View all publications

Annex 3: Detailed methodological information for other sectors

A3.1 The agriculture sector

New Zealand's methodology uses a detailed livestock population characterisation and livestock productivity data to calculate feed intake for the four largest categories in the New Zealand ruminant population (dairy cattle, beef cattle, sheep and deer). The amount of CH4 emitted is calculated using CH4 emissions per unit of feed intake. A schematic overview of the model is presented in the agriculture sector.

A3.1.1 Enteric methane emissions

Livestock populations

The New Zealand ruminant population can be separated into four main categories: dairy cattle, beef cattle, sheep and deer. For each livestock category, population models that further sub-divided the principle categories were developed. These models reflect New Zealand farming systems with regard to the timing of births, timing of slaughter of growing animals and the transfer of younger animals into the breeding population.

Animal numbers are provided by Statistics New Zealand from census and survey data. As shown in the agricultural worksheets in Annex 8, three-year rolling averages are used throughout the agricultural sector for population numbers.

For sheep, dairy cattle, non-dairy cattle and deer the three-year average populations are adjusted on a monthly basis to take account of births, deaths and transfers between age groups. This is necessary because the numbers present at one point in time may not accurately reflect the numbers present at other times of the year. Goats are also included in the analysis, but a separate model has not been developed. This is because goats represent only a very small proportion of the total animal population and numbers have dropped significantly in recent years.

Livestock productivity data

For each livestock category, the best available data are used to compile the inventory. These data are from Statistics New Zealand and industry statistics. To ensure consistency, the same data sources are used each year. This ensures that the data provide a time-series that reflects changing farming practices, even if there is uncertainty surrounding the absolute values.

Obtaining data on the productivity of ruminant livestock in New Zealand, and how it has changed over time, is a difficult task. Some of the information collected is robust i.e. the slaughter weight of all livestock exported from New Zealand is collected by the MAF on a census basis. This information is used as a surrogate for changes in animal liveweight. Other information is collected at irregular intervals or from small survey populations.

Livestock productivity and performance data are summarised in the time-series tables accompanying the agriculture sector (Chapter 6). The data include average liveweights, milk yields and milk composition of dairy cows, average liveweights of beef cattle (beef cows, heifers, bulls and steers), average liveweights of sheep (ewes and lambs) and average liveweights of deer (breeding and growing hinds and stags).

Dairy cattle: data on milk production are provided by Livestock Improvement Corporation (LIC, 2004). These data include the amount of milk processed through New Zealand dairy factories plus an allowance for town milk supply. Annual milk yields per animal are obtained by dividing the total milk produced by the total number of milking dairy cows and heifers. Milk composition data are taken from the LIC national statistics. For all years, lactation length was assumed to be 280 days.

Average liveweight data for dairy cows are obtained by taking into account the proportion of each breed in the national herd and its age structure based on data about breed and age structure from the LIC. Dairy cow live-weights are only available from 1996 onwards. For earlier years in the time-series, liveweights are estimated using the trend in liveweights from 1996 to 2003 together with data on the breed composition of the national herd. Growing dairy replacements at birth are assumed to be 9% of the weight of the average cow and 90% of the weight of the average adult cow at calving. Growth between birth and calving (at two years of age) is assumed to be linear, however the growth rates are not assumed to be constant as higher growth rates apply prior to weaning when animals receive milk as part of their diet. The birth date of all calves was assumed to be mid-August.

No data are available on the liveweights and performance of breeding bulls and an assumption was made that their average weight was 500 kg and that they are growing at 0.5 kg per day. This was based on expert opinion from industry data. For example, dairy bulls range from small Jersey's through to larger-framed European beef breeds. The assumed weight of 500 kg and growth rate of 0.5 kg/day provide an average weight (at the mid point of the year) of 592kg. This is almost 25% higher than the average weight of a breeding dairy cow but it is realistic given that some of the bulls will be of a heavier breed/strain (e.g. Friesian and some beef breeds). Because these categories of animal make only small contributions to total emissions e.g. breeding dairy bulls contribute 0.089% of emissions from the dairy sector, total emissions are not highly sensitive to the assumed values.

Beef cattle: the principle source of information for estimating productivity was livestock slaughter statistics provided by the MAF. All growing beef animals are assumed to be slaughtered at two years of age and the average weight at slaughter for the three sub-categories (heifers, steers and bulls) was estimated from the carcass weight at slaughter. Liveweights at birth are assumed to be 9% of an adult cow weight for heifers and 10% of the adult cow weight for steers and bulls. Growth rates of all growing animals between birth and slaughter are assumed to be linear, however the growth rates are not assumed to be constant as higher growth rates apply prior to weaning when animals receive milk as part of their diet.

Weights in slaughter statistics from the MAF do not separate carcass weights of adult dairy cows and adult beef cows. Thus a number of assumptions [Number of beef breeding cows assumed to be 25% of the total beef breeding cow herd. Other adult cows slaughtered assumed to be dairy cows. Carcass weight of dairy cattle slaughtered was estimated using the adult dairy cow liveweights and a killing out percentage of 40%. Total weight of dairy cattle slaughtered was then deducted from the national total carcass weight of slaughtered adult cows. This figure was then divided by the number of beef cows slaughtered to obtain an estimate of the carcass weight of adult beef cows. Liveweights are then obtained assuming a killing out percentage of 45%.] are made in order to estimate the liveweights of beef breeding cows. A total milk yield of 800 litres per breeding beef cow was assumed.

Sheep: livestock slaughter statistics from the MAF are used to estimate the liveweight of adult sheep and lambs, assuming killing out percentages of 43% for ewes and 45% for lambs. Lamb birth liveweights are assumed to be 9% of the adult ewe weight with all lambs assumed to be born on 1 September. Growing breeding and non-breeding ewe hoggets are assumed to reach full adult size at the time of mating when aged 20 months. Adult wethers are assumed to be the same weight as adult breeding females.

No within year pattern of liveweight change was assumed for either adult wethers or adult ewes. All ewes rearing a lamb are assumed to have a total milk yield of 100 litres. Breeding rams are assumed to weigh 40% more than adult ewes. Wool growth (greasy fleece growth) was assumed to be 5kg/annum in mature sheep (ewes, rams and wethers) and 2.5kg/annum in growing sheep and lambs.

Deer: liveweights of growing hinds and stags are estimated from MAF slaughter statistics, assuming a killing out percentage of 55%. A fawn birthweight of 9% of the adult female weight and a common birth date of mid-December are assumed. Liveweights of breeding stags and hinds are based on published data, changing the liveweights every year by the same percentage change recorded in the slaughter statistics for growing hinds and stags above the 1990 base. No within year pattern of liveweight change was assumed. The total milk yield of lactating hinds was assumed to be 240 litres (Kay, 1985).

Goats: enteric CH4 from goats is not a key category. There is no published data on which to attempt a detailed categorisation of the performance characteristics in the same way as has been done for the major livestock categories. New Zealand uses the IPCC CH4 emissions factor for goats (9 kg CH4/head/yr).

Dry matter intake calculation

Dry matter intake(DMI) for the classes (dairy cattle, beef cattle, sheep and deer) and sub-classes of animals (breeding and growing) was estimated by calculating the energy required to meet the levels of performance assumed and dividing this by the energy concentration of the diet consumed. For dairy cattle, beef cattle and sheep, energy requirements are calculated using algorithms developed in Australia (Standard Australian Livestock Tables, CSIRO, 1990). These are chosen as they specifically include methods to estimate the energy requirements of grazing animals. The method estimates a maintenance requirement (a function of liveweight and the amount of energy expended on the grazing process) and a production energy requirement - influenced by the level of productivity (e.g. milk yield and liveweight gain), physiological state (e.g. pregnant or lactating) and the stage of maturity of the animal. All calculations are performed on a monthly basis.

For deer, an approach similar to that used for cattle was adopted using algorithms derived from New Zealand studies on red deer. The algorithms take into account animal liveweight and production requirements based on the rate of liveweight gain, sex, milk yield and physiological state.

Monthly energy concentrations

A single set of monthly energy concentrations of the diets consumed by beef cattle, dairy cattle, sheep and deer was used for all years in the time-series. This is because there is no comprehensive published data available that allow the estimation of a time-series dating back to 1990.

Methane emissions per unit of feed intake

There are a number of published algorithms and models [For example Blaxter and Clapperton,1995; Moe and Tyrrel, 1975; Baldwin et al., 1988; Djikstra et al., 1992; and Benchaar et al., 2001 - all cited in Clarke et al., 2003.] of ruminant digestion for estimating CH4 emissions per unit of feed intake. The data requirements of the digestion models make them difficult to use in generalised national inventories and none of the methods have high predictive power when compared against experimental data. Additionally, the relationships in the models have been derived from animals fed indoors on diets dissimilar to those consumed by New Zealand's grazed ruminants.

Since 1996, New Zealand scientists have been measuring CH4 emissions from grazing cattle and sheep using the SF6 tracer technique (Lassey et al, 1997; Ulyatt et al, 1999). New Zealand now has one of the largest data sets in the world of CH4 emissions determined using the SF6 technique on grazing ruminants. A database has been constructed and is being systematically examined to obtain generalised relationships between feed and animal characteristics and CH4 emissions. As an interim measure, published and unpublished data on CH4 emissions from New Zealand were collated and average values for CH4 emissions from different categories of livestock obtained. Sufficient data were available to obtain values for adult dairy cattle, sheep more than one year old and growing sheep (less than one year old). These data are presented in Table A3.1.1 together with IPCC (2000) default values for percent gross energy used to produce CH4. The New Zealand values fall within the IPCC range and are adopted for use in this inventory calculation. Table A3.1.2 shows a time-series of CH4 implied emission factors for dairy cattle, beef cattle, sheep and deer.

Not all classes of animals are covered in the New Zealand data set and assumptions had to be made for these additional classes. The adult dairy cattle value was assumed to apply to all dairy and beef cattle irrespective of age and the adult ewe value was applied to all sheep greater than one year old. A mean of the adult cow and adult ewe value (21.25g CH4/kg DMI) was assumed to apply to all deer. In very young animals receiving a milk diet, no CH4 was assumed to arise from the milk proportion of the diet.

Table A3.1.1 Methane emissions from New Zealand measurements and IPCC defaults

  Adult dairy cattle Adult sheep Adult sheep < 1 year

New Zealand data (g CH4/kg DMI)

21.6

20.9

16.8

New Zealand data (%GE)

6.5

6.3

5.1

IPCC (2000) defaults (%GE)

6 ± 0.5

6 ± 0.5

5 ± 0.5

Table A3.1.2 Time-series of implied emission factors for enteric fermentation (kg methane per animal per annum)

Year Dairy cattle Beef cattle Sheep Deer

1990

70.8

51.0

8.9

21.0

1991

71.4

51.7

9.1

20.7

1992

73.3

52.7

9.2

20.6

1993

73.7

54.2

9.3

20.5

1994

73.9

53.6

9.4

20.6

1995

73.6

53.3

9.5

20.9

1996

74.4

52.6

9.6

21.3

1997

74.3

54.5

9.9

21.9

1998

75.1

54.5

10.0

22.2

1999

76.5

55.3

10.1

22.3

2000

78.7

55.1

10.2

22.4

2001

78.8

56.4

10.5

22.3

2002

79.0

56.1

10.5

22.3

2003

79.1

56.3

10.6

22.1

Previous fixed EF

76.8

67.5

15.1

30.6

A3.1.2 Uncertainty of animal population data

Details of the most recent surveys and census are included to provide an understanding of the livestock statistics process and uncertainty figures. The information documented is from Statistics New Zealand. Full details of the surveys are available from Statistics New Zealand's website http://www.stats.govt.nz/datasets/primary-production/agriculture-product....

2003/4 Agricultural Production Surveys

The target population for the Agricultural Production surveys is all businesses engaged in agricultural production activity (including livestock, cropping, horticulture and forestry) with the intention of selling that production and/or which owned land that was intended for agricultural activity during the year ended 30 June. The estimated proportion of eligible businesses responding to the 2003 Agricultural Production Survey was 85 percent. These businesses contributed 87 percent of the total agricultural output. The sample error and percentage imputed for 2003 are shown in Table A3.2.1. For the 2004 survey, 30,715 forms were distributed. The survey response rate was 87% or 91% of the estimated value of production. Interim animal number results from the 2004 survey have been used for the 2005 submission. The 2003 animal numbers have been updated with final animal number estimates.

2002 Agricultural Production Census

The target population for the 2002 Agricultural Production census was all units that were engaged in agricultural production activity (including livestock, cropping, horticulture and forestry) with the intention of selling that production and/or which owned land that was intended for agricultural activity during the year ended 30 June 2002. The target population also includes businesses and persons commonly referred to as 'lifestylers' engaged in agricultural production activity. The response rate was 81 percent. Statistics New Zealand imputes using a random 'hot deck' procedure for values for farmers and growers who did not return a completed questionnaire.

The 1999 Livestock Survey

The frame for the 1999 Agricultural Production survey was based on a national database of farms called AgriBase which is maintained by AgriQuality New Zealand Ltd (formerly MAF Quality Management). A sample survey was conducted to obtain estimates of livestock on farms and area sown in grain and arable crops for the 30 June 1999 year. Questionnaires were sent to approximately 35,000 farms. The overall response rate for the survey was 85.7 percent. The remaining units were given imputed values based on either previous data or on the mean value of similar farms. Table A3.2.2 gives the sample errors based on a 95% confidence level for the survey data collected in 1999.

Table A3.2.1 Provisional sampling error and imputation levels for the 2003 Agricultural Production survey

  Sample errors at 95% confidence interval(%) Percentage of total estimate imputed

Ewe hoggets put to ram

4

12

Breeding ewes 2 tooth and over

2

12

Total number of sheep

2

11

Total lambs marked or tailed

2

11

Beef cows and heifers (in calf) 2 years and over

2

12

Beef cows and heifers (in calf) 1 - 2 years

5

11

Total number of beef cattle

2

12

Calves born alive to beef heifers/cows

3

12

Dairy cows and heifers, in milk or calf

2

14

Total number of dairy cattle

2

14

Calves born alive to dairy heifers/cows

3

13

Female deer mated

4

9

Total number of deer

4

9

Fawns or calves weaned on the farm

4

9

Area of potatoes harvested

1

12

Area of wheat harvested

4

11

Area of barley harvested

4

13

Table A3.2.2 Agricultural sector sample errors based on 95% confidence level

Variable (total population) Survey design error (%) Achieved sample error (%)

Dairy cattle

1

1.0

Beef cattle

1

0.9

Sheep

1

0.7

Goats

1

1.5

Deer

1

1.4

Pigs

1

0.9

A3.2 Additional information for the LULUCF sector: The Carbon Monitoring System plots and the New Zealand Carbon Accounting System

Major ongoing work in the LULUCF sector includes research and implementing a monitoring system for the carbon stocks and fluxes in soils, shrublands and natural forests. This research was initiated by the MfE in 1996 and is being performed by two of New Zealand's Crown Research Institutes - Landcare Research and Forest Research. This five-year research project had the following objectives:

  • The estimation of carbon storage in soils, shrublands and natural forests in 1990.
  • The development of a national system to determine soil carbon changes associated with land-use change.
  • The development of an effective information system to manage the above information.

Provisional results are available from the work under the first objective. Hall et al. (1998) have estimated that in 1990 carbon stored in natural forests was 933 MtC, while 527 Mt C was stored in shrublands and other woody mixed-vegetation. Forest floor litter carbon is estimated separately, based on Tate et al.(1997), as containing 570 Mt C for all natural vegetation (i.e. both forest and scrub areas). These estimates are highly sensitive to both the accuracy of mapped areas and heterogeneity within mapped classes. Current (very provisional) estimates for soil carbon at soil depth intervals of 0-0.1, 0.1-0.3 and 0.3-1m are 1300±20, 1590±30 and 1750±70 Tg C respectively (Tate et al., 2003). Some soil cells are still poorly represented in the database and additional field work is being undertaken. Further information on this project and initial estimates of carbon stocks at 1990 are found in Coomes et al. (2002), Lawton and Barton (2002), Lawton and Calman (1999), and Hall et al. (1998).

In 1999, the soil and vegetation carbon monitoring systems (CMS) developed during the first three years of the project were reviewed by an international panel of forestry and soil experts. The panel's report concluded that the systems being developed for New Zealand's natural forests are consistent with current forest inventory practices in other countries. Furthermore, the soils that the system represented are measured in a significantly advanced methodology as compared with the IPCC default method (Theron et al., 1999). The international review of the system was held in time for the key recommendations of the review to be undertaken before the development phase was concluded.

The statistical design of the vegetation CMS provides for the establishment of 1400 permanent field plots on an 8x8 km grid across natural forest and shrublands for territorial New Zealand (Coomes et al., 2002). This includes the North and South Islands, Stewart Island, the Chatham Islands and other offshore islands. To provide continuity, and to build on previously collected data, about one-third of the plots are existing ones matched to nearby grid intersection points, and the rest are new plots established at unmatched grid intersections specifically for monitoring forest carbon pools. The plot measurements use the 20m by 20m quadrat method (Allen 1993) which has been used at various sites of interest in New Zealand but never on a statistically representative basis across all of the nation's natural forests and shrublands. Measurements are taken of above ground biomass such as tree heights and diameters, understorey vegetation, litter and coarse woody debris. These measurements will have use for other international forestry reporting obligations such as those required under the Montreal Process, the FAO Global Forest Resource Assessment and the Convention on Biological Diversity.

The soil CMS analyses soil samples to a depth of 0.3m for carbon content. One in every three of the vegetation plots is sampled for soils to reduce the uncertainty in some soil cells.

The CMS's for soil and vegetation are currently moving from design to implementation. The first year's fieldwork for the operational vegetation CMS commenced in January 2002 and was completed in early 2003. The second year's fieldwork began in March 2003. Fieldwork over at least three more years will be required to install the complete network of field plots. Following this, another five-year round of sampling will be required to validate the implementation and begin monitoring of any changes. The current intention is then to repeat these measurements every ten years.

For the soil CMS, 40 soil-paired plots will be established to monitor key changes in soil carbon when land-use changes i.e. scrub to grassland, grassland to Kyoto forest and vice versa. The first four paired plots were established in 2003.

The New Zealand Carbon Accounting System (NZCAS) is being developed. This system will account for human-induced carbon sources and sinks from New Zealand's land use, land-use change and forestry (LULUCF) activities which (a) is appropriate for annual UNFCCC greenhouse gas emission LULUCF sector reporting, (b) enables accounting and reporting under the Kyoto Protocol, and (c) underpins scenario development and modelling capabilities that support New Zealand's climate change policy development.

The most developed module of the NZCAS is for natural forests and shrublands (Coomes et al., 2002; Allen et al., 2003). This is based on the CMS plots. These unique natural forests cover 6.4 million hectares and have been either too remote or inaccessible for timber extraction and are now largely protected, often as national parks. New Zealand has not, until now, needed to establish a national forest inventory to cover its protected forests. This is in sharp contrast to the advanced system used for monitoring and forecasting future wood supply from its 1.8 million hectares of plantation forests.

A monitoring and modelling module is currently being designed for those areas where afforestation and reforestation activities have occurred since 1990, so-called 'Kyoto' forests. This will involve inventory measurements from permanent plots coupled with the use of existing allometric equations and/or forest volume and carbon models (Beets et al. 1999).

A3.3 Additional methodology for the LULUCF sector: the Land Cover Database

The LCDB1 was completed in 2000 using SPOT 2 and SPOT 3 satellite imagery acquired over the summer of 1996/97. A 1 ha Minimum Mapping Unit (MMU) was used and this was retained for LCDB2. LCDB2 used the Landsat 7 ETM+ sensor, with the imagery pan-sharpened to a 15m spatial resolution. All imagery for LCDB2 was acquired during summer 2001/02. Development of the final database involved several Government Crown Research Institutes, Ministries and companies. LCDB2 was released in July 2004.

A description of the process to create LCDB2 is shown in Figure A3c.1. A single set of polygon boundaries is used for both the attributes from LCDB1 and LCDB2. This removes the need for GIS overlay analysis to detect changes in landcover between databases. As part of the process of developing LCDB2, any errors identified in the LCDB1 were corrected and areas of apparent change confirmed.

The target classes used for LCDB1 and LCDB2 are hierarchical (and derived from eight first order classes at the highest level, with an increasing number of more detailed classes at lower levels). The first order classes are based on the physiognomy of the land cover (i.e. grassland, shrubland and forest etc). The following divisions are based on other characteristics, such as phenology (evergreen/deciduous) and floristic composition (broadleaved/needle leaved).

LCDB2 was developed from image processing supplemented by ancillary data such as vegetation surveys, plot data and aerial photography. The database was also subjected to intensive field checking to determine the following:

  • Whether the land cover types identified in the draft vectors are present on the ground.
  • Whether land cover types observed on the ground are captured and correctly labeled in the draft map.
  • Identify land cover classes with unknown or questionable spectral signatures.
  • Identify characteristic signatures of the target land cover classes to be used to train the classification in areas that cannot be field checked. Extrapolation of ground data was restricted to one New Zealand 260 map sheet (30 km x 40 km), as the spectral signatures of target classes can vary across a Landsat 7 ETM+ scene (185 km wide).

In assigning land cover to a specific class, the dominant cover rule was used. For example, a shrubland polygon with three or more main species (where further subdivision of the patch based on the 1 ha MMU is not possible), is classified according to the dominant species in the matrix. This procedure was maintained throughout the LCDB2 mapping project.

For LCDB1, overall user accuracy was assessed at 93.9%. A classification accuracy has not been established for LCDB2. The database was released to ensure that users have access to the updated national dataset for planning and monitoring purposes. However, users can be confident that the accuracy of LCDB2 will be equal to or higher than the User Accuracies for LCDB1, namely: bare ground (81%), natural forest (95%), mangrove (97%), planted forest (90%), horticultural (95%), pastoral (98%), scrubland (89%), tussock (95%), and wetlands (87%).

Table A3.3.1 Mapping of LCDB classification to the IPCC land-use categories

IPCC category LCDB class

Cropland

 

CM (perennial)

Orchard and Other Perennial Crops, Vineyard

CM (annual)

Short-rotation Cropland

Forest land

 

FM (planted)

Afforestation (imaged, post LCDB 1), Afforestation (not imaged), Deciduous Hardwoods, Forest Harvested, Other Exotic Forest, Pine Forest - Closed Canopy, Pine Forest - Open Canopy

FM (natural)

Natural Forest, Broadleaved Natural Hardwoods, Manuka and or Kanuka

Grassland

 

GM (low prod)

Alpine Grass-/Herbfield, Depleted Tussock Grassland, Fernland, Gorse and Broom, Grey Scrub, Low Producing Grassland, Major Shelterbelts, Matagouri, Mixed Exotic Shrubland, Sub Alpine Shrubland, Tall Tussock Grassland, Flaxland, Herbaceous Freshwater Vegetation, Herbaceous Saline Vegetation, Mangrove

GM (high prod)

High Producing Exotic Grassland

Other land

 

O

Alpine Gravel and Rock, Coastal Sand and Gravel, Landslide, Permanent Snow and Ice, River and Lakeshore Gravel and Rock

Settlement

 

S

Built-up Area, Dump, Surface Mine, Transport Infrastructure, Urban Parkland/ Open Space

Wetland

 

W (unmanaged)

Estuarine Open Water, Lake and Pond, River