use of remote sensing to validate existing vehicle emission … · 2016-04-07 · vehicle emission...
TRANSCRIPT
Use of Remote Sensing to Validate Existing Vehicle
Emission Models and to Develop New Models
CASANZ Conference, 21 September 2015, Melbourne Robin Smit, Elizabeth Somervell, Guy Coulson
The Hitchhiker’s Guide to the RSD
Robin Smit, Elizabeth Somervell & Guy Coulson
www.niwa.co.nz
• NIWA funded this study to scope out and test the best and most useful ways for validation
of the well-established New Zealand vehicle emission model VEPM (Vehicle Emissions
Prediction Model) using remote sensing data.
• This is of particular interest to New Zealand, as remote sensing data has systematically
been collected over several years in this country.
• The study reviewed 115 publications, and developed an innovative proof of concept hybrid
vehicle emission modelling approach that uses remote sensing data.
Study objectives
• Report is comprehensive,
and provides a ‘cookbook’
for RSD data analysis and
preparation
• Not possible to discuss all
relevant info in sufficient
detail in 15 min, so reader
referred to paper and report.
• Few aspects are highlighted.
• The use of remote sensing to measure vehicle emissions is not new, dating back to 1971.
Since then a number of remote sensing systems have been developed, in most cases
with very similar operating principles.
• Remote sensing has been regularly used around the world to compare model predictions
with observed values.
• However, comparing remote sensing data with vehicle emission predictions is not straight
forward several issues need to be considered/addressed before a valid comparison
can be made.
Vehicle emission model validation – remote sensing
• Laboratory measurements (engine/chassis dynamometer) using test cycles have
remained the prominent empirical base for vehicle emission models around the world.
• Other measurement methods such as remote sensing, tunnel studies and PEMS have
mainly been used for validation and other purposes (e.g. high emitter identification).
• It is often overlooked that all these methods have their own strengths and weaknesses.
Laboratory versus remote sensing
Don’t Panic!
• Nine fundamental differences that need to be considered and accounted for
in a validation study.
1. Different sampling strategies
2. Different measurement techniques
3. Different determination of emission factors
4. Different levels of detail in the on-road fleet mix
5. Different spatial and temporal resolution regarding driving conditions
6. Different meteorological (test) conditions
7. Different fuel quality
8. Different impact of ageing effects
9. Different impact of vehicle loading
Comparing models with remote sensing data
Don’t Panic!
• 1. Different sampling strategies
• 2. Different measurement techniques
• 3. Different determination of emission factors
• 4. Different levels of detail in the on-road fleet mix
• 5. Different spatial and temporal resolution regarding driving conditions
• 6. Different meteorological (test) conditions
• 7. Different fuel quality
• 8. Different impact of ageing effects
• 9. Different impact of vehicle loading
• Remote sensing typically collects ‘snapshot’ on-road emissions data from a relatively large
sample of moving vehicles, whereas laboratory testing collects drive cycle (trip) based
emissions data from a relatively small sample of both stationary and moving vehicles
under controlled conditions.
• Remote sensing set up can exclude specific vehicle classes (e.g. trucks with e.g. vertical
exhausts) or driving conditions (e.g. high speed highways).
Comparing models with remote sensing data
1. Different sampling strategies
2. Different measurement techniques
3. Different determination of emission factors
4. Different levels of detail in the on-road fleet mix
5. Different spatial and temporal resolution regarding driving conditions
6. Different meteorological (test) conditions
7. Different fuel quality
8. Different impact of ageing effects
9. Different impact of vehicle loading
• Laboratory emissions are measured with standard pollutant analysers
(FID, gravimetric filter method, chemiluminescence, NDIR)
• Remote sensing systems use the principle that the majority of gases will absorb light at
particular wavelengths (UV/IR).
Comparing models with remote sensing data
Don’t Panic!
• 1. Different sampling strategies
• 2. Different measurement techniques
• 3. Different determination of emission factors
• 4. Different levels of detail in the on-road fleet mix
• 5. Different spatial and temporal resolution regarding driving conditions
• 6. Different meteorological (test) conditions
• 7. Different fuel quality
• 8. Different impact of ageing effects
• 9. Different impact of vehicle loading
• Laboratory emission test results are directly converted into distance-based emission
factors (g/km).
• Remote sensing data are converted into emission factors expressed as grams per litre or
kg of fuel burned using a chemical mass (carbon) balance approach.
• Assumptions required regarding fuel carbon content, ‘NDIR to FID’ scaling factor for HC
and representative fuel consumption rates.
Comparing models with remote sensing data
1. Different sampling strategies
2. Different measurement techniques
3. Different determination of emission factors
4. Different levels of detail in the on-road fleet mix
5. Different spatial and temporal resolution regarding driving conditions
6. Different meteorological (test) conditions
7. Different fuel quality
8. Different impact of ageing effects
9. Different impact of vehicle loading
• Remote sensing provides detailed emissions and vehicle information for the vehicles it
measures.
• This is in contrast with emission models where emission factors are available for specific
vehicle classes only.
Comparing models with remote sensing data
Don’t Panic!
1. Different sampling strategies
2. Different measurement techniques
3. Different determination of emission factors
4. Different levels of detail in the on-road fleet mix
5. Different spatial and temporal resolution regarding driving conditions
6. Different meteorological (test) conditions
7. Different fuel quality
8. Different impact of ageing effects
9. Different impact of vehicle loading
• It is not possible to directly compare emission factors (g/ km) with remote sensing data
collected at fixed roadside locations.
• Remote sensing techniques measure concentrations and (approximately) associated
driving conditions at a particular point on the road (generally under slight acceleration) at a
high frequency for about half a second, whereas drive cycles used in laboratory testing
typically reflect 10 to 30 minutes of continuous driving over a range of driving conditions
(effectively a ‘journey’).
• Variables are required that quantify ‘driving conditions’ and link the two independent
databases.
Comparing models with remote sensing data
Don’t Panic!
• 1. Different sampling strategies
• 2. Different measurement techniques
• 3. Different determination of emission factors
• 4. Different levels of detail in the on-road fleet mix
• 5. Different spatial and temporal resolution regarding driving conditions
• 6. Different meteorological (test) conditions
• 7. Different fuel quality
• 8. Different impact of ageing effects
• 9. Different impact of vehicle loading
• Laboratory tests are typically conducted at specific temperatures and humidity, and
exclude aircon use.
• RSD measurements reflect actual ambient temperatures and humidity, and aircon use.
Comparing models with remote sensing data
• 1. Different sampling strategies
• 2. Different measurement techniques
• 3. Different determination of emission factors
• 4. Different levels of detail in the on-road fleet mix
• 5. Different spatial and temporal resolution regarding driving conditions
• 6. Different meteorological (test) conditions
• 7. Different fuel quality
• 8. Different impact of ageing effects
• 9. Different impact of vehicle loading
• VEPM is to a large extent based on overseas models, and therefore reflects the quality
and composition of the fuel used in overseas testing programmes, at the time of
measurement.
• The RSD data will reflect the local NZ fuel quality and composition.
Comparing models with remote sensing data
1. Different sampling strategies
2. Different measurement techniques
3. Different determination of emission factors
4. Different levels of detail in the on-road fleet mix
5. Different spatial and temporal resolution regarding driving conditions
6. Different meteorological (test) conditions
7. Different fuel quality
8. Different impact of ageing effects
9. Different impact of vehicle loading
• Ageing effects are normally simulated in emission models by generic mileage correction
algorithms, which are based on limited data.
• Recent research indicates that the accuracy of those algorithms in vehicle emission
models is low and requires attention.
• Ageing effects can be examined by relating remote sensing emissions to vehicle model
year and by conducting remote sensing campaigns in subsequent years and, ideally,
multiple sites.
Comparing models with remote sensing data
Don’t Panic!
• 1. Different sampling strategies
• 2. Different measurement techniques
• 3. Different determination of emission factors
• 4. Different levels of detail in the on-road fleet mix
• 5. Different spatial and temporal resolution regarding driving conditions
• 6. Different meteorological (test) conditions
• 7. Different fuel quality
• 8. Different impact of ageing effects
• 9. Different impact of vehicle loading
• Vehicle loading is recorded during laboratory testing, but the extent to which dynamometer
loading corresponds to real-world driving may be questionable.
• Actual vehicle loading is not measured with the RSD, although a few studies reported the
use of a weighing station before remote sensing measurements were carried out.
Comparing models with remote sensing data
• Proof-of-concept hybrid model created to illustrate how the rich New Zealand remote
sensing database can be used to develop a high resolution NZ emissions model.
• Refined spatial and temporal vehicle emission modelling is required to properly assess:
(reduction of) population exposure to air pollution and (health) risk
(local) management of traffic emissions (signal settings, dynamic speed limits, etc.).
• NZ does not yet have a vehicle emission model that is sensitive enough
to operate at the local level.
• High-resolution and real-time input data are becoming available and
can be collected cheaply for this type of emission modelling
(e.g. ATLAS II mobile phone app).
New hybrid micro-scale model
Don’t Panic!
1. For each vehicle class, choose a representative vehicle (fuel consumption) for which
there is CO2 emissions data.
2. Model CO2 emissions based on speed using the P∆P CO2 emissions algorithms.
New hybrid micro-scale model
3. Link these to mean RSD CO2:pollutant ratios based on VSP.
4. Apply using high-resolution/real-time drive cycle speed data.
New hybrid micro-scale model
• The hybrid model links binned remote sensing data with CO2 prediction algorithms
through an appropriate vehicle classification scheme and use of a proxy variable for
engine power (VSP).
New hybrid micro-scale model
Don’t Panic!
New hybrid micro-scale model
Don’t Panic!
• Remote sensing data is valuable in many ways.
• Provided that several data verification and preparation steps are properly conducted.
• It can be used to validate existing vehicle emission models.
• It can be used to create new high resolution vehicle emission models.
• Don’t panic; always carry a towel
Concluding remarks