The EMEP/EEA Emissions Inventory Guidebook Dr Chris Dore Chair of the TFEIP.

The EMEP/EEA Emissions Inventory Guidebook

Dr Chris Dore Chair of the TFEIP

Contents

1. Accuracy In Emission Inventories

2. Principles of Uncertainty

3. Uncertainty Tools

4. Conclusions

5. Discussion Points

Does it matter?!• Actually, it is not very important for demonstrating

compliance with targets• But key for trying to reflect the real world.

1. Accuracy

Some starting considerations…• Point sources vs area sources• Source/fuel mix• Activity data – trends with time vs absolute• EFs – variations across time series, applicability• Completeness vs guidebook• Completeness vs real world• Mapping emissions• Projections & scenarios

1. Accuracy

1. Accuracy

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

100%

NOx NMVOC SO2 NH3 PM2.5

Waste & Other

Agriculture

Processes & Product Use

Other Mobile

Road Transport

Domestic & Inst. Comb.

Industrial Combustion

Power Stations

2. Principles of Uncertainty Uncertainty analysis is generally used to represent

“accuracy”

Point sources- combination of random independent errors Area sources- one EF, prone to bias.

Propagation of Errors• Assign uncertainty to AD and EF

– from measurement, default ranges, expert judgement

• “,… root the sum of the squares…”

• Simple mathematical combination of elements to give an uncertainty for the total emission.

3. Uncertainty Tools (1)

Monte-Carlo Analysis

3. Uncertainty Tools (2)

• Uncertainty profiles, accounts for inter-dependencies...• Much better tool, but more challenging to use.

Trend Uncertainties• Standard tool used for assessing the uncertainty

in the trend included in the Guidance.

3. Uncertainty Tools (3)

Strengths• Methodologies common with GHGs (UNFCCC)• Standard mathematical approaches for assessing

uncertainty• Simple methods available.

3. Uncertainty Tools

Weaknesses• Low uncertainty does not necessarily mean good

accuracy!– incomplete inventory, use of inappropriate EFs etc.

• Uncertainty ranges applied to EFs are usually no better than a guess!– Not usually enough data points for a statistical analysis

• Error propagation analysis is too simple– Does not account for interdependencies/biases etc.

• Modellers want uncertainty on mapped emissions.

3. Uncertainty Tools

Development of New Tools• Moran’s co-efficient

– A mathematical metric of spatial autocorrelation (chess board = -1, random = 0, uniform = +1).

– Indicates adjacent grid cell dependencies

• Uncertainty of mapped emissions

3. Uncertainty Tools

Development of New Tools• Uncertainty of mapped

emissions• Combination of emissions

uncertainty with mapping uncertainty

3. Uncertainty Tools

Learning from the Past

3. Uncertainty Tools

0

200

400

600

800

1000

1200

1400

1600

1800

2000

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Em

iss

ion

s (

kto

nn

es)

NOx(1970-2011)

NOx SNAPSHOT

0

200

400

600

800

1000

1200

1400

2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

Em

iss

ion

s (

kto

nn

es)

SO2(1970-2011)

SO2 SNAPSHOT

Important Considerations• EF uncertainties are not robust enough• Error propagation analysis - too simple?• Uncertainty analysis does not indicate the ability to

represent the real world• Modellers want uncertainty on mapped emissions.

• … we need to improve what we are delivering!• … and in particular better explain what it

represents.

4. Conclusions

Some Questions1. Can we improve current EF uncertainties?

2. Should we all be using Monte-Carlo analysis?

3. Can we add to/adjust uncertainty results to give an indication of real-world representation?

4. Can tools be developed that better provide the information that users need?

5. What resources do we have to support this?

5. Discussion Points

THANK-YOU FOR

YOUR ATTENTION