Appendix C Country examples – Tier 3 integration
Stand-level methods (with empirical forest growth models and dynamic dead organic matter
Canada applies a Tier 3 methodology to estimate emissions and removals from its Forest Land. Canada’s National Forest Carbon Monitoring, Accounting and Reporting System (NFCMARS - Kurz & Apps, 2006) includes the CBM-CFS3 model (Kull et al., 2006; Kurz et al., 2009; Stinson et al., 2011). This model integrates forest inventory and yield curves with spatially-referenced activity data on Forest Management and natural disturbances (fires, insect infestations) to estimate forest carbon stocks, carbon stock changes, carbon dioxide emissions and removals and methane and nitrous oxide emissions.
The CBM-CFS3 is an example of a flexible integration framework that can implement both spatially-referenced and spatially-explicit approaches (both polygon and pixel-based) to simulate forest carbon dynamics. Moreover, the model can simulate a single stand, a region or several hundred million hectares of forests. And depending on available data, it can be scaled up from representing a small number of forest strata to representing many thousands of forest strata. The model has been applied in Canada, 26 European Union countries, Russia, Korea, Mexico, China and other regions. Because the model was developed more than 15 years ago, the main constraints in the toolbox arise from software and hardware limitations that make it difficult and impractical to scale the model to pixel-based approaches with millions of pixels. While some tools have been developed as interim solutions, work is under way to implement the scientific modules of the CBM-CFS3 on a new platform (FLINT).
The CBM-CFS3 model uses regional ecological and climate parameters to simulate carbon transfers among pools, to the forest products sector and to the atmosphere. The CBM-CFS3 model tracks emissions and removals as they actually occur over time. Harvesting and natural disturbance result in significant transfers of dead biomass carbon to the litter and dead organic matter pools. The model simulates the subsequent slow decay of the biomass that results in emissions for years or decades following the harvesting or natural disturbance, depending on the decay rates, as well as the removals that occur as forest stands regenerate after the disturbance.
As a result of this approach, which aims to estimate actual emissions and removals when they occur, the model is able to estimate more accurately the long-term impact of disturbances and provide accurate projections, as is required in the construction of a projected reference level. For further detail, see Chapter 7 and Annex 3.4 of Canada’s 2010 and 2011 National Inventory Reports .
Canada’s area under forest management (229 million hectares) covers about 66% of the country’s forests. The area subject to forest management is defined using an area-based approach as outlined by the IPCC (IPCC 2003) and includes:
- lands managed for the sustainable harvest of wood fibre
- lands under intensive protection from natural disturbances (e.g., fire suppression to protect forest resources)
- protected areas, such as national and provincial parks that are managed to conserve forest ecological values.
Canada’s monitoring system draws on the close collaboration among scientists and experts in different disciplines. It was recognized early on that the approaches, methods, tools and data that are available and most suitable for monitoring human activities in one land category are not always appropriate for another. Important differences exist in the spatial framework specific to each land category, with the risk that activity data and estimates become spatially inconsistent.
In managed forests, the analysis units considered in the development of the inventory are the management units found in provincial and territorial forest inventories. For the purpose of this assessment, managed forests were classified into some 523 analysis units across 12 provinces and territories. Analysis units typically result from the intersection of administrative areas used for timber management and ecological boundaries.
The most suitable spatial framework for GHG estimation on agricultural lands (Cropland category) is the National Soil Database of the Canadian Soil Information System and its underlying soil landscapes. A full array of attributes are used to describe a distinct type of soil and its associated landscapes, such as surface form, slope, typical soil carbon content under native and dominant agricultural land use, and water table depth.
The age-class distribution of the managed forest is captured by the forest inventory data and annual change information (due to harvesting, fire and insect infestations) used in the CBM-CFS3. The managed forest is composed of relatively old stands, with over half being 80 years or older in 2009. This age-class structure reflects past natural disturbances and management.
The input data for the CBM-CFS3 include information about forest growth rates for different forest types, site classes and regions. A description of how growth data by species and region are represented in the model and the source of the information can be found in Canada’s 2010 and 2011 National Inventory Reports (Chapter 7 and Annex 3.4), Kurz et al., (2009) and Stinson et al., (2011). The same growth and yield curves are used for both projected removals and for estimates of actual removals.
Canada’s managed forest is composed of substantial areas of slow-growing and relatively old stands. Harvesting decisions are determined according to provincial and territorial policies and regulations, taking into account the age of the forest, proximity to processing facilities, environmental considerations and other factors. Based on provincial and territorial input, CBM-CFS3 simulates harvesting at the appropriate age which varies by species and region and can include salvage logging of stands previously disturbed by fire or insects.
The following projected management activities are considered: clear-cut harvesting, selection harvesting, salvage harvesting, shelter wood harvesting, commercial thinning and slash burning. The proportion of the total harvest accounted for by the various harvesting methods is projected using the recent average proportion of harvest to total harvest. The impacts of other silvicultural activities, such as tree planting, fertilization, and pre-commercial thinning are not accounted for explicitly because these activities are rarely implemented (fertilization, pre-commercial thinning) or their impacts are implicitly accounted for in the growth and yield data used in CBM-CFS3 .
Canada reports the HWP pool using three categories of (sawnwood; wood panels, paper) and a Tier 2 approach utilising data from the FAO, and country-specific wood density factors. This information is converted to carbon using Tier 2 estimates of emissions from both exported and domestically produced and consumed HWP.
Canada’s forest is continental in scale: a forest of this size means that almost every year some portion of the forest is affected by severe natural disturbances (i.e. wildfire and insect infestations). Canada predicts with a high degree of confidence the minimum level of wildfire that will occur every year. The background value of 95,000 hectares of managed forest burned each year is based on data from the past 51 years (1959-2009) which show that at least this amount burned during 90% of the years. The effects of background endemic insect infestations are captured in forest inventory and increment data.
Emissions from the background level of wildfire are calculated using a direct wildfire emissions factor of 0.132 kt CO2e per hectare burned. This factor is derived from data underlying Canada’s 2011 National Inventory Report, and is the average emissions factor for wildfires in the managed forest during 1990-2009. Non-carbon dioxide emissions are substantial, amounting to 19% of the direct fire emissions.
Pixel based methods (with hybrid growth models, dynamic dead organic matter models and agriculture)
The land area of Australia is about 760 million hectares. About 25% of total human induced greenhouse gas emissions in Australia result from activities such as agricultural production and land clearing. Given the size of Australia, it is not economically feasible or logistically practical to measure emissions and removals of greenhouse gases over such a large area with the use of direct emissions estimation methods alone e.g. field sampling. Given these national circumstances, the design of Australia’s national inventory system for the land sector relies heavily on the use of a modelling framework, to estimate the carbon stock change in biomass (above and belowground), litter and soil carbon resulting from land use and management activities.
In 1998 Australia embarked on a program to develop a comprehensive system to estimate emissions and removals from Australia’s land based sector . The system integrates spatially referenced data with an empirically constrained, mass balance, carbon cycling ecosystem model (FullCAM) (Richards and Evans, 2000; Richards, 2001) to estimate carbon stock changes and greenhouse gas emissions (including all carbon pools, gases, lands and land use activities). FullCAM is an ecosystem model that calculates greenhouse gas emissions and removals in both forest and agricultural lands using a mass balance approach to carbon cycling. As a significant amount of emissions and removals of greenhouse gases occur during transitions between forest and agricultural land use, integration of agricultural and forestry modelling was considered essential. Currently the system supports Tier 3, Approach 3 spatial enumeration of emissions and removals calculations for the following sub-categories:
- Forest land converted to Cropland
- Forest land converted to Grassland
- Grassland converted to Forest land
- the agricultural system components of Cropland remaining cropland and Grassland remaining grassland.
Australia uses a combination of geographically explicit data to represent land areas, consistent with Approach 2 and 3 as described in IPCC 2006 Guidelines and the 2013 IPCC KP Supplement. Data on areas of forest management for Forest Land remaining Forest Land are drawn from Australia’s National Forest Inventory. Supplementary spatial information from the Land Use Mapping programme of Australia’s Bureau of Agricultural Resource Economics and Sciences is used to identify land areas in the Cropland remaining Cropland, Grassland remaining Grassland, Wetlands, and Settlements categories.
Spatial enumeration is achieved through the use of a time series (since 1972) of Landsat satellite data which is used to determine change in forest extent. The forest cover change information is used with time series climate data and spatially referenced databases of land management practices. Australia monitors forest cover using national coverages of Landsat satellite data (MSS, TM, and ETM+) across 25 time epochs (periods between dates for which remote sensing data are available) from 1972 to 2016 which have been assembled and analysed for change. These national maps of forest cover are annual from 2004 and are used to detect fine scale changes in forest cover at a 25 m by 25 m resolution. Where forest cover change is identified in an epoch, the actual date of forest cover change in each 25 m by 25 m pixel is randomly allocated within the sequence of satellite pass dates.
FullCAM models both biological and management processes which affect carbon pools and transfers between pools in forest and agricultural systems. The exchanges of carbon, loss and uptake between the terrestrial biological system and the atmosphere are accounted for in the full, closed cycle mass balance model which includes all biomass, litter and soil pools. Analysis and reporting includes all carbon pools (biomass, dead organic matter and soil), greenhouse gases (CO2, CH4 and N2O), and covers both forest and non-forest land uses. It is an integrated suite of the following models:
- A hybrid forest growth and biomass estimation (Brack et al., 2006; Waterworth and Richards, 2008):
- This uses 3PG - the physiological growth model for forests (Landsberg and Wareing, 1997; Landsberg et al., 2000; Coops et al., 1998; Coops et al., 2000) to develop a site index rather than predict biomass directly
- CAMFor - the carbon accounting model for forests (Richards and Evans, 2000a):
- This model is based on CO2Fix and allows for the inclusion of management and natural disturbance events
- It also models dead organic matter pools based on estimates of turnover and decomposition
- CAMAg - the carbon accounting model for cropping and grazing systems (Richards and Evans, 2000b)
- To account for agriculture, and the effects of management and natural disturbances
- It also models dead organic matter pools based on estimates of turnover and decomposition
- Roth C - the Rothamsted Soil Carbon Model – Roth C (Jenkinson et al., 1987; Jenkinson et al., 1991).
- This is applied for all soils.
To meet its objective of providing a comprehensive carbon accounting and projections capacity for land based activities, the National Inventory System (formerly known as the National Carbon Accounting System) has required strategic development of several key datasets and modelling and accounting tools. The system and underlying supporting data and science have been documented in many reports that are publicly available. Early reviews made it clear that approaches based on measurement were not feasible and that the calibration of relevant models would be required. The most significant value of FullCAM is that it allows for an ongoing evolution in the quality of any data inputs, be they for future accounting periods or improvements in fundamental input data or model calibration. Since 2014, FullCAM outputs have been downloaded into a SQL output database providing a ready tool to facilitate transparency, quality control and publication of important variables. Such ongoing improvements were not as readily made under the regional approaches envisaged formerly. FullCAM also provides for greater responsiveness to the various international reporting demands. The fine spatial resolution, activity-driven and time-based modelling provides a capacity to report at both project and continental scales, in response to specific activities, and with sensitivity to the timing of an activity.
Brack, C., Richards, G., & Waterworth, R.M. 2006. Integrated and comprehensive estimation of greenhouse gas emissions from land systems . Sustainability Science. 1:1:91-106.
Waterworth, R.M., Richards, G.P., Brack, C.L., & Evans, D.M.W. (2007). A generalised hybrid process-empirical model for predicting plantation forest growth. Forest Ecology and Management. 238:231-243.
Pixel-based (empirical models, emissions factors and Approach 3 time series data)
The Government of Indonesia is committed to ambitious greenhouse gas (GHG) emissions reduction targets. With a significant proportion of Indonesia’s total emissions generated by land-based activities, the sector is a major focus of mitigation efforts. Understanding the size and source of historical emissions is critical to planning efficient and effective interventions, as well as gauging the potential impact of alternative land management options on future emissions. This is why the Government of Indonesia has been developing the Indonesian National Carbon Accounting System since 2008, with the first national level, annual results publicly released in December 2015, covering forests and peatlands for the period 2001 to 2012. The results of the system along with all documentation on methods is now available in the INCAS website .
The system uses:
- A time-series of forest-non-forest data developed from remote sensing data to determine areas of change (using a time series algorithm)
- Maps of forest type based on existing maps
- Spatial and non-spatial forest management information
- Estimates of area burnt from hotspot mapping and manual mapping
- Empirical forest growth models developed from past measurements from Indonesia’s National Forest Inventory
- Allometric models based on Indonesia specific data
- Dynamic modelling of dead organic matter
- Tier 2 emissions factors for mineral soils and peat emissions
- Tier 1 emissions factors for non-carbon dioxide emissions, in particular from fire.
The data are integrated using a combination of FullCAM in empirical mode (for the forest growth and dead organic matter) and simple excel sheets for Tier 1 and 2 data. The results of these systems are then combined to produce annual GHG emissions and removals estimates for deforestation, forest degradation, sustainable management of forests and enhancement of forest carbon stocks.
The INCAS represents a good example of interagency collaboration, with the Indonesian National Institute of Aeronautics and Space (LAPAN) developing the time series forest/non-forest data, the Directorate General of Forestry Planning and Environmental Management providing forest inventory and forest type maps, with integration managed through the Forest Research, Development and Innovation Agency of the Ministry of Environment and Forestry (FORDA). A core feature of the INCAS has been the development of a small core team of experts within FORDA. Consistent and sustained support for these experts has allowed the system to rapidly develop in recent years and provides a good example of a successful management structure for other countries.
The INCAS is by no means finished. The release of the first national results represents the first step in INCAS development. Planned improvements include potential to move to a new integration tool (such as FLINT), full pixel based analysis for both land use change and fire, and continual progression towards Tier 3 methods for key emissions sources such as peat. Indonesia has also indicated an interest in developing the system to account for the entire land sector in the future.
Pixel-based full lands accounting (SLEEK, Kenya)
The System for Land-based Emissions in Kenya (SLEEK) has been under active development since 2013. The system differs from most others as it is attempting to develop emissions estimates for the entire land sector within a single system. In this case reporting for forests and REDD+ becomes a sub-set of the entire system. This has the advantage that all land areas are included as well as the transitions among land categories. Moreover this helps to ensure that REDD+ reporting is consistent with the methods and data used to develop biennial reports and other actions under the INDCs.
Initial runs of the system should be completed in mid 2016, with official runs occurring prior to the end of 2016. The current design uses the following data:
- Annual time series of land cover (9 classes) specifically developed by the Department of Remote Sensing of in collaboration with the Kenya Forest Service and Ministry of Agriculture.
- Semi-automated attribution data to determine cause of cover change to produce land use maps
- Empirical forest growth models derived from forest inventory and research site data
- WOFOST model to estimate crop growth
- Empirical pasture growth models
- Roth C soil carbon model
- Climate data
- Tier 1 modules where higher tiers have not been calibrated (for example, many soils types) or not available (for example, Wetlands).
These data are integrated within a single framework called the Full Lands Integration Tool (FLINT) . SLEEK has led the development of the FLINT as a generic tool that can be easily applied by other countries. The FLINT framework has also been used to guide the management of the program. This has allowed all the different agencies to work together around a single tool and approach, while still maintaining independence and ownership of their core work.
Several government agencies have come together to help develop SLEEK. This collaboration is a key achievement of the program and has helped prevent duplication of effort. Members from each organisation are represented on a series of Element Working Groups that help set the direction and plan work to deliver the required products. The organisations include:
- Kenya Forest Service, who are developing biomass estimates and forest growth curves and estimates of dead organic matter for plantations and natural forests
- Ministry of Agriculture, who are responsible for all soils monitoring and modelling (for all land uses), and for managing the crop growth modelling
- Department of Resource Surveys and Remote Sensing (DRSRS), who are leading the land cover mapping program
- Regional Centre for Remote Sensing and Mapping, who are supporting DRSRS to develop the time series land cover maps
- Kenyan Meteorological service, who are producing daily climate grids for key weather variables to support the models
- Kenyatta University and Embu University College, who are developing crop models and conducting field trials.
Pixel-based forest and deforestation accounting (empirical models and Approach 3 time series data; Mexico)
The application of the CBM-CFS3 to regions in Mexico is an on-going collaboration between the forest services of Mexico and Canada. Following the approach selected in Canada, the intersections of administrative (state boundaries) and ecological (terrestrial ecozones, Level 1) were used to define 94 spatial units. The model has been applied in a spatially-referenced Tier 3 approach in six states in Mexico to estimate past and projected future greenhouse gas emissions and removals. The model has also been applied using a spatially-explicit approach to a single Landsat scene in Mexico to demonstrate the approach and to quantify the impacts of four activity data sets derived from remote sensing products, each set with and without attribution of observed land-cover changes to specific disturbance types (Mascorro et al., 2015). The results demonstrate that in a heterogeneous landscape with frequent small-scale disturbances and rapid regrowth, remote sensing products based on 30-meter resolution and annual time steps perform better than products based on 250-meter resolution sensors or products that detect changes over multi-year periods (Mascorro et al., 2015; Kurz et al., 2015).
Analyses of past emissions and projections of future REDD+ scenarios for the entire Yucatan Peninsula in Mexico were based on land-cover change matrices developed from remote sensing products and time-series of change maps developed by Mexican agencies. The CBM-CFS3 was then applied to estimate past GHG emissions and removals as affected by human and natural disturbances. Future projections were implemented with business-as-usual disturbance rates (averages of the past 10 years) and alternative scenarios of reduced rates of deforestation and degradation (Kurz et al., 2015). One of the lessons learned in the application of the CBM-CFS3 in Mexico is that data on fire wood collection, a potentially important human impact on biomass and dead organic matter carbon stocks in forest ecosystems, are very difficult to obtain.