Arwyn Jones, Marc Van Liedekerk and Alan Cooper*
This project examines the potential role of geographic information systems, such as ArcInfo, for processing information on radiation deposition for large areas (e.g. continental Europe). From the outset, a GIS was seen as being an important tool on the analysis of the data and presentation of this information in the form of maps since all the measuring points are geo-referenced. The project aims to process the sampling data to a state whereby isolines of caesium deposition can be calculated at a number of levels: on a European scale, for individual countries and for areas of high deposition (e.g. 'hot spots').
The paper describes and discusses the methodology used to pre-process and analyse the data, focusing in particular on the various possibilities, and limitations, for the generation of isolines using functions contained within ArcInfo's TIN and GRID modules.
An equally important element of the project is the presentation of the final results of the analysis. The deposition data will be combined with other thematic datasets (e.g. roads, rivers, political boundaries) to produce plot files that will form the basis of a serie of maps. For this aspect, the paper outlines an innovative routine for the transfer of ArcInfo datasets into an Apple Mac environment for the generation of high quality value-added graphics prior to printing.
The paper presents a summary of the work to-date together with a discussion on a novel method for possible access to the ArcInfo files directly from the WWW.
The Chernobyl nuclear power plant accident on the 26 April 1986 resulted in the contamination of large parts of the European continent by a number of radioactive substances. These elements included Caesium-137, Strontium-90, Plutonium-238, 239, 240 and americium 241. After the accident various compilations were made by various countries or regions in Europe of the contamination resulting from the radioactive material released during the accident. These compilations were made for different purposes and, consequently, there are significant differences in their resolution and quality. To date, no attempt has been made to compile the necessary data to produce a comprehensive presentation of the contamination over the whole territory of Europe, the continent on which, by far, the majority of the released material was deposited. In many cases, improved data have since been, and continue to be, obtained through more refined and extensive monitoring. This is particularly so in those areas where contamination was greatest.
As a result, the time is now opportune to prepare maps of the radioactive deposition of the whole of the European territory consequent on the Chernobyl accident. The publication of such maps by the tenth anniversary of the accident would have wide public and scientific interest. The compilation of such maps is being undertaken within the European Commission's (EC) Environment Institute at the Joint Research Centre at Ispra. Such maps aim:
The radiological data, provided by participating scientific institutes and competent authorities of the countries of the European Union are being integrated in an Oracle Relational DataBase Management System (RDBMS) at the Joint Research Centre (JRC) at Ispra. After data validation and inter-comparison exercises, a series of maps will be compiled from the data by means of an information management platform. In the case of this project, such an environment being used to develop the information platform is the ArcInfo Geographic Information System (GIS).
The scope of this paper can be subdivided into three basic sections. The first section highlights the advantages and disadvantages of the different methods that exist within the various elements of ArcInfo to generate isolines; the second part of the paper presents a brief discussion on a novel and effective technique for the generation of high quality hardcopy output; and, finally, in the third section, the paper presents a brief summary of how the maps might be made available to the wider audience. In particular, the last section outlines a novel and exciting possibility to provide direct access associated ArcInfo datasets from the Internet and World Wide Web environments.
The preparation of the maps can be divided into two main parts: a) the collection, validation of radioactivity and geographic datasets and b) the production of the maps.
As a basis for the maps, the information available in the REM database was taken (EC 1994). The REM (Radioactivity Environmental Monitoring) databank was set up by the JRC-Ispra in 1988 to bring together and store in a harmonised way, environmental radioactivity data produced in the aftermath of the Chernobyl accident. The information held by the bank cover data from the member states of the European Union as well as other European countries. The deposition vales come from a variety of sampling techniques, air-gamma spectrometry survey, deposition sensors, analysis of soil, water, milk, meat and vegetables. For this project, the data fields of interest are the locational information (latitude and longitude), the values for caesium and the data for the collection of the sample. The latter is important as all the caesium vales will be recalculated for the 10th May 1986.
Current effort is underway to collect additional information for a number of regions for where there are no data in the REM databank and for areas where there is only limited data.
The cartographic detail for the European and Country scale maps will be provided by information contained in the Digital Chart of the World (DCW) and from the EC Eurostat Office in Luxembourg. The DCW is an established dataset of assorted digital cartographic features for the world at a scale of 1:1,000,000. Investigations into the possible sources of cartographic information for the 'hot-spot' areas are still in-hand.
One of the important elements of this project is the use of a GIS for the preparation and production of the maps. The system being used in this project is ArcInfo, in particular, intensive use is being made of the GRID and TIN modules.
The information on radioactive deposition from the collaborating laboratories comes in the form of point data which can be geographically located by a latitude and longitude coordinate. This information is used as input to the GENERATE function of ARC. This creates a point coverage of the sampling locations for the whole of Europe. Additional information, such as caesium level and any other attribute information, can then be attached directly to the PAT file or they can be related through the point identification code. In the long-term, an a supplementary project will utilize Arc Info's Database Integrator to make a direct link from the coverage to the attribute information stored in the Oracles tables within the REM databank.
The deposition data are then transformed from the latitude and longitude base to a suitable map projection which allows cartographic data to be overlayed (e.g. coastline for checking that the values fall on land). The maps will utilise an equal area map projection such as the Lambert-Azimuth, to display the information.
However, the objectives of the project require the generation of isolines of deposition to be displayed. To achieve this task, some degree of interpolation and generalisation of the data must be undertaken. Within the ArcInfo package, there are numerous methods for generating isolines which range from various techniques for the generation of grids or lattices to the generation of surfaces based on triangulated irregular networks.
The basic objective for such maps is the production of isolines from the point datasets of sampling locations. The basic step is to generate a representation of a surface from the data which can then be contoured. Unfortunately, no single function exists within the ArcInfo software suit to generate isolines directly from a point dataset. The isoline routines mainly work on grided or cell-based datasets. It should be remembered that these maps will be produced at a coarser resolution that the density of many of the sampling locations, which in itself is variable over the continent of Europe, so there is a need to aggregate the number of sampling points in someway. In addition, the processing may wish to fill in 'holes' in the dataset. To this end, the use of mathematical interpolation techniques have to be considered.
Within ArcInfo, there are three of possible approaches. These include GRID functions (such as POINTGRID, IDW, SPLINE), the use of TINs or the use of statistical routines to produce new datasets (e.g. KRIGGING).
The results from all these techniques can be modified by the use of additional functions within GRID (e.g. focal filtering). In addition, once the data are in the grid format, many further operations can be utilised. The basic functions in GRID are:
POINTGRID which creates a grid from point features in an ArcInfo coverage. The value of the cell is dependant on the value of the point attribute specified; IDW performs an inverse distance weighted interpolation on a point dataset. This function can take account of floating-point data and can assign the value of a cell on the basis of neighbouring points (not necessarily within the cell). IDW can be used to fill holes in datasets; SPLINE performs a two-dimensional minimum curvature spline interpolation on a point dataset resulting in a smooth surface that passes exactly through the input points.
The TIN method uses a linear relationship between the sampling points to establish a surface based on a triangulated irregular network. A tin connects a set of irregularly spaced xyz locations by means of triangles. Tins are useful for representing surfaces that are highly variable and contain discontinuities. Tins can be used to provide interpolations over large areas which are uniform or have few sample points. It is possible in TIN to specify regions of constant values (e.g. lakes) or areas which have no data. It is possible to generate a lattice from a TIN and vice versa. In the case of these test maps, the point dataset for caesium and polygon coverage of the coastline were used as input to the CREATETIN function.
There are two possible approaches within this section: to use the KRIGING function within Arc or to export the data to an external statistical packages.
KRIGING interpolates a lattice from a set of variably spaced points using semi-variograms and various kriging routines.
External statistical packages could perform a statistical evaluation of the data and, on the basis of that evaluation, generate and apply the appropriate sampling framework of the dataset. In actual terms, this would probably take the format of a grid or lattice which could then be re-imported into Arc. There are numerous statistical packages that can perform such analysis. For example, SPLUS, STATSCI, GSLIB and SURFACE+ are a few.
The project is currently investigating the potential of the last approach.
Contours or isolines can be generated by three functions within ArcInfo: LATTICECONTOUR, LATTICEPOLY and TINCONTOUR.
Both LATTICECONTOUR and LATTICEPOLY require datasets to be in a lattice or cell based format before they can operate. Both routines work in a similar manner although their outputs are markedly different. LATTICEPOLY converts a lattice to a polygon coverage classified on the basis of the ranges specified in pre- defined INFO lookup table. In the case of these maps, the polygons have an attribute code which is specified in the look-up table against an actual values for caesium (i.e. the PAT file contains a code for the deposition value). Alternatively, LATTICECONTOUR converts a lattice to a coverage containing isolines. In this case, the value of caesium deposition is attached to the arcs through the AAT file.
In both cases, the results of the isoline generation can be enhanced (in both a positive and negative manner) by the use of GRID FOCAL techniques and the Arc FILTER command. If the generalisation of information is required, these functions can remove small areas which may be considered as speckle or 'noise' within the grided data. For cartographic applications where a degree of generalisation of the data is often required, the use of such tools has proven useful in initial testing.
TINCONTOUR, in a similar manner to LATTICECONTOUR, converts a TIN to a line coverage containing isolines.
Although all the methods described above produce reasonable output, there are a number of problems that need answering. The fundamental weakness of the interpolation and contouring routines contained in ArcInfo are as follows:
a) the lack of any functionality to assess the effectiveness and accuracy of the interpolated dataset to the original input;
b) the interpolation techniques themselves are relatively weak compared to alternative statistical interpolation procedures (for example, there is a basic assumption that the datasets are single populations, there are no methods for assessing possible trends in the data, and there are no functions to handle non-linear sampling);
c) there are no tools within the standard Arc environment to statistically evaluate the input dataset and to help the user assess suitable inputs to variables in the existing interpolation functions. The fact that these routines do provide credible results at time is probably a function of the general characteristics of the dataset.
The POINTGRID function is virtually useless in this case as it has been designed to deal with integer values. Where a cell contains several points with precision to several decimal places, such as the caesium values in this dataset, the value assigned to the output cell is virtually a random selection of one of those points. It would be possible to generate a lattice which is sufficiently fine to capture all the sampling points. This lattice could then be resampled within GRID. However, in cases such as this, data are being interpolated more than once, thus distancing the final value with the original measurement. IDW produces output that can be related to the input dataset. Unfortunately, the cell values, as a result of the underlying statistical expression results in a smoothing of the output dataset. Extreme upper and lower values are removed. IDW can be used to interpolate output in areas of low sampling density using appropriate radius values (i.e. to bring data points from a location removed from the grid cell). SPLINE produces very complex surfaces as the effect of all points are taken into account. In addition, there is a degree of 'black- box processing' as the user has to set a number of variables within the routine which influences the surface being produced and in turn reflects the input dataset. However, there are no tools to help the user assess the possible variability of the dataset in terms of the function. In the same vein, Kriging has too may variables to set without the user being given adequate tools to assess the dataset in light of these requirements.
Tin produces a surface that no trends in the data and assumes linear relationships between the sampling points. In many instances, the results from TIN produces acceptable maps. However, these maps appear to have the highest accuracy where the sampling density is relatively dense. Again, there are too many variable which the user has to define on a qualitative basis. An additional problem that affects some of the datasets for Europe is the TINCONTOUR function cannot handle isolines which are composed of large numbers of arcs. In these cases, the TIN model is converted to a grid using TINLATTICE and the conventional contouring routines with ARC are applied (see previous section and comments below). The consequence of this action is that the data are subjected to multiple interpolations (i.e once from TIN and once from the griding routine).
Regarding the production of isolines, there are profound difficulties with all the techniques. Both LATTICECONTOUR and TINCONTOUR use a series of baselines and interval functions to define the isolines. This works well with datasets that use only integer values for the specification of the isolines or if the sub-division is linear (i.e. one isoline every 100 intervals). The production of non-regular isolines, such as from a logarithmic scale, is not feasible without the use of AML programs to run the routines several times with a large interval and the baseline value corresponding to the required isoline.
The major problem with LATTICEPOLY is the that the caesium level is attached to the polygon labels and not to the arcs.
An important element of the project is the presentation of the final results of the analysis. The deposition data will be combined with other thematic datasets (e.g. roads, rivers, political boundaries) to produce a series of plot files that will form the basis of these maps, it is crucial that the output is of the highest quality and that all steps in the generation of the plots can be controlled.
For this aspect, the JRC is collaborating with Lovell Johns Ltd. This is a British company with a long tradition of publishing maps and atlases and now specialising in digital cartography. Lovell Johns have developed an innovative routine to transfer ArcInfo coverages into an Apple Macintosh Desk Top Publishing (DTP) environment such as Freehand for the generation of high quality value-added graphics prior to final printing.
Briefly, the methodology works in the following manner. Topologically organised coverages containing polygons coded with a predefined item value in the PAT or coded arcs with appropriare attributes in the AAT of the coverage are generated. An attribute based feature-sorting Arc Macro Language (AML) program is used to split the arcs of the dataset into coverages of a given value. For example, all arcs with an identifier of '1', would be assigned to a new coverage 1. This would allow optimization of a given identifier layer suitable for transfer to the AppleMac environment.
Each coverage would then be appended together for transfer to the Mac as filled polygons in Postscript format. This transfer is achieved using the Lovell Johns MapScript software. Once the in the Mac DTP environment, the identification on the arcs appears as individual layers, which can be arranged appropriately. Once in Freehand, the information can be manipulated as required in a very user friendly and powerful way with the use of mouse-based operations. Polygons can be filled, lines can be modified where necessary, text can be placed on the map or as supporting documentation which is embedded in the graphics.
On completion of the pre-printing stage, the files will be sent to a filmwriter to produce the mastercopies for the printing stage.
It is proposed that the maps of caesium deposition will be compiled in a 'conventional' book-format publication. In addition, the JRC is investigating the possibility of making such digital datasets available on a CD-ROM. Such a development would permit users to use the published maps as inputs into their studies or models.
In addition, the JRC is planning to develop a user friendly menu-based interface to the data held in ArcInfo by means of the AML. Menus provide a simple to use and highly visual means of providing a customised front-end to specific projects such as this. An initial draft of the interface is being developed.
An interesting approach is to make the datasets available to a wider audience through a combination of networking technologies, which are widely available, together with the analytical power of ArcInfo. Through the World Wide Web (WWW), easy access to any kind of information has become reality. The vision of the World Wide Web is a collection of programs that can understand the numerous different information-retrieval protocols (e.g. FTP, Telnet, WAIS, Gopher) currently in use on the Internet as well as the data formats of those protocols (e.g. ASCII, GIF, JPEG, MPEG, Postscript,DVI, TeXinfo) and provide a single consistent user-interface to them all. In addition, these programs would understand a new protocol (HTTP) and a new data format, the Hyper Text Mark-up Language (HTML), which are both geared toward hypertext and hypermedia. Hypertext is a term that describes a computer interface to text which allows information to be cross-referenced by clicking with a mouse on a 'highlighted' cross-referenced phrase. This action would bring up the document at the 'other end' of the cross-reference. Hypermedia is the extension of this to include graphics and audio as themes which can be selected or viewed. The effective use of the HTTP protocol, the HTML language and implementations of suitable client/server systems, such as NCSA's Mosaic or the Netscape Navigator, have provided sufficient ground to reconstruct a similar information system based on such a client/server architecture with only slightly reduced functionality over the AML interface. Such an interface could allow users to view various datasets, query the underlying data and generate processed new datasets from within the environment of the network browser.
For further information on this approach, refer to the paper by Van Liedekerke and Jones in these proceedings.
The paper presents a summary of the work to-date on the production of a series of maps that will show the caesium contamination from the Chernobyl nuclear power plant accident for the whole of Europe. The paper describes and discuss the methodology used to generate isolines using functions contained within ArcInfo's TIN and GRID modules. Initial experimentation on a sample of the deposition data at the JRC suggested that these techniques produce credible but varying results. These maps of caesium contamination based on these tools are currently being reviewed and validated by appropriate experts.
Although all the methods described above produce output, there are a number of fundamental problems . Basically, the core ArcInfo environment lacks functionality to assess the effectiveness and accuracy of the interpolated dataset to the original input, the interpolation techniques themselves are relatively weak when addressing complex populations and there are no tools to statistically evaluate the input dataset in order to provide meaningful values to the variables demanded by some Arc functions. A possible approach is to use an external statistical package to provide the required interpolation and re-import the data into Arc for the production of the required graphics.
In summary, the paper describes an approach being developed at the JRC for the direect access of ArcInfo datasets (i.e coverages, tables and graphics) directly from a network browser. Such an approach has exciting possibilities for the publication of information on a world wide basis.
Arwyn Jones, Marc Van Liedekerke
Commission of the European Communities
Joint Research Centre
Environment Institute
21020 Ispra (Va)
Italy
Tel: +39 332 789162
Fax: +39 332 789256
arwyn.jones@jrc.it
Alan Cooper
Lovell Johns Ltd
St Asaph Business Park
St Asaph, Clwyd
UK
Tel: +44 1745 582248
Fax: +44 1745 582592