Publish or perish

I guess the old academic adage of “publish or perish” still holds true, although some appear to perish despite the publish, as I have written about before, in my post on the Catch 22 of academic publishing.  Obviously, fame doesn’t come instantly, but good things come to those who wait.  In this case it was Christopher Mason of  Hyder Consulting who dug up what I had done and published, and almost forgotten, so I guess I owe Chris a big one for not letting me perish.

My favorite

Christopher didn’t just dig me up. His main literature source is actually one of my favorites when it comes to transportation vulnerability, it’s Katja Berdica’s 2002 article on road vulnerability, the paper that got me started  in this direction and that I reviewed on this blog some time ago.

Terms and definitions

In the report, Mason lists the different terms and definitions that Berdica and I use in our approaches towards vulnerability:

Looking back it’s interesting how I defined reliable as able to operate under any circumstance, while I saw vulnerable as not able to operate under certain circumstances. Not a very sophisticated definition, I must admit that, but it did the trick, apparently.

Shocks aka risks aka vulnerabilities

While Berdica sort of has the upper hand on the definitions, when it comes to what can affect transportation networks adversely, Mason  uses my three types of risks and adds a forth type of accidents and attacks:

• Structure-related (the way the transport network is built and the attributes of the network itself, its physical characteristics such as geometry, width, curvature, gradient, weight restrictions, height restrictions, presence of bridges, tunnels)
• Nature-related (topography and the terrain the road traverses and natural incidents such as floods, snow, ice, fog, climate change and so on)
• Traffic-related (traffic flow and attributes causing change such as peak periods, special events, maintenance operations, accidents)
• Other (terrorism, international events, protests, industrial incidents)

The fourth “shock” as Mason calls it, was added in by him, while the first three are taken directly from a paper I presented at three different conferences in late 2004 and early 2005. Said paper was titled reliability and vulnerability versus benefits and costs and discussed the notion that reliability and vulnerability apparently do play a major role in transportation, yet are seemingly omitted from cost-benefit analyses of transportation projects, something I intended to rectify in my research. I never did, but that’s another story. That said, Mason makes extensive use of my three categories of risks together with his fourth category in determining the transportation hazards that the UK road network may be exposed to.

Resilience factors

Another interesting takeaway from Mason’s report are the 10 factor’s that make up transportation resilience, taken from a 2006 article by Pamela Murray-Tuite:

• Redundancy – the transport system contains a number of functionally similar components which can serve the same purpose and hence the system does not fail when one component fails (for example, a number of similar routes are available with spare capacity).
• Diversity – the transport system contains a number of functionally different components in order to protect the system against various threats (for example, alternative modes of transport are available).
• Environmental Efficiency – a transport system which is environmentally efficient will be more sustainable, and capacity is less likely to be constrained due to environmental reasons.
• Autonomy – the components of the transport system are able to operate independently so that the failure of one component does not cause others to fail (for example, can the transport system operate safely in the event of a power cut?).
• Strength – the transport systems ability to withstand an incident (for example, how extreme a flood event can the system cope with?).
• Adaptability – or flexibility, can the transport system adapt to change and does it have the capacity to learn from experience (for example, an area-wide traffic management system can adapt to differing traffic conditions).
• Collaboration – information and resources are shared among components and/or stakeholders (for example, contingency plans in the event of an emergency and the ability to communicate with system users).
• Mobility – travellers are able to reach their chosen destinations at an acceptable level of service.
• Safety – the transport system does not harm its users or expose them, unduly, to hazards.
• Recovery – the transport system has the ability to recover quickly to an acceptable level of service with minimal outside assistance after an incident occurs.

Compare these to the resilience taxonomy developed by Andrew Cox et al. (2011). That is a very different picture, but there are still many commonalities. Personally, I prefer this framework here. Mason uses these factors to identify and map vulnerabilities and hotspots in the transportation network, and I am really impressed by the level of detail in the report.

Google Scholar

I would never have known about my rise to fame were it not for that I from time to time use Google Scholar to check for articles referencing my works. I like Google Scholar, because it gives me a much wider range of results, as already noted by  Howland et al. (2009). I must admit that I do like to see who uses what parts of my work and where. Firstly, because I may find further ideas for expanding my work, based on what others have done with it. Secondly, because I may find more people to connect with, and many of my LinkedIn connections were made after “stumbling upon myself “, so to speak.

Reference

Mason, C. (2010) Network Resilience and Adaptation – Phase 1 Final Report. Report no.  3003-GD31094-GDF-02, Hyder Consulting.

Author link

• linkedin.com: Christopher Mason

Related link

• hyderconsulting.com: “Outstanding people – exceptional solutions”

Download

• eeda.org.uk: Network Resilience and Adaptation (pdf)
  File size: 13MB. Right-Click and Save-As may better than just clicking the link.
• husdal.com: Literature review (4 pages from the above report)

Related posts

• husdal.com: Reliability and vulnerability versus costs and benefits
• husdal.com: Road vulnerability as defined by Berdica

If you can’t use your head, use your overhead

If you, like me, went to school in the late 1980s, you will probably still remember teachers using overhead projectors showing copies of textbook pages with no or little images and reading them verbatim rather than trying to explain things properly? I’m sure the three guys  behind this book do not fall into that category, and I know I’d love to be in their classroom. This book fully explains the concepts of transportation, spatial interaction, accessibility, location theory, graph theory and what not better than anything I have seen. Why? because it is so well and to-the-point illustrated. Hardly any concept related to transportation is  left out, and hardly any mentioned concept is not well illustrated with rich examples, i.e. self-explanatory figures.

Transportation and supply chains

Transportation networks are the main arteries of society, ensuring a 24/7 operation of  the community we live in. Transportation systems are also essential to supply chains. Without a proper transportation system supply chains cannot function properly. However,  the menial day-to-day operations related to transportation are often overlooked in supply chain management. This book goes a long way in bridging this gap, as it touches upon topics such as freight transport and locations of terminals. It also extends the term supply chain by introducing a new term: the commodity chain.

Transportation and GIS

If you have read my About Me page, you will probably realize that my fascination for this book stems from my background, as I hold a degree in Geographic Information Systems GIS), focussing on various applications of GIS in transportation (GIS-T). GIS-T only takes up a small part of the book, but it is an excellent chapter on GIS in transportation, as it is co-written by Shih-Lung Shaw, who also co-wrote a book wholly dedicated to GIS-T together with Harvey Miller, one of my professors from the University of Utah.

Companion website

The book comes with a companion website that shows you the main contents of the book. This is where you can see first-hand how brilliantly it is illustrated.  If that doesn’t convince you to buy the book, I really don’t know what would. And while most of the illustrations in the book are grayscale only, on the website they are all in full color.

Reference

Rodrigue, J.P., Comtois, C. & Slack, B. (2009) . The Geography of Transport Systems , 2nd edition. New York: Routledge

Author links

• hofstra.edu: Jean-Paul Rodrigue
• umontreal.ca: Claude Comtois
• concordia.ca: Brian Slack

amazon.com

• Buy the book: The Geography of Transport Systems

Related

• husdal.com: GIS for Transportation
• husdal.com: Corridor analysis
• husdal.com: How to make a straight line square

Showcase after showcase

The book is published by ESRI Press, and describes very well what you can do with ArcView GIS, in showcase after showcase, with stunning illustrations. However, there is not much analysis or explanations as to how the results came about.

For a GIS Analyst working behind the computer scenes this book is almost worthless, for the GIS Manager this book is almost priceless, showing in detail all that can be done (and that you may want to ask your GIS Analyst to do).

One big commercial

After reading the book I feel like having watched an ArcView commercial. I’m thrilled at the possibilities, but did I learn how to use GIS for transportation? No.

Reference

Lang, L. (1999) Transportation GIS. ESRI Press.

Author links

• JMR Consulting: Laura Lang

amazon.com

• Buy this book at amazon: Transportation GIS

Related

• husdal.com: Book Review: GIS for Transportation – Principles and Applications
• husdal.com: GPS for Vehicle Routing and Tracking
• husdal.com: GIS for Transportation

Solid

Mind you, this is not for the fainthearted, this is solid academic work and presumes some academic knowlegde prior to reading this book. It is specked with references that are hard to get, and you are likely to spend more time in the library reading up on the bibliography than digesting the actual text. Still, if GIS-T is your line of research, you cannot avoid having this book. It is by far one of the most comprehensive I have seen. It is clear that the authors posess solid knowledge and have covered a wide field and left nothing out. It may have a rather inhibitive price; in hindsight it was well worth the money spent.

Reference

Miller, H., Shaw, S.L. (2001). Geographic Information Systems for Transportation. New York: Oxford University Press.

Author links

• geog.utah.edu: Harvey Miller
• utk.edu: Shih Lung Shaw

amazon.com

• Buy this book at amazon.com: GIS for Transportation

Related

• husdal.com: Book review: GIS for Transportation

Arcview Network Analyst

The ArcView Network Analyst (AVNA) extension module allows the user to solve 3 categories of network analysis problems; Find Best Route, Find Closest Facility and Find Service Area. Find Best Route problems involve finding the “least cost impedance” path on the network between two or more stops. Find Closest Facility pertains to finding the distances from an event to the nearest facilities, or vice versa, finding the distance from a facility to one or more events. Find Service Area determines the area that a particular facility can serve within a given time or cost frame.

Link

• The Tutorial: ArcView Network Analyst

JavalancheTM
an application for analyzing hazards next to roads
Jan Husdal, University of Utah, 2002.

View the Javadoc documentation here

Introduction

Java was chosen as programming environment not only because of the author’s familiarity with this particular programming language, but also because it allowed for quick conversion of the analysis methodology into programming steps and because it would allow for a later extension of the application within a Windows or GUI framework. In essence, this is a standalone application, albeit it has the potential for interfacing with a commercially available GIS. The application is designed for input/output from/to raster GIS. In this case, MFworks was used as the chosen GIS software for output and illustration. The application imports a raster surface as a tab-delimited text file, applies the analysis methods, and exports raster surface as a tab-delimited text file for import into a raster GIS for visualization of the result. This application was designed in a typical bottom-up manner, yet still trying to preserve a top-down perspective by allowing for future improvements and additions.

Purpose of application

Background for developing application

The purpose of this application is assisting in analyzing the threat of natural hazards to roads.

Traditionally, in studying the effect of hazards on roads, a hazard map is prepared based on the hazard in question, the contributing factors and then overlaid with a road map. If the road or a buffer around its vicinity intersects hazard areas, these areas constitute a potential threat.

In the approach used in this procedure, imagine traveling along the road and looking to either side for hazards. The neighborhood that needs to be examined is dependent on the hazard and its contributing factors. This neighborhood search then produces a hazard map that is directly related to the road.

The main conceptual difference with this approach as opposed to the overlay method is the use of a variable-shaped hazard neighborhood that is a function of location along the road. A traditional buffer is a fixed neighborhood, whereas this buffer is dependent on both hazard and location. A fixed neighborhood can be really unsatisfying in some cases because it might look too far and thus overestimate the hazard threat or simply not look far enough and thus miss some actual hazards.

The application to be developed should thus be general enough that it with simple steps can incorporate a number of natural hazards, all based on the neighborhood search approach. Currently the application is built for analyzing avalanche-prone terrain near a road. However, the application is structured such that other hazards can be analyzed without totally redesigning the application, videlicet by just adding another module.

Using the application for analyzing avalanche threat

The procedure for finding the potential for avalanche hazard to a given road segment is based on the assumption that snow, similar to water, will choose the path of least resistance from its initial starting point above the road, and then slide towards the road. Thus, the presence of avalanche terrain within a watershed, or basin, adjacent to a road segment can serve as an indicator for the avalanche threat pertaining to that road segment.

In short, the generic description of the procedure is as follows: With the road or buffer line being a line of neighboring individual cells,

For every cell along the road, determine all upstream cells that make out a basin adjacent to the road line itself or a buffer at a certain distance form the road.
For every basin, determine the “amount” of avalanche threat based on slope, vegetation and other contributing factors within the basin.
For every basin, determine the stretch of road it affects.
For every affected stretch of road, determine the threat potential based on the “amount” of threat in the basin that is linked to this particular stretch of road.

Usage

The application was put to use, together with MFWorks as the GIS software, for developing a spatial framework for analyzing hazards to transportation lifelines. The result was presented at AAG 2002.

Conceptual workflow within the application

Input

The following input is required as tab or comma delimited text files:
1) DEM surface,
2) slope surface *
3) road line or buffer edge, and
4) non-vegetated areas.

* The slope surface can easily be derived within the application, but is in this case found via commercially available GIS.

Bear in mind that every cell in the surface is assigned a unique identifier. By our convention, this is done by counting from top left to bottom right, row by row, with 1 as the first cell ID. This cell ID makes it easy to relate cell positions in different surfaces to each other.

Step by step

The text files are read into the application and converted to Raster objects, on which the analysis is conducted.

First, derive the flow direction from DEM. The flow direction follows the convention of Tomlin’s directional identifiers and this information is later used in delineating the upstream basins adjacent to the road line/ buffer edge.

A unique ID identifies each road/ buffer cell. Being the presumably lowest cell within a potential basin, this unique cell ID also denotes the basin ID for the basin adjacent to this particular road/buffer cell.

Derive the upstream basin for every road/buffer, starting with the lowest cells within the road or the buffer and working your way up the hill. Every cell in the basin is identified by its unique cell ID. Now, every cell in every basin is known. The basin ID coincides with the cell ID of the road/buffer cell it is adjacent to. The basin cells cell ID makes it possible to link every basin cell or other properties such as slope, vegetation, risk etc.

Based on slope and non-vegetation, the risk potential for every cell is calculated and given a value between 0 and 1, as described in the procedure. With the risk for every cell known, the total risk for every basin is calculated.

Based on the basin ID, the lowest basin cell is found, and the downstream cells from this point make out a potential runout zone. The runout zone adjacent to the road delineates the possible impact area of the upstream basin, and the road segment is assigned a risk value based on the risk found in the upstream basin.

Structure of application

Overview

In principle, the application is divided into 4 parts:
the main class, RasterAnalysis, which runs the application, and accesses methods in
the object class Raster and
the analysis modules, FlowDirection, BasinModule, RiskModule;
the helper classes for file and keyboard handling, KeyboardInput, RWFile, FileIO, and
the visualization module, GraphicsModule, which is to ensure correct output.

* The flow direction could have been found with commercially available GIS and then used as an additional input file. However, for the learning experience, we chose to develop this class as part of the application.

The key idea behind this structure is that the application can be extended with more modules as it is developed for different hazards.

RasterAnalysis class

The RasterAnalysis class runs the application. It creates instances of the modules and classes and invokes the necessary methods in these.

KeyboardInput class

The KeyboardInput class is essentially a helper class designed to facilitate keyboard input at the designated screen prompts. It accepts double precision, floating point, integer values and String as its input.

RWFile class

The RWFile class is a helper class that reads from and writes to a tab-delimited text file. The methods in this class convert a raster surface from a tab-delimited text file to/from a Raster object that is used in the analysis.

FileIO class

The FileIO class is similar to the RWFile class in that it reads from a text file. In this case, it is simply designed to read lines of text from an ASCII text file. This class was created to help display instructions and other information by reading it from separate text files rather than envoking it using System.out.print() within the other classes.

Raster class

The Raster class is the primary object in this application and represents the raster surface that is to be analyzed. The class contains methods for getting and setting data at a) a given cell ID, or b) a given row and column. In doing so the class also contains methods for converting a cell ID to a row-column position. In addition there are methods for setting and getting the raster dimension, which is the number of rows and columns in the raster.

FlowDirection class

The FlowDirection class takes a DEM Raster as input object and creates a Flowdirection Raster object as output. This Raster object is then later used to find upstream cells from any given cell.

BasinModule class

The BasinModule class is what this application was all about initially, videlicet delineating the basins along a road that may or may not be prone to avalanches. The basin is represented by an object of type ArrayList that contains the Cell IDs for each cell in the basin, which are then aggregated into a second ArrayList containing all basins with all individual cell IDs. This class not only contains methods for finding the basins, but also other basin related procedures, such as converting the basin(s) from ArrayList to Raster for output to a text file.

RiskModule class

The RiskModule class contains methods for handling risk-related procedures, such as converting slope values to risk values, masking out non-vegetated cells, and finally, a method for populating a Raster object with the basin cells that are associated with risk pertaining to slope and non-vegetation.

GraphicsModule class

The GraphicsModule class is not so much an analysis class, but rather a helper class designed to convert the Raster, TreeMap and ArrayList objects gained from the analysis into text files that can be imported into a GIS for visualization.

How to use the application

Input

As mentioned above, the following input is required as tab or comma delimited text files:
1) DEM surface,
2) slope surface *
3) road line or buffer edge, and
4) non-vegetated areas.

* The slope surface can easily be derived within the application, but is in this case found via commercially available GIS.

Remember that every cell in the surface is assigned a unique identifier. By our convention, this is done by counting from top left to bottom right, row by row, with 1 as the first cell ID. This cell ID makes it easy to relate cell positions in different surfaces to each other.

You may want to keep these text files in separate directories. Currently both input and output are kept in the same directory. It is not difficult, however, to assign a different directory to input and output simply by changing the working directory statements in the RasterAnalysis and GraphicsModule classes.

The raster surfaces are straight-forward tab-delimited text files, with no header information, where each row in the surface translates to one line in the text file. On import into the application these text files are converted to Raster objects, containing a 2-dimensional array [ ] [ ].

Working directory and file naming

Currently, the application is adapted to fit the directory structure on the programmer’s university network and home PC. Change this to fit your needs. You will also need to change the text file names in the RasterAnalysis and GraphicsModule classes.

Screen output

As it is, only minimal information is given as screen prompts. The Java code does contain a number of System.out.print() statements that have been “commented out”. Most of these whese were statements used for controlling that the application worked as intended and may be made visible again if desired.

Future improvements

Data Structure

The application is a bottom-up creation, and thus, the data structure evolved along with the application and may not be the most efficient. Some benefits may be gained from reassessing the utility classes used which then may eliminate some of the conversion methods in the analysis and visualization classes. An overall revision of the process may also lead to an elimination and merger of some classes.

Workflow

As modules were added on, this increased the back-and-forth interaction between the individual classes and methods. An overall revision may lead to an elimination and merger of some classes.

User interface

Currently, only minimal information is given as screen prompts, because the main aim of the application was developing a functional application that suited the understanding and needs of the analyst. If the application is to have a commercial value, the development of a user interface, either as simple command prompts or the embedding the application with a windows framework, would add considerable value. Some effort has been made though in that the FileIO class allows for reading text onto the screen where necessary. Adding more information in this manner can provide a first step at making the application more user-friendly.

Java documentation

View the Javadoc documentation here

Useful literature

Java books that have helped me:

• Beginning Java 2 – JDK 1.3 edition, Ivor Horton, Wrox Press
• Java 2 – The complete reference, Herb Schildt, McGraw-Hill Professional Publishing
• Java 2 – How to program 4th edition, Harvey Deitel, Paul Deitel, Prentice Hall

Related

• husdal.com: Analyzing hazards to transportation lifelines

GISRUK 2002

This is co-authored paper presented at GISRUK 2002, University of Sheffield, Sheffield, UK, 3-5 April 2002

Read online

Read online

Reference

Sherlock, R., Moone, P.,  and Winstanley, A. (2002) Shortest Path Computation: A Comparative Analysis. Paper presented at GISRUK 2002, University of Sheffield, Sheffield, UK, 3-5 April 2002

Javalanche

While MFworks was used as a GIS tool, the analysis and delineation of avalanche basins was done using Javalanche, a Java application developed for this purpose.

Slides

The slides below were presented at the AAG Annual Meeting, Los Angeles, 19-23 March 2002. The idea behind the research which was part of my (now discontinued) PhD in Geography was to develop a new spatial framework for modeling hazards to transportation lifelines. MFworks was used as GIS software.

The traditional approach uses a buffer  for overlay/intersection. This research uses a neighbourhood search function. The case study was done for avalanches onto roads in Little and Big Cottonwood Canyon in Utah. The method can also be applied to other spatial hazards, such as wildfire or rockfall.

The study is part of a larger research project undertaken by the National Consortium for Remote Sensing in Transportation, and the focus is to develop analytical tools and methods that can be uniformly applied to identify, map and assess a variety of hazards to transportation systems.

The traditional method for assessing spatial hazards to roads involves overlaying map layers for intersection, or by creating a buffer. However, this does not take into account the specific characteristics of the hazard, it may also lead to over- or underestimation, it does not take into account the impact on the road.

Our method uses the notion of object fields, where every location along atransportation network has a potential spill plume or watershed.

The method searches the neighbourhood from a given location and generates the watershed during the search, thus defining the area that impacts the given location.

For avalanches, the method uses the slope as the main parameter.

The avalanche basin is found by using Landsat TM data, along with USGS DEM and USGS DOQ.

The procedure is as follows: 1) For every cell along the road, determine all upstream cells that make out a basin adjacent to the road or adjacent to a buffer a ceratin distance from the road. 2) For every basin, determine the amount of avalanche threat based on slope, vegetation and other contributing factors within the basin. 3) For every basin, determine the stretch of road the particular basin effects. 4) For every affected stretch of road, determine the threat potential based on the amount of threat in the basin that is linked to this particular stretch of road.

Input: Road layer, DEM layer, NDVI layer.

Step 1: Calculate avalanche runout basins.

Step 2: Calculate risk in non-vegetated areas.

Calculate runout.

Result: Road segments affected by avalanches.

This method can be applied to other hazards., e.g. wildfires: Backtracking the spread of fire towards the road. How long will it take for a wildfire to reach the road?

This method can be applied to other hazards., e.g. roackfall: Defining runout zones from given or likely starting points.

Conclusion: The neighborhood search is a feasible alternative to the traditional buffer/intersection approach. It uses a variably shaped neighborhood instead of a fixed-shaped buffer. The neighborhood is delineated by hazrad parameters and environmental parameters, and delineates the exact stretch of road that is affected. It can be applied to a variety of hazards and a variety of infrastructure objects.

However, there are some issues: There may not be suffcient data available for each type of hazard. Remotely sensed data may not have the scale required. And hazard and/or neighborhood parameter estimation is largely based on expert judgement.

Reference

Husdal, J. (2002) A spatial framework for modeling hazards to transportation lifelines. presented at the AAG Annual Meeting, Los Angeles, 19-23 March 2002.

Related

• husdal.com: The reliability and vulnerability of transportation lifelines
• husdal.com: Javalanche

Introduction

We live in an advanced and seemingly peaceful society. We have tamed nature the best we can, but natural disasters still happen. We are surrounded by state-of-the-art safety technology, but major accidents still happen.

The ever increasing complexity of today’s modern society, coupled with its interwoven and interdependent technological infrastructure, where even a small-scale man-made or natural disaster can affect much more than just the immediate vicinity of the disaster, has given rise to the field of risk assessment, aimed at quantifying risks and their probability, and thus managing risks.

General use of risk assessment can provide a basis for preventing and limiting the consequences of accidents, thus enabling the risks to be dealt with in a coherent way (DCDEP, 2000). This is also supported by the United Nations Environment Program APELL (Awareness and Preparedness for Emergencies at the Local Level), which lists the identification of risks that can pose a potential threat as one of its main goals. Similarly, FEMA (FEMA, 2001) has its PROJECT IMPACT, aimed at building disaster-resistant communities. In order to gain better preparedness for disasters, possible risks must be identified. Visualization may here provide a valuable tool for both identifying, exploring and communicating risks.

Geographic visualization has emerged as a tool for searching through huge volumes of data, for communicating complex patterns, for providing a formal framework for data presentation, and for exploratory analysis of data (Gahegan, 2000). Data on risks and hazards often tend to heterogeneous, complex, inter-dependent, not directly comparable, and correlated in ways that may not be apparent without the use of visualization technology. In a paper on the use of GIS to assess natural hazards, Coppock (1995) noted visualization as important not only in the development of GIS generally, but also as a tool to improve reliability of hazard assessment, thus decision support, and also to improve the ability of non-experts to take advantage of the information presented. A statement on the need to improve the representation of risk and vulnerability is also found in Radke et al. (2000), noting that the average GIS is not able to represent the depth and richness of the dynamic nature of risk and vulnerability.

This paper will address the issue of visualization as a means of risk communication as well as risk exploration, highlighting the different approaches that need to be used in explorative visualization versus communicative visualization.

Risk and vulnerability

Objective versus perceived risk
In its simplest form risk may be seen as the product of probability and consequence.

R = P x C (Risk = Probability x Consequence)

In this sense, both high-probability low-consequence incidents and low-probability high-consequence incidents may have the same risk value: It is quite probable that I may stumble and fall when walking along a forest trail; the consequences in most cases are minor. It is much less probable that I will be hit by a falling tree when walking the same trail; the consequences however may be dire. Even though the objective risk numbers may be the same, we humans will rank one risk above the other. Which is the greater risk?

This makes it clear that in dealing with risk, it is often necessary to distinguish between objective risk and perceived risk. Objective risk is a measure of the scientifically quantifiable risk, often derived from probability calculations. Perceived risk is a measure of the imminent danger a person feels he or she is in. Perceived risk can also be a measure of what consequences the individual person is willing to accept before he or she considers a probable incident to be a risk.

Vulnerability
Risk often pertains to a factual event; vulnerability often pertains to a system, such as power supply, telecommunications, infrastructure or society as a whole. Vulnerability is a measure of how well a system can cope with or sustain a risk. The average risk of a long-term power outage may be relatively small in our modern world, except for example during severe winter storms. Yet, even though the risk is minor, the vulnerability is extremely high. This makes it eminent that any risk needs to be linked to its consequences to have a meaning. Consequences in narrative form are one form of “visualization”. Since vision is the dominant sense in sighted humans (Gahegan, 2000), a visual presentation of consequences will often generate a better insight.

Risk experts versus public opinion
It may seem oversimplified to phrase it this way, but one could say that the risk expert is mainly concerned with finding the quantitative risk, that can be stated in mathematical figures, because that is what most technological decisions are based on. The general public or the individual person on the other hand is more concerned with the qualitative risk that affects him or her personally. Consequently, the communication of risk becomes either an obstacle for widening or a bridge for closing this gap between the expert and the layperson. Risk visualization can play a major part in doing so.

Risk communication
As a consequence of the understanding about the divergence in perceptions of risk between the public and the experts, and the ensuing debate over the acceptability of such risks, a whole new area of study has developed, called risk communication. Risk communication is the process of developing and delivering a message from the risk expert to the general public (Cutter, 1993). This process may be aimed at communicating about the danger in a pending emergency or in general to inform about the risk in an issue that the public may or may not perceive differently from the experts.

Risk visualization can, in fact, be viewed as one form of risk communication, which can be further separated into explorative visualization and communicative visualization. The first approach emphasizes exploration, meant to study and to analyze phenomena or events that can be considered a risk, the latter emphasizes communication, meant to inform and to raise awareness among the public in general.

One of the main issues that needs to be addressed in risk communication is the question of how to communicate risk and the question of to whom the communication is directed. Going further into area of risk communication will be beyond the scope of this paper; readers are referred to Lundgren (1994) or Gutteling et al. (1996) for further study.

Issues in visualization of risk

Visualization versus cartography
Visualization of spatial data has close links with cartography. Cartography is the art or science of making maps (Webster) or the art and science of graphically representing a geographical area, usually on a flat surface such as a map or chart; it may involve the superimposition of political, cultural, or other nongeographical divisions onto the representation of a geographical area. Cartography is an ancient discipline that dates from the prehistoric depiction of hunting and fishing territories (www.britannica.com).

Maps have been used for centuries to visualize spatial data. They help their users to better understand spatial relationships. From maps, information on distances, directions and area sizes can be retrieved, patterns revealed and relations understood. (Kraak and Ormeling, 1996)

Past presentation-oriented cartographic research has emphasized the use of static maps designed for public consumption with the emphasis on extracting specific pieces of information (MacEachren and Kraak, 1997). Visualization, which has emerged and gained momentum during the last decade, is less concerned with presentation created by a supposed expert, and more concerned with exploration by the individual.

MacEachren (1994) introduced a map-use based approach to visualization, presented as a cube, which was later modified together with Kraak (MacEachren and Kraak, 1997), see figure.

Goals of map use. Sequence of goals that can be facilitated by visualization methods are arrayed in a map-use cube. From: MacEachren and Kraak, (1997).

Exploration versus communication
Even though this map-use cube identifies 4 speparated stages in visualization, it may be better to see visualization as a continuuum ranging from exploration, that is looking at the data, via analysis, that is discovering hidden relationships, to presentation, that is communicating one or more realizations of the data.

As mentioned above, past communication-oriented cartographic research has emphasized the use of static maps designed for public consumption with the emphasis on extracting specific pieces of information (MacEachren and Kraak, 1997). Unfortunately this still holds true in the realm of risk mapping, where the lay user of the map has to rely on the information that the risk expert chooses to put on the map. With little or no understanding of how the map was derived or how to interpret it properly, risk perception becomes a highly personal matter.

Uncertainty
The very word risk implies uncertainty. Conversely, if there exists an uncertainty whether a hazard exists, there remains a probability that it does and therefore there exists a risk. Quantitative analysis of uncertainty and variability is receiving growing acceptance in risk acceptance. Thus, it is the responsibility of the risk assessor to use whatever information us available to obtain a number between zero and one for a risk estimate, with as much precision as possible, together with an estimate of the imprecision. (Wilson and Shylakter, 1997)

Davis and Keller (1997) explore the modeling and visualization of multiple spatial uncertainties, using slope stability and landslide susceptibility. Davis and Keller argue for a dynamic, rather than static display, because an important aspect of visualizing uncertainty should be the ability to view the various realizations rather than the parameters. This is particularly true when viewing the results of a process model. The static display of standard deviation values would be of little use to decision makers. Visualizing the implications of variance is far more important. This has direct bearings to risk visualization where one static display of a certain set of parameters does not give a full picture of the variability in the data. In lay terms, exploring different combinations of risk related to the certainty of the derived risk may give a different view than the risk value alone. If the risk is high, but its certainty is low, then is the risk actually high or is it not?

A similar approach can be discerned in Ehlschlaeger et al. (1997), where uncertainty in elevation data creates a number of diverging realizations of a possible corridor between two locations. Using the analogy of corridor analysis, and substituting the elevation surface with a risk surface, this could be applied to road design in finding the least hazardous path over a given terrain, avoiding terrain with a high potential for wildfires, landslides, flash flooding in gullies, or analyzing an existing right-of-way corridor with relation to its hazard potential.

Correlation
Tobler (1979) championed the phrase “everything is related to everything else, but near things are more related than distant things”. One way of visualizing this is through scatterplots. Scatterplots are perhaps the most well-known visualization technique for exploring spatial relationships Gahegan (2000), as well as dependency between variables in a risk analysis (Vose, 1997). The major difference between the two is that while risk analysts and statisticians confine themselves to investigating the dependencies between two variables, spatial analysts using visualization techniques can work in 3 or more dimensions:

Display of spectral training data from classification exercise using remotely-sensed imagery. From: Gahegan (2000)

In general, the scatterplot is often used to graph variables directly against each other, effectively removing locational information. In this way, the data is transformed from its spatial reference to an arbitrary non-spatial scale, devised by the user (Gahegan, 2000). With the spatial dimension removed, the viewer can focus on examining the relationships independent of the location.

Cartographic methods
For detailed descriptions on which cartographic techniques that are most appropriate, the reader is referred to Bertin (1993), MacEachren (1995) and Ormeling and Kraak (1996) to mention but a few.
In this paper, it will have to suffice to point out some important issues. Figure 3.3 shows nine possible map types for displaying the same data. Which one that is most suitable for risk mapping, is difficult to tell, that will depend both on the risk data itself and the way the risk assessor would want it to be displayed. In general though, it must be warned against using choropleth maps that distribute a risk evenly over a surface when the risk in fact is not homogenous to such an extreme degree that it follows the choropleth boundaries.

Transformation in maps

Transformation possibilities among maps. From: Ormeling and Kraak (1996)

One example of such misleading risk mapping can be seen in the FEMA’s HAZUS maps (Fig. below). Here, the user can select a hazard and calculate his or her own damage estimation. Since this estimation is based on census data, it naturally follows the census tracts. In the figure, ground acceleration during an earthquake is used to calculate potential damage, and assumed to be uniform over the whole census tract, in order to generate a damage estimate. Not only may this confound the ignorant user of this map, it clearly distorts the factual damage.

Excerpt from map output from HAZUS, showing damage estimate based on ground acceleration during an earthquake (Courtesy of Lorraine Nelms)

More persuasive are the ground-shaking intensity maps developed by Perkins (1987) in the late 1970s, and described in Monmonier (1997). On Shaky Ground (Perkins, 1987) includes risk maps for three types of construction and the potential accumulated damage associated with each building type for a given area. Even though the maps contain no information on existing land use, they are much easier to read and understand than the map in figure above.

Examples of visualization of risk

Risk analysts are experts in their particular field, and exceptions granted, not necessarily educated in cartography or visualization techniques, given their preoccupation with mathematical calculable deterministic or probabilistic models. Consequently, their rendering of spatially related risk may be inflicted with serious flaws.

This chapter does not attempt to provide an exhaustive overview of “good” or “bad” examples of risk mapping. Even Monmonier (1997) only scratches the surface, but supplies enough examples to raise a few eyebrows. Here Monmonier looks at how well America maps its natural, technological, and social hazards, concluding that hazard maps not only show how well we understand hazards, but more often, how little we understand of hazards. The interested reader is referred to this literature for wider and more conclusive study on risk mapping and visualization.

The following is thus more a supplement to Monmonier than a separate collection of examples. Looking back at what has been stated on visualization in the previous chapter, it is left up to the reader to decide for themselves whether these are good visualizations or not.

Drought

Predicting a wildland fire season is difficult, at best. And the earlier a fire season is forecast, the more unknowns there are. For example, how much or how often rain will fall in the spring may mean the difference between few small fires or many large ones. Still, weather patterns and fuel conditions can be important indicators of the potential for a busy, or slow, fire season.

The Climate Prediction Center, CPC, together with the U.S. Department of Agriculture and the National Drought Mitigation Center in Lincoln, Nebraska, issues a weekly drought assessment called the U.S. Drought Monitor. The Monitor provides a consolidated depiction of national drought conditions based on a combination of drought indicators and field reports. Not only indicative of wildfire risk, it presents a general overview of the situation related to drought, making potential drought- related risks apparent.

This map shows levels of concern about wildfire in Utah, ranging from low (green) to high (red) based on a combination of population density, past fire occurence and vegetation. Source: Utah Bureau of Land Management.

This figure shows the same image as above, but in a grayscale. Note that the dark green areas are no longer discernible from the red or orange areas.

Wildland fire
In order to ensure that visualization achieves its goal, the data should be encoded to the available visual attributes bearing in mind the perceptibility of the attributes (Gahegan, 2000). This is especially true in communicating risks. For example, classifying danger zones was red and safe zones as green poses a challenge for red-green color-blind persons. Color graduation yellow-orange-red is preferred and still discernible even in a black-and-white (grayscale) photocopy.

Aggregated risk in small-scale maps
This figure shows a map depicting the overall catastrophic risk in the United States. No information is given how the risk index is derived or how it is aggregated. Nonetheless, it still indicates which is the dominating risk in the given region.

Evaluation and conclusion

As Molak (1997) points out, earlier forms of risk analysis and risk communication tended to overemphasize the role of the expert in “proving that something is NOT dangerous”. However, as he states further, the most important issue is to always make risk assessment transparent to the public with all the assumptions and parameters clearly stated. The thought process that goes into evaluating a particular hazard is more important than the application of some sophisticated mathematical technique or formula, which often may be based on erroneous assumptions or models of the world. Visualization is an excellent tool to overcome these limitations, because it stimulates thought and because it can be used to display uncertainties and the variability of the parameters that influence risk.

In all this praise for explorative visualization, one should not overlook the limitations of this technology, as Gahegan (1999) remarks and Uhlenküken et al. (2000) give excellent examples of. No visualization can overcome the deficiencies in the data or deficiencies in the software itself. What lacks most is the interoperability of systems for scientific visualization, image processing, GIS, database engineering, statistical analysis and other methodology.

Explorative visualization should be the main impetus for the risk analyst, in an attempt not to just to quantify the risk best possible, but to identify correlations, uncertainties and hidden information. Decision-making based on risk assessment does not rest on a crisp finite analysis, but on an evaluation of all contributing factors and uncertainties.

Future outlook

In the search for relevant applications of visualization techniques for risk analysis, it was discovered that most literature, with a few notable exceptions, only employs static 2D techniques, maybe covering a few variations of risk factors at the best. Risk analysis seems preoccupied with numbers rather than figures, more concerned with modeling risk accurately than allowing room for uncertainty and exploration. Unfortunately, this seems to be the nature of risk analysis; in the end, the public wants certainty whether there exists a risk or not, vagueness does not seem to be an option here.

Risk analysis is a crucial element in emergency preparedness. Visualization is one methodology that should be part of risk analysis. Visualization is closely linked to GIS. The number of limitations, challenges and possible improvements in GIS with respect to using it in emergency preparedness, as highlighted by Radke et al. (2000), can serve not only as a guideline for future GIS research for emergency preparedness, but also point to rewarding research avenues for visualization of risk and vulnerability.

In conclusion, it is hoped that visualization techniques find their way into risk analysis, thus not only enhancing, but also de-mystifying the number-and-probability-crunching that is currently prevailing. As this paper has shown, the potential for improvement is fairly large; something this author definitely aims at contributing his share to.

References

Alexander, D., 1995, A survey of the field of natural hazards and disaster studies, In: Carrara, A. and Guzzetti, F. (Editors), Geographical Information Systems in Assessing Natural Hazards, Kluwer Academic Publishers, Dordrecht, The Netherlands, 1-19.

APELL, 1999, United Nations Program on Awareness and Preparedness for Emergencies at Local Level, http://www.unepie.org/apell/

Bertin, J., 1983, Semiology of graphics, University of Wisconsin Press, Madison, WI

Coppock, J. T., 1995, GIS and natural hazards: An overview from a GIS perspective, In: Carrara, A. and Guzzetti, F. (Editors), Geographical Information Systems in Assessing Natural Hazards, Kluwer Academic Publishers, Dordrecht, The Netherlands, 21-34.

Cutter, S. L., 1993, Living with risk, Routledge, Chapman and Hall Inc., New York, 215p

Davis, T. J. and Keller, P., 1997, Modelling and visualizing spatial uncertainties, Computers and geosciences (special issue), 23, 4, 397-408

DCDEP, 2000, Risk Assessment in Europe, A summary of the EU workshop on Risk Assessment, Oslo, Norway, 25-26 November 1999, published by the Directorate for Civil Defense and Emergency Planning (DCDEP), Norway, http://www.beredskapsnett.no

Ehlschlaeger, C. R., Shortridge, A. M. and Goodchild, M. F., 1997, Visualizing spatial data uncertainty using animation, Computers and geosciences (special issue), 23, 4, 387-395

FEMA, 2001, Project Impact, http://www.fema.gov/impact/impact00.htm

Fuhrmann, S., Kuhn, W. and Streit, U. (Editors), 2000, Geoscientific Visualization, Computers and Geosciences (special issue), 26, 1

Gahegan, M., 1999, Four barriers to the development of effective exploratory visualisation tools for the geosciences, IJGIS, 13, 4

Gahegan, M., 2000, Visualization as a tool for geocomputation, in: Openshaw, S., and Abrahart, R.J. (Editors) Geocomputation, Taylor and Francis, London, 253-274

Gutteling, J. M. and Wiegman, O., 1996, Exploring Risk Communication, Kluwer Academic Publishers, Dordrecht, The Netherlands, 221p.

International Cartographic Association (ICA), 2001, Commission on Visualization and Virtual Environments, http://www.geovista.psu.edu/icavis/index.html

Kraak, M. J. and Ormeling, F. J., 1996, Cartography: Visualization of spatial data, Addison Wesley Longman Ltd., Harlow, England, 222p.

Lavakare, A., unknown, GIS and risk assessment, In: www.gisdevelopment.net

Lundgren, R., 1994, Risk communication: A handbook for communicating environmental, safety and health risks, Battelle Press, Columbus, Ohio, 176p.

MacEachren, A. M., 1994, Visualization in Modern Cartography, Pergamon Press, London, 345p.

MacEachren, A. M., 1995, How maps work: representation, visualization, and design, Guilford Press, New York, 513p.

MacEachren, A. M. and Kraak, M. J., 1997, Exploratory cartographic visualization: advancing the agenda, Computers and geosciences (special issue), 23, 4, 335-343

Miller, H., 2001, Geographic Visualization, Lecture notes from 6010 Geocomputation, University of Utah, spring semester 2001

Molak, V., 1997, Introduction and overview. In: Molak, V. (Editor), Fundamentals of risk analysis and risk management, CRC Press Inc., Boca Rotan, Florida, 1-10.

Monmonier, M., 1996 How to lie with maps, University of Chicago Press, 207p

Monmonier, M., 1997 Cartographies of danger, University of Chicago Press, 378p

Perkins, J. B., 1987, The San Francisco Bay Area – On Shaky Ground, Association of Bay Area Governments, Oakland, CA

Radke, J., Cova, T. J., Sheridan, M. F., Troy, A., Lan, M., Johnson, R., 2000, Application challenges for geographic information science: implications for research, education, and policy for emergency preparedness and response, URISA Journal, 12, 4

Slovic, P., 1997, Risk perception and trust. In: Molak, V. (Editor), Fundamentals of risk analysis and risk management, CRC Press Inc., Boca Rotan, Florida, 233-246.

Tobler, W. R., (1979). Cellular Geography, D. Reidel Publushing Company, 379-386

Uhlenküken, C., Schmidt, B. and Streit, U., 2000, Visual exploration of high-dimensional spatial data: requirements and deficits, Computers and Geosciences, 26, 1.

Vose, D., 1997, Monte Carlo Risk Analysis Modeling. In: Molak, V. (Editor), Fundamentals of risk analysis and risk management, CRC Press Inc., Boca Rotan, Florida, 45-66.

Wilson, R. and Shlyakter, A., 1997, Uncertainty and variability in risk analysis. In: Molak, V. (Editor), Fundamentals of risk analysis and risk management, CRC Press Inc., Boca Rotan, Florida, 33-43.

Reference

Husdal, J. (2001). Can it really be that dangerous? Issues in visualization of risk and vulnerability. Unpublished working paper. University of Utah, Salt Lake City, USA.

Related

• husdal.com: The use of air photos in emergency management

MFworks has evolved from MAPFactory, originally designed by C. Dana Tomlin, the father of map algebra.  Conducting network analysis in MFworks comprises iterative steps that lead to a functioning network. These steps will convert map layers with square cells into linear elements that are linked together as lines, with directional flows assigned to each cell, and map layers containing cost variables. This tutorial, developed by husdal.com in 2002, is a showcase on network analysis in MFworks, with step by step instructions and a summary of the theory behind it.

MSc in GIS

This tutorial builds on my thesis for my MSc in GIS, where I explored the topic of network analysis in raster GIS, using MFworks as example software, investigating current algorithms, procedures and network modelling techniques and finding some odd artefacts along the way.

MFWorks

Tutorial

Click here for the tutorial.

Related

• husdal.com: Network analysis in raster versus vector GIS
• husdal.com: How to make a straight line square (MSc in GIS)
• husdal.com: Corridor analysis – a timeline of evolutionary development