GIS – WHAT CAN GO WRONG?

ABSTRACT

This presentation discusses experiences and especially problems faced with GIS operations in the region. The author has been working as a GIS and Remote Sensing consultant to the Forest Cover Monitoring Project (FCMP), to the Sustainable Management of Resources in the Lower Mekong Basin Project (SMRP), and to a number of other GIS related projects in the region operating at various levels and scales.

The first part of this presentation deals with technical aspects, i.e. scale issues, geo-referencing, data generation, quality control, data base maintenance, and system configurations

The second part of this presentation discusses human resources aspects, i.e. staff selection, qualification requirements, introductory training, and upgrading of knowledge.

The third part of this presentation deals with data use and institutional aspects, i.e. data ownership, data distribution, metadata bases, data pools, and research and practice.

The main problems encountered are highlighted.

Since the problems faced with GIS are numerous, the issues reported here can not be discussed in great detail. This presentation does not provide readily made solutions. It merely intends to focus attention and to encourage discussions on the issues reported, which may eventually help finding better solutions.

WHAT IS GIS USED FOR?

In Theory …

There are several definitions of what GIS is and what it should be used for. The following two definitions are probably the most common ones:

The IMAP Model

• Input
• Management
• Analysis
• Presentation

"A GIS is a system composed of hardware, software and procedures for the

• capture
• storage
• manipulation
• analysis
• modeling
• output

of spatial data to solve complex planning and management problems."

Looking at what remains of these definitions in practice in the region, one often finds the following situation:

The ImP Model

(++ much used, + used, - little used, -- rarely (or not at all) used)

"A GIS is a system composed of hardware, software and procedures for the

of spatial data to solve complex planning and management problems."

WHAT IS GIS USED FOR?

Conclusions

Looking at what GIS is used for in practice in the region, one arrives at the following conclusions and questions:

• Not only data input or capture (from existing sources), but DATA GENERATION probably constitutes the most important component of GIS applications.
• Using GIS for Data Generation is perfectly justifiable given the lack of accurate and / or up-to-date base data. The analytical capacities of GIS can not be used unless one has the necessary base data in place, or, as Sherlock Holmes put it: "It is grave mistake to draw conclusions before one has data."
• A frequently read GIS guidline states: "In the 1970s, one could, with reason, have argued that GIS was essentially an extension of cartography with other tools."

INTRODUCTION: WHAT DO WE EXPECT FROM GIS DATA?

Accuracy

GIS data should reflect the situation on the ground.

• Positional: the location of objects on GIS should depict their location on the ground as accurately as possible.
• Thematical: the classification of objects on GIS should depict their character on the ground as accurately as possible.

Up-To-Dateness:

GIS data should be recent.

Availability

GIS data should be available

TECHNICAL: PIXEL SIZE AND INFORMATION CONTENT

PIXEL = smallest unit of an image

(Information Content is to be understood as the theoretical maximum given infinite object detail.)

One should keep in mind:

• The maximum scale depends on the pixel size.
• The information content does not grow proportionally, but with the square of the resolution (or scale) ratio.

TECHNICAL: DIGITIZING TIMES AT VARIOUS SCALES

Common approaches towards GIS are:

This does not consider the huge amount of work which GIS operations may cause. The consequences are:

The workload of data generation is usually heavily underestimated.

The initial project (unit) design is dramatically under-dimensioned.

The project (unit) never reaches its objectives

TECHNICAL: GEO-REFERENCING (I)

• The process of referencing spatial data to a map coordinate system.
• This process is of crucial importance for the positional accuracy and the spatial compatibility of data.
• Criterion used to evaluate the quality of geo-referencing: the RMS Error (Root / Residual Mean Square).
• Simplified: RMS Error is the average distance between the actual location of a point (after geo-referencing) and its desired location.
• Guideline: the lower the total RMS Error, the more accurate the geo-referencing.
• In practice: RMS Error is reduced by subjectively altering coordinates. This may result in very ill fitting data layers.
• Example: interpreted satellite images (1:250,000), positional errors of up to ± 500 meters.

TECHNICAL: GEO-REFERENCING (II)

• In some cases geo-referencing is done by manually transferring data to base maps (e.g. from aerial photos).
• This works best in flat terrain with sufficiently dense infrastructure or hydrographic features, if these features are used as references.
• In practice: This technique is applied in any terrain. Little use is made of reference features. This may result in very ill fitting data layers.
• Example: interpreted aerial photos (1:25,000), positional errors of up to ± 250 meters (1 cm at source scale!).

TECHNICAL: DATA GENERATION / INTERPRETATION

• GIS data generation is time-consuming, strenuous and often tedious work.
• "Some people feel their qualifications render the routine work to be beneath their dignity (of interpretation). In these instances, such work is given to newcomers or less-qualified staff."
• Supervision is not done properly.
• Disastrous effects on data quality in terms of both positional and thematic accuracy.
• Example I: Satellite image interpretation (1:100,000).
  In a full coverage repetition experiment only 64 % of the area was assigned to the same class in both rounds.
  Even if the interpretation was simplified to only two classes (Forest and Non-Forest), only 88 % of the area were assigned to the same class in both rounds.
• Example II: Satellite image interpretation (1:100,000).
  In a sample repetition experiment simplified to only 2 classes (Forest and Non-Forest) only 52 % of the area was assigned to the same class in both rounds.
  This almost amounts to random assignment!

TECHNICAL: DATA BASE MAINTENANCE

• GIS data bases can be huge (thousands of files, several gigabytes).
• GIS data bases are commonly used and updated by various operators simultaneously.
• Operators tend to keep the data sets they are currently working on in a rather unorganized way.
• Data sets are not named after their content, but after the operator (or his daughter …).
• Different updates are simultaneously applied to copies of the same data sets.
• Data sets are split for map printing purposes.
• Insufficient or no documentation is kept.
• It becomes extremely difficult to retrieve information. Enormous time has to be spent on organizing data before the fabled analytical capacities of GIS can be put to any use.

TECHNICAL: SYSTEM SETUP

Over-Dimensioned Systems

• Example: UNIX based Systems, ArcInfo Systems
• Expensive Hardware & Software
• Complicated System Maintenance
• Local Support often not available
• Too much time is spent learning, understanding (and playing with) system & software -> less time spent on production
• High risk of individuals gaining hidden executive power due to their system management abilities

Under-Dimensioned Systems

• Example: Slow (Standard Office) Computers, Insufficient Storage Space, Outdated Software (PC ArcInfo), No Networks
• Frequent Mistake: Cheapest Computers are purchased. Education-oriented systems (IDRISI) are used in a production environment.
• Too much time is spent on routine tasks and data swapping ® less time is spent on production

HUMAN RESOURCES: STAFF SELECTION

• It is a frequently told (and believed) myth, supported by the industry, that GIS can be used by anyone.
  This may be true if the objective is just running a couple of queries on a ready-made database.
  The situation is entirely different if the objective is data generation.
• GIS is a powerful tool. However, one can not expect that it can be operated just like any other computer software.
• Using GIS efficiently does require a certain degree of mathematical, geometrical and logical understanding plus organizational skills.
• GIS staff often are (have to be) selected from an existing pool of available staff.
  This may be no problem if a GIS is set up in an organization with a geographical background (such as cartography).
  The situation may be entirely different if a GIS is set up in e.g. agriculture or forestry ("the tree identification specialist turned geographer").
• Background qualifications of available staff can make or break the success of a GIS installation.

HUMAN RESOURCES: INTRODUCTORY TRAINING

• Introductory training is often held in national or regional training centers.
• These institutions will train people in basic operations.
  They will train people in the software packages they have available on the training site.
  The training is normally done without regard to specific (institutional or project) requirements.
  Result: staff normally have to be re-trained on-the-job later. This normally happens with a considerable delay, so skills acquired during the introductory training have already been forgotten.
• People are rarely, if ever, trained to read (the Help).
• Introductory training is not planned carefully. In many cases it does not produce the expected results. Expensive follow-ups become necessary.

HUMAN RESOURCES: Skills Upgrading

• The life span of software packages is not more than 2 - 3 years.
• With every (major) new software generation, productivity increases significantly (example: PC ArcInfo - ArcView).
• GIS staff, however, have a tendency to stick to procedures once established, thus giving away major advantages that could be gained from software upgrades.
• It is frequently done, but normally not sufficient to buy software upgrades without providing additional training.
• Additional training, if provided, is often held in national or regional training centers. Frequently, this is just a repetition of the basic training.
• Upgrading of skills is mandatory. It must be tailored to the institutions / projects specific requirements. It should best be held in-house.
• Skills upgrading is essential to increasing productivity and maintaining technology at stat-of-the-art levels. Training must be tailored to the institutions’ / projects’ specific requirements. It should best be held in-house.

DATA USE: OWNERSHIP / DISTRIBUTION

DATA USE: META DATA BASES

DATA USE: DATA BASES

What happens?

• The time frame set for this activity is rather short.
• The project is under-staffed.
• Data collection from all possible available sources is started.
  These sources provide mainly data sets at small scales, not useful for anything but overviews.
  No or few quality checks are done on the data collected.
• Occasionally, data generation at small scales is started to fill perceived gaps. (Which may not exist at all. Sometimes people are just not aware that data sets already exist.)

What is the product?

• An impressive, flashy looking "new" database, perhaps even on the Internet.
• This database merely contains copies of data sets that existed earlier.
• One may end up with several duplicate data bases, which differ only in names and layouts, but not in their contents.

DATA USE: RESEARCH AND PRACTICE

• One finds quite a number of cooperations between projects and universities, wherein the universities provide the GIS know-how.
• Advantage: university know-how commonly comes at a lower price than the services of permanent advisors or private companies.
• Disadvantage: university know-how is often rather theoretical and a means in itself.
• There is much fascination with technique, but little knowledge as to what information is required.
• Their research approach may also differ widely from a approach oriented towards production.
• Example:
• Various vegetation classifications carried out in overseas universities with little or no ground truthing.
• "Scientifically" designed forest inventories.
• Consequences
• Widespread dissatisfaction among institutions and projects as far as GIS is concerned.
• The assistance offered by universities often constitutes the first GIS experience of an institution or project.
• Understandably, this may result in a "No thank you" attitude towards GIS in the future.

SO, WHAT DO WE NEED?

• clearer identification of the user’s information requirements
• better explanations of what GIS can do and what not
• better explanations of how information content is related to scale
• sound estimates of how much time and effort GIS operations may require
• more production-oriented approaches
• technical guidelines and procedures instead of repetitious data bases
• better quality control
• more structured training and upgrading of knowledge
• better formalized exchanges of metadata
• better coordination among institutions that generate data
• clearer definitions and more transparency of data ownership
• more open data distribution policies
• clear commitments from the donor side that if the generation of GIS data is funded, the products are expected to be used publicly

PUBLIC MONEY SHOULD CREATE PUBLIC DATA!