FLINTpro Global Run

Summary of Design


Country level estimates of carbon fluxes associated with tree cover loss and gain were estimated using a ‘Tier 1’ modelling configuration developed within FLINTpro.

The implementation of FLINTpro for this work uses four key processes.

The output represents the first global run using the FLINTpro system. The run demonstrated that it is now possible to rapidly and reliably conduct detailed analyses at the global scale. The limitation of the Global Run is the input data with main issues being:

Improving the input data used will improve the outputs. Where organisations require more reliable estimates is it simple to add better local data where available. We are actively collaborating with several groups to create such data and would welcome input.

Data Requirements and Workflow

The 2006 IPCC Guidelines for National Greenhouse Gas Inventories: AFOLU, provide guidance on methods and approaches for estimating emissions and removals from the land sector.  Within this guideline, there are three Approaches related to the collection of activity data. Approach 1 provides data that only represents land use at discrete points in time and therefore does not provide information on land use change. Approach 2 for data collection is more advanced in that it provides data on land use change between time-periods, but it is not capable of tracking land use change over time or space. Approach 3 refers to spatially and temporally specific data capable of tracking land use and land use change through time and space.
Similarly, there are three Tiers related to the methods for estimating emissions associated with specific land use and land use changes. Tier 1 methods use IPCC default emission factors and carbon stock factors based on global regionalisation. Tier 2 improves on Tier 1 through country-specific emission and carbon stock factors. Finally, Tier 3 takes a further step and introduces more advanced empirical and process models, or repeated inventories. Depending on the Measurement, Reporting and Verification (MRV) system, different combinations of Tiers and Approaches are used. For full definitions of Tiers and Approaches, see the IPCC Guidelines (2006). 
For the purposes of this exercise, the Tier 1 implementation for FLINTpro uses Tier 1 emission factors from the 2006 guidelines and Requena Suarez et al (2019) in conjunction with Approach 3 type activity data from Hansen et al. (2013). The implementation is limited to forest gain and forest loss at a national scale, covering Above Ground Biomass (AGB), Below Ground Biomass (BGB) and Dead Organic Matter. While Soil Carbon results were also produced, additional work would be required to improve the robustness of the estimates before publishing.

Triggering Events


FLINTpro estimates emissions and removals of CO2 from the land sector but tracking fluxes of carbon between pools (such as aboveground biomass) and the atmosphere. These fluxes are created when there is an event as well as through processes. Events cause a movement of carbon between pools as a specific point in time, for example a forest clearing event, while a process is constant when certain conditions are met, for example, forest growth when a forest is present.
As FLINTpro is a spatially explicit system (Approach 3), this requires identifying and setting a sequence of events and processes for each Simulation Unit (pixel) across the simulation area (the total area being assessed). For the global run, Spatial Triggers were used to determine the dates of events.
A Spatial Trigger is one that is triggered using the land cover product. Where a pixel transitions from forest land to non-forest, there is a ‘forest clearing event’, for example. To assign a date for an event with a Transition Trigger, the FLINTpro assigns a date between the Landsat pass dates of the land cover products where a transition was detected. For example, if a simulation unit (pixel) transitioned from non-forest to forest between 1 January 2001 and 1 January 2012, the ‘plant forest’ event would be a random date between 1 January 2001 and 1 January 2012.

Tier 1 Forest Carbon Stock Estimation Module 

The Tier 1 Forest Module used for these simulations estimates the annual change in AGB and BGB guided by the data and methods outlined in the IPCC Guidelines (2006) and  Requena Suarez et al (2019). The module operates on each individual pixel. Where a pixel has tree cover at the start of the simulation (in the year 2000), FLINTpro determines the ecological zone of the pixel based on the FAO’s GEZ, and based on this, applies the corresponding IPCC Tier 1 Forest Biomass value (the maximum biomass a pixel can have). Where a pixel gains tree cover during the simulation, it is assumed to grow at a linear rate. For this, the corresponding Tier 1 AGB value and BGB ratios are used accordingly with the IPCC age classes, GEZ, BGB to AGB class. The Tier 1 values create a ‘broken stick’ growth curve for each forest type (Figure 1). The pixel will continue to sequester biomass until it reaches the maximum Forest Biomass value.  Where the pixel has tree cover loss, a clearing event is triggered within the FLINTpro simulation, where it is assumed that there is instant oxidization. Instant oxidation means that the carbon from the forest pools will all move to the atmosphere in the year of clearing.  All values in the databases can be updated through SQLite Studio. While this approach was used with the Tier 1 data, Tier 3 forest growth models, such as Chapman Richards curves, could alternatively be applied with FLINTpro.
If the land cover product does not provide a precise date (DDMMYYYY) of tree cover gain (or loss), a date is randomly allocated between the time-series observation period (using a uniform distribution). For the Hansen et al (2013) Version 1.5, this means the loss event would occur randomly within a year that loss was identified (e.g. 2012). For gain, this will occur at a random date between 2000 and 2012.
Dead Organic Matter (DOM) was applied using a relative frequency for the potentially applicable forest types. The effectively created an ‘average’ between coniferous evergreen and deciduous broadleaf forest types at the national scale. It was assumed that any pixel with tree cover at the start of the Run, the DOM was in equilibrium (no loss or gain). Where there was a loss in tree cover, dead organic matter was instantly oxidised, where there was tree cover gain, the DOM increases at a linear rate over 20 years to the new equilibrium point.

FLINTpro Global Run

Figure 1: Example of ‘broken-stick’ growth curves (AGB) for four forest types in Global Region - Africa. Growth was assumed to be linear from the year of forest gain until the maximum biomass was achieved.


FLINTpro for the EPI

In support of the Environmental Performance Index (EPI), which is produced by Yale University and Columbia University in collaboration with the World Economic Forum, FLINTpro was used to generate an estimate of carbon stock and stock change for 180 countries associated with tree cover change. As a global model, simplifying assumptions were made; however more refined data could also be used to provide more accurate country-level estimates.

Input Data

The Tier 1 implementation relies on the global datasets defined in the IPCC Tier 1 methods, together with other global datasets to help draw more meaningful results from the data. The datasets that are used for or support the IPCC Tier 1 implementation are described in Table 1. While IPCC methods and data were used for creating the estimate, given the absence of separation of anthropogenic and natural events, the results are not suitable for national reporting.

Table 1 – List of the data that is used or can be used in FLINTpro IPCC Tier 1 simulations. Default data is provided for the basic implementation, and additional data can be added by the user.

Data Required

Data Source


Administration boundary (project, national, district)

Open Street Maps

Administrative Boundaries are used as a ‘location’ input in the modelling system and help to control where the simulation is conducted. This includes determining the appropriate emission factors to be applied (based on the relevant continent). Runs were completed for regions or continents (i.e. Africa), then disaggregated into countries. Administrative boundaries were sourced from Open Street Maps.   

Climate Zone


Climate Zones represent a grouping of comparable or similar climatic variables. These may be regions with a similar annual rainfall and temperature patterns (e.g. Tropical Dry). Climate is a necessary input for determining the Tier 1 SOC and DOM values.  

Forest Cover Loss & Gain

Hansen et al

Hansen et al (2013) Forest land Version 1.5 cover change data has three products. One includes year of forest loss data covering 2000 to 2017, a second that includes if forest gain occurred between 2000 and 2012, and a third for forest extent in 2000. Where forest gain occurred, a random date was selected using an equal probability distribution.  

Spatial differentiation of forests


The IPCC Tier 1 data relates to global ecological zones. In order to simulate these in a spatially explicit manner, Global Ecological Zones (which relate climate and forest type information) where used.   

Forest Biomass


A combination of IPCC  2006 and 2019 Tier 1 data are used for estimating the emissions from forest loss or increases in carbon stock from forest gain (IPCC 2006; Requena Suarez et al 2019). This includes for AGB, BGB, and DOM. The majority of the biomass values related to the 2006 guidelines, with the only updates being the 2019 growth values for selected forest types.


Gap Filling Tier 1 Data

There are different levels of disaggregation of data presented within the IPCC guidelines, creating issues with the input data processes. For example, Table 4.4 in the IPCC Guidelines Vol. 4 (2006) for BGB, disaggregates the data into categories beyond the global ecological zones. To correct for this, a straight average across all the forest types, and biomasses was completed. For example, Subtropical Mountain systems do not have BGB estimates associated with them, an estimated value was calculated from the other subtropical ecological zones. Similarly, the global ecological zones do not always have a corresponding IPCC class within the guidelines. For example, the Desert and Shrubland ecological communities do not have a Tier 1 forest value. In the circumstance that an ecological zone does not have a direct corresponding IPCC value, an estimate was made based on proximity. For example, Tropical Desert was given the value of Tropical Dry Forest, and so on. Where regions were not specified in Table 4.7 of the IPCC Guidelines Vol. 4 (2006), more generic values from Table 4.12 were used. Growth updates available in Requena Suarez et al (2019), these were applied where applicable.


As with all modelling systems, any limitations that are present in the input will be present, and at times exacerbated, in the outputs. There are a number of known limitations with the input data used for creating the Global Run.  The benefit of FLINTpro is that these limitations are quickly identified, and can be corrected where a fix is available.
The activity data used for the run is the Hansen et al. (2013) tree cover data (Version 1.5). While a valuable data source for global information, there are known limitations with the data in this application at local levels (Hansen et al. 2013).  The data indicates where there has been tree cover loss and if there has been tree cover gain, but there is no indication as to the cause of that change (anthropogenic or natural). In the absence of a secondary attribution layer, this limits the ability to compare the outputs of this Run with published data from national inventories that show anthropogenic emissions. Further, as a global product, it is also expected that the definition of tree cover applied will not always conform with national governments.  Further, the Gain product available in V1.5 only represents gain from 2001 to 2012. While areas that gained first between 2001 and 2012 will continue to grow up until 2017, there are no new tree cover gains after 2012. This will likely cause an overestimate of emissions from 2012 to 2017.
The interactions of the other datasets also introduce uncertainties in the outcomes. The total simulated area for a country will be the area of intersection of all valid data in the included layers (GEZ, climate zone and the country national boundary). Where these three data layers do not align, areas will be omitted from the simulation. Countries that have ocean borders will have areas omitted from the analysis due to the data layers used to determine the ecological and climate zone.
The FAO GEZ data layer (2010) has simplified country borders along coastlines which excludes land areas that do not intersect this data layer. In some regions, the GEZ data layer is misaligned with the projection of the country boundaries again excluding some land areas that don’t fall within the boundary.
The climate zone data layer used also excludes portions of the country area from the simulation due to the coarseness of the dataset. The raster only provides a valid (non-zero) climate zone value for pixels that contain a certain threshold of land area inside them. This means that coastal borders are coarse and small islands can have large portions of the total land area excluded from the simulation.
At the global scale, this resulted in the exclusion of approximately 1% of the total area. However, for some countries, the combined effect of these data errors can be significant. In the most extreme example, the Marshall Islands national boundary layer has an area of 20,849ha. However, the simulated area is the intersection of this layer with the GEZ layer and any climate zone non-zero values, which results in a total simulated area of 5ha due to a combination of the errors described above. The effect of these errors on the country results will reduce in significance as the coastline length and complexity reduces relative to country size.

Areas for Improvement 

The estimate of emissions from tree cover loss and gain is in an initial estimate using global data. There are areas for improvement, including using secondary attribution data to assign the cause of tree cover change, differentiating from anthropogenic and non-anthropogenic emissions. This would vastly improve the estimates of anthropogenic emissions and would reduce the emissions in heavily forested countries such as Canada, Russia, and the United States. Using a tree cover dataset that was fully time-series consistent and identified gains and losses for a pixel through time would also remove false trends in the outputs.
The run would also be improved by moving to the full 2019 IPCC updates, or to higher tiers of methods, such as Tier 2 or Tier 3 if the data is available. This would improve the accuracy of the outputs and account for differences in national circumstances. Finally, using land cover data that is time-series consistent, and shows multiple gain and loss events would likely dramatically improve the accuracy of the outputs.


Over a relatively short period of time, using FLINTpro allowed the development of a 17-year time series of carbon stock and stock change for 180 countries. The outputs are valuable for reviewing trends in datasets, and in considering the role of non-anthropogenic events on forest systems. As FLINTpro has been explicitly designed to allow for incremental improvement, as the input data is improved, it is possible to rapidly improve the outputs such a global run.


Hansen, M. C., P. V. Potapov, R. Moore, M. Hancher, S. A. Turubanova, A. Tyukavina, D. Thau, S. V. Stehman, S. J. Goetz, T. R. Loveland, A. Kommareddy, A. Egorov, L. Chini, C. O. Justice, and J. R. G. Townshend. (2013). “High-Resolution Global Maps of 21st-Century tree cover Change.” Science 342 (15 November): 850–53. Data available on-line from: http://earthenginepartners.appspot.com/science-2013-global-forest

Intergovernmental Panel on Climate Change, (IPCC), (2006)., 2006 IPCC Guidelines for National Greenhouse Gas Inventories Volume 4: Agriculture, Forestry and Other Land Use. Available at:  https://www.ipcc-nggip.iges.or.jp/public/2006gl/vol4.html

Requena Suarez, D, Rozendaal, DMA, De Sy, V, et al., (2019)., Estimating aboveground net biomass change for tropical and subtropical forests: Refinement of IPCC default rates using forest plot data. Glob Change Biol.; 25: 3609– 3624. https://onlinelibrary.wiley.com/doi/full/10.1111/gcb.14767