GOBLIN version 1.0: a land balance model to identify national agriculture and land use pathways to climate neutrality via backcasting

The Paris Agreement commits 197 countries to achieve climate stabilisation at a global average surface temperature less than 2 C above pre-industrial times using nationally determined contributions (NDCs) to demonstrate progress. Numerous industrialised economies have targets to achieve territorial climate neutrality by 2050, primarily in the form of “net zero” greenhouse gas (GHG) emissions. However, particular uncertainty remains over the role of countries’ agriculture, forestry, and other land use (AFOLU) sectors for reasons including the potential trade-offs between GHG mitigation and food security, a non-zero emission target for methane as a short-lived GHG, and the requirement for AFOLU to act as a net sink to offset residual emissions from other sectors. These issues are represented at a coarse level in integrated assessment models (IAMs) that indicate the role of AFOLU in global pathways towards climate stabilisation. However, there is an urgent need to determine appropriate AFOLU management strategies at a national level within NDCs. Here, we present a new model designed to evaluate detailed AFOLU scenarios at national scale using the example of Ireland, where approximately 40 % of national GHG emissions originate from AFOLU. GOBLIN (General Overview for a Backcasting approach of Livestock INtensification) is designed to run randomised scenarios of agricultural activities and land use combinations within biophysical constraints (e.g. available land area, livestock productivities, fertiliser-driven grass yields, and forest growth rates). Using AFOLU emission factors from national GHG inventory reporting, GOBLIN calculates annual GHG emissions out to the selected target year for each scenario (2050 in this case). The long-term dynamics of forestry are represented up to 2120 so that scenarios can also be evaluated against the Paris Agreement commitment to achieve a balance between emissions and removals over the second half of the 21st century. Filtering randomised scenarios according to compliance with specific biophysical definitions (GHG time series) of climate neutrality will provide scientific boundaries for appropriate long-term actions within NDCs. We outline the rationale and methodology behind the development of GOBLIN, with an emphasis on biophysical linkages across food production, GHG emissions, and carbon sinks at a national level. We then demonstrate how GOBLIN can be applied to evaluate different scenarios in relation to a few possible simple definitions of “climate neutrality”, discussing opportunities and limitations. Published by Copernicus Publications on behalf of the European Geosciences Union. 2240 C. Duffy et al.: GOBLIN


Introduction
Article four of the United Nations Framework Convention on Climate Change (UNFCCC) Paris Agreement (UNFCCC, 2015) states that in order for parties to achieve long-term temperature goals, peak greenhouse gas (GHG) emissions must be reached as soon as possible. Parties must strive to "achieve a balance between anthropogenic emissions by sources and removals by sinks of GHGs" (UNFCCC, 2015). The Agriculture Forestry and Other Land Use (AFOLU) sector incorporates both agricultural activities, such as animal husbandry and crop production, and land-use, land-use change & forestry (LULUCF) activities. As such, it contains important GHG sources and sinks, making a net contribution of 24% to global GHG emissions (Smith et al., 2014). However, LULUCF is regarded as a major potential carbon dioxide (CO2) sink that will be central to any future balance between emissions and removals (IPCC, 2019b;Smith et al., 2014). Lóránt and Allen (2019) emphasise the central role that the AFOLU sector will play to reach climate neutrality, through mitigation of current emission sources, reduced emissions intensity of agricultural production linked with increased efficiency, production of bio-based products to substitute more carbon-intensive products, and carbon sequestration.
An increasing number of countries have established ambitious national "climate neutrality" targets for 2050 in legislation (Oireachtas, 2021;Reisinger and Leahy, 2019;UK CCC, 2019). These targets pose a particular challenge for countries with high per-capita GHG emissions and a high percentage land occupation with ruminant livestock production, such as Ireland (Duffy et al., 2020c) and New Zealand (NZ-MftE, 2021)because of the difficulty of reducing ruminant livestock emissions of methane (CH4) and nitrous oxide (N2O) (Herrero et al., 2016), and the large carbon dioxide (CO2) sinks needed to offset remaining CH4 and N2O based upon the 100-yr average global warming potentials (GWP100) recommended for national inventory reporting (UNFCCC, 2014). Furthermore, meeting climate neutrality targets is likely to require AFOLU sectors to be better than climate neutraland to provide net GHG offset to compensate for difficult-to-mitigate residual emissions in other sectors, such as aviation (Huppmann et al., 2018).
Hitherto, most national or AFOLU-specific plans for climate neutrality by 2050 have been based on achieving a balance between GHG emissions and removals in terms of GWP100 equivalents (Schulte et al., 2013;Searchinger et al., 2021;UK CCC, 2019). However, the warming effect of stable but continuous CH4 emissions is approximately constant, whilst the warming effect of continuous CO2 and N2O emissions is cumulative (Allen et al., 2018). Consequently, global climate modelling indicates that biogenic CH4 reductions of 24-47%, relative to 2010 are sufficient to achieve climate stabilisation at a global mean surface temperature 1.5 degrees centigrade above pre-industrial times (Rogelj et al., 2018a). A modified version of GWP100, termed GWP*, has been proposed to evaluate future climate forcing effect considering the recent change in CH4 emissions, which is more consistent with global climate modelling used to identify climate stabilisation pathways (Huppmann et al., 2018;Rogelj et al., 2018b). However, GWP* diverges from current inventory reporting, and effectively discounts attribution of recent warming caused by existing methane emissions, posing challenges for attribution and questions for international equity if applied to determine climate neutrality at national level (Rogelj and Schleussner, 2019). Furthermore, the Paris Agreement specifically mentions the need to safeguard food security and end hunger (UNFCCC, 2015). Thus, there is considerable debate and uncertainty regarding the broad suite of agricultural and land-use activities compatible with climate neutrality at individual country level, strongly depending on GHG aggregation metric (e.g. GWP100 or GWP*), and/or various approaches to downscale global emissions and sinks from particular scenarios compatible with climate stabilisation (Huppmann et al., 2018;Rogelj et al., 2018b), and the particular impacts of GHG mitigation on food production in different countries (Prudhomme et al., 2021) .There is an urgent need to explore implications of different definitions for national AFOLU sectors.
Ireland's AFOLU sector provides an excellent case study to explore the implications of different definitions of, and pathways towards, climate neutrality because it sits at the international nexus of livestock production and climate mitigation. In 2019, agriculture contributed ~34% to national GHG emissions (Duffy et al., 2021b) owing to a large ruminant sector producing beef and milk largely (90%) for international export. Somewhat unusually within Europe, Ireland's LULUCF sector is a net source of GHG emissions owing to over 300,000 ha of drained organic soils emitting approximately 8 million tonnes of CO2 eq. annually, compared with a declining forestry sink of approximately 4.5 million tonnes of CO2 annually (Duffy et al., 2020c). In 2018, the entire AFOLU sector made up ~ 40% of the Irish national emissions profile (CCAC, 2021). Methane accounts for circa 60% of agricultural GHG emissions, and LULUCF emissions of CH4 could increase if organic soils are rewetted to reduce CO2 emissions. The future shape of climate neutrality in Ireland's AFOLU sector, and the amount of beef and milk that can be produced within associated emission constraints, is thus particularly sensitive to CH4 accounting (Prudhomme et al., 2021). Nonetheless, it is clear that achieving climate neutrality will require dramatic changes in agricultural and land management practises, not least because AFOLU emissions have been increasing over the past decade (Duffy et al., 2020c). The debate about future land use has implications for livelihoods and cultural norms (Aznar-Sánchez et al., 2019), and is therefore highly sensitive. In such a context, pathways to climate neutrality cannot be objectively identified through extrapolation of recent trajectories nor stakeholder "visions", invoking the need for a backcasting approach to first establish what a climate neutral AFOLU sector could look like. This paper presents a new biophysical model capable of identifying broad pathways towards climate neutrality in Ireland's AFOLU sector, "GOBLIN" (General Overview for a Backcasting approach of Livestock INtensification). GOBLIN integrates, with sensitivity analyses, key parameters that influence agricultural production, GHG fluxes, ammonia (NH3) emissions and nutrient losses to water, using methodology aligned with Ireland's UNFCCC reporting. The model is designed to be run repeatedly with randomly varied, biophysically compatible combinations of parameter inputs in order to identify specific combinations of agricultural production and land use that achieve climate neutrality by the target year. In the following sections, we will describe the scope, model architecture, implementation and functionality of GOBLIN, ending with discussion on its suitability for intended application and conclusions.

Model classification, scope & description
Scenario analysis is one of the major methods utilised in research on the impacts of agriculture (Kalt et al., 2021). Noszczyk (2019) highlights some of the popular modelling approaches to land-use change which include, statistical and econometric, spatial interaction, optimisation, and integrated models. GOBLIN can be classified as an integrated land-use model, given that it provides links between human (including inputs and outputs) and natural land-use changes. Global examples of the integrated land-use change models include LandSHIFT (Schaldach et al., 2011) andCLUMondo (Van Asselen andVerburg, 2013). Exploratory scenarios describe plausible, but alternative, socioeconomic development pathways (Rounsevell and Metzger, 2010). Forecasting scenarios can fail to give a clear indication as to the impacts of policy implementation (Brunner et al., 2016). While, backcasting is a complementary approach to scenario development that starts with the definition of a desired future state, and then determines various pathways that will achieve that future state (Brunner et al., 2016;Gordon, 2015). The GOBLIN model embraces this backcasting approach by randomly running scenarios that are screened against a specific target (e.g. climate neutrality by 2050). Model input parameters are randomised for 100s of model runs, so that unbiased scenario outputs can then be filtered according to the pre-defined target. Crucially, these results are not limited or biased by preconceived notions of "feasibility" or "plausibility". As such, all calculated potential options for achieving the defined target are identified.
The scope of GOBLIN is currently confined to national AFOLU boundaries (Fig. 1), accounting for the main AFOLU sources and sinks reported in national inventory reporting (Duffy et al., 2020), inter alia, CO2 fluxes to and from (organic) soils and forestry, CH4 emissions from enteric fermentation, manure management, wetlands and other sources, and direct and indirect losses of nitrogen (N) from animal housing, manure management and fertiliser application, in the form of N2O, ammonia (NH3) and dissolved forms (e.g. nitrate, NO3) (Duffy et al., 2020). GOBLIN applies a gross-net approach to calculate absolute emissions and removals. This differs from recent LULUCF accounting in European Union policy that has used a net-net approach to determine changes in the GHG flux from LULUCF. Fig. 1 highlights the main sources and sinks accounted for in GOBLIN, alongside related sources and sinks that will be accounted for in subsequent life cycle assessment (LCA) through coupling and/or integration with related models (Forster et al., 2021;Soteriades et al., 2019;Styles et al., 2016Styles et al., , 2018. Key emissions sources and sinks critical to the determination of "climate neutrality" in Ireland's AFOLU sector accounted for in GOBLIN (white), alongside linked upstream-and downstream-sources and sinks to be included in subsequent life cycle assessment (LCA) modelling to determine wider climate mitigation efficacy.
In the form of a global sensitivity analysis (Saltelli et al., 2009), GOBLIN varies key uncertain parameters within the AFOLU sector to calculate emissions and removals, associated with linear rates of land use change up to the initial "target year" for neutrality. The year 2050 has been selected for this model illustration given its relevance to Irish reduction ambitions, however it is not fixed as a target year, given that various definitions of climate neutrality involve GHG flux trajectories beyond 2050. The backcasting approach used in GOBLIN makes explicit the linkages across biophysical constraints, relating model outputs (emission reduction targets) with model inputs (parameters defining production systems and land management). These explicit linkages enable GOBLIN users to better understand complementarities and trade-offs across AFOLU activities with respect to the climate neutrality objective, based on transparent and objective scenario construction. A primary aim of the model is to ensure consistency of scenarios in terms of land use (e.g. within available areas for grazing and carbon sequestration), associated agricultural production potential within land constraints (related to key production efficiency parameters), and associated GHG fluxes. The model allows scenarios to be built based on standardized sampling methods for key input parameters, avoiding sampling bias introduced by screening methods (Saltelli et al., 2000). The model is designed to run a large number (e.g. 100s) of times to generate a suite of results representing different land use scenarios to 2050 (and beyond), and time series of emissions and removals up to 2120. Scenarios can then filtered to identify which ones comply with climate neutrality based on different definitions and metrics, e.g.: (i) net zero GHG balance based on GWP100 (IPCC, 2013); (ii) no additional warming based GWP* (Allen et al., 2018;Lynch et al., 2020); (iii) compliance with a specific CH4 target downscaled from Integrated Assessment Models (IAMs) combined with a GWP100 balance across CO2 & N2O fluxes. Climate neutrality can be determined at one point in time (e.g. 2050), and/or as a time-integrated outcome over the second half of the century as per the Paris Agreement (UNFCCC, 2015). Filtered scenarios enable identification of input combinations compatible with climate neutrality as an objective evidence base for stakeholders to elaborate more detailed pathways towards climate neutrality considering wider socio-economic factors (Clarke et al., 2014).
A key feature of GOBLIN is its relation of complex interactions across livestock production, grassland management and emissions offsetting within the AFOLU sector to a few simple input parameters used to define a plethora of possible scenarios. Reflecting the dominance of bovine production within Ireland's AFOLU sector, primary input data to initialise the model are national herd sizes (derived from milking cow and suckler-cow numbers) and average animallevel productivity (e.g. milk yield per cow) to determine feed energy intake, fertiliser application rates and grass utilisation rates to determine stocking densities and production outputs, followed by proportions of any spared grassland (relative to the baseline year) going to alternative land uses. In v1.0, alternative land uses are limited to fallow or commercial or conservation forestry and rewetting of drained organic soils (bioenergy cropping and anaerobic digestion can be readily integrated for coupling with downstream energy models). Subsequent iterations and model coupling will account for upstream effects of e.g. fertiliser and feed production and extend downstream value chains to consider e.g. energy and material substitutions, taking a full LCA approach (Fig. 1). Activity data and emission coefficients are largely based on those used in Ireland's National Inventory Report (NIR) (Duffy et al., 2021b), which are in turn based on IPCC (2006) and IPCC (2019a) good practice guidelines for national GHG reporting at Tier 1 level for soil emissions, Tier 2 level for animal emissions and Tier 3 level for forestry carbon dynamics.

Modelling architectural overview
GOBLIN incorporates seven modules, displayed in a dataflow diagram (Pressman, 2010) in Fig. 2, some of which are derived from previous models on national grassland intensification (Mc Eniry et al., 2013), farm LCA (Jones et al., 2014;Styles et al., 2018) and forest GHG fluxes (Duffy et al., 2020a). The flow of data is represented by arrows between interlinked modules (brown rectangles), processes (purple circles) and data stores (green, open-ended rectangles) (Fig. 2). The scenario, herd, grassland, livestock, land-use, forestry, and integration modules included in GOBLIN reflect initiation and synthesis functions, along with data on the main activities and emissions arising within the AFOLU sector. The modules are run in sequential order, with subsequent modules relying on the output generated by previous modules.
Initially, the scenario generation module (1) varies the key input parameters utilised in the sub modules. The cattle and sheep livestock herd module (2) computes the national cattle herd and ewe flock from milking and suckler cow numbers and upland and lowland ewe numbers (input parameters) based on coefficients derived from the average national composition (Donnellan et al., 2018)see Table 3. The grassland module (3) computes the energy (feed) requirements of each animal cohort within the national herd, fertiliser application and subsequently the area of grassland needed (depending on concentrate feed inputs, fertiliser application rates and grass utilisation rate) and the grassland area free for other purposes ("spared grassland"). Emissions related to livestock production are computed in the livestock module (4) and rely on inputs from the cattle herd (2) and grassland (3) modules, based on a Tier 2 IPCC approach (Duffy et al., 2020c;IPCC, 2019a). Once the grass and concentrate feed demand has been calculated (detailed in subsequent sections), using the herd and grassland modules, the land-use module (5) computes the remaining emissions from land uses related to forest, cropland, wetlands and other land. The remaining LULUCF categories related to forest are captured in the forest module (6) and are utilised by the land-use module (5). The scenario generation module provides the proportion of spared grassland to be converted to each alternative land-use (forestry, rewetting, etc.). GOBLIN does not yet include a harvested wood products module, but the architecture anticipates this being included in subsequent versions, based on harvestable biomass outputs from the forest module related tree cohort (species, yield class and age profile) and management practises. The sequential resolution of these modules allows for an accurate representation of biophysically resolved land-use combinations in terms of land areas, production (meat, milk, crops and forestry) and emissions.

Modelling Application
Grass feed requirements are calculated based on the Tier 2 IPCC (IPCC, 2006) net energy requirements for livestock (NEfeed) related to animal cohort (c) and productivity (p), minus net energy received from supplementary (concentrate) feeds (NEsupp.) and grass net energy density (DNE-grass) (Eq. 1). Subsequent calculation of N excretion (Nex) from animals and share of time indoors (IPCC, 2019a) enables average organic nutrient loading to grassland to be calculated. Organic nutrient loading is then combined with average synthetic fertiliser application rate (exogenous variable) to determine total N inputs (Ninput) and average grass yield (Ygrass) based on the grass yield function reported by Finneran et al. (2012). According to the grass utilisation coefficient (Ugrass), calibrated for baseline (2015) animal grass feed requirements and grassland area (A-BLgrass), the calculated required area of grassland is then subtracted from the grassland area reported in the baseline year (2015) to calculate spared grass area (A-Sgrass).
Spared grassland area is apportioned to various alternative land uses based on exogenous inputs via the scenario module. The GOBLIN integration module then combines outputs from the grassland, livestock, forest and land-use modules to calculate relevant GHG fluxes. Table 1 gives a brief description of the modules and their purpose. The following sections will elaborate on scenario generation, cattle herd building, grassland management, land balance, emissions and forestry sequestration calculations. Table 1.

Module Function Details
Scenario Module The production of randomised scenario parameters.
Samples input variables from predefined maximum ranges (technical potential) with a Latin Hyper Cube algorithm to build each of the scenarios.

Herd Module
The generation of dairy, cattle, upland and lowland sheep national herd/flock numbers.
Utilises herd/flock coefficient data derived from Donnellan et al (2018) to create the national herd based on milking-and suckler-cow numbers and ewe numbers (from Scenario module).
Grassland Module Calculation of grassland area required for livestock production and calculation of nutrient application to grassland area.
Utilises IPCC (2006) guideline tier 2 functionality to calculate grass land area required based on: (i) nutritional requirements of the national herd (see Eq. 1); (ii) organic N returns to soil; (iii) average fertiliser application rates, linked with grass productivity fertiliser response curve.

Livestock Module
Calculation of agricultural emissions and nutritional requirements related to livestock production.
Includes tier 2 functionality for the estimation of nutritional requirements of livestock.
Land-use Module Calculation of emissions related to land-use and land-use change Algorithms for emissions of methane CH4, N2O, NH3 and CO2 to air based on IPCC (2006) and IPCC (2019a) methodologies.
Land-use calculations relate to forested lands, wetlands and grasslands.

Forestry Module
Calculation of emissions and sequestration related for afforestation.
Calculation of forest sequestration based on IPCC (2006), IPCC (2019a) and Duffy et al (2021b). Past sequestration is estimated as well as projected future sequestration. Other emissions associated with management of soils under forestry are also calculated here.

GOBLIN Module
Coordination and integration of the program modules and production of final results.
Management module utilising tools and functions from previous modules to produce the final results.

Scenario Generation
There are 65 input parameters included in the global sensitivity analyses that influence the outputs of GOBLIN. Table 2 outlines the definitions, baseline values and scenario ranges of the key input parameters. Categories related to productivity increases are designed to reflect efficiency gains resulting from adoption of mitigation technologies. The objective of the GOBLIN model is to identify which combinations of input variables are compatible with climate neutrality in the target year. With this number of input parameters (65) and the complexity of the relationships between them, it is impossible to study all combinations of parameters. To reduce the number of simulations while keeping a broad and unbiased exploration of the possible value ranges for these parameters, a Latin Hypercube sampling algorithm is utilised (McKay et al., 2000). This established sampling method allows the values taken by the input parameters in the scenarios to be distributed across plausible (technically possible) ranges and then utilised by downstream modules to generate results. The proportion of grass production consumed by livestock via grazing and feeding on conserved grasses (silage and hay).

57%
50% -80% Afforested area The proportion of spared grassland area on mineral soils that will be utilised for forest.
NA 0 -100% of spared mineral soil area

Cattle herd model
Calculation of national livestock numbers relies on coefficients relating animal cohorts to the numbers of milking and suckler cows (Donnellan et al., 2018). In terms of cattle production, dairy (milking) and beef-suckler cow numbers are exogenous parameters bounded between floor and ceiling values (in this use case, 0 and 1.43 and 0 and 1.55 million head respectively). A calving rate of between 0.81 and 1 for dairy cows, and between 0.8 and 0.9 for suckler cows, is used to derive the number of 1st year and second year male and female calves (48 % of male calves under 1 year, 44% of male calves between 1 and 2 years and 46% of male calves over 2 years). The dairy and suckler heifers are then derived with a replacement rate of, respectively, 0.23 and 0.15. Finally, the number of bulls is computed as a share of suckler cows. The dairy and beef herd are thus recomputed for different dairy and suckler cow numbers. Table 3 shows the coefficients utilised in the computation of national cattle and sheep herds for 2015, based on the number of milking and suckler cows, and upland and lowland ewes.  (Soteriades et al., 2018). The same is assumed in relation to beef cattle with the exception year 1 and 2 heifers whose live weights range from 275 to 322 and 430 to 503kg/head, respectively. Increased beef liveweights are based on the Teagasc sectoral roadmap (Teagasc, 2020a). Live weights, live weight gain and milk yield, are used to calculate net energy requirements for specified animal cohorts (IPCC, 2006).

Grassland management module
The purpose of the grassland module is to estimate the required area of land necessary to maintain the scenario-specific herds and flocks at a given yield and utilisation rate. National average grassland utilisation rate is calibrated at 57% of grass productivity based on calculated grass uptake and total grassland area utilised in baseline year (2015). The calibrated rate is between the average rate of 60% reported by McEniry et al. (2013), and a rate of 53% deduced from average grass dry matter (DM) utilisation report by Creighton et al. (2011) divided by average DM production reported by Donovan et al (2021). The estimation of grassland area is contingent on establishing the energy requirements of herd/flock and grassland fertilisation rates, as described above. Fig. 3 shows the data flow within the grassland module. Grassland production is computed per major soil group (Gardiner and Radford, 1980;McEniry et al., 2013), from group 1 (highest productivity potential) to group 3 (lowest productivity potential). Each grass type has a different yield class (YC) based on its soil group. GOBLIN's grassland module deduces the area required to satisfy the livestock grass demand for each category of grass (pasture, silage, hay) for each YC (1,2,3) and year. The basic equation is as follows: , , , Where refers to area demand, refers to grass type, refers to grass YC based on soil group, and refers to year. The parameter refers to the grass supply, while refers to the grass yield.
GOBLIN allocates the silage, hay and grazed grass requirement at the year t ( , ) between soil group based on the share the soil group in the grass production at the reference year (2015) Where ( ) refers to the maximum yield response to fertiliser nitrogen rate from Finneran et al. (Finneran et al., 2012) in experimental fields, given as: where is the manure excretion on pasture and , is the manure excretion on pasture in the reference year. This term considers the influence of the livestock stocking rate on pasture fertilization. For grassland other than pasture (Hay and grass silage), , = 1.
represents the nitrogen application (manure and synthetic application).
The remaining elements of equation 4 are ∝ , and ∝ , , where ∝ , refers to the yield efficiency of each YC category (0.85, 0.8 and 0.7 for respectively YC 1,2,3), and ∝ , refers to the utilisation rate (calibrated as described above).
Once land-use demand has been satisfied, the area available for land-use change ( Once the spared area ( , ) has been determined, it can then be allocated to alternative land uses.

GHG fluxes
The GOBLIN integration module coordinates the livestock and other agricultural emissions with LULUCF fluxes. The following subsections will elaborate on each of these in turn, beginning with the estimation of livestock and other agricultural emissions

Livestock emissions
This module utilises an adapted farm LCA model developed in previous studies of UK livestock systems (Soteriades et al., 2018(Soteriades et al., , 2019bStyles et al., 2015) to estimate environmental footprints. Algorithms for emissions of CH4, N2O, ammonia (NH3), and CO2 to air were applied to relevant activity data inputs. Enteric CH4 and manure management CH4 and N2O emissions were calculated using IPCC Tier 2 equations (IPCC, 2006(IPCC, , 2019a and Tier 2 calculation of energy intake and Nex according to dietary crude protein (CP) intake. Enteric fermentation is based on a methane conversion factor (Ym) value of 6.5% (4.5% for lambs) applied to gross energy intake calculated by cohort as previously described, and an average feed digestibility of 730 g/kg for Irish cattle (Duffy et al., 2020c). Soil N2O emissions are derived from Nex during grazing, and the application of synthetic fertiliser (as urea or calcium ammonium nitrate) and manure spreading. Indirect emissions of N2O were calculated based on NH3 emission and Nleaching factors from the most recent national emission inventory (Duffy et al., 2021b).
Emissions of CH4, NH3 and direct/indirect N2O from housing and manure management were calculated from total Nex indoors based on the proportion of time animals are housed, housing type, and manure management system specific emission factors (IPCC, 2019). The fraction of time spent indoors for milking cows, suckler cows, heifers, female and male calves, bullocks and bulls are respectively, 0. 43, 0.39, 0.36, 0.48, 0.07 and 0.43 (O'Mara, 2007). Manure storage NH3-N EFs of 0.05 and 0.515 of total ammoniacal N (TAN) for tanks (crusted) and lagoons were taken from (Misselbrook et al., 2010), assuming 60% of N excretion is TAN (Webb and Misselbrook, 2004)applied to 92% and 8% of managed cattle manures, respectively (O'Mara, 2007).

Soil emissions
Emissions from agricultural soils originate from mineral fertilization, manure application and urine and dung deposited by grazing animals. The average annual mineral N fertilization rate across all grassland is 70 kg ha -1 in the baseline (McEniry et al., 2013). Direct N2O emissions for manure spreading are calculated based on IPCC (2006) using an emission factor of 0.01 kg N2O-N/kg N. The NIR (Duffy et al., 2021b) utilises country specific disaggregated emissions factors from N2O-N in relation to direct emissions from faeces and urine, which in aggregate equate to 0.0088 of Nex, 56% lower than that of the IPCC (2006), but 55% higher than the IPCC (2019a) refinement. A country specific 10% leaching of fertiliser residue and grazing N inputs to water is also applied (Duffy et al., 2021b). However, it should be noted that while this leaching factor is considered "representative of Irish conditions" (Duffy et al., 2021b), this fixed factor does not allow for variation according to N loading rates. In addition, an NH3-N emissions factor of 0.06 was applied to grazing TAN deposition (Misselbrook et al., 2010). Indirect N2O-N emissions were calculated as per (IPCC, 2019a): 0.01 of volatilized N, following deposition, and 0.01 of leached N. Other sources (residues, cultivation of organic soils, mineralization associated with loss of soil organic matter) are kept constant in this version of the model, as these represent minor emission sources. NIR (Duffy et al., 2021b) country specific emissions factors relating to synthetic fertiliser direct emissions were applied. These emissions factors correspond to: 0.014, 0.0025 and 0.004 kg N2O-N/kg N applied, respectively for CAN, urea and urea + n-butyl thiophosphoric triamide. The fraction of synthetic fertiliser N that volatilises as NH3 and NOx (kg N volatilised (kg of N applied) -1 ) is also disaggregated by type (0.45, 0.097 and 0.02 corresponding to urea, urea + n-butyl thiophosphoric triamide and CAN, respectively). These values are based on updated IPCC Misselbrook and Gilhespy (2019).

Land-use module
The land-use module coordinates a range of emission calculations and allocation of spared land between different land-uses based on input parameters defined in the scenario module, as outlined in the subsections below.

Land-use allocation
Spared land is computed in the grassland module. The proportion of spared area that is organic or mineral soil is defined by the scenario input parameters. The proportion of spared area that is organic is limited by the total organic grassland area in 2015. Any spared area that exceeds the area of organic grassland soil is deemed mineral soil by default. The spared organic and mineral soil areas are then assigned various land uses. Drained organic soils are either rewetted or converted to fallow (drainage maintained) depending on scenario input regarding fraction of spared organic soils rewetted. On spared mineral soil areas, the proportion of area afforested is determined by the scenario input values. Spared area that has not been allotted to afforestation is said to be left in "farmable condition", in line with subsidy incentives. Fig. 4 summarises the apportioning of spared area in GOBLIN.

Forest emissions
Additional land-use emissions not accounted for in the forest sequestration module are calculated in the land-use module. These emissions relate to drainage and rewetting of organic soils, biomass burning, land use conversion and deforestation. The CO2, N2O and CH4 emissions from drained organic forest soils and drain ditches are based on the IPCC good practice guidelines (IPCC, 2006) and the 2013 wetlands supplement (Hiraishi et al., 2014). In addition, the NIR (Duffy et al., 2020c) breaks these organic soils into nutrient-rich and nutrientpoor organic soils. The default emission factor of 2.8 kg ha -1 yr -1 N2O-N is applied to nutrientrich organic soils, however, Duffy et al (2020c) utilise a country specific emission factor of 0.7 kg ha -1 yr -1 N2O-N on organic soils classed as poor. The CH4 emissions from drained organic soils and drained ditches are also based on default emission factors from the IPCC wetland supplement (Hiraishi et al., 2014) and country-specific parameters were derived from the NIR (Duffy et al., 2020c).

Grassland Emissions
Grassland emissions accounted for in the land-use module relate to drainage and rewetting of organic soils, biomass burning and land use conversion. A Tier 1 methodology from the IPCC (2006) is used to estimate the direct carbon loss from drainage of organic soils. The default emissions factor of 5.3t C ha -1 y -1 for shallow drained managed grassland soils for cold temperate regions is derived from the 2013 wetlands supplement (Hiraishi et al., 2014). The estimation of emissions from the drained inland organic soils derives from the 2013 wetlands supplement (Hiraishi et al., 2014). The default emission factor of 4.3 kg N2O-N yr -1 for nutrient poor, drained grassland from the 2013 wetlands supplement (Hiraishi et al., 2014) is utilised. Tier 1 IPCC (2006) methodology is used to estimate CO2 removals (from the atmosphere) via uptake by soils, CO2 losses from dissolved organic carbon to water, and CH4 emissions. Emissions factors are again derived from the 2013 wetlands supplement (Hiraishi et al., 2014). Finally, emissions of CH4 and N2O from the burning of biomass are estimated utilising the IPCC (2006) Tier 1 approach.

Wetland Emissions
Wetland emissions include CO2 from horticultural peat extraction, drainage and rewetting and burning, CH4 and N2O from drainage and burning, and CH4 from rewetting. The NIR (Duffy et al., 2020c) includes emissions related the extraction and use of peat products under the category of "horticultural peat". Data related to the quantities of exported peat are reported by United Nations Commodity Trade Statistics Database (UN, 2016). To calculate off-site emissions from peat products, GOBLIN utilises a Tier 1 methodology (IPCC, 2006) to estimate carbon loss by product weight.
Carbon stock changes in biomass are determined by the balance between carbon loss due to the removal of biomass when preparing for peat harvesting, and the gain on areas of restored peat lands (Duffy et al., 2020c). Non-CO2 emissions related to drainage and rewetting are CH4 and N2O. CH4 emissions are estimated in accordance with the 2013 wetlands supplement (Hiraishi et al., 2014) and require data on the area impacted by drainage and the density of drainage ditches. Annual direct N2O-N emissions from drained organic soils are estimated utilising a Tier 1 approach based on the IPCC (2006) methodology and a default emission factor of 0.3 kg N2O-N yr -1 .
GOBLIN also calculates emissions from CH4 and N2O from biomass burning. The value used in the NIR (Duffy et al., 2020c) to represent the mass of fuel available for burning is 336 t ha -1 DM. The emissions factor values utilised for CO2, CH4 and N2O correspond to 362 g kg -1 , 9 g kg -1 and 0.21 g kg -1 DM burned, respectively.

Cropland Emissions
Cropland emissions are estimated utilising a Tier 1 approach (IPCC, 2006). CO2 emissions include emissions related to land-use transitions from grassland or forested land to cropland and from biomass burning. N2O and CH4 are also related to biomass burning. Emissions of CO2, CH4 and N2O from the burning of crop biomass are also estimated utilising the IPCC (2006) Tier 1 approach.

Forest management
Irish forest cover accounts for about 11% of total land area (DAFM, 2018). Conifers make up over 71% of the forest estate, the main species being Sitka spruce (Picea sitchensis (Bong.) Carr.) (SS) comprising over 50% of total forest land area. In 2017, broadleaf species made up almost 29% of total forest land area (DAFM, 2018;Duffy et al., 2020bDuffy et al., , 2020a. However, given that the historic rate of broadleaf inclusion within afforestation was less than 10% for significant periods (DAFM, 2020b), GOBLIN utilises an aggregate value of 20% broadleaf inclusion to represent historic afforestation. Given the complexity of representing the current forest estate, and simulating future afforestation/reforestation, the forest module is split into two containers: the old forest container (OFC) and the new forest container (NFC). The OFC estimates sequestration from afforestation from 1922 until 2025 and is used to determine the age profile of standing forest. After 2025, the OFC no longer adds area to the model, but continues calculation of growth (carbon sequestration) and harvest (terrestrial carbon removal) in pre-existing forested area until the end of the simulation has been reached (2050 in our example).
From 2025 onwards, sequestration from afforestation is calculated in the NFC utilising annualised afforested areas derived from the target-year spared area calculated in the grassland management model and shares of that area going to forest types (scenario module). The NFC computes sequestration from afforestation from 2025 to the end point (target year) of the simulation. The results of the OFC and NFC are added together to calculate total net sequestration in forests. The purpose of this two-step calculation is to save system resources. Net sequestration in the existing forest estate only needs to be calculated once as it remains the same across different scenarios, irrespective of changes in the afforestation rate. As such, we utilise the OFC a single time, adding the static results to the variable output from each scenario generated in the NFC. Fig. 5 illustrates the flow of data through the forest model. The brown rectangles represent entities, mainly conifer and broadleaf, for old and new forest. The purple circles represent processes, while the green rectangle represents a common data store. The old and new forests are kept in separate containers before being aggregated. To estimate the various elements (sequestration from biomass, organic and mineral soil emissions, dead organic matter, etc.) for the forest estate, a matrix approach is adopted. For each element in the forest model, a value matrix is established based on the age of the forest stand. Stand age is then utilised to establish the total biomass, dead organic matter and emissions from organic soils. Once the final matrix has been established, it is aggregated into a single vector with a single cell per year. At this point, any further annual additions or subtractions that need to be made are factored into the model. For further detail on the calculation of biomass increment, DOM, organic and mineral soil emissions refer to Duffy et al (2020a).

Model validation
The main purpose of the GOBLIN model is to provide an evidence base for climate action in Ireland's AFOLU sector, aligned with existing GHG accounting procedures that will ultimately be used (with refinements through time) by policy to track progress towards climate neutrality. Acknowledging the significant scientific uncertainty around many AFOLU fluxes, the most appropriate manner to validate GOBLIN in relation to its core purpose, is to test how well it replicates NIR fluxes from the same activity data. Largely, these activity data are input to GOBLIN in the same format as for the NIR, with some differences relating to the simulation sequence, most notably for animal cohort numbers, which are derived from milking cow, suckler cow and ewe numbers. Therefore, to validate national cattle herd estimations (accounting for the vast majority of livestock emissions), outputs from the herd module derived from Donnellan et al (2018) coefficients, were compared with NIR activity input data from 1990 to 2015 (Fig. 6). The coefficients utilised in GOBLIN are derived from recent data, so the accuracy of total cattle number estimations increases through time, converging in 2015.

Average cattle livestock population (lines) and standard deviation among sub-groups over time (shaded areas) inputted to the national inventory report (NIR) and generated by the GOBLIN herd module from milking-and suckler-cow numbers, respectively.
GOBLIN applies a range of IPCC default and Ireland-specific emissions factors in line with the NIR. The EPA has implemented a detailed quality control and assurance procedure for Ireland's NIR reporting. This includes auditing and external reviews of the agriculture sector and the Emissions Trading Scheme (Duffy et al., 2021b). Table 4 shows the complete list of Irish specific emissions factors utilised. To assess whether or not GOBLIN has achieved its goals, validation of emission and removal calculations for livestock production and land use (change), as well as forest biomass calculations were carried out utilising real-world activity data supplied by the Central Statistics Office (CSO). These activity data are also input to the NIR (with some minor differences relating to derived variables for simulation purposes), so that GOBLIN should generate almost identical time series of emissions and removals as the NIR using past input data. GOBLIN outputs over 1990 to 2015 were compared with NIR outputs over the same time period, using CRF files dating back to 1990. Fig. 7 and 8 illustrate the validation of GOBLIN's replication of NIR flux accounting across major emissions and removals sources.
Beginning with land-use and land-use change (Fig. 7), solid lines represent CO2, CH4 and N2O emissions modelled in GOBLIN, while the dashed lines represent equivalent emissions reported in the NIR. Absolute emission levels and trends calculated by GOBLIN very closely match those of the NIR, with the most notable deviation arising for forest sequestration (representing the complex Tier 3 modelling of fluxes, sensitive to compound estimates of stand age profiles across hundreds of land parcels). Fig 8. shows validation of agricultural emission sources. Enteric and manure management CH4 from GOBLIN and the NIR are almost identical, while CO2 and N2O emissions levels and trends are very similar. This validation specifically indicates that emission factors, land area calculations, forestry increments and harvest removals, and animal feed intake calculations derived from raw input data are in line with NIR methodology, providing confidence in scenario extrapolations based on variations in these input data.

Example of Model Output
To demonstrate and explore the critical functions of GOBLIN, several scenarios were analysed to reflect national level GHG reductions within the AFOLU sector (Table 5). As set out in Ireland's Climate Action Bill (2021), Ireland must achieve a 51% emission reduction by 2030. Given that agriculture makes a significant contribution to the national emissions profile (DAFM, 2020a), the illustrative scenarios produced as part of this model summary reflect potential emissions reduction pathways. In terms of animal numbers, all scenarios reflect reductions in dairy, beef and sheep numbers of 10%, 50% and 50%, respectively, by 2050. In terms of land-use, all scenarios, with the exception of scenario 4, assume at least the baseline (recent average) afforestation rate continues to 2050 (the average afforestation rate was 6,664 ha yr -1 between 2006 and 2017 (Duffy et al., 2020a)). All annual afforestation rates continue to 2050, with zero afforestation assumed after 2050, and are based on a 70:30 conifer:broadleaf mix. •

100% of drained organic soil under grassland rewetted
• Remaining area assumed to be mineral and afforested.
• Remaining organic area taken out of production 26086 4 Animal reduction and increased production • Dairy, Beef and sheep herd numbers reduced by 10%, 50% and 50% by 2050, respectively.
• Milk output increased by 14% per cow • Beef live weight + 20% 0 5 Animal reduction, increased production, afforestation and wetlands • Dairy, Beef and sheep herd numbers reduced by 10%, 50% and 50% by 2050, respectively • Milk output increased by 14% per cow • Beef live weight + 20% • 100% of drained organic soil under grassland rewetted • Remaining area assumed to be mineral and afforested.
• Remaining organic area taken out of production Fig. 9 and 10 present the main AFOLU GHG fluxes. Firstly, the agricultural emissions ( Fig. 9) illustrate the results for CH4 emissions from enteric fermentation and manure management, N2O results from manure management and other direct and indirect N2O emission pathways, and finally, CO2 emissions from fertiliser application to soils. Emissions related to livestock are slightly higher in scenarios that have increased production related to milk and beef output than scenarios with default production estimates. Fig. 10 illustrates land-use emissions related to CH4, N2O and CO2. Firstly, we examine CH4 emissions from land-use and land-use change. The changes relative to the baseline year are as a result of a decrease in grassland area and changes in forest and wetland areas. Changes in grassland CH4 result from reduction in animal numbers, rewetting of organic soils and removal of production from organic soils. Relative to scenario 0, the straight animal reduction scenario, there is a 19, 20 and 22% increase in CH4 emissions in scenarios 1, 3 and 5, respectively largely owing to rewetting of drained organic soils. These increases are largely observed in the grassland category, with additional emissions in the wetland and forest categories. In the wetland and cropland categories, an increase is observed relative to the baseline year. This is explained by the utilisation of a multi-year average to estimate the burned area, this average is higher than the baseline year, as such emissions related to burning in the target year are higher.
Secondly, we examine N2O emissions related to land-use and land-use change. Relative to scenario 0, we can observe a 3-4% increase in emissions for scenarios 1, 3 and 5. The increases in emissions from wetland areas are related to the rewetting of previously drained soils. Again, we can see that cropland emissions increase, however, this is, again, a reflection of burned area assumptions. The next noticeable difference is in terms of grassland N2O emissions which appear to fall dramatically. Past N2O emissions in this category are driven largely by conversion of modest amounts of forested land to grassland. As the model assumes land is converted from grassland to other uses, and not the other way around, the emissions in this category drop significantly. Relative to scenario 0, emissions in scenarios where rewetting takes place increase by 20%. As there are no changes to cropland, emissions remain constant among scenarios, the increase relative to the baseline year is again explained by assumptions regarding the burned area.
Finally, Fig. 10 presents the CO2 emissions from land-use change. Emissions related to grassland drop to less than 0.1% relative to scenario 0 where rewetting has taken place (scenarios 1, 3 and 5). Regarding forestry, Fig. 10 highlights the expected value in 2050, drawing a line linearly from 2015 to 2050. As expected, sequestration potential is greater at higher levels of afforestation. The entire time series is explored in more detail in Fig. 10. Wetland emissions increase, relative to scenario 0, by 4 to 5% in scenarios in which rewetting takes place. Lastly, we have assumed no emissions changes for cropland.
To further elaborate the forestry modelling, Fig. 11 shows the forest sequestration time series for each of the scenarios. As can be seen, scenarios 0, and 1 reflect the average afforestation rate, or the "business-as-usual" land-use change, while scenario 4 has no afforestation. Scenarios 2, 3 and 5 increase sequestration potential significantly. Scenario 2 assumes that all spared area is mineral soil and as such this scenario has the highest afforestation rate, and the highest sequestration potential. Scenario 3 assumes that all drained areas are rewetted, and the remaining land area is mineral and afforested. Lastly, scenario 5 assumes the same, however, there is less land area available as a result of increased production output from animals. The time series also inherently factors in the harvesting rates. All scenarios assume that afforestation, if applicable, takes place up to 2050, with zero thereafter.

GOBLIN scenario GHG balance through time based on CO2e aggregation using GWP100 (blue line) and GWP* (black line)
Finally, Fig. 12 represents aggregated GHG emissions from the AFOLU sector for each scenario using either GWP100 or GWP* to equate warming potential to CO2e. The calculation of GWP* is based on Lynch et al. (2020). The aggregated emissions are presented net of forest sequestration in order to present a final emissions balance. As can be seen, the reduction in animal numbers drives both emissions reductions. The rewetting of previously drained land provides an easy win in terms of emissions reductions despite additional CH4 and N2O emissions. However, the potential to offset remaining emissions, in terms of carbon sequestration, comes by utilising spared land for afforestation. Both organic soil rewetting and higher rates of afforestation are needed to reduce the GWP100 emissions balance, which in the best case (scenario 3) is reduced by circa 73% from the 2015 balance. Fig. 13 presents an extended time series for forest sequestration to 2120. Specifically, Fig. 13 illustrates afforestation to 2050, with zero afforestation thereafter. A forest conservation approach is considered for all new forest, assuming a 0% harvest rate. This conservation approach does successfully avoid the so called "carbon cliff" in scenarios 2, 3 and 5. However, the marginal gains are reduced over time as trees reach maturity. Ongoing model development will enable longer-term mitigation associated with harvested wood use to be represented.

National AFOLU models for climate policy
The AFOLU sector is central to global efforts required to stabilise the climate, and will need to shift from being a net source to a net sink of emissions by 2050 in order to constrain anthropogenic global warming to 1.5 o C (Masson-Delmotte et al., 2019). Such a shift will require widespread and rapid deployment of appropriate mitigation options to reduce the emissions intensity of agricultural production whilst maintaining food security, alongside food demand management and actions to realise emissions removals via forestry and bioenergy (Huppmann et al., 2018;IPCC, 2019b). The GOBLIN model described here was developed as a tool to quantify long-term (circa 100 year) GHG emission fluxes associated with different AFOLU scenarios representing changes in land-use over the next three decades. The intention is to bridge the gap between hindsight representation of national emissions via UN FCCC reporting (Duffy et al., 2020c) and global IAMs (Huppmann et al., 2018) that are broad in scope but lack (sub)national detail. IAMs global pathways towards climate stabilisation involve many assumptions and are difficult to downscale to national targets. Whilst a number of countries have set national "net zero" GHG emission targets for 2050 (UK CCC, 2019), there remains considerable uncertainty about the role of distinct national AFOLU sectors, particularly with respect to appropriate targets for CH4 emissions and CO2 offsetting within NDCs (Prudhomme et al., 2021). Ireland provides an excellent case study country to explore possible trade-offs between food production and various definitions of climate neutrality owing to high per capita GHG (including CH4) emissions from the AFOLU sector, both from ruminant food production destined for export and from land management (Duffy et al., 2020c).
GOBLIN has been calibrated against Ireland's NIR (Duffy et al., 2020c) to align outputs with GHG reporting methodologies, but is novel in its integration with a land balance approach to determine future combinations of emissions sources and sinks related to animal feed energy requirements and grass production under different fertilisation and grazing (utilisation efficiency) regimes. Through integration of animal energy demand functions and grass fertiliser response curves, the model is able to vary areas needed to support different combinations of livestock systems at the national level. This functionality enables critical aspects of livestock production efficiency to be explicitly varied within scenarios, providing deep insight into interactions between livestock production, including sustainable intensification trajectories (Cohn et al., 2014;Havlík et al., 2014) that represent implications for future food production, and biophysically compatible levels of organic soil rewetting and sequestration across forest types. The latter functionality derives from integration of aforementioned livestock system modelling with detailed representation of the complex carbon dynamics of existing and "new" forests. This represents a significant advance in national AFOLU GHG modelling capability, and will build on modelling of livestock emissions displacement with forestry offsets recently calculated in Duffy et al (2020a) to provide a solid evidence base for development and implementation of NDCs.
Crucially for a national AFOLU sector so far from complying with any definition of climate neutrality, fully randomised scenario simulations within GOBLIN will generate new evidence on which biophysically coherent combinations of agricultural activities and land uses satisfy particular definitions of climate neutrality. The combination of randomisation and a backcasting approach to filter climate neutral scenarios can inform objective comparison of trade-offs and may also help to elicit more constructive and focussed stakeholder engagement on a complex and sensitive topic. The small number of scenarios modelled in this paper were designed simply to demonstrate the technical potential of the model, but ultimately, GOBLIN provides a platform to support participatory modelling (Basco-Carrera et al., 2017) or systematic analysis of alternative land-use choices (Loucks and Van Beek, 2017). Combining the biophysical outputs of GOBLIN with socio-economic assessment will be crucial to determine effective climate policy at national level.

Defining "climate neutrality"
When model development began in 2018 it was assumed that achieving "net zero" GWP100 balance would be the primary objective for GOBLIN scenario modelling. Such an approach remains valid and in line with UN FCCC reporting, and is applied for other countries' 2050 climate targets (Lóránt and Allen, 2019; UK CCC, 2019). Since then, there has been significant debate about how to combine the short-term warming effect of CH4 with the long-term cumulative warming effect of CO2 and N2O (Cain et al., 2019;Prudhomme et al., 2021) . An important but initially unanticipated use of GOBLIN will therefore be to explore the implications of various possible definitions of "climate neutrality", underpinned by different value judgements. It is clear from the small selection of indicative scenarios analysed in this paper that choice of GHG aggregation metric and definition of climate neutrality profoundly alters the mix of agricultural production and land-use (change) compatible with climate neutrality in 2050 and beyond. None of the scenarios meet climate neutrality in the traditional GWP100 sense. However, a "no further warming" definition, represented by a zero balance for GWP* (Lynch et al., 2020), is achieved (or surpassed) by 2050 among four of the six indicative scenarios explored here, whilst "net zero GHG", represented as a zero balance for GWP100 (IPCC, 2013), is not achieved across any of the scenarios by 2050. For example, reducing the dairy herd by 10%, and beef cattle and sheep numbers by 50%, could result in "no further warming" (GWP* balance) climate neutrality in 2050 assuming all organic soils are rewetted and recent rates of afforestation (just under 6,700 ha yr -1 ) are maintained. However, the same scenario brings the AFOLU sector only half way towards net zero GHG emissions (GWP100 balance) by 2050. Separate calculation of each major GHG within GOBLIN will enable a wider range of climate neutrality "filters" to be applied beyond these simple GWP balance examples, such as a separate target for CH4 combined with a GWP100 balance across N2O and CO2. Over half of global CH4 emissions come from food production (Saunois et al., 2020); detailed modelling of ruminant food production compatible with various approaches to determine territorial climate neutrality could contribute significantly to policy formulation on separate CH4 targets, e.g. the EU Methane Strategy. Additionally, cumulative GWP* and GWP100 can also be applied as neutrality filters.

Model limitations and development priorities
GOBLIN examines rewetting of drained organic soils and forestry as the primary mechanisms of emissions mitigation and offset within Ireland's LULUCF sector, reflecting the "main levers" that can be pulled to achieve climate neutrality. Additional land-use-technology interactions that could realise significant GHG mitigation by 2050 include, for example, bioenergy crop production, such as willow and miscanthus for electricity, heat or advanced liquid biofuel chains, and manures or grasses for biomethane production (Englund et al., 2020;Van Meerbeek et al., 2019). GOBLIN can be adapted and coupled with existing downstream energy emissions models to explicitly represent AFOLU consequences of such options, as well as to illustrate inter-sectoral mitigation pathways (Fig. 1). In this regard, it is important to note that the forestry element of GOBLIN is relatively sophisticated, representing forest composition in terms of broadleaf and conifer species mixes, differing forest management practises and harvest rates. This provides interesting possibilities to link AFOLU mitigation with future use of harvested wood products, possibly in cascading value chains that store carbon in wood products before end-of-life use for bioenergy carbon with capture & storage (BECCS) that can transform forestry CO2 sequestration into potentially permanent offsets (Forster et al., 2021). One of the first applications of GOBLIN will be to couple AFOLU forestry outputs with downstream LCA modelling of wood value chains in order to generate robust projections of CO2 offsetting out to 2120, providing new insight into the post-2050 longevity of various climate neutrality scenarios. Additionally, cropland areas are kept constant, reflecting the minor role of crop production in Ireland's current agri-food system and GHG emission profile. Nonetheless, future versions of GOBLIN should allow cropping area to be changed, reflecting potential increase in demand for plant-based proteins, in place of animal protein (Tilman and Clark, 2014). Finally, whilst GOBLIN has been extensively validated against the NIR for current management practises, components such as fertiliser-response curves for grass productivity could be altered by new grass varieties or mixed grassclover swords, or updated to be more spatially explicit in relation to soil and land categorisations (O'Donovan et al., 2021). There is potential to adapt this (and other) components of GOBLIN to represent specific mitigation options. Acknowledging that there are still important developments related to, inter alia, management of harvested wood products and bioenergy production to be included in future iterations of the model, GOBLIN represents a powerful tool for academics and policy makers to better understand what is required to reach climate neutrality within Ireland's AFOLU sector (and indeed other national AFOLU sectors dominated by livestock production). Crucially, GOBLIN decouples scenario generation from preconceptions of what pathways to climate neutrality could look like by enabling randomised scenarios to be generated and filtered in a backcasting approach. Although such modelling on its own cannot provide all the answers, it does establish a range of biophysically plausible targets which stakeholders can select from and choose to navigate towards, considering important factors such as delivery of wider ecosystem services, and socio-economic and cultural feasibility. Future iterations of the GOBLIN model will seek to explicitly model the effect of land-use change on a wide range of ecosystems services via the inclusion of a broader set of LCA impact categories and ecosystem service indicators.

Global Transferability
GOBLIN is parameterised utilising emissions factors and land-use characteristics related to Ireland's AFOLU sector, in line with specific national climate neutrality modelling objectives. However, the model is based on the IPCC GHG accounting framework, and refactoring for wider spatial applicability was considered from the outset. In this regard, each module contains its own database of emissions factors. The source country is utilised as the primary key, and the relevant country for the scenarios can be selected upon initialisation of the model run. This does not mean that GOBLIN is currently ready to deliver international results. Significant refactoring would be required across various country-specific functions, such as grass fertiliser response curves and grass utilisation efficiency. Livestock intensive, temperate contexts will be significantly easier to parameterise owing to similar biophysical characteristics and EFs. For example, the model is currently being adjusted to include Scotland as an output country. However, contexts that differ a great deal from that of Ireland will require significantly greater refactoring. Modules related to land-use and land-use allocation will potentially require the most detailed refactoring depending on how much they depart from the Irish context. In addition, the forest module, being Tier 3 at present, would need to be rebuilt for each country (or at least agro-ecological region) of application. Additional livestock categories and cohorts would also be necessary for specific regions. The modular nature of the model allows for "plugin" of new modules, or "plug-out" of unnecessary modules depending on user needs. This adds flexibility and simplifies integration of new components in future iterations. Thus, the value of GOBLIN lies in its regional specificity to explore climate neutrality pathways aligned with much coarser resolution IAMs projections, and this currently limits applicability to Ireland, but with high potential for application in other livestock intensive, temperate contexts following modest adaptations.

Conclusion
The AFOLU sector is both a source and a sink for GHG emissions. The sector will play a key role in mitigation of emissions via reduced agricultural emissions intensity and increased carbon sequestration and other off-setting/displacement activities. GOBLIN is a high resolution integrated "bottom-up" bio-physical land use model for Ireland's AFOLU sector.
The novelty of GOBLIN lies in its integration detailed land requirements and GHG emissions associated with different levels of livestock intensification and grassland management on the one hand, and sophisticated representation of forestry carbon dynamics on the other, alongside other important land-use emission sources and sinks. GOBLIN is aligned with, and validated against, Ireland's inventory reporting methodology for GHG emissions, including a Tier 2 approach for livestock emissions and a Tier 3 approach for forestry. By calculating GHG flux trajectories towards (randomised) future (2050) scenarios of agricultural activities and landuse (change), GOBLIN is able to provide new insight into the biophysical boundaries associated with different definitions of climate neutrality. This could help ground an increasingly polarised debate around the role of AFOLU in ambitious national climate policy. Detailed representation of current and future forestry combinations (species, management and harvesting mixes) also provides a powerful platform for future downstream modelling of harvested wood product uses in the bioeconomy. This could be complemented by integration of bioenergy uses for spared land through further model development and/or coupling with existing bioenergy models and will enable the evaluation of long-term (to 2120) GHG fluxes in order to determine more enduring climate neutrality actions. Following model development and validation, GOBLIN will be used to provide a unique, impartial and quantitatively rigorous evidence base on actions and strategies needed to achieve climate neutrality across Ireland's AFOLU sector.

Code Availability
The exact version of the model used to produce the results used in this paper is archived on Zenodo (Duffy et al., 2021a) and freely available for download.