bayesian network model for evaluation of ecological river construction m. arshad awan

30
Bayesian Network Model for Evaluation of Ecological River Construction M. Arshad Awan

Upload: ernest-sharp

Post on 16-Dec-2015

216 views

Category:

Documents


3 download

TRANSCRIPT

Bayesian Network Model for Evaluation of Ecological River Construction

M. Arshad Awan

Bayesian NetworkA probabilistic graphical model

that represents a set of random variables and their conditional dependencies via a directed graph (DAG), e. g.,

EcologyThe study of the interactions of

living organisms with each other and with their environment.

General River ManagementFlood Control

◦Embanking◦Waterway management

Water resource management◦Irrigation◦Drinking water supply◦Industrial water supply◦Hydraulic power generation

New Demands in River ManagementEnvironment-friendly

◦Landscape, temperature, humidity, oxygen

Ecological healthiness◦Species diversity, balance of food

chain◦Abundant number of species◦Habitats for animals

Water-friendly activity◦Exercise, rest, walking, picnic,

fishing, learning, observation

Ecological River ConstructionNature-shaped river

◦Recover the natural environments as close as possible (shallows, swamp, tree, grass, etc.)

◦Within the limit of flood controllability

◦Ecological system recovery◦Sustainability

Supply the area for water-friendly activity◦Rest area, shelter, walkway, sports

area◦Accessibility

Successful Ecological RiverHow to evaluate?Possible variables

◦Sufficient water-quantity◦Clean water-quality◦Good landscape◦Secure structure of nature-recovery◦Convenient facility◦Sufficient space, etc.

Research DefinitionGoals

◦To develop a model to evaluate the ecological river construction

◦To find the required/desired plan quantitatively

Technical tool◦Bayesian Network Model

Expected effects◦Evaluation of existing rivers◦Evaluation of results on investment◦Provide the suggestion to reconstruct and

manage the facility◦Provide the guideline for the new project

Progress in term projectSurvey:

◦Ecological river engineering◦Bayesian belief networks (BBN)

Selection of input variables for BBN

Tool to develop BBN◦Netica

Development of proposed BBN

Input variables 1 Water Quantity - sufficient water quantity is one of the most significant

factor to characterize a river. - but too much water in a urban river is not always

good in the aspect of flood control, safety issue,

maintenance cost, and etc. - perceptions on how much water is sufficient are very

subjective.

10 20 30 40 50 60 70 80 90 100

lack sufficient Too much

Input variables 2 Water Quality - People are very sensitive on the water quality. - The more clean and clear, the better - It costs a lot to maintain the desired water quality. - The desired water quality of river is not necessarily to

be high as the quality of drinking or industrial water - perceptions on the desired water quality of river are

very subjective.

1 2 3 4 5 6 7 8 9 10

dirty clean Very clean

Input variables 3 Ecology - One of main goals of stream restoration is ecological

balance and soundness. - It can be measured by biodiversity, the number of a

species, ecological system service, habitat areas for wild lives, and etc.

bad average

good

1 2 3 4 5 6 7 8 9 10

Input variables 4 Landscape - Landscape of a river is composed of many factors - trees, plants, forest and wetland, riparian corridor with

built environment, bank, and etc. - perceptions on landscape are very subjective and may

be characterized by 3 linguistic terms: excellent, good, ordinary.

ordinary

good excellent

1 2 3 4 5 6 7 8 9 10

Input variables 5 Stream shape (Fluvial geomorphology) - Stream shape is very important to ensure the self-

purification of water and the sustainability of ecosystem by supplying various aquatic environments.

- Stream shape should be restored as close as possible, but must not decrease the flood controllability.

- replacement of shore protection, islands, shoals, pools, fish-ladder, removal of artificial facilities such as water steps and small dams, etc.

artificial

natural

Too natural

1 2 3 4 5 6 7 8 9 10

Input variables 6 Facility - people want to do some activities near a river - Although artificial facilities may not be good for the

ecological system, the least amount of facilities to provide people with accessibility and water-friendly activities are necessary

- shelter, rest area, walkway, exercise facility, road, parking lot, etc.

- In some cases, too many facilities are constructed. - In some cases, people ask more facilities. - How many facilities are reasonable?lack sufficient Too many

1 2 3 4 5 6 7 8 9 10

Bayesian Belief Network (BBN)Structure

◦Connection of nodes (DAG)Inference

◦Infer the value of variablesLearning

◦Training examples

Building BBN Structures

Netica (BBN Tool)

Netica (BBN Tool)

Proposed BBNTo evaluate a river, a set of nodes are

connected: ◦ based on the combination of 6 input

variables

The output of evaluation can be differentiated based on the criteria which uses different sets of variables◦ comprehensive evaluation : 6 inputs◦ aquatic environment evaluation:

quantity, quality, ecology◦ land environment evaluation:

landscape, stream shape, facility◦ Balance/successful evaluation : 6 inputs

comparison

Ecological River Construction

Network report

Aquatic Environment (CPT)

Land Environment (CPT)

Ecological River Const. (CPT)

A random training sample

Learning AlgorithmThere are three main types of

algorithms that Netica uses to learn CPTs: ◦Counting, ◦Expectation-maximization (EM), and ◦Gradient descent.

Counting is: ◦Fastest, simplest, and can be used

whenever there is not much missing data, or uncertain findings for the learning nodes or their parents.

References Woo, H., Trends in ecological river engineering in

Korea, Journal of Hydro-environment Research (2010), doi:10.1016/j. jher.2010.06.003.

Finn V. Jensen and Thomas D. Nielsen, “Bayesian Networks and Decision Graphs”, February 8, 2007, Springer.

Judea Pearl, “Probabilistic Reasoning in Intelligent Systems: Networks of Plausible Inference”.

Marcot, B. G., J. D. Steventon, G. D. Sutherland, and R. K. McCann. 2006. Guidelines for developing and updating Bayesian belief networks applied to ecological modeling and conservation. Canadian Journal of Forest Research 36:3063-3074.

McCann, R., B. G. Marcot, and R. Ellis. 2006. Bayesian belief networks: applications in natural resource management. Canadian Journal of Forest Research 36:3053-3062.

References Marcot, B. G., R. S. Holthausen, M. G. Raphael, M. M.

Rowland, and M. J. Wisdom. 2001. Using Bayesian belief networks to evaluate fish and wildlife population viability under land management alternatives from an environmental impact statement. Forest Ecology and Management 153(1-3):29-42.

The Anticipated Impacts of the Four Rivers Project (ROK) on Waterbirds (Birds Korea Preliminary Report).

Workshop on hydro-ecological modeling of riverine organisms and habitats, ecological processes and functions (6th to 7th of June 2005, The Netherlands).

http://www.gleon.org/ (Global Lake Ecological Observatory Network).

http://en.wikipedia.org/. Sandra Lanini, “Water Management Impact Assessment

Using A Bayesian Network Model”, 7th International Conference on Hydroinformatics, HIC 2006, Nice, FRANCE.

Thanks!