5.2 The RIAT+ System
The RIAT+ system was developed during the EU OPERA project (www.operatool.eu)
and it is intended to help regional decision makers select optimal air pollution
reduction policies that will improve the air quality at minimal (industrial and/or
external) costs. To achieve this, the system incorporates explicitly the specific features
of the area of interest with regional input datasets for the:
• Precursor emissions of local and surrounding sources;
• Abatement measures (technical and non-technical) described per activity sector
and technology with information on application rates, emission removal efficiency and cost;
• The effect of meteorology and prevailing chemical regimes through the use of
site-specific source/receptor functions.
The system runs as a stand-alone desktop application and can be downloaded
from the OPERA project website (http://www.operatool.eu/download/). The
package is distributed with a personal, non-exclusive and royalty-free license and
has been applied in various regions, such as Emilia-Romagna (Carnevale et al.
2012) and in Alsace (Carnevale et al. 2014).
The RIAT+ software implements both the possible decision pathways introduced in Chap. 2 in the light of the classical DPSIR (Drivers-Pressures-StateImpact-Responses) scheme, adopted by the EU:
• Scenario analysis, where emission reduction measures are selected on the basis
of expert judgment or Source Apportionment and then tested through simulations of an air pollution model.
• Optimisation, where the set of cost effective measures for air quality
improvement are automatically selected by solving a multi-objective optimization problem.
To allow both approaches to be implemented in a fast and handy way (i.e. to be
able to support a real-world discussion about possible options) a key feature of
RIAT+ is a S/R model used to relate emissions (pressures) to a suitable air quality
indicator, AQI (state). In principle, such a S/R model should be a full complex
chemical transport model, but in practice this would be impossible for the computational burden that such models imply. So, within RIAT+, the relations between
emissions and air quality indicators are expressed by means of Artificial Neural
Networks (ANNs), that, in turn, are tuned to replicate the results of deterministic air
quality model. ANNs are often referred to in this context as “surrogate models”.
The reason for this choice is that neural networks are known to be suitable to
describe a nonlinear relationship between data, such as those theoretically involved
in the formation of air pollution. Their identification procedure requires two steps:
(1) the definition of the specific structure, and (2) the calibration of the parameters
to the specific application. The selected structure of the ANNs must be able to retain
what are considered to be the essential features of the original model. As the value
86
C. Carnevale et al.
The RIAT+ system was developed during the EU OPERA project (www.operatool.eu)
and it is intended to help regional decision makers select optimal air pollution
reduction policies that will improve the air quality at minimal (industrial and/or
external) costs. To achieve this, the system incorporates explicitly the specific features
of the area of interest with regional input datasets for the:
• Precursor emissions of local and surrounding sources;
• Abatement measures (technical and non-technical) described per activity sector
and technology with information on application rates, emission removal efficiency and cost;
• The effect of meteorology and prevailing chemical regimes through the use of
site-specific source/receptor functions.
The system runs as a stand-alone desktop application and can be downloaded
from the OPERA project website (http://www.operatool.eu/download/). The
package is distributed with a personal, non-exclusive and royalty-free license and
has been applied in various regions, such as Emilia-Romagna (Carnevale et al.
2012) and in Alsace (Carnevale et al. 2014).
The RIAT+ software implements both the possible decision pathways introduced in Chap. 2 in the light of the classical DPSIR (Drivers-Pressures-StateImpact-Responses) scheme, adopted by the EU:
• Scenario analysis, where emission reduction measures are selected on the basis
of expert judgment or Source Apportionment and then tested through simulations of an air pollution model.
• Optimisation, where the set of cost effective measures for air quality
improvement are automatically selected by solving a multi-objective optimization problem.
To allow both approaches to be implemented in a fast and handy way (i.e. to be
able to support a real-world discussion about possible options) a key feature of
RIAT+ is a S/R model used to relate emissions (pressures) to a suitable air quality
indicator, AQI (state). In principle, such a S/R model should be a full complex
chemical transport model, but in practice this would be impossible for the computational burden that such models imply. So, within RIAT+, the relations between
emissions and air quality indicators are expressed by means of Artificial Neural
Networks (ANNs), that, in turn, are tuned to replicate the results of deterministic air
quality model. ANNs are often referred to in this context as “surrogate models”.
The reason for this choice is that neural networks are known to be suitable to
describe a nonlinear relationship between data, such as those theoretically involved
in the formation of air pollution. Their identification procedure requires two steps:
(1) the definition of the specific structure, and (2) the calibration of the parameters
to the specific application. The selected structure of the ANNs must be able to retain
what are considered to be the essential features of the original model. As the value
86
C. Carnevale et al.