Cartographic Modelling
Cartographic Modelling
CARTOGRAPHIC MODELLING
TABLE OF CONTENTS
1 Introduction....................................................................................................................................... 2 2 What is cartographic modelling?.......................................................................................................3 2.1 Modelling and model inversion................................................................................................. 4 2.2 Modelling with spatial data layers............................................................................................. 5 2.3 Map algebra modelling conventions.........................................................................................! e"#utorial $%ercise 1................................................................................................................... & 2.3 'atural language........................................................................................................................ & 3 (artographic models and )I*............................................................................................................+ 3.1 Implementing a cartographic model.......................................................................................... + e"#utorial $%ercise 2................................................................................................................. 12 e"#utorial $%ercise 3................................................................................................................. 12 3.2 (onditions ,or applying cartographic models.......................................................................... 12 3.3 -le%ible modelling in practice................................................................................................. 14 4 . ,inal word.....................................................................................................................................15 /eriodic *el, .ssessed $%ercise 1.............................................................................................10
CARTOGRAPHIC MODELLING
1 Introduction
This section looks at Cartographic Modelling. Fortunately, the concepts are not as frightening as the terminology and the application of cartographic models is delivering benefits to spatial analysts worldwide. Why? ecause cartographic modelling offers a rigourous procedure for integrating map layers and designing analysis schema for spatial data !Figure "#. This section will re$uire you to demonstrate some straightforward logical reasoning % and a basic understanding of mathematics % as we work through the basic methods of cartographic modelling. Figure ". &'ample of applying a cartographic model to calculate the percentage change in value between two dates.
(ource) erry !*++,# -t is difficult to be precise about the origin of cartographic modelling because it is derived from a collection of old ideas, such as sieve mapping and -an Mc.arg/s Design with Nature !"010#, that have been organised, augmented and e'pressed in terms amenable to digital data processing. -t is, however, the work described in Tomlin !"023# as 4Map 5lgebra6 and erry !"027# as 4Map8ematics6 which established cartographic modelling as an accepted methodology for the processing of spatial data. Furthermore, by e'panding map algebra as a high8level /natural/ computational language cartographic modelling 1 2225 'igel #rodd
3 researchers have greatly influenced the 9-( and geocomputation communities by, for e'ample, adding time, dynamic processing and the third dimension to the traditionally static, two8dimensional spatial data analysis. 5s you develop your interest in the topic there is a bibliography in the resources which provides reference to papers on cartographic modelling.
:ana Tomlin
;oseph erry
4 addition, many of these provide toolkits to build models that draw heavily on map algebra, where map algebra is a formal language for analysing spatial data sets. This includes -drisi, 5rc9-( and FM&. 5 cartographical model is a graphic representation. -t uses pictures linked by arrows in a flow chart. -ts purpose is to help the analyst organiCe the necessary procedures as well as identify all the data needed for the study. -t also serves as a source of documentation and reference for the analysis. -n this way it can support communication between colleagues and contribute to the metadata that accompanies the outputs from spatial analysis.
5 of the phenomena of interest. Wouldn/t it be great if we knew what the weather would be like ne't weekend or in France ne't summer so that we could choose the best two weeks for a holiday. -f we built a model that shows how different factors % local and global 8 influence weather patterns then we could predict what is likely to happen and also what might happen if there were changes in one or more factor % the classic /What if?/ simulations. 5n alternative use of models is to strive to achieve a predetermined obDective. -n forward modelling we establish the relationship between an input factor and the output and can then re8run the model using different input values. -n a model where 5 E see that if we increase the value of we might vary F C we can by * then the new value of 5 !denoted 5/# will
increase by * i.e. 5/ E ! F *# F C. The alternative use is to fi' 5 and then e'amine how and C to achieve that obDective. For e'ample, we might seek to maintain a minimum discharge from our river catchment at all times. 5s we cannot control the flow of water directly we could do this by making changes to one or more of the factors that influence discharge. (ome things we cannot change % rainfall, catchment siCe and relief % but others we can. For e'ample, we could test what changes in vegetation cover would have to be made to maintain a more uniform flow, or investigate the possibility of increasing the siCe of a lake e.g. by dredging or building a dam. -f we were to introduce additional criteria, such as cost, we could attempt to identify an optimal solution based on multiple criteria. y focusing our analysis on changes to the input factors we have inverted the model and these type of applications are known as /model inversions/. We have learned that modelling can support /simulations/ and /optimisations/ and here lies the real strength of cartographic modelling as a method of analysis. Bf course, before we get carried away you must remember that the accuracy of your predictions will depend greatly on the $uality of your original model of reality and the $uality of the data that you use to populate the model.
! and proceed backwards in a step by step manner toward the e'isting data. This process guards against the tendency to let the available data shape the final product. The procedure begins with the definition of the final product. What information is re$uired? What values will the product have? What will those values represent? We then ask what data are necessary to produce the final product and define each of these data inputs and how they might be measured or derived. The following e'ample illustrates the process)
(uppose we wish to produce a final product that shows those areas with slopes greater than *+ degrees. What data are necessary to produce such a map?
To produce a map of slopes greater than *+ degrees, we will first need a map of all slopes. -s a map of all slopes present in our database? -f not, we take one step further back and ask another $uestion) what data are necessary to produce a map of all slopes?
5n elevation layer can be used to create a slope map. :oes an elevation layer e'ist in our database? -f not, what data are necessary to derive it? The process continues until we arrive at e'isting !raw# data. The e'isting data may already be in digital form, or may be in the form of paper maps or tables that will need to be digitiCed. -f the necessary data are not available, you may need to develop a way to use other data layers or combinations of data layers as substitutes.
Bnce you have the cartographic model worked out, you may then proceed to invert the model, run the inverted model and generate the output data layers.
0 as the normal range of algorithms. Bnce we have built our model then we need to invert the model before we apply it. The computations are made for each independent element according to mathematical convention. Iet us consider a model with one dependent term, 5, and 3 independent terms, , C and :) 5E F !C J :#
and we make calculations in the following order) multiply C and : to generate the first element add the first element to !the second element#
-t is important to respect the priorities given by use of brackets and the fact that some arithmetic functions might not be commutative e.g. 5 8 does not e$ual 8 5.
What are some of the conventions in cartographic modelling? &ach different data file type is represented by a different shape !Figure *#. ?aster layers are represented by rectangles, vector layers by triangles, values files by ovals, and tabular data by a page with the corner turned down. File names are written inside the symbol. Bperations are shown as parallellograms and linked to input and output data layers with arrows. When an operation re$uires the input of two layers, the arrows from those two layers are Doined, with a single arrow pointing to the operation !Figure 3#. Figure *. :ata file types.
Figure 3 shows a model in which two raster layers, administrative areas and population, are used with an overlay operation and a division calculation to produce a new raster layer of population density. The original model is that population density is a function of population and area. Figure 3 illustrates the inverted model that is used to make the calculation. Figure 3 5 cartographic model to generate population density.
&
.ow would the graphics differ if the input and output layers were in vector format? $'Tutoria% E($rci $ 1 Time) "K minutes Construct a cartographic model of the slopes e'ample in sub8 section *.". The obDective is to map /steep/ slopes and you can assume that an elevation map e'ists in the database. .int) first construct a model of the relationship between steep slopes, all slopes and elevation and then invert the model to show how you would make the calculations. 5n answer is given in the 5nswers to e8Tutorials. -n the same way as algebraic operations can be combined to form a comple' e$uation, cartographic models can be constructed to model comple' spatial relationships. developing a cartographic model. ut before we consider these procedures we will look at how natural language is used in
+ representing a map layer, to create obDect nouns, or new map layers. ?eferring to the e'ample in Figure 3, a population map !subDect noun "# is overlaid !verb# on a map of administrative areas !subDect noun *# to create a population density map !obDect noun#. Table " lists some verbs of spatial operations used in Tomlin6s approach. Aote that, unfortunately, not everyone adopts the same terminology and you should be alert to differences between analysts. These inconsistencies and ambiguities are also, regrettably, prevalent in different 9-( software. Table ". &'amples of natural language verbs. Operation a linear data set -ntersect two polygon BL&?I5G networks (elect according to a &MT?5CT condition Verb Description ?enumber all loci with a value reflecting their distance from a given starting point or line Iay two polygon networks over each other and produce new polygon net (elect specified values and @ or ranges of values from one layer to make a new layer Make a corridor from (>?&5:
12 siteOstatus riverOmap roadsOmap (tage two is completed by describing, in natural language, a scheme of spatial operations re$uired to identify potential sites for the supermarket. Figure , shows how stage three is completed by forming a flow chart to represent the logic in a 9-( proDect. -t is sometimes easier to visualise this with thumbnails of the data layers !Figure K#. Aote how the appropriate verb !describing a spatial operation# has been added from Table ". The e$uation numbers, &$", &$*,... &$2, relate to e$uations in Table *. Table * presents four of the e$uations it would be necessary to solve as part of the process of finding a suitable site for the supermarket.
Figure ,. Flowchart of the operations needed to create a map identifying suitable locations for a supermarket. Figure K. Flow chart with thumbnails.
11
Table *. 5lgebraic e$uations from Figure K. From I5A:ON(& 4e'tract6 ?&(-:&AT-5I &$ " a % b E c where) a E landOuse map b E non residential Cone c E residential From (-T&O(T5TN( 4 e'tract6 FB?O(5I& &$ * d % e E f where) d E siteOstatus map e E sites not for sale f E sites for sale 4Bverlay6 ?&(-:&AT-5I and FB?O(5I& &$ 3 c J f E g where) g E residential sites for sale
From the above e'ercise it should be apparent that the analytical power of cartographic modelling lies in the ability to combine a series of e$uations by using the results obtained from one e$uation as the input for the ne't. -n this way a comple' spatial problem can be tackled by breaking down the individual components of the problem into a series of
12 $'Tutoria% E($rci $ 2 Time) 3+ minutes ?eferring to Figure 3 and Table * complete the missing e$uations for steps ,, K, 1, 7 and 2 in the modelling process. 5nswers are given in the 5nswers to e8Tutorials. smaller solvable e$uations. The final stage in the modelling process is to annotate the flow chart with the appropriate commands from the 9-( package in which it is intended to perform the analysis. -n our e'ample these commands are represented by verbs from Tomlin6s Map 5nalysis >ackage !&MT?5CT, BL&?I5G and (>?&5:#. 5s was mentioned previously, however, different software applications translate these generic statements into their own terminology. Whilst making the running of training courses more profitable this localised terminology does nothing for spatial analysis and makes learning a new package more time consuming. These are trivial issues, however, compared to the benefit to be gained from implementing cartographic models. $'Tutoria% E($rci $ ) Time) 3+ minutes 5nnotate the flow chart in Figure * with the appropriate commands from a 9-( software package you are familiar with. For this e'ercise you should assume the available sites are represented as area entities.
13 raster and vector data layers in a seamless analytical framework. This is because rasterisation and vectorisation are not clean, they might add something or change something in a data layer, and the effects of conversion are unpredictable. (o, - am still re$uired to prepare data in a consistent data structure. -n general, there are more spatial operators coded to process raster data than vector data and so the raster data model may be preferred by data analysts. The second condition, the iterative processing environment, logically se$uences map analysis operations and involves the following) ". ?etrieving one or more map layer from a database, *. >rocessing the data as specified by users, 3. Creating a new map containing the processing results, ,. (toring the new map for subse$uent processing. -t is important that each new map is automatically georegistered to the other maps in the database. The output from one operation can then form the input to a later stage of processing. (uch cyclical processing provides a fle'ible structure similar to Pevaluating nested parentheticalsP in traditional math. Within this structure, you first define the values for each variable and then solve the e$uation by performing the mathematical operations on the values in the order prescribed by the e$uation. For e'ample, the e$uation for calculating percentage change in population starts with population at time + !M# and time +F" !G#. The difference is calculated by subtraction and then stored as an intermediate solution. The intermediate solution is divided by the initial population !M# to generate another intermediate solution that, in turn, is multiplied by "++ to calculate the solution !5# 8 the percentage change value. -nput values) M, G -ntermediate solutions 5bsolute change in value) MGdiff E G % M >roportionate change in value) MGproportion E MGdiff @ M (olution) 5 E MGproportion ' "++Q
The same mathematical structure provides the framework for computer8assisted map 1 2225 'igel #rodd
14 analysis. The only difference is that the variables are represented by mapped data composed of thousands of georeferenced values. Figure 1 shows a similar solution for calculating the pattern of percentage change in animal activitybut the calculations are performed for each grid cell in the study area. The result is a map that identifies the percent change at each location. Figure 1. 5n iterative processing environment, analogous to basic math, is used to derive new map variables !red tones indicate decreased animal activityR green tones indicate increased activityR the e'ample location shows a 2.K"Q decrease#.
Map analysis $uantifies the nature of change i.e. magnitude and direction of change in the thematic attribute, and the location of change using the spatial attribute. The characteriCation of PwhatP and PwhereP provides information needed for further environmental analysis, such as determining if areas of large increases in animal activity are correlated with particular cover types or near areas of low human activity.
15 social concerns e.g. pro'imity to high housing density, visual e'posure to houses, as well as purely engineering concerns e.g. steep slopes, water bodies. Combining physical and socio8economic concerns as part of an integrated spatial solution is another significant benefit. Furthermore, the ability to simulate various scenarios e.g. steepness is twice as important as visual e'posure, and pro'imity to housing is four times more important than all other considerations, provides an opportunity to embed geospatial information into the decision8making process. y noting how often and where the proposed route changes as successive runs are made under varying assumptions, information on the uni$ue sensitivity to siting a highway in a particular locale is described. Compare this to a non8model based planning process. -n the old environment, decision makers attempted to interpret results bounded by vague assumptions and system e'pressions of a specialist. Cartographic modelling, however, engages decision makers in an analytic process, because it documents the thought process and encourages interaction. -t is the e$uivalent of a Pspatial spreadsheetP that encapsulates the spatial reasoning of a problem and solves it using digital map variables.
* A +ina% ,ord
What we hope you will have discovered from this section is that cartographic modelling provides a structured approach to 9-( design. 5t present the biggest drawback to cartographic modelling is the lack of standards in the use of algebraic construct and natural language terminology between 9-(. Therefore, developing a cartographic model in -drisi, for e'ample, re$uires a different set of statements in a different order to those necessary to perform the same operation in another raster 9-(. Gou could probably perform a similar analysis using a vector based 9-(, but it might take many more steps and re$uire a different set of commands. :espite these problems cartographic modelling is probably one of the key areas of 9-( which will be refined by future developers. Bne of the reasons for this is the relentless desire to converge 9-( with mainstream -T. -t should not go unnoticed that up until now many 9-( analysts have done their own thing... and not neccessily adopted industry standard development tools and procedures. -n the past decade, Nnified Modelling Ianguage !NMI# has emerged as the software 1 2225 'igel #rodd
1! blueprint language for analysts, designers, and programmers alike. -t is now part of the software trade. The NMI gives everyone from business analyst to designer to programmer a common vocabulary to talk about software design. 5s the orland software house e'plains) /$%& is applica le to o 'ect(oriented pro lem solving. )nyone interested in learning $%& must e familiar with the underlying tenet of o 'ect(oriented pro lem solving ( it all egins with the construction of a model. ) model is an a straction of the underlying pro lem. The domain is the actual world from which the pro lem comes/. 5t the center of NMI are its nine kinds of modelling diagrams and several of these diagrams provide a ready8made framework that can be mapped onto cartographical modelling !Table 3#. y learning a new vocabulary 9-( developers might find it easier to integrate with other information technologies. Table 3. 5nalogies between NMI and cartographical modelling. Diagram Relationship to cartographical modelling Nse case diagrams Class diagrams describe what a system does. its classes and the relationships among them BbDect diagrams show instances instead of classes. They are useful for e'plaining small pieces with complicated relationships (e$uence diagrams Collaboration diagrams (tatechart diagrams 5ctivity diagrams show how operations are carried out in time convey the same information as se$uence diagrams, but they focus on obDect roles show the states of the obDect and the transitions that cause a change in state focus on the flow of activities involved in a single process 1 2225 'igel #rodd -terative processing environment Flow chart Flow chart -dentifies obDectives cartographic model give an overview of a system by showing The conceptual
10 Diagram Relationship to cartographical modelling Component diagrams are physical analogs of class diagram -mplementation in a 9-( software and hardware :eployment diagrams show the physical configurations of
$d E($rci $ 1
". &'plain the relationship between cartographical modelling and map algebra. Which authors have been responsible for developments in these areas? *. What is natural language? Why is it important in processing spatial data? 3. What are the four steps re$uired to implement a cartographic model in a 9-( software application? ,. -dentify three benefits of using cartographical modelling in 9-(. Iook back through the te't to check your answers.