Eliciting and Representing Knowledge to support Flood Risk Management
Research Project
Proposed Concepts for Domain Knowledge
The project produced a number of conceptual maps that used more general concepts of flood event. The knowledge engineers in this case, studied information sources on the domain and built the maps based on their conceptual understanding of the text. This allowed us to scope the functionality of the tools as well as the domain with the aim of locating a focus area that is of interest to the solution of flood-related issues – here to demonstrate their applicability and effectiveness having been successful in various other knowledge domains and related ones, such as weather forecasting and droughts.
The Drainage Unit – one of 8 departments within the MPT in Barbados is responsible for various aspects of flood management. Maintenance and planning of drainage facilities is the fundamental responsibility of the DU that involves different kinds of analyses to manage the impacts and mitigate the risks of flood events. The areas discussed below as candidates in the development phase of the project, focus on sub-concepts within this wider issue.
Flood Risk Management Knowledge Model
Maintenance of Drainage Structure - “Suck” Wells
This method of disposal of storm water is central to the effectiveness of the drainage system in Barbados. There are an estimated number of 3500 suck wells located along highways or on plantation grounds. The DU maintains a Well Inventory database and uses GIS software to produce maps identifying attributes and locations of wells. Well data and maps are used at different levels of management and decision making. These are categorized into operational, middle and top management to support the processes of disaster response, maintenance, and planning, respectively. Maintenance tasks are carried out routinely to clear debris from accumulating or to respond to complaints. Reports from maintenance and inspection work are used to update the Well database, as well as the data collected from storm water surveys that is an ongoing exercise. This enables the organization to monitor performance, carry out various kinds of assessments, such as drainage impact, maintenance planning and budgeting, and to analyze the effectiveness of the structure for flood mitigation - the overall goal.
The decisions to maintain an existing structure or design new measure – dig a new well, new technique or method, uses the explicit knowledge (data and analyses) and experiential or tacit knowledge of the persons making them. A common task of Well inspectors and engineers is to assess a drainage problem and recommend maintenance works. The skill, experience, and information source the individual uses and applies – their conceptual model, is of fundamental value to the knowledge assets of the organization that often remain locked in the minds of those experts and lost after they leave the job. The methods proposed for use in our project are designed to capture and represent this often ‘elusive’ knowledge in a format that allows it to be shared and extended as new information becomes available. The knowledge model - sets of interconnected conceptual maps and resources, provide task experts with a tool to verify and update their domain knowledge, and the organization to preserve its operational knowledge. These concepts maps are considered case-bases of those experts conceptual activity or a memory of their experiences. The toolset utilize techniques to build a case library from the concept maps and support users to introduce new concepts and provides reasoning mechanisms on new situations using previous cases.
Storm water Runoff Coefficient
Flooding, in addition to climate and people intervention, is also caused by the drainage system being unable to cope with the volume of surface runoff. Flood plain mapping is carried out in Barbados at road crossings, minor (nuisance) and major flood plains. The data on drainage facilities, watershed and rainfall is used to assess the capacity of the system at peak storm flow conditions and determine the potential flood hazard. The information from this analysis is provided to interested parties and used by the DU to identify priority locations for maintenance. Observations and data collected during and after flood events are used to assess the impact and validate estimated drainage capacity. This enables the verification and update of models used in flood mapping and prediction, and measures used for mitigation.
As a process, flood hazard mapping requires various sets of data to give an estimate of the impact of an actual flood event on a community and keep track of change in flood patterns. Estimations of flood levels inform critical decisions in extreme situations. The complexity and uncertainty of the factors that influence the accuracy of predictions of a flood stage are considered in the context of this project as weather related and the condition of the catchments areas. This involves analyses of rainfall data (amount and duration) in relation to drainage infrastructure to produce information about how the system coped (runoff figures). Domain(s) experts employ knowledge on the attributes and their relationships to design the models used to calculate runoff and update that knowledge on the basis of new domain theories and the observations they make on change in the flood behaviour (tacit). They use explicit knowledge on the history of flood events to monitor this change. This produces information used in maintenance and risk awareness and other flood impact studies.
Studies carried out following a major flood event form multiple case bases. Cases of flood stage estimates (episodes) and cases that describe the event in terms of impact on drainage such as proposed maintenance works, information on secondary hazards, and limitations on the drainage system (catchments drainage cases). The reports by the inspectors concerning issues of constraints in catchments areas such as vegetation or dumping practices, new development, and those related to the effectiveness of existing flood mitigation measures. This provide the attributes for drainage cases as well being a potential case base as more studies are carried out in problematic areas. The areas covered by the cases bases are complementary with regard to the fact that they model one or a series of flood events for a single area on the FHM. Other areas may also be viewed as distinct cases in terms of local factors that allow identifying patterns of system mis-use or design/planning fault that are useful for legal and engineering decisions. Though complementary, they must be viewed as distinct in the context of the MPhil project and a separate knowledge modeling exercise.
Community Perception
The DU receives information from community members, national disaster bodies and other discussion formats on existing and new problems. The concerns raised in the complaint/report are investigated; a solution proposed and carried out. The data collected in the investigation include general information on the task (location, date, officer) and the problem (photograph, sketch, interview with complainant). The problem reported can be simple and solvable through scheduling a routine maintenance task. More complex issues require deliberation over alternatives and selection of the most appropriate for the situation.
Most of the data described above (excluding digital photographs) exist in paper form, such as complaint sheets , in addition to some records on assessments and recommendations made. Storing the general information on the complaints received in a database allow frequency listing of problem, at location, etc. It is not clear (at this stage) whether the database will incorporate a column describing the outcomes of the investigation and/or the solution to the problem in the complaint record. It is possible however, to aim at building a case base of a selected set of complaints (where data on assessments and recommendations is available).
Analysis of frequencies raises questions about why they occur at that rate, for instance the relationship between fault scenario and location characteristics (soil, land use, drainage facility). An expert uses their domain knowledge and experience to select features (concepts) and propositions that describe associated significance values (weight). The conceptual mapping approach offer means to validate, update and store maps of expert perceptions. The tool offers, as discussed in previous section, reasoning mechanisms using a concept map (expert perception) as cases. This accumulates knowledge about change in feature set and highlights significant and less so ones, warranting legislation, accommodation, or do nothing.
Information Sources used in domain focus descriptions
Outline of research methodology
Knowledge Elicitation and Representation:
- Identify key personnel from whom knowledge can be elicited.
- Obtain knowledge on concepts and prepositions that are used in the domain.
Knowledge Model Evaluation:
· Carry out an empirical study into the impact of the produced knowledge model in learning and operations. These perceptions can be captured using a survey that targets the task experts involved in the exercise, new staff and specialists on other drainage structures.
· Analyze survey results and make recommendations with regard to the usefulness of the approach to represent, use or integrate different types of organizational knowledge (operational and academic). This involves evaluation of expert perception on how well the tools allowed them to validate or share their individual domain knowledge. Other areas the conclusions can address are the usefulness of the approach to support organizations to model these valuable pieces of knowledge incrementally.
Other Information Sources and useful links
Busby
L. A. (2003) General Data Challenges Facing The Caribbean in the Context of
Sustainable Development. In Proceeding of Workshop on Using Information For
Decision-Making on Sustainable Development Issues and Challenges for Carribean
SIDS. St. Lucia, 27 – 28 May 2003. http://www.un.org/esa/sustdev/natlinfo/indicators/idsd/workshops/workshop1pdf/eclac.pdf
Coffey,
J. W., R. Hoffman, Canas A. J. and Ford K. M. (2002). A Concept Map-Based
Knowledge Modeling Approach to Expert Knowledge Sharing. IKS 2002- The
IASTED International Conference on Information and Knowledge Sharing, November
2002, Virgin Islands.
Evans
M. A, McMullen R., Leake D., Bogaerts S (2004). Remote Collaborative Help:
Using Cases and Concept maps to Coordinate and Contextualize Knowledge. In
UbiComp 2004 Workshop Proceedings. http://www.xrce.xerox.com/competencies/work-practices/ubicomp_workshop/evans_ubicomp04_w5_final.pdf
Douglas E. (2003) Flood Hazard Management, Mapping and Early Warning Systems in Jamaica. http://www.ewc2.org/upload/downloads/Errol_Douglas.doc
Leake
D. (1996). CBR in Context: The Present and Future. In case-based
Reasoning: experiences, Lessons, and Future Directions, Leake D. (Ed.). Menlo
Park: AAAI Press/MIT Press.
Leake D. B. and Canas A, J. et al. (1999). Aiding Knowledge Capture by Searching for Extensions of Knowledge Models. AAAI Workshop Technical Report WS-99-10: Exploring the Synergies of Knowledge Management & Case-Based Reasoning, AAAI Press, Menlo Calif.
Leake
D. et al. (2002) Capture, Storage and Reuse of Lessons about
Information resources: Supporting Task-Based Information Search. http://www.cs.indiana.edu/~leake/papers/p-00-02.pdf
Lewis D. R. (1998) An Assessment of the Impact of Surface Flooding and the Response of Residents of Different Socio-Economic Status in
the Kingston Metropolitan Region. Cave Hill Main Library.
Nurse,
L., McLean, R., Suarez, A., Ali, M., Hay, J.E., Maul, G. and Sem, G. (1998).
Small island states. In: R. Watson, M. Zinyowera, R. Moss and D. Dokken (eds.). The
Regional Impacts of Climate Change: An Assessment of Vulnerability. A
special report for IPCC Working Group II. Cambridge University Press, 331-354.
Nurse,
L.A., Sem, G. (with Hay, J.E., Suarez, A.G., Wong, P.P., Briguglio, L., and S.
Ragoonaden), 2001. Small island states. Chapter 17 in J.J. McCarthy, O.F.
Canziani, N.A. Leary, D.J. Dokken and K.S. White (eds.). Climate Change 2001:
Impacts, Adaptation, and Vulnerability. Intergovernmental Panel on Climate
Change (IPCC), Third Assessment Report. Cambridge
University Press, Cambridge, p. 843-875.
OAS. Primer on Natural Hazard Management in Integrated Regional Development Planning. 1990.
OAS. Disasters, Planning and Development: Managing Natural Hazards to reduce loss. 1990.
Penning-Rowsell,
Parker D. J., Harding D. M. Floods and Drainage (1986) – British Polices
for Hazard Reduction, agricultural improvement and wetland conservation.
Published in The Risks & Hazards Series 2.
Safayeni
F., Derbentseva N., Cañas A. J. (2003) Concept Maps: A Theoretical Note on
Concepts and the Need for Cyclic Concept Maps, Manuscript submitted for
Publication. http://cmap.ihmc.us/Publications/ResearchPapers/Cyclic%20Concept%20Maps.pdf