• About this site|
• Site index|
• Contact us|
• Jobs

Review of New Zealand's Net Position

Executive summary

1 Introduction

2 Sector reviews

3 Impacts of New Zealand's Climate Change Policies

4 Compilation and processes used for the Net Position report 2005

5 Conclusions and recommendations

6 References

Appendix A: Additional analysis of the New Zealand SADEM Energy Model

Appendix B: Priority ranking of SADEM Energy Model improvements

2 Sector reviews

2.1 Energy and industry

2.1.1 SADEM modelling approach and use in a climate change context

This section gives a brief overview of our understanding of the SADEM model and some general comments on its appropriateness for projecting greenhouse gas emissions in New Zealand. We are aware that COVEC have produced a detailed report for the Ministry of Economic Development (MED) reviewing all aspects of SADEM from an energy modelling perspective (Covec 2005). Not wishing to repeat this exercise, the focus of our review has been on the use of SADEM in a climate change context, although we provide some comments on COVEC's recommendations in Section 2.1.9.

SADEM is a partial equilibrium model of the entire New Zealand energy system, solving in five-year increments to 2025. The model has been developed over a number of years by the MED. Physically, it is a collection of relatively straightforward spreadsheets which describe energy supply by fuel and energy demand by sector, together with a more complex electric power simulation tool. The figure below (Figure 2) illustrates the SADEM model with the arrows depicting the upward movement of prices and the downward movement of useful energy demands respectively.

The figure shows that oil and coal supply are assumed to be perfectly elastic (arrows only point up). The prices for these fuels are fixed by assumption as New Zealand is assumed to be a price taker in these markets. Similarly, heavy industries (forestry, basic metals and petrochemicals) are assumed to have inelastic demand i.e. their demand is fixed by assumption and does not respond to price.

Therefore, SADEM's equilibrium for supply and demand only really applies to the natural gas, electricity, light industry/commercial demand and residential demand modules. The equilibrium process for these four modules involves setting the natural gas resource and price, building up electricity plant dispatch based on variable cost, adding new electricity investment if required, and finally equating this supply with energy demands based on econometric analysis on historical data.

SADEM does not include any feedbacks from energy supply to aggregate or sectoral economic output and so, while it can give insights into direct changes relating to the energy sector, it cannot give any insights on the impact of these changes on the rest of the economy. Instead of technological detail (outside of the electricity sector), the model uses econometric equations to link future demands to changes in key parameters such as GDP and prices, based on their historical relationship. While this has the advantage that it anchors changes in energy demand to recent experiences (very relevant when the focus is CP1), it does not take account of any different responses that might occur if prices or the policy environment were very different from those seen in the past (e.g. responses to a sustained price signal from a carbon tax could be very different from those observed when instability in energy markets gives rise to short-term high prices). The lack of technology details also makes it difficult directly to evaluate such topics as technology or fuel efficiency standards, cost and performance improvement in technologies, and an understanding of potential (non-price) constraints on the growth of energy supply and demands (e.g. saturation effects).

A potential further problem for longer-term projections is the static nature of the electricity model (solving for each discrete time period based on marginal costs) and the impact this has on plant entry and exit. Our feeling is that this could potentially lead to discontinuities in the results (plants being built and then not used if prices change in subsequent periods), which we understand are avoided at the moment by user intervention. It may be worth considering whether some kind of dynamic element to the model would be an improvement over the current situation (this would largely depend on how frequently such user intervention is necessary).

In conclusion, although simple in its approach, we feel that SADEM gives a reasonable overall depiction of the specifics of the New Zealand energy sector. The model emphasizes key features of the energy supply sectors, (e.g., the heavy reliance on hydro-electricity and the retirement of the existing Methanex gas-methanol plant). Most of the model complexity relates to its treatment of the electricity sector. While developments in the electricity sector will be important for future trends in greenhouse gas emissions, there are other sectors such as transport and industry that are very significant sources of greenhouse gas emissions and which therefore also deserve a more sophisticated treatment. Given that the first commitment period is a relatively short time away, this simple approach on the demand-side should not give rise to many problems in projecting the net position over CP1. However, we would have more concerns about any longer-term projections made using the current formulation of the model. In addition, the lack of bottom-up detail on the demand side makes it difficult to directly model the impacts of some key policies in the New Zealand Climate Change Programme. This means that the savings from some policies have to be calculated off-model and then subtracted from the projections. While this approach is not uncommon with respect to international practices (for instance, the effects of some policies within the UK climate change programme are handled in this way), it does mean that particular care needs to be taken to ensure consistency and transparency between the assumptions used in the projection model and any off-line calculations. This topic is discussed further in section 3.

Later sections in this chapter outline our recommendations for improving SADEM both in terms of projecting emissions over the longer term and its ability to model policy impacts. We also make some suggestions about using other modelling approaches as a complement to SADEM.

2.1.2 Reasonableness of input assumptions/'reality check'

The major input assumptions to SADEM, including GDP, exchange rates and fossil fuel prices appear reasonable. GDP data have been taken from official Treasury forecasts of economic growth and, while we cannot comment on the robustness of these forecasts, this source would seem entirely appropriate [We note that the 2005 Treasury forecasts for GDP have now been published and are somewhat different to the previous year‘s. However, when averaged over the period to the CP1 these differences appear unlikely to have a significant impact on the energy projections.]. Oil prices have been based on forecasts from the International Energy Agency, which is again one of the most authoritative sources for such data. Coal and gas prices, at least over the period to CP1, are based on the particular characteristics of the New Zealand's indigenous and imported sources and discussions with MED indicate that a considerable amount of thought has been given to these.

Other important assumptions include the treatment of taxation on energy services, the consideration of degree data from four New Zealand cities and its weighting by projections of population, the discount rate, and the methodology for translating wholesale to final energy prices for the various end-use sectors. None of these give us any great cause for concern, although in relation to the final item we note that this is calculated based on prior estimates of the annualised costs of infrastructure investment and operation and results in a significant uplift to costs (e.g. an average cost for residential electricity of NZ¢ 5.87/kWhr) relative to impacts such as the NZ¢ 0.6/kWhr cost reduction resulting from the PREs. Even small percentage changes in infrastructure costs could therefore be significant and this suggests to us that it may be worth considering more explicit depiction of infrastructure elements in the model (including for electricity, refineries and LNG terminals).

The model tracks emissions of CO₂ from fuel use and processes (hydrogen, cement and lime production). We understand that the emission factors for fuel combustion are the same as those used in the emissions inventory, which itself is subject to international review, and so we have not looked at these in detail. However, one question that came up in discussions with MED was whether different emission factors were needed for indigenous and imported coal for electricity production as currently a single emission factor appears to be used. Projections for CO₂ emissions from processes are linked to forecasts of output, essentially assuming a constant emission factor. As long as the output forecasts are robust, then this approach would seem sensible.

Non-CO₂ greenhouse gas emissions from fuel combustion are not modelled directly. Rather they are estimated at 5 % of the total CO₂ emissions based on the historical relationship between CO₂ and non-CO₂ emissions in from energy and industrial process sectors. While this source of emissions is relatively small, we feel that this approach could be improved upon by applying the relevant emission factors for methane and nitrous oxide to the projections for the different fuels. Such an approach would be consistent with the UNFCCC Tier 1 methodology and would be able explicitly to take account of changes in the fuel mix, which might cause the historical CO₂/non-CO₂ emission relationship to breakdown.

2.1.3 Electricity and renewables

As noted earlier, this is the most complex part of model. Electricity supply is modelled firstly by assuming a simplified load curve (6 periods broken out by 4 seasons). Available hydro resources are calculated stochastically based on a triangular distribution of seasonal rainfall, and ensuring that remaining water storage is in accordance with existing operating practices (i.e., a water contour table based on which plant type is dispatched first). Thermal and renewable (with overall constraints on supply) sources are dispatched next, all in order of short run variable cost, until pumped storage schemes can meet the peak load duration curve. Then, if energy demands are not met, new investments are made in increments of electricity plant. As with the rest of the model, there is no geographical consideration of units.

Such electricity dispatch models based on variable costs are common, although we have a number of observations on this approach. Firstly, there appears to be no automatic check that plants will cover their avoidable fixed costs. Secondly, the model assumes that if the firms cannot cover their avoidable costs, they will not operate - but our understanding is that the model does not distinguish between those that might be mothballed in such a situation, as against being retired permanently. Thirdly, there is no information on the age of electricity plants embodied in the model, although we understand that manual intervention can be used to retire a plant for non-variable price reasons. While for short-term projections none of these issues is necessarily a particular problem, we would suggest that for longer-term projections (say greater than 10 years) a better depiction of plant entry/exit would be of most use.

2.1.4 Natural gas

The derivation of natural gas prices is an important determinant of fuel use and switching and hence overall carbon emissions. The price path for natural gas is determined by a forecast of the year at which domestic gas resources are exhausted and interpolates between current prices and international LNG prices as a backstop. However, the methodology for defining the price path appeared somewhat arbitrary, and does not appear take into account any analysis of expected risk and market strategy of the monopolistic supplier. Although the chosen price path may not be unreasonable, it seems to us that a number of alternate price paths could be just as easily defined and it may be interesting to see whether alternative specifications would have any material effect on emissions over the short or long term.

In addition, given that New Zealand will become so heavily dependant on LNG imports, more in-depth analysis of this market will clearly become relevant. For example, whether there is likely to be a global LNG market in the short to medium term, or more likely that regional markets will predominate. Finally, as noted previously, more detailed infrastructure modelling would help look at questions relating to the need for future LNG terminal capacity.

2.1.5 Exogenous fuel supplies

The exogenous fuel supply modules for oil and coal appear reasonable in the context of New Zealand being a price taker for oil imports and having large domestic coal reserves in an isolated market. As with other sectors we would be interested in some discussion of the role of future infrastructure constraints/new capacity and its impact on final energy costs to end-users.

2.1.6 Equilibrium demand sectors

Econometric estimation of energy demands by sector and iteration with the modules for available natural gas and electricity supply represents the final step of the equilibrium algorithm.

Residential

In the residential sector a two-step approach is used to first estimate overall demand and then to attribute this demand to individual fuels. Generally the signs and magnitudes of the independent variable coefficients appear reasonable (further discussion of the econometric estimation is provided in Appendix A). However, we believe that the inclusion of degree-days may not be particularly helpful. Firstly, the formulation gives equal weighting to both cooling and heating degree days, without distinguishing between them, whereas in reality the former would presumably necessitate electric air conditioning while the latter may require other fuels for space heating. Secondly, the degree day variable is not significant at any reasonable level, indicating that it is not a major driver of energy demand. While we understand that the main reason for the including the parameter is to reflect the changing geographical distribution of the population, we would suggest that its use is reviewed.

We are aware that COVEC have already made some recommendations as to possible improvements (see Section 2.1.9), but in addition we suggest that some comparison (if not incorporation) with a bottom-up approach to modelling energy in households could add valuable insights on future trends and aid policy development. We note that BRANZ is developing HEERA (Household Energy Efficiency Resource Assessment), a bottom-up scenario type model, which is supported by a database containing information on dwelling types, end-use and appliance levels. Such a model should provide a physical interpretation for trends in energy demand and so identify where saturation effects may lead to the breakdown of historical relationships between energy demand and other factors such as GDP. In addition, it would be able explicitly to model policy options relating to technology performance, such as efficiency standards and insulation levels.

Other Industry and Commercial

The econometric estimation process for the light industry and commercial energy demand is also a two-step process (overall demand and then fuel shares), although in this sector only a simple fuel share trend backcast from 2025 is used in the second step. Once again all coefficient signs and magnitudes appear reasonable, although as with households there seem to be some weaknesses with including degree days as an explanatory variable and we would suggest this is reviewed.

It was somewhat surprising that GDP is not a more significant parameter in this model. However, we suspect that this is because an aggregate measure of GDP is not an accurate representation of the changing economic patterns within this sector and hence energy use. We would agree with COVEC that a more disaggregated structure would be helpful in projecting energy demand, particularly over the longer term. In particular, separately identifying key industrial sectors and commercial activities would be useful as these sectors are likely to be rather heterogeneous in terms of their energy intensity. We would also suggest looking at how the fuel shares could be made endogenous so that the model is able to fuel switch in response to changing relative fuel prices, such as those arising from the imposition of a carbon tax.

2.1.7 Non-equilibrium demand sectors

Transport

Transport is the most significant sector in New Zealand for energy-related greenhouse emissions and yet its depiction within SADEM is relatively simplistic. Projections are either econometrically estimated using only GDP and input prices (road transport), or are simply extrapolated from current trends (non-road transport).

Focussing on the estimation of road demand (as the largest transport component), the SADEM petrol model seems to give sensible results, but we have a number of issues with the diesel model. Firstly, there is no split in the model between freight and passenger demand for diesel and so no information about the underlying reasons for the increase in diesel demand. Secondly, aggregate diesel demand is trended upwards via use of a dummy variable to better track historical trends. This formulation gives us some problems, as it is clearly significant, but imparts no information as to what is driving this post-1993 change. Dummy variables are of considerable use in modelling, as long as they clearly relate to alternate states of the world and can provide insight into what is causing a change. We would therefore recommend that this variable is better defined or the model be respecified to shed actual insights into the underlying causes of trend changes. We would also have preferred to see an explicit link between the petrol and diesel sub-models, as one could expect that rapidly increasing diesel use might at least partially be compensated by a substitution away from petrol consumption.

Transport, as with households, is a sector where we feel a bottom-up approach to projecting energy demand should at the very least be used for comparison with the econometric approach, if not used as a future replacement. Currently, the econometric estimation does not take into account fuel efficiencies, vehicle utilization, current fleet structure or penetration rates for new vehicles in the market - all types of information that a bottom-up method would need to consider and which could be useful in modelling future policy initiatives (e.g. on biofuels). We are aware that MED and MOT are discussing updating MOT's bottom-up transport model and we would strongly encourage greater integration of this and the SADEM model as a matter of priority.

Heavy industry

At present, projections of heavy industry energy demand are determined exogenously to SADEM, via detailed consultation with industry associations. While this does approach is useful in capturing specific policy and industry events and trends, it has the disadvantage that this sector is not treated in a consistent way with others in the SADEM model in relation to the impacts of price changes and policy impacts e.g. response to a carbon tax. We understand that MED has commissioned COVEC to look in more detail at the treatment of the heavy industry sector and would hope that this study can suggest ways of bring the treatment of this sector more into line with the rest of the model.

2.1.8 Treatment of uncertainties

Three different energy scenarios have been identified for the New Position report: most likely, high emissions, and low emissions, which are differentiated based on GDP forecast and oil and coal prices. While such scenario analysis is useful in exploring a range of future developments, it only considers one aspect of uncertainty. Recently, the IPCC have issued a guidance note on the treatment of uncertainty to its lead authors of the forthcoming revised greenhouse gas inventory guidelines. This guidance identifies three main sources of uncertainty:

• Unpredictability
• Structural uncertainty
• Value uncertainty

Table 1. A simple typology of uncertainties [Taken from ’Guidance Notes for Lead Authors of the IPCC Fourth Assessment Report on Addressing Uncertainties‘ , Intergovernmental Panel on Climate Change, July 2005.]

The scenario analysis undertaken with SADEM seeks to explore the element of unpredictability. The global economic system is extremely complex and its impact on New Zealand GDP and fossil fuel prices is inherently unpredictable and so it is quite reasonable to explore alternative assumptions for these parameters.

However, when using different sets of variables (rather than changing a single variable) as inputs into a scenario, it is important but difficult to disentangle the relationships between these selected variables. For example, while high emissions scenario assume high GDP forecast coupled with low prices of oil and coal, it might equally be argued that strong global economy growth leading to high oil prices be linked with high rates of New Zealand GDP, especially for such an export orientated economy. It may therefore be better in future to conduct a large number of sensitivity runs on individual and sets of parameters in order to get a better understanding of which changes have the most impacts on the results. These could then be used in a Monte Carlo type analysis to obtain a distribution of results.

It may also be helpful to consider the other two types of uncertainty identified by the IPCC. For instance, structural uncertainty could be addressed by comparing the results of key sectors in SADEM with bottom-up models using the same input assumptions and some indication of value uncertainty could be obtained from looking at the results of the econometric estimation [In this context it is interesting to look at the experience with the UK energy model. In order to get a first estimate of the uncertainties introduced into UK projections by the uncertainty in the values of parameters fitted to historical data, the DTI used a simplified model of energy consumption, which postulated aggregate energy demand as a function of price, income and temperature. They considered that the parameter uncertainty in the simplified model may give some indication of the potential uncertainty in the wider model. The uncertainty associated with the fitting process was calculated from standard output statistics and suggested a standard error for the projections indicating a forecast interval of the order of ±6% at two standard deviations in 2010 (DTI, Energy Projections for the UK, Energy Paper 68, 2000, ISBN 0115154965 ).].

2.1.9 Reasonableness of COVEC comments and proposals for future work

The COVEC team appear to have carried out a thoughtful review of the SADEM model and made a number of sensible suggestions (Covec 2005). It is worth remembering that their review focused on considering SADEM as an energy model to support MED activities, rather than its particular application to produce greenhouse gas emissions for the net position report. Therefore while many of their comments are in line with our own review, in other places there may be a difference of emphasis.

In particular we strongly endorse the following COVEC conclusions and recommendations:

• Bottom-up methodologies are useful for policy analysis, particularly for non-price instruments. These can be undertaken off-line and integrated with the business-as-usual results from SADEM.
• Revisit usage of the degree day variables in residential and OIC sectors (is its estimation stable?)
• Explore disaggregation of sectors in OIC, possibly at the 1-digit ANZSIC level.
• Liase with Ministry of Transport to consider roles and responsibilities for transport energy demand modelling going forward
• Consider the use of a modular demand function for estimating petrol and diesel demand (i.e., fuel demand = vehicle fleet x utilisation x fuel efficiency)

We are less convinced, from the point of view of projecting greenhouse gases, that some of the suggested improvements in the electricity sector should be a priority. This part of the model already seems to have a good representation of the New Zealand situation and while improvements are no doubt possible (we have suggested some ourselves), on balance we would prefer to see more attention paid to an accurate and integrated depiction of heavy industry.

2.1.10 MED Priority Ranking of Model Improvements

MED have provided a priority listing of their future planned work programme with the SADEM model. We are pleased to see the 'A' ranking given to further analysis of PRE's and the heavy industrial demand project. Of the 'B' ranked projects we strongly support the proposed work on the vehicle fleet model and work to integrate the demand models and fuel shares models for OIC and residential and recommend that this also include reviewing the use of degree days. Additionally, we would hope that some comparative work could be performed with the BRANZ HEERA residential energy model. We would also like to see some work on the explicit representation of energy infrastructures, or at least clearer explanation of how these costs are attributed to the various fuels in different sectors. Further comment on MED's list of priorities for the SADEM model is given in Appendix B.

Finally general consideration should be made to two overview questions. Firstly, what is the coverage of this model? As a whole systems energy model, should it accept off calculation inputs and constraints from more detailed sectoral models and industry stakeholders or should these be integrated where possible? Secondly, what is the time frame of this model? Is it focused on near-term trends or is it a mid to long-term energy model? This has important implications for consideration of structural economic changes and new technology penetration and use and may point to whether further bottom-up detail needs to be incorporated.

2.2 Land use, land use change and forestry

2.2.1 General comments on the LULUCF sector

The descriptions of methods and data in the NZ National Inventory Reports show that the LULUCF calculations are likely to be well founded. However, the (summary) information on calculations of "Removal units from sinks" described in Section 5 of the Net Position report (MfE 2005a) is poorly described. A single document providing the details of how the projections were made should be prepared and maintained as future revisions are made.

In NZ the order of magnitude of removals from sinks is the same as emissions from other sectors. This is probably unique in the international context. Sensible improvements to methods and clarifications to "Kyoto rules" and subsequent revisions of removal estimates are therefore likely to have greater apparent importance than for other countries where LULUCF is a small component of GHG emissions and removals.

It is good to see that the NZ government is investing in a purpose built system for Carbon Accounting. Uncertainty in estimating emissions and removals due to land use, land-use change and forestry can be considerable. A unified system to make the appropriate measurements will ensure that the required credits will be eligible under the Protocol.

It is a common experience to UNFCCC Annex 1 countries that as scientific knowledge increases the magnitude of net emissions or removals due to land use, land-use change and forestry often tends to become smaller. It is therefore not surprising that the projected removal units in NZ for CP1 have fallen from 105 Mt CO₂ eq estimated in 2003 to 71 Mt CO₂ eq estimated in 2005.

The largest change between the estimates of 2004 and 2005 is due to recognition that some of the forests planted since 1990 are not eligible to be used under the Kyoto Protocol (KP) Article 3.3 i.e. they are not truly "Kyoto Forests" because they have been established within existing areas that may be deemed to be forestland. The correction to exclude removals associated with these ineligible forests appears to be presented as an emission in "Removal units from sinks" described in Section 5 of the Net Position Report (MfE 2005a). This may highlight the issue, but is confusing in that the situation is of a smaller removal rather than combined removals and emissions.

Emissions offsets under CP1 for NZ have been limited to those under Article 3.3 of the Kyoto Protocol. Article 3.4 Forest Management may provide another small but useful sink credit. The amount that NZ can claim is limited to 3.7 Mt CO₂ eq over CP1 but given the importance of credits this could still be useful.

A simple assessment of the relative risks to estimates of future net removal units from sinks due to a range of issues in presented in Table 1.

2.2.2 Comments and review of approaches used for LULUCF sub-sectors

Forest planting since 1990

The methodologies, drivers and estimated removals associated with forest planting are good.

The "Most likely" scenario for the number of removal units over the first commitment period (CP1) is estimated to be 95.3 Mt CO₂ eq. This value appears to be well founded. The methodology used is scientifically sound and the data used as input to the method is taken from a reliable source.

These units are however for all forests planted since 1990. Under Article 3.3 of the Kyoto Protocol (KP) eligible forest must be established on land that has not recently been "forest". The definition of "forest" under the KP is such that new forest planting on NZ shrublands would need to be excluded. Accurate estimates for the area to be excluded will not be available until the NZ Carbon Accounting System (e.g. MfE 2005c) is in place but pilot studies at Marlborough (Trotter et al 2004) have provided some preliminary estimates of the fraction of new planting that may be ineligible. In the 2005 estimate of projections 16% of all removals are therefore assumed to be ineligible. This is 14.7 Mt CO₂ eq over CP1 and is presented separately as a negative value (implying an emission) in the report. It would seem more appropriate to report the 'Most Likely' eligible sink credit as 80.6 Mt CO₂ eq (i.e. 95.3 less 16%).

The pilot study highlights the uncertainty in this fraction of ineligible planting in that it may be between 8% and 26% for the Marlborough area. These limits have been used to assess 'Pessimistic' and 'Optimistic' scenarios for fractions of forest planting sinks that should not be included. When more accurate information is available it would be preferable that this is built into the modelling of the uptake of carbon by post-1990 afforestation rather than as a fractional reduction in claimed credits. The most accurate approach would be if data were available on the timing, as well as the area, of non-eligible planning, since carbon uptake rates for forests vary significantly over the first 20 years.

There has also been a downward revision of estimated sink credits from afforestation due to a lower assumption of future planting rates (10,000 ha per year in 2005 calculations compared to 20,000 ha per year in the 2004 report) This reduction reflects observed reductions in new planting rates in recent years. For future scenarios annual new planting of 20,000 ha per year is taken as the 'Optimistic' scenario with zero new planting as the 'Pessimistic' scenario. This latter assumption may be contentious but given the present difficult commercial circumstances for forest owners and developers it is not unrealistic. There is anecdotal evidence from NZ forestry industry groups that indicates planting rates for the present year (2005) will be significantly below the 10,000 ha level. The rate of new planting is a known risk. If no further planting occurs between now and the end of CP1 then 3.4 Mt CO₂ eq of sink credits would be missing in comparison with continued new planting at the present 10,000 ha per year. This estimate assumes that no new forest would be established on shrublands that may meet the Kyoto forest definition.

The methodology used assumes that all new planting in NZ will follow the growth rate of the average Radiata pine plantation. This assumption is unlikely to cause any major error in the national carbon removal rate due to post-1990 afforestation.

Loss of soil carbon after afforestation

A loss from soils of 2.2 Mt CO₂ eq over CP1 due to afforestation of grassland is estimated. (Note this is wrongly described as 'loss of soil carbon converting "grassland" to pasture' in the text of Section 5 of the Net Position report (MfE 2005a)). The 'Optimistic' scenario for projections assumes these emissions will be zero whilst the 'Pessimistic' scenario estimates a loss of 8.6 Mt CO₂ eq. The scenarios are probably based on uncertainty of estimates in scientific papers but we have been unable to identify and validate their basis exactly.

The values estimated are based on research described in several scientific publications. However loss of carbon from soils after afforestation is a dynamic process so the time since planting, as well as soil characteristics, will strongly influence emission rates. Depending on the scale and timing of future afforestation the losses from soil may be lower during the CP1 than has been included in the 2005 estimates. The NZ Carbon Accounting system should provide data to reduce uncertainty but further research into the variation in emissions from soils with time after planting may also be useful.

Removals by natural scrub

This activity is no longer included in the projections. A description has been provided of how the estimate in the 2004 projections was made. This shows that removals to scrub are unlikely to be eligible for sink credit and have therefore been correctly removed from the projections. We note that the previously estimated value of 3.75 Mt CO₂ eq has been retained for the 'Optimistic' scenario.

Planted forest not meeting KP definition

This is a key activity as its inclusion in the 2005 estimates makes the largest contribution to the net reduction in removal units compared to the 2004 estimates. Its value and presentation is discussed under 'Forest planting since 1990' above.

CO₂ released through scrub clearance (burning and decay)

Emissions estimated under this heading are due to site preparation for subsequent new planting. These emissions will need to be included as part of the "afforestation" process for post-1990 forest planting under Art 3.3. The approach used to make this estimate is reasonable given the lack of specific data.

The estimated emission (1.25 Mt CO₂ eq over CP1) is small compared to the total estimated removal.

It is not clear from the provided descriptions if scrub that meets the Kyoto forest definition has been included. If this were the case then the emission to be included would be 15% smaller on the 'most likely' scenario based on the pilot study to estimate planting of shrublands meeting Kyoto forest definition.

Deforestation

The projected area of deforestation during CP1 is based on recent observations that 2-3% of harvested forests are not replanted. This implies that 9000 ha will be deforested over CP1. Emissions from these clearances have been estimated assuming that the areas will contain biomass and carbon of typical mature Radiata pine forest and all that carbon will be emitted as CO₂ in the year of clearance. This is a reasonable basis for a first estimate but emissions of other greenhouse gases should also be made. Changes in soil carbon stocks after forest clearance have not been included but could be significant.

The estimates are directly dependent on the assumed clearance rates but future intentions of forest managers are very uncertain. In the 'Optimistic' scenario it is assumed that the same rate of clearance as for the 'Most likely' scenario will apply. This reasonably implies that the recent rate of clearance will be a minimum unless there is a major change in drivers of forest management policy. The 'Pessimistic' scenario is based on clearance of 30,000 ha over CP1. This value appears to have been set by the 'cap on CO₂ liability resulting from deforestation' of 21 Mt CO₂ eq set by the NZ Government. It is not clear as to the Government's policy response should this cap be breached. In terms of uncertainty calculation, a value of -21 Mt CO₂ eq is used to truncate the Monte Carlo triangular probability distribution. We recommend that consideration might be given to sampling beyond this limit for the pessimistic scenario.

Methods for assessing future deforestation have been reviewed and this work illustrates the difficulty of predicting the future intentions of forest managers.

Defining deforestation as land that is not replanted after harvest is rather narrow. Deforestation of natural forests and shrublands that may meet the Kyoto forest definition are not included in this approach. The drivers for deforestation may include for example perceived needs for pasture or housing development and will not simply depend on the narrow views of forest managers on the economics of commercial forestry. Further work is required to investigate where and why permanent deforestation occurs in NZ.

We note that measurements of actual deforestation from the NZ Carbon Accounting System should provide appropriate data on this activity during CP1.

2.2.3 Interpretation of KP rules

The application of the Kyoto Protocol rules for the LULUCF sector appears acceptable in general but further consideration is required on how to present 'Planted forest not meeting KP definition' including any associated emissions from scrub clearance as described in the preceding sections.

We also note that New Zealand has not yet set a definition of 'width of forestry' e.g. that might exclude shelter-belt plantings and riparian vegetation.

2.2.4 Treatment of uncertainties

The concept of "optimistic" and "pessimistic" scenarios is correct. These scenarios are for the input drivers and may be due to uncertainty in future activity e.g. planting rates or uncertainty in existing activities because of lack of information e.g. deforestation rate. This scenario analysis combines an exploration of two of the main sources of uncertainty (see Section 2.1.8) i.e. unpredictability (in terms of the projections of future activity) and value uncertainty for some parameters (e.g. uncertainty in existing activities such as fraction of ineligible planting). However, this approach does not describe statistical uncertainties in the methodologies. Such uncertainties are not discussed in the documentation received.

Table 2 shows the risk to the projections of net removal sinks arising from uncertainty in key components and drivers. The potential size of effect and the quality of the methodology are rated separately, on scales from 1 for small potential or good knowledge to 3 for large potential or poor knowledge. The priority for further work is given by the product of these two rates. A higher product indicates a greater priority for further work.

Table 2. Risk to projections of net removal units from sinks due to uncertainty in components and drivers.

View the risk to projections of net removal units from sinks due to uncertainty in components and drivers (large table).

2.3 Agriculture

Agriculture is a major source of emissions in New Zealand, accounting for 49.4% of all greenhouse gas emissions in 2003; emissions have also shown steady growth (of 15.6%) since 1990 (MfE, 2005b). The main sources in the sector are emissions from enteric fermentation, which accounted for almost two-thirds (63.4%) of emissions, and agricultural soils, which accounted for just over one-third (34.9%). Other minor sources are manure management (1.6%), and prescribed burning of savannah and field burning of agricultural residues (0.1%). The increase in emissions since 1990 is due to increased dairy production (leading to increased enteric fermentation emissions from diary cattle) and increased N₂O emissions from soils due to increased nitrogenous fertiliser use and increases in nitrogen excreted by dairy cattle.

The methodologies used in the projections of agricultural emissions are summarised in Table 3. The two key elements are the forecasts of numbers of key livestock species using the Pastoral Supply Response Model (PSRM) and the use of linear extrapolation of historic implied emission factors to derive values for enteric methane emissions per head and nitrogen excreted per head in 2010.

The PSRM model is a time series econometric model, which describes New Zealand's main farming production systems using aggregate agricultural production and price data in a system of equations (Forbes and Gardiner, undated; NZIER, 2003). It is designed to be representative of the biological constraints and investment decisions made by New Zealand farmers, and has recently been redeveloped to reflect large changes in New Zealand's agricultural sector and the economy. It projects annual inventory numbers as at 30 June, births and livestock numbers for slaughter, and is used by MAF as the basis of its contribution to the Treasury's economic and fiscal updates as well as to meet internal needs, obligations to international agencies such as FAO, and to provide information for the emissions projections. The model is driven by exogenous forecasts of prices for commodities.

PSRM provides projections of the livestock numbers of dairy cattle, beef cattle sheep and deer. However post-model adjustments are required to be made to the model outputs, to capture the impact of other non-price factors. For the projections these were:

• The removal of South Island high country leasehold areas from livestock farming;
• The conversion of grazing land to forestry rises from 10,600 ha (annually) in 2004 to 18,700 ha in 2010 (displacing sheep);
• Deforestation of 34,000 ha between 2005 and 2010 with land going into dairy;
• Total stock units are assumed to remain approximately constant (after allowing for displacement by conversion to forestry) to ensure aggregate feed demand is in balance with supply.

It appears that the assumptions about deforestation and afforestation are not consistent with the assumptions made in projections of LULUCF (see Section 2.2). Afforestation is estimated as 10,000 ha per year (20,000 ha per year under an optimistic scenario) and deforestation is assumed to be at the historical (2-3%) rate which is equivalent to 9000 ha in total across the years of CP1. We recommend that the consistency of the assumptions used in these two sectors is checked and that a consistent approach is used in future.

MAF has confirmed that back-casting ('in sample' testing) of PRSM has previously been performed in order to validate the model. We recommend checks continue to be performed in the future to confirm the on-going accuracy of the model.

Table 3 Methodology used for projection of agricultural emissions

View methology used for projection of agricultural emissions (large table).

2.3.1 Enteric Fermentation Projections

Future enteric fermentation emissions are estimated by combining the estimated number of animals in 2010 from PSRM (after the post-model adjustments) with a future implied emission factor which is derived from a historical time-series of implied emission factors (used in compiling the inventory) using linear extrapolation. The implied emission factors for the inventory are calculated using an approach consistent with IPCC Tier 2; the model uses a monthly time step and comprehensive sub-classification of animals, and information on performance, energy requirement and dietary composition to estimate total digestible matter intake (DMI). All values are three year rolling averages. For each animal species an empirical relationship has been derived for the amount of methane produced per unit of DMI, based on measurements of emissions in the field using an SF6 tracer technique.

It is reported that this method was chosen because of its simplicity and because the linear model chosen fitted the data very well. There is an awareness of the need to ensure that an linear extrapolation for a biological system does not lead to an unfeasible outcome and it is pointed out that the productivity of New Zealand animals is low by developed country standards, so there is plenty of scope for increased productivity, and that the increases are consistent with industry productivity goals. A recalculation of the inventory was considered but rejected, due to the need to estimate changes in a large number of input data sets, (e.g. milk yield, lambing percentage, live-weight at slaughter).

The historic implied emission factors are plotted in Figure 3a and b. There is a strong linear trend in the factors, and the correlation factors are high enough to suggest that extrapolation based on a linear trend is reasonable for use in projections for 2010. It is recommended that this situation is kept under review however, as the graphs do suggest a levelling off in the implied emission factor for deer, and the dairy statistics appear to show two periods of increase in the factor followed by relatively 'flat' trends. This may be due to the rapid expansion of the dairy herd leading to a higher proportion of young animals with lower emissions factors in the herd. As the model used to derive the implied emissions factors is based on a rolling three year average for all parameters, this will tend to dampen down the trend, and produce fewer 'outliers'. Three year rolling average values are presumably used to give an implied emission factor consistent with the use of a three year rolling average for animal livestock numbers when producing the emissions inventory. (The latter is recommended as good practice by the IPCC to help prevent bias in the event that the base year 1990 was not representative of normal activity level). It might be useful to examine whether the correlation is as strong for implied emission factors derived using data for parameters for a single year.

The increasing implied emission factors reflect the increasing productivity of New Zealand's animals (although the magnitude of this increase is not explicitly estimated). It is assumed that the PSRM model also makes assumptions about increasing productivity [If no increases in productivity were assumed in the PSRM model then this could potentially mean that enteric emissions are overestimated, as presumably higher livestock numbers are forecast which will then multiplied by an emissions factor which is appropriate to a more productive animal. We did not receive confirmation concerning this issue from MAF.]. Ideally these two sets of assumptions should be consistent and it might be useful to check the increase in emissions factor which the model gives if the assumptions in the PSRM model are used, even if other input factors are held constant.

2.3.2 Agricultural Soils

N₂O emissions from agricultural soils are projected using the inventory methodology combined with forecasts of implied emission factors for nitrogen excreted from animals and of nitrogenous fertiliser use. The forecasts of excreted nitrogen are (in a similar way to enteric methane emissions) based on the forecasts of animal livestock numbers from the PSRM model and a linear extrapolation of historic nitrogen excretion implied emission factors (Figure 4a and b) calculated for use in the inventory. The same general comments on the use of this linear extrapolation apply as for the enteric emissions estimates.

The forecast of nitrogenous fertiliser use is taken as an average of the forecast provided by Fertiliser Manufacturers Research Association (FMRA) (best value for 2010 of 408.5 kt) and a value of 433.7 kt produced by linearly extrapolating fertiliser use since 1990 (2003 use was 331.5 kt). While there is a high correlation factor (of 0.96) for the trend in past fertiliser use, it would seem important to understand the main drivers for this rapid increase in use, and make a judgement as to whether they are likely to continue before extrapolating forward. It does seem likely that the future changes limiting nitrogen use in some catchments (MfE, 2005a) would reduce this growth rate. In many European countries nitrogenous fertiliser consumption has fallen in recent years (e.g. EU 15 consumption fell by 5% between 1997 and 2001), partly as a result of the introduction of action to prevent nitrate leaching into vulnerable water courses. The FMRA estimate takes into account future exchange rates, agricultural commodity process, shipping costs and general projected economic circumstances for agriculture. It should be checked whether the FMRA estimate takes into account the effect of measures to limit fertiliser use. The linear trend value (433.7 kt) is close to the FMRA high scenario of 436.8 kt. While using an average of these two forecasts is acceptable, it is not ideal. Although N₂O emissions from fertiliser use account for a relatively small proportion of agricultural emissions, they are rising rapidly (even based on the lower FMRA projection) and in the future, it would be useful to consider the projection of fertiliser use in more detail. It may not be appropriate to consider a linear trend extrapolation.

2.3.3 Minor Sources

For some small emissions sources (enteric fermentation and manure management for non-key sources, and burning of savannahs and agricultural residues,) a projection is not made and emissions are assumed to remain at 2003 levels. This seems a reasonable approach given the size of the sources. For manure management for key animal species, the projected livestock numbers from the PSRM model are combined with the 2003 emissions factors. Again this approach seems reasonable given the size of emissions from this source as long as no significant changes in the amounts of manures going into different manure management systems due to changes in farming practice are foreseen.

It is noted that the UNFCCC ERT recommended a number of potential improvements to some emissions factors and suggested that attempts were made to derive country specific emissions factors (e.g. for N₂O emissions from mineral fertilisers) (UNFCCC 2004). Such improvements to the inventory will (as they are fed through into the projections methodology) also improve the accuracy of the projections.

2.3.4 Treatment of Uncertainty

For the agricultural sector, a high and a low forecast have been made on the basis of upper and low bounds for animal numbers, the nitrogen excreted and enteric methane emissions per head, and upper and lower bounds for estimated fertiliser use.

The upper and lower bounds for animal numbers have been generated by assuming that the 2004 livestock numbers change by plus or minus 5% and then running PSRM, and in the post model adjustments by changing the maximum carrying capacity. It is not clear whether this is intended to provide an indication of structural uncertainty (see Section 2.1.8) (i.e. the uncertainty in future conditions affects the demand for animal products and hence livestock numbers) or value uncertainty with respect to livestock numbers. If it is intended to look at structural uncertainty, which would seem appropriate, then it might be more useful to run the PSRM model with changes in the main drivers, to see the effect that this has on the animal numbers. If, in contrast, the aim is just to reflect the uncertainty in the animal livestock numbers then this is estimated to be only about 2% [According to information received concerning the overall sampling error in the NZ livestock number surveys.] i.e. not as large as the plus or minus 5% range used. The backcasting ('in sample' testing) which has been carried out on PSRM could be used to give an indication of the value uncertainty associated with the PSRM forecasts.

The upper and lower bounds for the enteric methane and excreted nitrogen parameters are based on the 95% confidence intervals for the emissions factors which was generated by a statistical package based on the fit of historic data to a linear trend. They are thus an estimate of value uncertainty. This seems a reasonable approach for estimating this uncertainty (the absolute uncertainty of the estimate is much higher due to the high uncertainty in the emission factors).

For nitrogenous fertiliser use, the difference between the FMRA baseline estimate and their high and low scenarios were used to give the variation round the projected value. This can be considered to address the unpredictability element of uncertainty. It should be noted that the use of an 'average' value for the baseline scenario means that the upper and lower bounds are both above the original high and low FMRA scenario values.

2.4 Waste

In New Zealand, emissions of waste arise from the sector activities of solid waste disposal on land (CH₄) (comprising 81% of the emission from the waste sector), and wastewater handling (CH₄ and N₂O). Emissions from waste incineration are not calculated in New Zealand as emissions are regarded as negligible and are therefore not included in the national greenhouse gas inventory (MfE 2005b). We note that although emissions from this source are recognised as being small, the UNFCCC inventory expert review team (ERT) has previously encouraged New Zealand to estimate and report these emissions to help improve the completeness of the national inventory (UNFCCC 2005).

In 2003, the New Zealand National Greenhouse Gas inventory emissions from the waste sector totalled 1.75 Mt CO₂ eq and represented 2.3% of all greenhouse gas emissions (MfE 2005b). Since 1990, emissions from this sector have decreased by 29.2%, a significant decrease that has been ascribed to successful initiatives to improve solid waste management practices.

Projections of CH₄ from solid waste disposal have been calculated using the methodology described in the national greenhouse gas report (MfE 2005b) involving a combination of country-specific and IPCC default variables, and (projected) population data obtained from Statistics New Zealand. In contrast, emissions from waste water treatment (comprising 19% of waste emissions) have been estimated using a simple linear extrapolation of the inventory emission trend 1990-2003 for this sub-sector.

The Net Position Report indicates that average annual emissions over the first commitment period have a most likely value of 1.1 Mt CO₂ eq. This equates to a further reduction of almost 40% from the 2003 emission level. The main reasons provided for this future reduction are the anticipated impacts arising from policy implementation in the waste sector, specifically the New Zealand Waste Strategy and the National Environmental Standard for Landfill Gas collection. Comment on the reasonableness of the magnitude of the emission reductions allocated to these two policies is provided in Section 3 (Impacts of New Zealand's Climate Change Policies).

In general the methodology and input assumptions (based on the New Zealand greenhouse gas inventory) that have been used to calculate projected emissions for the solid waste disposal sub-sector appear sound. We note the recommendations from the UNFCCC ERT concerning possible areas for future improvement of this inventory sub-sector. The benefits from this work would also of course be realised in terms of an improved confidence in the projected emissions.

One point that we have noted concerning the projections from the solid waste disposal sub-sector is that, as described above, the emission calculations include population data obtained from Statistics New Zealand (2004 base). We have noted that this dataset differs from the earlier population dataset used by MED in the SADEM energy model in terms of both the year base and the dataset projection scenario (which encompasses the various fertility, mortality and migration assumptions). We recommend for consistency purposes that a common (and as up to date as possible) population dataset is used across all sectors.

The methodology used to estimate the projected emissions from wastewater handling is more simplistic. While the fitted line through the points 1990-2003 does result in an acceptable r² value reflecting the goodness of fit, a potential problem with this approach is that it does not take account of the potential for the current upwards trend to plateau in future years. Wastewater treatment is derived from both domestic and industrial sources, and so one question is whether it might be more suitable to again extend the inventory methodology to model these emissions directly in future years. A second but less preferable option would be to scale emissions using a suitable proxy i.e. population growth or perhaps GDP. Although we recognise in terms of total greenhouse gas emissions that the wastewater handling sector contributes only a small amount to the total greenhouse gas emissions (ca. 0.4% in 2003), this would still equates to around 0.35 Mt CO₂ eq of the projected emission value for 2010. This is the same order of magnitude as many of the anticipated emission reductions arising from the New Zealand climate change policies, and so should not be regarded as totally insignificant.

2.4.1 Treatment of Uncertainty

The pessimistic and optimistic scenarios indicate average annual emissions of 1.2 and 1.0 Mt CO₂ eq respectively. The difference between scenarios relates only to the assumptions made on the possible impact of one of the two policies relevant to this sector (i.e. the Waste Strategy). Comment on the range of values associated with this policy is made in Section 3.2.4.

In this respect the waste sector uses a different interpretation of the concept of 'optimistic' and 'pessimistic' scenarios compared with other sectors. Specifically, in the waste sector the scenarios are not based on different assumptions concerning main input drivers or uncertainties in future activities within the sector (i.e. concerned with unpredictability), but on the likely outcomes of just one policy. It would be possible to construct different scenarios based on different assumptions for key input variables where data is available to justify any range of values selected e.g. the percentage of total solid waste disposed to landfills, the recovered methane rate etc. This recommendation is reinforced by the national greenhouse gas report (MfE 2004b) which recognises these data areas as key sources of uncertainty. As it is possible that improved data on these aspects will become available in future years it would appear appropriate to build a measure of uncertainty of these variables into the optimistic/pessimistic scenarios.