These solutions do not take advantage of the natural borders existing in the city, but overrules these with their regularly spaced network of lines.
Fig.1 Polygons from administrative map Fig. 2 Polygons from regular grid net gives homogeneous aggregation areas ensures reproducibility over time and between nations Illustration shows the residential area "Kartoffelrækkerne" in Copenhagen with it’s semidetached row-housing surrounded by areas of primary apartment-blocks mixed with one-family-houses and shops. The approach using the existing map, in this case administrative city units, have the advantage over the regular grid-net in it’s likeliness to give more homogeneous areas. Other methods not depending on additional data layers have primarily the Thiesen-polygon approaches been considered. This approach was discarded since it always produces polygons with exactly one building en each, and therefore required additional processing before allowing aggregation. To ensure primarily portability, grid nets have been selected for the present urban land use map. Though being sub-optimal in homogeneity it has the advantage of being unbiased. In this way the method can be reproduced in other countries, and the resulting maps have
a better chance of being comparable, also over time, enabling monitoring of urban change. 2.2
Definition of output classes
One of the considerations of major importance when developing or choosing a classification system is the motivation. Different motivations and different goals leads to conflicting interests [11]. We would like to think of “classes” as different areas used for different land use, illustrated by different colours on a map. When considering colouring maps, we must additionally require each instant of a class to have a certain minimum area, in order to be visible. The system, presented here, is operating on two spatial levels: point and polygon. The point level is represented in the individual building, while the polygon level is represented in “city-block” areas. Classification values are to be applied to the classes so that a single value represents one house or one city block, respectively. A “City block” can be the area between city streets, i.e. if you consider the streets as rivers, the “city-blocks” are the associated islands. In this project a city-block can also be an administratively defined area like counties or parish, or it can be a generically constructed areas e.g. Thiessen polygons or regular grid cells, as described above. In the Hierarchic Urban Land Use Class System (HULUCS) nomenclature developed in this project, it is specifically intended that the city-block classes must be specified using the building level classes, as tokens in their definition. For example: It there is no city block class for “lake”, since no group of buildings could form such a class. The system has also two hierarchical levels, a coarser and a finer. Which one to use depends on the quality of the input data and the geographical scale of the output product. With scarce input data or when producing maps of large areas the coarser class system (A,B,C…Z. and 1,2,3…6 for points and areas respectively) may be preferable. While having more detailed input data and the need to produce detailed output, the more detailed class system (AA,AB…BA,BB,…ZZ and 1-1, 1-2, 1-3…60) might be a better choice. The finer classes can always be collapsed into the coarser since they are constructed as true subdivisions of those. To ensure that the sum of a number of finer classes always adds up exactly to one coarser class it is necessary to include a misc. class in some of the coarser class. Those misc. classes have all been given a "*Z" or "*-0" (dash-zero) name to secure homogeneity throughout the classes. It is important to stress the difference between, e.g. class “2” (Business) and class “2-0” (misc. Business). The first is the coarser class containing all “2-*” subclasses, while the latter is the misc. class containing items that classify as “2” but not in any of the ordinary “2-*” subclasses.
HULUCS - Hierarchic Urban Land Use Class System Building unit classes A
Residential AA AB AC AD AE AZ B Business BA BB BC BD BE BF BG BH BI BJ social security BK BL BM activities BN BO BZ C Recreation CA CB holiday camp, etc. CC CD CZ Z Miscellaneous ZZ
Farmhouse Detached house, one family house Semi detached houses, terraced or cluster houses Block of apartments Dormitory Other permanent residential building Primary production Manufacturing Electricity, gas and water supply Construction Sales and repair Hotels and restaurants Transport, storage and communication Financial intermediation Real estate, renting and business activities Public administration and defence; compulsory Education Health and social work Other community, social and personal service Private households with employed persons Extra-territorial organizations and bodies Other Business, trade or industry activities Private. Part time residences, summerhouses, etc. Residential/recreative housing, youth hostel, Allotment, garden Sports Other recreation Misc. misc.
Table 1. HULUCS - The Building level
HULUCS - Hierarchic Urban Land Use Class System City-block unit classes 1
Residential 1-1 Detached houses 1-2 Semidetached houses 1-3 Blocks of flats 1-0 misc. Residential 2 Business 2-1 Farming 2-2 Industry 2-3 Sales & Administration 2-5 Hotels & Restaurants 2-6 Education 2-7 Health & social work 2-8 Culture 2-9 Infrastructure 2-0 misc. Business 3 Mixed Residential & Business 3-1 Mixed Residential & Service 3-2 Mixed Residential & Industry 3-0 misc. Mixed Residential & Business 4 Recreation 4-1 Summer houses 4-2 Allotment gardens 4-3 Sport 4-4 Park and Playground 4-0 misc. Recreation 5 Underdeveloped 5-0 Underdeveloped 6 Misc. 6-0 Misc. Table 2. HULUCS - The City-block level 2.3
Defining the rules for the classification process
A part from deciding on which classes, into which to divide the world, we also have to define the classes. That is, if we have a class A, we need to specify exactly when and when not something is of class A.
2.3.1 The classification process Each parcel is classified separately. The process assumes the existence of a list of all the buildings located inside the parcel in question, as well as assuming that each building have a known land use code according to the HULUCS single building level and additionally holds information on the buildings total area. This is described below in the chapter “implementing and data”. To classify a given parcel, the program creates an intermediate table accounting for the area of each of the building. From this table is accumulated area of each class, totalled over all the buildings on the parcel, by means of a area based cross tabulation. These totals are the basis of the classification into "Parcel unit classes" (1-1,1-2,13,etc.) An example of such intermediate tables is shown here, though during the program run they exist only in the computer memory.
Fig.3 The classification procedure. (circles are red, squares green and triangles blue.) Build-up versus Non-build-up check First considered is the total building area compared to the parcel ground area. If the building area totals to less that 2% of the parcel area the parcel is considered "Undeveloped", meaning that though there are, at least, one building on the parcel it is to little to be taken into account. In this case the parcel is left with class 5-0, and no further checks are made. Clean class check Now knowing that the parcel is build up, it will first be checked if one land use class are dominant enough to entitle the parcel to be classified as a clean class parcel. A clean class is a "Parcel unit class" that is described by a single "Building unit class", i.e. a waste majority of the buildings on the parcel have the same building class. In HULUCS a parcel is considered clean if 75% or more of the building area belongs to the same class. e.g. On a parcel with three houses totalling 500 m2 , two of the houses are AB (detached house) and totals the 480 m2 while the third building is type BG
(transport, storage and communication) and coves 20 m2. This parcel have 96% (480*100/500) of its area in one (AB) class, and is accordingly considered clean. In this case the parcel is left with class 1-1, and no further checks are made. Mixed classes check Pre-defined mix of two different main classes. If none of the building unit classes sums-up enough area to justify classifications as a clean class, then it is checked if the parcel can meet the demands of one of the predefined mix classes. There exists, presently, three pre-defined mixed classes, i.e. 3-1 (Mixed Residential & Service), 3-2 (Mixed Residential & Industry) and 2-1 (Farming). The difference between 3-1 and 3-2 is based on potential pollution, where 3-1 are considered to be the types of business that do not spread smoke, sewage water, and noise into the surrounding environment, while 3-2 potentially can do that, and therefore requires other considerations for where to locate. The demands to meet for a mixed class are similar to those of a clean class. There have to be at least 75% of the building area supporting the classification. In the case of mixed classes joining the areas from two classes can collect these 75%, and non-of the two must be represented with less than 25% of the parcel’s area, i.e. minimum 25% of one class, plus minimum 50% of another. Mix within the same main class. Mixed classes of this type is used if non-of the clean classes and non-of the predefined mixed class demands could be satisfied. A mix within one main class can be satisfied if all the buildings in a main class can present a total of 75% or more of the parcel area. E.g. if all buildings with a B* building unit class (Business) can sum their areas to a minimum of 75% of the parcels area, then the parcel are classified as 2-0 (Misc. Business). Estimating the quality of the classification The quality parameter used is quite simple, yet helpful. The value is the percentage of the total m2 that supports the HULUCS result. In the above example with three houses totalling 500 m2, two AB that totals 480 and one BG at 20 m2. 96% (480*100/500) of it’s area are AB, and the parcel is accordingly considered a clean 1-1 and 0.9600 is transferred as a quality estimate. In the case of mixed classes only the area off the dominant class supporting the classification are counted. For this reason some of the misc. classes tend to show relatively low quality. This is not considered an error, but could lead to misinterpretation, if the generation process is not known. It is under consideration to change this in feature versions. 2.4 Implementation and data The system to produce urban land use maps based on register data was based primarily on an NT platform. Due to undocumented features in the windows system, heavy tasks would bring the computer to stop when having used up all its memory. Those
Fig. 4 Flow of the map production [16]
processes were completed on a UNIX platform. It is beyond this paper to explain the shortcomings of the window system, and from hereon it is just noted that ArcView can perform the necessary tasks, given the adequate conditions. The implementation followed the general scheme illustrated by the flow char. Though it was necessary to make some workarounds primary to save computer run time. The three input data are BBR (the Danish Building and Dwelling Register), DAV (Danish Address and Road database) and GridDK 100 (a 100 meter grid net covering Denmark5). The 2.4 mill. BBR entries was geocoded using DAV6 since BBR do not include the buildings coordinates.
Fig. 5 The “DAV” road network used for Fig. 6 Road net and the “BBR” buildings. geocoding of the “BBR” buildings Colours explained below. The classification of each cell initiates by identifying which BBR points fall inside the given polygon. This spatial selection requires considerable computer power/time with large data sets. To optimise performance when searching through the geo-data, the data set was indexed using ArcView’s indexing methods, but this had no major effect on the run time of the Avenue code that carry out the classification. A workaround was found. In stead of importing the BBR points into ArcView the geocoded BBR points were converted to a standard database file (Dbase IV), containing X-coordinate, Y-coordinate, Usage code and building area. A separate program was developed, in Borland Delphi, to read and process this BBR information. The spatial selection equivalent was handled by introducing a new field with a number generated from the X- and Y-coordinates. First replacing the two last (least significant) decimals in the Xand the Y-coordinates with zeros, and then concatenating them to one long number generated this new number. Example. X-coordinate 12345 and Y-coordinate 98765 becomes 12300 and 98700 respectively and are then concatenated to 1230098700. 5
The GridDK has lower left corner identical with UTM zone 32 and follows the UTM zone 32 grid. 6 The geocoding was performed by Per Reupert Kristensen, Energistyrelsen, Denmark
This number uniquely describes the lower left coordinate of the 100 m by 100 m square containing the point in question, and has been named the Square-ID. The “spatial selection” is now easily performed, simply by picking all points having the same Square-ID. The entire data set was sorted according to Square-ID field prior to further process. This approach only requires a single sorting off the 2.4 mill. records, while traditional GIS based spatial selection, square by square, requires considerable more looking through and may be at least one sorting of the entire data set per square, if the data are not indexed. Picking all the records with the same Square-ID are quickly done, when utilising the fact that we know that they are all ready sorted. The program then considers the picked records and proceeds with the classification process as described above, in the chapter "The classification process". The output is another database file holding one record for each Square-ID found in the input data. This file contains: The Square-ID, it’s lover left X- and Y-coordinates, the classification result in HULUCS code, a classification accuracy estimate and finally a building density. Transferring both the Square-ID and the two coordinates are redundant but save the effort to regenerate the one from the other. The Square-ID is not used any further and the density are only used for an urban density map that is beyond the scope of this paper. This file is then feed into ArcView for final generation of the map. Using Avenue and the X- and Ycoordinates to identify the lover left corner position ArcView generates a 100m by 100m square. The HULUCS code, the quality parameter and the density are then associated with the polygon as three attributes. The colouring etc. is performed using standard ArcView facilities.
Fig. 7 The “BBR” buildings and the Fig. 8. The classified “GridDK 100” and “GridDK 100”100meter grid net the “BBR” buildings [10] This new approach reduced the time needed for the entire classification process from app. 1000 house to app. 30 minutes, on the same computer. This confirms yet again
the importance of planning data structure7 and algorithms to optimally fit the task as well as each other.
3
Results
The immediate result of the work is the knowledge obtained through method development and of cause the output map of urban land use in Denmark. Since generic methods and systems compatible with a broad verity of statistical bureaus, have been used, the methods should be applicable with a homogeneous result in most countries. Method development have indicated a number of lessons It is very important to give great considerations to the classification categories when selecting or constructing urban land use systems. Special effort should be taken to assure that the classes are exhausting and exclusive, i.e. that they leave no wholes and do not overlap. The single most important factor when deciding on land use classes was found to be the motivation behind the entire investigation. Economic considerations call for one set off classes while environmental may call for an entirely different set. Additionally it turned out to be important that the categories have categories compatible with existing statistical data, since these have to fit as input in order for the operation to be successful. We have tried to limit the “describers” in the classification to include parameters likely to be found in public registers for most European countries. Finally it seems important that the resulting output map follows general mapping traditions as well as makes intuitively sense to potential users. We have attempted to make e.g. the colouring scheme comply with this.
Maps Based on data from the 2.4 mill. buildings in the Danish Building and Dwelling register (1997) and a grid net, a map of the country have been generated, using the described process. The resulting map is a mosaic of squares, here coloured in accordance with the prevailing land use. Fig. 9 shows the city of Odense with suburbs (approximately 15 by 15 Km) on a light yellow background with blue water and black roads. The city centre is dominated by purple 7
Data structures see [1],[20] and [4]
Fig. 9 The Final map. Here the city of Odense. Approximately 15 x 15 Km
“Mixed Residential & Service”, whilst large suburban areas are dominated of the red colours of “Detached houses” and “Semidetached houses” and the brown of “Blocks of flats”. Round the urban fringes can be found “Industry” areas with blue and some purple squares. The green of “Farming” is only found alongside rural roads. The isolated yellow “Recreation” dots are mainly club houses or sports facilities. Mind that a building has only one address and even larger industry or sports complexes are therefore inherently confined to a single square. Statistical outputs When generating the HULUCS map the routine also output simple statistics. For each cell is generated a HULUCS class result together with a quality parameter, both described above. Additionally a density value is associated with each cell. The density value is the accumulated floor square meters of all buildings in the cell, divided by the cells own area, and given as %. According to the roles describe above, areas with less than 2% are not considered build-up. These statistical results have been summarised by HULUCS class for a national average. The table below shows the summarised national values.
Fig. 10 Summary table of the statistical outputs from the classification process. Fig. 10 shows summarised values for each HULUCS class for entire Denmark. The column “Count” is the number of cells with the given value. The column “Relative”, is “Count” in percentage of the total. The column “aisqua” is the previously described quality parameter. Here summarised as minimum, maximum and average for each class. The “densit” column is the density parameter described above. Here summarised as minimum, maximum and average for each class. Interesting are to see that the pre-defined values like the lower 2% limit and the 75% criterium explained above are reflected by the density and quality parameters. No doubt that the total counts would change had different cut-off parameters been
selected. The quality parameter shows that miscellaneous and mix classes are the only ones to allow quality below 75%, as described in chapter 2.3.1. The impressive minimum 98% in class 5-0 is indirectly by definition since all cells with more than 2% density are allocated to other classes.
Fig. 11 The resulting map for the major part of Denmark [12]
4
Conclusions
It was found important to give great concerns to motivation when selecting or constructing systems to describe urban land use classes. Different motivations behind the operation call for fundamental different sets of classes. Spatial routines, like spatial selection, are slow GIS operations. To save time when handling large data sets they ought to be limited to a minimum use. Sorting of the data before starting may in some cases reduce the problem, if the program are build to make use of the apriory knowledge that the data are all ready in a sorted order. Porting parts of the data heavy handling out of slow, though elegant GIS macro languages into speedy languages like Pascal or C can show beneficial. Perspectives One natural next step for the aggregating register data and statistical data into thematic maps, using GIS, would be the possibility to handle other dimensions. Aggregating points by lines, could be used to calculate how many houses are on each road segment, or the average income, aggregated by street. Aggregate
INPUT
OUTPUT
Point Line Polygon
Point
Line 1 4 7
2 5 8
Polygon 3 6 9
Table 3. Aggregation combinations theoretically possible. Theoretically there are no problems in aggregating any of the three basic primitives (point, arc and polygon) by each other. This creates nine potential aggregating combinations i.e. Point by polygon, point by line, point by point, line by polygon, line by line... polygon by polygon. To the best of our knowledge no GIS presently have all nine possibilities as standard functionality, though they seem elementary at first glance. We suggest further investigation in these GIS techniques.
References 1. Burrough, P.A. and McDonnell, R.A., Prinsiples of Geographic Information Systems, Oxford University Press, New York, 1998. 2. Chorley, R.J. and Haggett,P., Models in Geography, Methuen & Co, Ltd, 1967. 3. Balstrøm, T., Jacobi, O. and Sørensen, E.M. Ed. GIS i Danmark (GIS in Denmark) Teknisk Forlag, Copenhagen, 1994. 4. Decker, R., Data Structures, Prentice-Hall, Inc., 1989. 5. Ekstrand, S, and Hansen, C., Operational Urban Mapping using Digital Aerial Photography Integrated with a High Resolution DEM. 1, 212-219. 1997. ERIM international inc. Third International Airborne Remote Sensing Conference and Exhibition. 7-6-1997. 6. European Environment Agency, Europe's Environment - The Dobris Assessment, Jonathan Sinclair Willson, Copenhagen, 1995. 7. Garnier, B.J. Practical work in Geography. Edward Arnold (pub.) LTD. London, 1963. 8. Hansen, H.S. (Ed.). PSSD Planing Systems for Sustainable Development - The Methodical Report. NERI Technical Report No. 351. 2000. 9. Larsen, H.H., Ejendomsdatabaser, in [3]. 10. Hvidberg, M. Mapping of Register data, using standars GIS. Papers and SAMS Case Study Summaries, 98-102. 4-5-2000. Stockholm, Sweden, Boverket & Naturvårdsverket. How to integrate environmental aspects into spatial planning. 4-5-2000. 2000. 11. Hvidberg, M., Setting up indicators for monitoring of urban environment, based on central registers. - Hierarchic Urban Land Use Classification System - HULUCS, in 2.3.3.b1 - Final Report - Level 4 task, INTERREG II C - Planning Systems for Sustainable Development (PSSD), 2000. 12. Hvidberg, M., Integration and aggregation of central registers and planning/topographic zones - An Aggregation Machine, in 2.3.2.d1 - Final Report - Level 4 task ,INTERREG II C - Planning Systems for Sustainable Development (PSSD), 2000. 13. Jensen, S.F., Application of GIS in Air Pollution Exposure and Health Studies. Dept.of Atmospheric Environment, NERI. Nordisk AM/FM, GIS-Konferanse 1998. 1998. 14. Madsen, H.B., Nørr, A.H. and Holst, K.A., Atlas over Danmark (Atlas of Denmark) serie 1, vol. 3. Den Danske Jordklasificering (The Danish Soil Classification), Det Kongelige Danske Geografiske Selskab (Royal Danish Geographical Society) Copenhagen, 1992. 15. Mikkonen, H., PSSD web site - www.pssdtoolbox.net. 2000. 16. Nielsen, K. et. al. Areal informations Systemet (The Area Information System) Miljø og Energiministeriet, DMU. ISBN 87-7772-567-0. 2000. 17. Olsen, B,Ø., AIS web site - ais.dmu.dk . 2000. 18. Skov-Petersen, H., Spatial Aggregation Strategies - Applications in land use mapping. 1999. Aalborg, Denmark. Proceedings of ScanGIS ´99. Stubkjær E.and Hansen, H. S. 19. Sjernholm, M., Mielby, S. and Platou, W., Danske og internationale arealdatabaser (Danish and International Land Use data bases), in [3] 20. Tomlin, C. D., Geographical Information Systems, Prentice-Hall, Inc., 1990. 21. Wu, F. and Webster, C.J., Simulating artificial cities in a GIS environment: urban growth under alternative regulation regimes, Int.J.Geographical Information Science, 14, 625-648, 2000.
Biography Martin Hvidberg is Geographer M.Sc from inst. of Geography, Univ. of Copenhagen. He has worked with GIS and Remote Sensing (RS) since 1994, on Danish as well as on international level. Martin Hvidberg has the last two years worked as a GIS/RS researcher here at Danish Forest and Landscape Research Inst. Ministry of Environment and Energy, Denmark. Trademarks ArcView GIS is a registered trademark of Environmental Systems Research Institute, Inc. Avenue is a registered trademark of Environmental Systems Research Institute, Inc. Borland Delphi is a registered trademark of INPRISE Corporation. IRIX is a registered trademark of Silicon Graphics, Inc. UNIX is a registered trademark of The Open Group