Data Gaps

CGE Training Materials

National Greenhouse Gas Inventories

Addressing Data Gaps

Version 2, April 2012

Consultative Group of Experts (CGE)

Training Materials for National Greenhouse Gas Inventories

Target Audience and Objective of the Training Materials

 These training materials are suitable for people with beginner to intermediate level knowledge of national greenhouse gas (GHG) inventory development.

 After having read this presentation, in combination with the related documentation, the reader should:

 Have an overview of how to address data gaps

 Have a general understanding of the methods and tools available, as well as of the main challenges of GHG inventory development in that particular area

 Be able to determine which methods suits their country’s situation best

 Know where to find more detailed informatio n on the topic discussed.

 These training materials have been developed primarily on the basis of methodologies developed, by the IPCC ; hence the reader is always encouraged to refer to the original documents to obtain further detailed information on a particular issue.



2

Acronyms

 EF

 LULUCF

Emission Factor

Land Use, Land-Use Change and Forestry



3

Problems

 What do we do when there are gaps in the data?

 We only have data for 1995 and 2000.

 We want to switch to a Tier 2 method, but we only have disaggregated livestock data starting last year.

 The Energy Ministry stopped collecting data on natural gas flaring. What do we do?



4

Time Series Consistency

 Inventories can help you understand emissions/removals trends.

 These trends should be neither over nor underestimated, as long as can be judged.

 The time series should be calculated using the same method and same data sources in all years.

 In reality : it is not always possible to use exactly the same methods and data for entire time series.



5

Dealing with Reality

 Data gaps may occur because :



A new emission factor (EF) or method is applied for which historical data are not available

 New activity data become available, but not for historical years

 There has been a change in how the EF is developed or activity data are collected…

 … or activity data cease to be available

 A new source or sink category is added to the inventory, for which historical data are not available

 Errors are identified in historical data or calculations that cannot be easily corrected.

 These problems can be especially a challenge for agriculture and LULUCF sectors.



6

Emission Factors

 Recognize : using a constant EF does not ensure time series consistency.

 For some emission processes, emission rates may vary over time due to technological or other changes:

 No: For stoichiometric processes

 Yes: For many biological and technology-specific processes.


Training Materials for National Greenhouse Gas Inventories 7

Data Availability

 Changes and gaps in data:



More disaggregated or other improvements in data collection (e.g., better surveys in future years)

 Missing years or data no longer collected.

 Periodic data:

 Data collection only every few years or on regional rolling basis (i.e., each year a different region surveyed)

 Common for the LULUCF sector (e.g., forest inventory only done every five years).

 No data?



Splicing and Gap-filling Approaches

 Splicing : combining or joining more than one method or data series to form a complete time series:



Addresses a change in method (e.g., when Tier 2 method can only be applied to new data but Tier 1 is still used for historical data)

 Fills gaps due to the collection of periodically collected data.

 Use surrogate or proxy data to “create” data that are otherwise missing.



Splicing and Gap-filling Approaches

 Overlap

 Surrogate data (i.e., correlated proxy data)

 Interpolation/extrapolation

 Trend extrapolation.



Overlap Approach

 Calculate emissions or collect data using both old and new methods/systems for several years:

 Can be used with 1 year overlap, but should be done with great caution.

 Investigate the relationship between old and new time series for years of overlap.

 Develop a mathematical relationship and use it to recalculate historical data to be consistent with new methods/systems.



11

Overlap - Consistent Relationship



In this example, it is acceptable to use the overlap adjustment.



Overlap - Inconsistent Relationship

 In this example, it is not acceptable to use the overlap approach because there is too much variability between the relationship.



13

Overlap Approach

 Where there is a consistent relationship, the default is to use a proportional adjustment of old estimates/data to be consistent with new.



14

Overlap Approach

 Example 1: Use the overlap approach to estimate GHG emissions for years 2001 –2003, using the data below.

Tier 1 quantified

Tier 2 quantified

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

4,000 4,000 4,100 4,200 4,800 4,900 5,000 4,800 4,900 5,000

4,035 4,598 4,410 4,500 4,320 4,513 4,790



Overlap Approach

 Example 1: Step 1

Tier 1 quantified

Tier 2 quantified

Ratio Tier 2 : Tier 1

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

4,000 4,000 4,100 4,200 4,800 4,900 5,000 4,800 4,900 5,000

4,035 4,598 4,410 4,500 4,320 4,513 4,790

0.96 0.96 0.90 0.90 0.90 0.92 0.96

For each year, calculate the ratio between Tier 2 and Tier 1

E.g. for year 2010:

4790/5000 = 0.96



Overlap Approach


Tier 1 quantified

Tier 2 quantified


2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

4,000 4,000 4,100 4,200 4,800 4,900 5,000 4,800 4,900 5,000

4,035 4,598 4,410 4,500 4,320 4,513 4,790

0.96 0.96 0.90 0.90 0.90 0.92 0.96

Calculate average and standard deviation

Average = 0.93

Standard deviation = 0.027

Low variability



Overlap approach seems appropriate



Overlap Approach


Tier 1 quantified

Tier 2 quantified


2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

4,000 4,000 4,100 4,200 4,800 4,900 5,000 4,800 4,900 5,000

3,713 3,713 3,806 4,035 4,598 4,410 4,500 4,320 4,513 4,790

0.96 0.96 0.90 0.90 0.90 0.92 0.96

Apply average to calculate missing data:

Year 2001: 4,000 * 0.93 = 3,713

Year 2002: 4,000 * 0.93 = 3,713

Year 2003: 4,100 * 0.93 = 3,806



Overlap Considerations

 Remember, it is crucial to have multiple years of overlap to apply properly

 This method should not be applied blindly . You should do your best to understand the relationship between the old and new methods:

 E.g., why does the old method consistently give results that are 10 to 15% less than the new method?

 If you cannot explain the difference then you are not sure that the new method is actually better!

 Just because a method/model is more complicated does not mean it is more accurate!



Surrogate Data Approach

 Find a surrogate (i.e., proxy ) variable that is well-correlated with missing data:

 Can be used for missing activity data, for EFs (that change each year) or for emission estimates:

 Example: Automobile license payments may be well-correlated with petrol use.

So license data may serve as a surrogate for petrol consumption.

 This approach builds on techniques used in statistical (e.g., econometric) analysis:

 Regression techniques are valuable to identify potential surrogate parameter(s)

 Correlation analysis.



20

Surrogate Approach Steps

 Identify potential surrogate/proxy variables.

 If you have some actual data, calculate simple correlation coefficients:

 You should have more than one year of actual data to establish a relationship with the surrogate parameter.

 If the correlation is not obvious, then consider more sophisticated regression techniques to see if a relationship between actual and surrogate parameter can be found.

 If you have no actual data , then you will need to justify why the surrogate parameter is a legitimate proxy for actual variable(s).



21

Surrogate Approach

 This formula assumes a simple proportional relationship between the surrogate and actual variables.



22

Surrogate Approach

 Example 2: Using number of vehicles as a surrogate, estimate

CO

2 emissions for the variables below .

Target variable

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

Road transportation CO

2

Known surrogate variable

2001

Number of road vehicles in circulation (000s)

2002

3,520 3,520

2003 2004 2005

3,60 3,696 4,224

2006 2007 2008 2009 2010

4,31 4,400 4,224 4,312 4,400

Vehicle-related data from different studies:

•Transportation study 1 

2009 CO

2 emissions for average car = 190 g/km, average km per year = 13,000

•Transportation study 2  2008 CO

2 emissions for average road vehicle = 4,410 kg CO

2 per year


2 emissions for average passenger vehicle = 220 g/km, average km per year = 16,000

•Transportation study 4  2008 freight vehicles are 5% of all road vehicles and emit on average 550 g/km.



23

Surrogate Approach



Example 2: Step 1

Vehicle related data from different studies


2 emissions for average car = 190 g/km, average km per year = 13,000


2 emissions for average road vehicle = 4,500 kg CO

2 per year


2 emissions for average passenger vehicle = kg CO

2

3520 per year

•Transportation study 4 

Average for all road vehicle is more

2008 freight vehicle are 5% of all road vehicles and emit on average 550 g/km

Assess potential surrogate appropriate when focusing on road transportation emissions as a whole parameters

Year km per annum average km/year

Average emission factor gCO

2

/km

Average emissions per vehicle kgCO

2

/vehicle

All road vehicles

2008

4,410

All passenger vehicles

2007

14,000

200

2,800

Cars only

2009

13,000

190

2,470

Additional data collection: Traffic study 5

 Average km travelled per year by freight vehicles = 65,000 – 74,000 km

[ 4410 – 2800 * ( 100% – 5%) ] / 5% = 70000 i.e.`, if freight vehicles on average travel 70,000 km per year, both data on “all road vehicles” and “all passenger vehicles” are accurate.


24


Surrogate Approach



Example 2: Step 1 Use surrogate variable and parameter for calculation

Known surrogate variable

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

Number of road vehicles in circulation (000s) 3,520 3,520 3,60 3,696 4,224 4,310 4,400 4,400 4,410 4,450

Apply surrogate parameter (4,410 kg

CO

2

/vehicle) and calculate emissions

E.g. 4,224,000 vehicles * 4,410 kg

CO

2

/vehicle / 1000 = 18,628,000 t CO

2

Target variable CO

2 emissions in thousand metric tons road vehicles emissions

2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

15,523 15,523 15,911 16,299 18,628 19,016 19,404 19,404 19,448 19,625



25

Interpolation and Extrapolation

 Interpolation : Filling gaps in existing time series.

 Extrapolation : Filling gaps at end or beginning of time series.

 Techniques :

 Linear or nonlinear, justify choice

 Should not be used for variables that have large variability from year to year.



Interpolation Example



27

Extrapolation Example






Example 3: Using the interpolation technique, estimate the

GHG emissions for years 2004 –2006

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

GHG emission source x 3,800 3,920 4,030 4,135 4,235 4,655 4,770 4,880 4,975






Example 3: Step 1

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010


Analyze data and assess applicability and type of interpolation technique desired

Linear interpolation seems appropriate for this data set






Example 3: Step 2

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010


Calculate difference in GHG emissions between last year before the gap and first year after the gap

 4,655 – 4,235 = 420






Example 3: Step 3

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010



 4655 – 4235 = 420

Calculate length of the gap

 2007 – 2003 = 4 years






Example 3: Step 3

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010



 4655 – 4235 = 420


 2007 – 2003 = 4 years

Calculate average change in emissions per gap year

 420 / 4 = 105






Example 3: Step 3

GHG emission source x

1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

3,800 3,920 4,030 4,135 4,235 4,340 4,445 4,550 4,655 4,770 4,880 4,975


 4655 – 4235 = 420


 2007 – 2003 = 4 years

Calculate average change in emissions per gap year

 420 / 4 = 105

Calculate total emissions for gap year(s) by adding the average change per year

2004 emissions = 4235 + 105 = 4340

2005 emissions = 4340 + 105 = 4445

2006 emissions = 4445 + 105 = 4550



Splicing and Gap-filling Summary

Approach

Overlap

Applicability Comments

Data necessary to apply both old and new method must be available for at least one year, preferably more.

Only use when overlap shows pattern that appears reliable

Surrogate data Missing date is strongly

Trend extrapolation correlated with proxy data

Interpolation For periodic data or gap in time series

Beginning or the end of the time series is missing data

Should test multiple potential proxy data variables

Linear or non-linear interpolation.

Only use where data shows steady trend

Only use where trend is steady and likely to be reliable. Should only be used for a very few years



Summary Comments

 Preferred approaches are overlap and surrogate, because:



They are based on actual data

 Interpolation and extrapolation are effectively projections that assume certain trends in the absence of data.

 Similarly in research, it is not good practice to simply apply a gap-filling method blindly :

 You should understand why your approach is justified and be able to explain it transparently.

 Ask yourself: will what I am doing stand up to peer review in a technical journal?



36

Thank you



Data Gaps

CGE Training Materials

National Greenhouse Gas Inventories

Addressing Data Gaps

Example 2: Step 1

Example 2: Step 1 Use surrogate variable and parameter for calculation

Example 3: Using the interpolation technique, estimate the

GHG emissions for years 2004 –2006

Example 3: Step 1

Example 3: Step 2

Example 3: Step 3

Example 3: Step 3

Example 3: Step 3

Thank you

Related documents

Products

Support

Data Gaps

CGE Training Materials

National Greenhouse Gas Inventories

Addressing Data Gaps

Example 2: Step 1

Example 2: Step 1 Use surrogate variable and parameter for calculation

Example 3: Using the interpolation technique, estimate the

GHG emissions for years 2004 –2006

Example 3: Step 1

Example 3: Step 2

Example 3: Step 3

Example 3: Step 3

Example 3: Step 3

Thank you

Related documents

Add this document to collection(s)

Add this document to saved

Suggest us how to improve StudyLib