OPTAIR:     Multi-criteria optimization for air quality
management and emission control

5. Time table and scheduling

The 14 work packages (WP) of the proposed project are:

1. WP: International project coordination, involving the WEBAIR partners in all 17 partner countries where test applications are planned in principle, and the more closely integrated partners that develop optional components for the OPTAIR proposal.
2. WP: Requirements and documentation: the work package spans the entire project duration, with a gradual progression from requirements to documentation and manuals.
3. WP: End user (DM) participation: maintain user contacts and develop tools for the elicitation of end user inputs to the valuation of impacts.
4. WP: Instruments: a data base of technological, regulatory, planning, and economic instruments with their associated cost functions.
5. WP: Emission modeling: a central component of any emission control strategy and the related air quality modeling is the description of dynamic emissions.
6. WP: Basic simulation models: linkage and integration. Multiple alternative but closely linked models will be used for efficient optimization.
7. WP: Assessment, performance criteria: A key of the multi-criteria optimization is the description of systems performance by scalar variables as the basis for the optimization
8. WP: Optimization framework, satisficing. The primary step in the multi-criteria optimization is the generation of feasible alternatives against the set of constraints defined by the representative stakeholders identified in WP 3 in terms of the criteria from WP 7.
9. WP: Computational framework, cluster and grid computing. Provides the necessary performance to generate sufficiently large samples of feasible solutions for the subsequent reference point methodology
10. WP: Knowledge representation, machine learning and adaptive heuristics. Requires first results and an operational computing environment as well as realistic test cases from WP 8 and 9 to start beyond theoretical preparations.
11. WP: Discrete multi-criteria DSS. The second step in the optimization is a discrete multi-criteria DSS implementing a reference point methodology.
NOT ACTIVE IN PROJECT YEAR 1
12. WP: Uncertainty analysis, robustness: climate change scenarios.
NOT ACTIVE IN PROJECT YEAR 1
13. WP: Test cases: demonstrated improvements: The test cases are the central work package, spanning most of the proposed project duration,
14. WP: Dissemination strategy: planned over the entire duration of the project.

The planned scheduling of the work packages is shown in the table below, with the first year schedule highlighted:

to

WP	Work package name, short description	from
WP01	Project coordination and management	1	36
WP02	Requirements and documentation	1	36
WP03	End user participation (stakeholders, DM)	1	15
WP04	Instruments of emission control, cost functions	2	16
WP05	Dynamic emission modeling incl. dust entrainment	3	19
WP06	Simulation models: integration, alternative models	4	27
WP07	Assessment: performance criteria: compliance, economics	7	28
WP08	Optimization framework, phase I: satisficing	7	30
WP09	Computational experiments: cluster and grid infrastructure	7	33
WP10	Knowledge representation, heuristics, GA, machine learning	10	35
WP11	Optimization framework, phase II: Discrete multi-criteria DSS	-	-
WP12	Uncertainty analysis, robustness, GCM scenarios	-	-
WP13	Test cases: implementation, performance, validation	7	36
WP14	Dissemination activities	1	36

The workplan for the first project year will see all work packages with the exception of WP 11 and 12 running or at least started.

WP01: Project coordination and management
This work package spans the entire project duration at a low level of effort, and includes the maintenance of a common project web server as the core of the electronic communication and coordination system between partners to manage and synchronize development and test cases. A major part of the activities will be the management of all common project documents (see also WP 02) in a web based repository with a number of web based management tools.

WP02: Requirements and documentation
The initial set of technical and functional specifications is gradually replaced with progressing implementation turned into the corresponding reference and user manual pages, implemented as on-line hypertext documents integrated in the overall system.

WP03: End user participation (stakeholders, DM)
economic valuation as a central component in the multi-criteria optimization requires direct involvement of representative users, decision makers, stakeholders, major actors to obtain these unavoidably subjective values. At the same time, finding a compromise solution from multi-criteria alternatives requires a preference structure that is also subjective. The basic tools are a stakeholder data base to manage the necessary contacts in each test case, and on-line questionnaires on issues, objectives, criteria, constraints that define preference structures, as well as procedure for effective stakeholder involvement.

WP04: Instruments of emission control, cost functions:
While the data base is generic, and will support the selection and automatic download of instruments into the optimization scenarios, the specific parameters have to be set at the level of the individual scenario. Instruments can be defined:

• By domain, globally (with, however, potentially different effects on individual sources or source classes);
• By emission source class or category, e.g., industry, households, traffic.
• By geographical or administrative groupings
• By individual emissions sources.

• investment costs (either annualized or computed as NPV from total investment, project life time, and discount rate); and
• operating costs, that may be time or activity dependent.

WP05: Dynamic emission modeling incl. dust entrainment
The emission data used for air quality modeling and emission control optimization are either based on direct measurements (for major point sources), or generated from activity and fuel/technology data, where each of the parameters used to estimate emissions is subject to modification by any one of the emission control instruments. For the specific case of VOCs, fuel and technology specific speciation of chemicals will be applied. For biogenic emissions, land use/land cover (vegetation) maps will be used together with the basic meteorological parameters of temperature, humidity, and solar radiation. The most complex emission estimates are for particulates not originating from combustion processes: dust entrainment by wind will be generated by a separate dynamic emission model that uses:

• Detailed short-term high resolution wind fields derived from MM5 meteorological simulations, The proposed method will use a frequency distribution of bottom layer wind speeds around the average value estimated by the model to account for the non-linearities of the wind driven entrainment (exponential threshold function);
• Soil characteristics, in particular size distribution and specific weights;
• Soil moisture as generated by MM5;
• Economic activities such as construction and mining, trucking.

WP06: Simulation models: integration, alternative models
According to the proposed methodology of multiple models to improve optimization performance, the underlying detailed air quality simulation models are grouped into two sets according to the processes covered:

• Conservative substances SO2, NOx, PM10/2.5; these will be represented by the regulatory Gaussian model system AERMOD, and the dynamic 3D nested grid Eulerian model CAMx; due to the restrictions of the Gaussian approach (assuming homogeneous meteorological fields) the geographic scope of applicability is limited; for larger areas, several parallel AERMOD domains will be used corresponding to one CAMx master domain with nested sub-domains.
• Photochemical processes: O3, NOx: represented by the photochemical box model PBM and the detailed, dynamic 3D Eulerian model CAMx, that offers alternative chemical mechanism for the photochemical reaction kinetics.

WP07: Assessment: performance criteria: compliance, economics
A key of the multi-criteria optimization is the description of systems performance by scalar variables as the basis for the optimization. These must be extracted, aggregated or integrated from the dynamic, spatially distributed, multi-parameter models results and directly address regulatory targets (compliance), policy objectives beyond compliance, public health and environmental impacts, as well as economic (cost) criteria. Cost functions will be based on piecewise linear approximation, and include annualized (NPV) investment and activity proportional operating costs (e.g., Bromley, 1995; McAllister, 1980).
See also: the EEA Core Set of Indicators: http://themes.eea.europa.eu/IMS/CSI.

WP08: Optimization framework, phase I: satisficing
The primary step in the multi-criteria optimization is the generation of feasible alternatives against the set of constraints defined by the representative stakeholders identified in WP 3 in terms of the criteria from WP 7. The basic optimization framework will consist of:

• The interfaces and language to describe and configure an optimization scenario in terms of model domain, time periods, meteorology, pollutants and constraints;
• The management of the underlying models and input data required, with emphasis on computational performance (see WP 10 below).
• An open, modular interface to control and configure the strategy for the generation of alternatives (sets of decision variables).

WP09: Computational experiments: cluster and grid infrastructure
Provides the necessary performance to generate sufficiently large samples of feasible solutions for the subsequent reference point methodology can also, or in addition, be based on brute computational power. The most promising solution are based on parallel and cluster computing, since the problem is perfectly task parallel at a very high level of granularity (individual optimization scenarios) which leads to a prefect scalability of parallel implementations with near linear scaling performance. One of the approaches to be tested in this work package is distributed or GRID computing utilizing the distributed computational resources of the WEBAIR consortium.

WP10: Knowledge representation, heuristics, GA, machine learning
Adaptive heuristics are designed to increase the efficiency of computational experiments for the optimization. The three main strategies are:

• Analysing the cross variance, correlation or contingency of instruments, we can identify groups that together have a higher probability to lead to successful model runs. These groupings (as synthetic allele) can be systematically exploited with standard genetic algorithms, while retaining their interdependency.
• Machine learning algorithms like ID3/ID5R (Quinlan 1979), or simulated neural networks can be used to identify efficient search strategies. They provide guidance for local heuristic search around any random test case by modifying the a priori probabilities of instruments given the pattern of constraints violations.
• Domain specific heuristics are a "simpler" version of the same principle, but defined a priori by domain experts rather than extracted a posteriori from the observed system behaviour: using first order production rules, we can formulate strategies that will guide the design of a new computational experiment based on the observed pattern of performance and/or violation of constraints.

WP11: Optimization, phase II: Discrete multi-criteria DSS:
not active in year 1.

WP12: Uncertainty analysis, robustness, GCM scenarios:
not active in year 1.

WP13: Test cases: implementation, performance, validation
They provide the real world data and feedback on all the methods and tools planned. At the same time, they provide the possibility to establish, and in fact measure, the effectiveness of the optimization process by comparing baseline scenarios with efficient solutions. The test cases will take advantage of the ongoing case studies in the WEBAIR E! 3266 project. The primary test cases will be:

• The Republic of Cyprus, national scale system with several nested grid city domains including Limassol and its harbor area;
• Seoul and Gyeonggi-do, South Korea, with one nested megacity domain for the Seoul metropolitan area.
• The Greater Tehran Area, Iran.

WP14: Dissemination activities
Dissemination strategy of OPTAIR is based on

• Extensive use of the Internet as the very implementation medium of the system;
• The contacts, networks, and distribution channels of the WEBAIR project partners in 17 countries.