This study employs an atmosphere-only configuration of the United Kingdom Earth System Model vn1.1 (UKESM1.1; ) at a resolution of N96L85 (1.25° × 1.875° horizontal resolution; 85 vertical levels up to 85 km), with the addition of several new or updated processes that are relevant to biogenic emission impacts on atmospheric chemistry and aerosols. The impact of these processes, which are detailed in the following paragraphs, has been previously assessed in a present-day climate simulation . The Global Atmosphere 7.1 science configuration of the Met Office's Unified Model (MetUM) is used to simulate the physical atmosphere, while the United Kingdom Chemistry and Aerosols (UKCA) model with the Global Model of Aerosol Processes (GLOMAP-mode; ) are the chemistry and aerosol components.

The vegetation is represented by 13 plant functional types (PFTs), the distribution of which is prescribed based on a pre-existing coupled transient historical simulation of UKESM1.1. In this pre-existing simulation, the vegetation was interactive. Outputs from the same simulation are used to prescribe the canopy height and leaf area index (LAI) values. The PFT distribution and LAI drive BVOC emissions, for which the interactive biogenic volatile organic compound (iBVOC) scheme is used . A key difference in the model set up here compared to the standard UKESM1.1 is the use of updated isoprene and monoterpene emission factors .

Two mechanisms for fully interactive stratospheric and tropospheric chemistry are used in this study. StratTrop vn1.0 (ST; ) is the standard chemistry mechanism in UKESM1.1, while CRI-Strat 2 (CS2; ) is a more complex scheme, which includes HO_x recycling during isoprene oxidation, the separation of lumped monoterpenes into α-pinene and β-pinene, and a more explicit oxidation scheme for both monoterpenes that includes the formation of trace gases, predominantly oxygenated VOCs. The schemes are referred to as the standard and complex mechanisms, respectively. In both mechanisms, monoterpenes and isoprene contribute to aerosol mass through the formation of condensable vapours (Sec_Org_MT and Sec_Org_I) which condense onto pre-existing particles. We further include organically-mediated boundary layer nucleation (BLN), through which Sec_Org_MT can contribute to new particle formation in the boundary layer.

GLOMAP-mode simulates four aerosol types: sulphate (SU), sea-salt (SS), black carbon (BC) and organic matter (OM). These are combined with the Coupled Large-scale Aerosol Simulator for Studies In Climate (CLASSIC) aerosol scheme for mineral dust (DU). We apply a correction to the hygroscopicity values of the different aerosol types, which includes an increase in the hygroscopicity of OM. Previous research has found that the response in ACI associated with increased OM are atypical in UKESM1.0 compared to other ESMs, which is attributed to relatively inefficient cloud nucleation . This response may be affected by the change in hygroscopicity. Due to the coupling between the gas-phase chemistry and aerosols, the formation and growth of aerosol particles responds to changes in the simulated trace gases that are aerosol precursors, as well as changes in oxidants (e.g. ).

Calculation of the land use ERF is based on the comparison of a PI control (piClim-control*) and an experiment where the only change is the use of land cover representative of the year 2014, assumed to be PD (piClim-lu*) (Table ). The difference between the experiment pairs (perturbation - control) represents the impact of changing land use. This isolates the impact of changing vegetation, although emissions associated with the expansion of agriculture, such as NO_x from fertiliser applications, are excluded. Each experiment was run for 45 model years. The results presented are based on the latter 30 years of model simulations, following the protocols from the Radiative Forcing Model Intercomparison Project (RFMIP) and recommendations from . In all experiments, the radiative balance is sensitive to changes in aerosols, ozone, methane, cloud properties, water vapour and land surface properties. While the plant functional type, canopy height and leaf area index, as well as methane concentration, are prescribed, the other radiative feedbacks are prognostic.

We make use of a pre-existing experiment pair from UKESM1.1 (piClim-control and piClim-lu; ) and run two new key experiment pairs, which use two different chemistry mechanisms. Two simulations (piClim-control_S and piClim-lu_S) include the additional processes described in the previous subsection (Sect. ), except for the complex chemistry mechanism. As the complex chemistry mechanism has been found to have a substantial influence on BVOC impacts , it is included alongside other processes in the second experiment pair (piClim-control_C and piClim-lu_C). Two additional experiments (the pre-existing piClim-CH4 and new piClim-CH4_S) were used to calculate the methane (CH₄) feedback factor on its own lifetime. Further, the piClim-CH4_S experiment can be used to calculate the CH₄ effective radiative forcing (ERF_CH4) associated with land use change.

The net simulated land use effective radiative forcing (ERF_LU) is calculated based on the change in the net incoming (downwelling - upwelling) top of atmosphere (TOA) fluxes in both the shortwave (SW) and longwave (LW) between the land use perturbation experiment (piClim-lu_S or piClim-lu_C) and the respective control (piClim-control_S or piClim-control_C) simulations. The methodology to quantify the forcing associated with a particular gas, aerosol effects or surface albedo varies; the approaches used are described in more detail in the paragraphs below.

The recommendations of are followed to calculate the instantaneous radiative forcing (IRF) from aerosol-radiation interactions (IRF_ARI), as well as the cloud forcing from changes in cloud-radiative effects (ΔCRE): 1IRFARI=Δ(F-Fclean)2ΔCRE=Δ(Fclean-Fclear,clean) where F is the TOA radiative flux (downwelling - upwelling), Fclean is the TOA radiative flux excluding scattering and absorption by aerosols and Fclear,clean is the TOA radiative flux excluding cloud effects and scattering and absorption by aerosols. The calculation of radiative fluxes multiple times with different forcings included is referred to as the double call method. The aerosol-free fluxes are used to calculate the CRE, as aerosols above clouds also interact with the radiation and can bias the forcing diagnosis at TOA .

A double call to the radiation scheme is also used to calculate the instantaneous radiative forcing driven by changes in ozone concentrations (IRF_O3). For each simulation, the radiative fluxes are calculated once using a reference O₃ distribution and once using the simulated O₃ concentrations. The same reference distribution data, in this case monthly mass mixing ratios for the year 2014, are used for all simulations, so that the difference between the reference and simulated O₃ radiative fluxes can be compared between simulations to calculate IRF_O3 .

The surface albedo forcing (RF_Alb,surf) is calculated following : 3RFAlb=ΔFsurf,clear,clean×SWDW,surf,cleanSWDW,surf,clear,clean where Fsurf,clear,clean is the surface flux excluding scattering by aerosols and clouds, and (SWDW,surf,clean/SWDW,surf,clear,clean) is the ratio of the downwelling SW clean (excluding scattering and absorption by aerosols) to SW clear-clean (excluding scattering and absorption by aerosols and clouds) fluxes at the surface. This method is used so that the albedo effect is not amplified by the removal of the cloud and aerosol effects, which permits a higher proportion of SW radiation to directly reach the surface/TOA. However, the RF_Alb,surf is quantified at the surface, whereas the aerosol and O₃ forcings are calculated at the TOA. The presence of SW absorbing trace gases and aerosols in the atmosphere means not all of the reflected radiation would reach the TOA.

Two alternative surface albedo forcing calculations, which are valid at the TOA, are also explored to test the sensitivity of this result to the methodology. The simplest method makes the assumption that the change in SW TOA fluxes excluding scattering and absorption by aerosols and clouds is predominantly due to changes in surface albedo , so that 4RFAlb,clear,clean,TOA=ΔFSW,clear,clean. This method is expected to overestimate the albedo forcing, as the masking effect of clouds and aerosols is completely removed. Finally, the forcing can be estimated from the clear-sky change in surface albedo, the incident SW radiation at the top of atmosphere and the clear-sky atmospheric transmissivity (0.7) following and : 5RFAlb,clear,TOA=-0.7×ΔAlbedo×SWDW,TOA. This method includes a correction for the effects of gases and aerosols in the atmosphere that interact with the reflected SW radiation and provides an estimate for the TOA surface albedo forcing. However, the masking effect of clouds is not included.

The impacts of changes to CH₄ and CO₂ are not included in the net ERF_LU calculated from the land use perturbation simulation pairs, as the concentrations of both greenhouse gases are prescribed. Two methods are used to calculate the CH₄ forcing for the standard chemistry mechanism, but only method one is applied for the complex chemistry mechanism due to the high computational cost of the second approach. The first method follows a multi-step process to estimate the RF_CH4 based on the change in CH₄ concentration expected from the difference in CH₄ lifetime between the simulation pairs. The CH₄ lifetime for each simulation is initially calculated against removal by OH, but adjusted for stratospheric and soil sinks, following . Next, the change in CH₄ concentration based on these lifetimes is calculated as in multiple previous studies e.g.: 6[CH4]piClim-LU=[CH4]piClim-control×τpiClim-LUτpiClim-controlf where [CH4] is the CH₄ concentration, τ is the CH₄ lifetime and f is the feedback factor accounting for the feedback of CH₄ on its own lifetime . Changes in CH₄ affect OH concentrations, which in turn impact CH₄ oxidation and lifetime . We note the CH₄ feedback on its own lifetime varies depending on the Earth system model and background state of the atmosphere. We first calculate f based on outputs from pre-existing UKESM1.1 simulations (piClim-control and piClim-CH4) and the following equations: 7f=1/(1-s) where: 8s=Δln⁡τΔln⁡[CH4]. Next, the direct radiative effect from changes in CH₄ concentrations is calculated following : 9RFCH4init=aM‾+bN‾+0.043×M-M0 where a and b are constants (a=-1.3×10-6 W m⁻² ppb⁻¹ and b=-8.2×10-6 W m⁻² ppb⁻¹), M is the concentration of CH₄ (ppb) that would arise following the land cover change if the modelled concentration was free to evolve, M0 is the prescribed PI concentration, while M‾ is the mean of the two concentrations of CH₄. N‾ refers to the PI concentration of N₂O (ppb) in this case. Finally, RF_CH4init is scaled to account for methane's indirect radiative effects due to its impacts on O₃ and stratospheric water vapour production, as well as the cloud radiative effect. Previous research based on multi-model assessments suggested a scaling of 1.31 . However, this scaling may not be representative of UKESM. Instead, the decomposition of the CH₄ forcing specifically in UKESM1.0 is used to estimate the scaling. This UKESM-specific scaling is between 1.43 and 1.96 based on the ratio of the total PI-to-PD ERF_CH4, which includes all indirect effects, to the PI-to-PD ERF from changes in CH₄ concentrations alone. The range represents the uncertainty in the simulation results. The indirect radiative effects of CH₄ are assumed to scale linearly with the magnitude of the change in CH₄ concentration for this calculation.

The second method to calculate the CH₄ forcing is based on the additional experiment piClim-CH4_S introduced in the previous section, for which the CH₄ concentration calculated in method one was prescribed. The difference in TOA fluxes between piClim-CH4_S and piClim-control_S represents the ERF_CH4 when the standard chemistry mechanism is used. We also use the comparison between these two simulations to calculate the CH₄ feedback on its own lifetime in our modelling set up; we find that the difference between UKESM1.1 and the model set up used here with the standard chemistry mechanism is not substantial (1.31 vs. 1.29). We were not, however, able to calculate a separate estimate for the feedback when the complex chemistry is used with the available simulations.

Finally, the CO₂ forcing (RF_CO2) from the land use change is calculated following : 10RFCO2=a1(C-C0)2+b1|C-C0|+c1N‾+5.36×ln⁡(C/C0) where a1, b1 and c1 are constants (a1=-2.4×10-7 W m⁻² ppm⁻¹; b1=7.2×10-4 W m⁻² ppm⁻¹, c1=-2.1×10-4 W m⁻² ppb⁻¹), C is the concentration of CO₂ (ppm) that would arise following the land cover change if the modelled concentration was free to evolve, C0 is the prescribed PI concentration, while N‾ refers to the PI concentration of N₂O (ppb). The change in CO₂ concentration is calculated based on the historical land use carbon emission data provided for the Coupled Climate–Carbon Cycle Model Intercomparison Project (C4MIP) . An airborne fraction of 0.5 for the carbon emissions is used to calculate the resulting change in atmospheric CO₂.

The PI to PD land use change is dominated by the loss of forest cover, both broadleaf and needleleaf, in favour of grasslands (Fig. ). Regions of particularly substantial change include central and eastern North America and the southern hemisphere (SH) (sub)tropics, where some areas lose over 50 % of their forest cover. The northern hemisphere (NH) mid-to-high-latitude changes involve both broadleaf and needleleaf forests, while predominantly broadleaf forests are diminished at low latitudes (Fig. e). Smaller scale changes occur in southern Africa and central Eurasia. Conversely, in the European Alps, grasses are replaced by shrubs and needleleaf forests. The changes in land cover are reflected in the leaf area index (LAI), which decreases where forests are converted to grasslands (Fig. S1 in the Supplement). Overall, the PI to PD land use change results in a loss of 14 % of the global forest cover (Table ), which drives a decrease in the global mean LAI by 0.16 (-8 %).

Consistent with the decrease in high-BVOC-emitting forests in favour of grasslands, the global emissions of isoprene and monoterpenes decrease by around 115 Tg (C₅H₈) yr⁻¹ (-18 %) and 20 Tg (C₁₀H₁₆) yr⁻¹ (-16 %), respectively, as a result of the land cover change (Table , Fig. ), which is comparable to the decrease in BVOC emissions estimated in previous studies of PI to PD land cover impacts . The spatial pattern of decreased emissions closely follows that of forest cover loss (or, conversely, grass cover gain). However, the magnitude of the decrease in BVOC emissions is greatest in the tropics. This is due to the higher emission factors of broadleaf forest PFTs, which form a greater proportion of the deforestation at low latitudes, as opposed to needleleaf forest PFTs (compare Figs. e and e). The decreased emissions lead to decreases in the global burden of isoprene and monoterpenes by 20 % to 25 % (Table ).

The global burdens of both isoprene and monoterpenes are much lower for the simulations with more complex chemistry (Table ), despite very similar BVOC emission magnitudes. This is driven by the higher oxidising capacity of the atmosphere when the complex mechanism is used , driving a 58 % (21 %) decrease in isoprene (monoterpene) lifetime compared to the standard mechanism in the PI simulations. However, the relative change in BVOC burden and lifetime due to land use is largely consistent. For both chemistry mechanisms, the land use-driven decreases in burden are driven by the lower BVOC emissions combined with the 5 % to 8 % shorter lifetimes in the land use perturbation experiments. We attribute these shorter lifetimes to the greater availability of oxidants near source regions when BVOC emissions are decreased following land use changes; the surface level OH concentrations increase by 3 % to 4 % (Table ).

Figure  shows the net simulated effective radiative forcing from PI to PD land use change (ERF_LU) for the standard and complex chemistry experiment pairs. The global mean ERF_LU, which excludes the CO₂ and CH₄ contributions, is -0.080±0.047 W m⁻² with the standard chemistry scheme, while the magnitude of the forcing increases for the complex chemistry scheme to -0.116±0.041 W m⁻². These values include the impacts of changes in aerosols, ozone, cloud properties and surface albedo. The regional nature of land use change results in significant and substantial regional differences in TOA fluxes, despite the relatively small global mean ERF_LU. The dominant latitudinal feature is, on average, the -0.5 or -1.0 W m⁻² forcing at around 50° N for the standard and complex chemistry mechanisms, respectively (Fig. b, d), which is consistent with the substantial loss of forest cover at these latitudes. For both mechanisms, the negative forcing is particularly substantial over eastern Europe and central North America, where it regionally reaches -8 W m⁻². Significant impacts are also found on the other continents, for example the average ERF_LU for South America is -0.57±0.10 W m⁻² for the standard chemistry mechanism and -0.52±11 W m⁻² for the complex chemistry (Table S2). The greater magnitude of the NH forcing when employing more complex chemistry is driven by changes in northeast Asia (Fig. c); a negative forcing in this region is not present with the standard chemistry scheme.

The other substantial difference between the simulations with different chemical mechanisms occurs in the southern high latitudes, where the latitudinal mean ERF_LU rises to +0.4 W m⁻² when using the standard scheme but is strongly negative (below -0.5 W m⁻²) for the complex chemistry scheme. This latitudinal signal is driven predominantly by regional differences in East Antarctica, which we connect to changes in O₃ (discussed in more detail in Sect. ). The results tend to be more consistent at the lower latitudes, especially in the SH, where the ERF_LU values lie between -0.25 and 0.25 W m⁻², despite the substantial deforestation in the (sub)tropical SH.

Figure a–c shows the mean instantaneous radiative forcing associated with the aerosol direct effect (IRF_ARI) from land use change. The global average IRF_ARI is +0.057±0.015 and +0.058±0.011 W m⁻² for the standard and complex chemistry mechanisms, respectively. In both cases, the shortwave forcing dominates (+0.058 and +0.056 W m⁻² for the standard and complex mechanisms, respectively; Table ). IRF_ARI values tend to increase closer to the equator, although there is a secondary peak in the forcing in the NH mid-to-high-latitudes (Fig. c). On a continental scale, the greatest magnitude average forcing occurs in sub-Sahel Africa (over +0.20 W m⁻²; Table S2).

The positive IRF_ARI values are driven by a decrease in aerosol burden, which results in less scattering and reflection of SW radiation. Global organic matter (OM) burdens decrease by 3 % to 4 % (Table ). Figure d–f shows how the decreases in total column OM occur in the same regions as the positive IRF_ARI. The decreases are also greatest near the equator, where the BVOC emissions decrease by the greatest amounts (Sect. ), and where the highest magnitude of IRF_ARI is found. The magnitude of the decrease in OM mass is slightly greater for the standard chemistry mechanism than for the more complex mechanism, particularly in the SH. As the atmospheric oxidation capacity is lower when the standard mechanism is used, trace gases have longer lifetimes and can be transported further from source, resulting in the changes secondary aerosol occurring over a greater area (compare Fig. d and e).

While the global mean IRF_ARI values are positive, we note there are some regions of significant negative forcing, particularly in Central America. These regional impacts can be connected to an increase in sulphate aerosol in these regions (Fig. g–i). The decrease in BVOC emissions and burden under land use change increases the availability of the hydroxyl radical (OH), one of the main oxidants of sulphur dioxide (SO₂). As a result, aerosol formation and growth reactions occur closer to source regions, leading to greater sulphate aerosol burdens in these areas. Additionally, a slight shift occurs in the sulphate formation and growth pathways; there is an < 2 percentage point increase in the relative contribution from reactions with OH (Table S1). The OH nucleation and condensation fluxes increase, regionally by over 50 %, where BVOC emission fluxes have decreased and the availability of OH in the lower troposphere has increased (compare Figs. 2, S2 and S3). These changes potentially lead to a shift to more numerous but smaller sulphate particles . The impacts on sulphate aerosol are greater for the standard mechanism, where the oxidation capacity is more sensitive to changes in BVOC emissions .

The magnitude of the change in the global mean cloud radiative effect (ΔCRE) associated with pre-industrial (PI) to present-day (PD) land use change is very similar for both the standard chemistry mechanism (+0.087±0.033 W m⁻²) and the complex mechanism (+0.082±0.025 W m⁻²). However, the effect is almost limited to changes in the SW for the standard mechanism (ΔCRESW=+0.086±0.030 W m⁻², Fig. a), whereas LW fluxes are more affected for the complex mechanism (ΔCRESW=+0.047±0.027 W m⁻², ΔCRELW=+0.035±0.019 W m⁻²; Table ), although the decomposed ΔCRE still overlap due to the substantial standard error.

The ΔCRE tends to be positive over regions of substantial deforestation, such as central North America (regional average +0.50±0.17 and +0.91±0.18 W m⁻² for the standard and complex mechanisms, respectively) and eastern South America (+0.72±0.13 and +0.60±0.12 W m⁻²). The positive ΔCRE_SW is likely due to the changes in aerosols, introduced in the previous section (Sect. ), driving a decrease in the cloud droplet number concentration (CDNC) as the aerosol burden is reduced. The CDNC at 1 km altitude tends to decrease over regions of deforestation (Fig. d–f), in turn leading to an increase in the effective radius of the cloud droplets (Fig. g–i). These changes decrease the amount of incoming radiation reflected by the cloud, resulting in a positive forcing (i.e. the so-called Twomey effect is weaker; ).

The connection between lower organic aerosol loading and fewer cloud droplets is particularly apparent in central North America and eastern South America and is analysed in more detail here (see regions highlighted on Fig. a). The decrease in OM burden globally in response to land use change is greater in magnitude than the change in sulfate burden (Fig.  and Table ). Table  shows that OM burdens decrease by 7.7 % to 14.1 % in the identified regions. This is associated with a decrease in cloud condensation nuclei (CCN; aerosol particles > 70 nm in diameter) by 3 % to 4 %, which in turn drives a decrease in CDNC (from -3.0 % in South America to -9.6 % in North America; Figs. S4–S6). This decrease in CDNC is the likely mechanism behind the positive regional ΔCRE_SW, which is greater than 1 W m⁻², except for the South America region when using the complex chemistry mechanism. The lower magnitude of the radiative impacts in South America suggests other mechanisms may also be relevant in that region.

The method for calculating the ΔCRE used in this study will also capture differences due to changes in cloud amount and/or the location and lifetime of clouds, all of which may also be influenced by changes in water vapour or atmospheric dynamics. While the impacts of these changes are not investigated separately in detail here, and they may contribute to noted difference in ΔCRE_SW, the spatial pattern of ΔCRE_LW closely resembles the changes in average cloud amount at 10 km altitude (Fig. ). This high-altitude cloud amount increases over a 3 % larger area when the complex chemistry mechanism is used. The greater change in ΔCRE_LW with the complex mechanism could be driven by these differences in high-level clouds. An increase in cirrus-type clouds reduces the amount of outgoing LW radiation and leads to a positive forcing . However, the uncertainty in the results, particularly for ΔCRE_LW, is substantial and more in-depth analysis of the upper-tropospheric cloud formation is required for a robust understanding of the forcing mechanism.

The mean O₃ instantaneous radiative forcing (IRF_O3) associated with the pre-industrial (PI) to present-day (PD) land use change is shown in Fig. a and b. While the global mean values are weakly negative (-0.003±0.012 and -0.015±0.009 W m⁻² for the standard and complex chemistry mechanisms, respectively), the standard errors are substantial and there are regions of positive forcing, particularly in regions associated with the NH polar jet stream. The major difference between the two chemistry mechanisms is the sign of IRF_O3 over East Antarctica, which seems to be driving the difference in ERF_LU between the two mechanisms in this region. For the standard chemistry mechanism, the IRF_O3 is positive (< +0.1 W m⁻²) here, whereas the complex chemistry results in a negative forcing of a similar magnitude.

The changes in total column O₃ are mostly insignificant, though some significant differences closely resemble the patterns in IRF_O3 at latitudes over 30° (Fig. c, d). Looking at the O₃ vertical distribution allows for more insight into changes in O₃ and subsequent IRF_O3 impacts. While the O₃ volume mixing ratio tends to decrease in the troposphere, some increases in O₃ mixing ratios occur at the tropopause and in parts of the stratosphere (Fig. e, f). The tropospheric O₃ decrease may be driven by the loss of BVOCs, which are O₃ precursors. Monoterpenes only act as an O₃ precursor in the complex chemistry mechanism, but not in standard scheme , which would drive a greater magnitude impact on tropospheric O₃ for the complex scheme; the O₃ production fluxes decrease by 1.9 % and 2.9 % in response to land use change for the standard and complex scheme, respectively. However, the small magnitude of the changes in O₃ volume mixing ratios and their high sensitivity to the local chemical environment preclude the identification of a definite cause for the tropospheric variations. We suggest the higher altitude changes in O₃ mixing ratios are driven by dynamical processes. To assess the dynamical influence, the changes in the age of air are shown in Fig. g, h. These highlight that the same locations around the tropopause that are associated with substantial differences in O₃ mixing ratios are affected by changes in transport timescales between the troposphere and stratosphere.

The experimental set-up includes prescribed methane (CH₄) concentrations, therefore limiting the impact land use change can have on the contribution to ERF_LU from CH₄ forcing. However, as changes in the abundance of BVOCs can affect CH₄ lifetime through changing the availability of OH, we can estimate the change in CH₄ concentration due to land use change . We find the CH₄ lifetime against loss by OH decreases by 4 % and 3 % for the standard and complex mechanisms (Table ) as a result of land use change, leading to lower equilibrium CH₄ concentrations by 37 and 23 ppbv, respectively. The impact is greater for the standard chemistry scheme, which is consistent with the greater sensitivity of atmospheric oxidants to BVOC emissions for this mechanism.

Following the calculated decrease in CH₄ concentrations from PI to PD land use change, the CH₄ effective radiative forcing (ERF_CH4) is negative. The comparison of the additional simulation piClim-CH4_S (see Sect.  and ), which is set up with a prescribed global mean CH₄ concentration of 740 ppbv consistent with the calculated CH₄ decrease of 37 ppbv, and piClim-control_S suggests the global mean ERF_CH4 is -0.015±0.043 W m⁻² (Table ). This forcing estimate overlaps with the ERF_CH4 values calculated using the equations and additional scalings (see Sect. ) when considering the substantial uncertainty. Based on these calculations, the impact is of a greater magnitude for the standard chemistry (-0.046±0.008 W m⁻²) than the complex chemistry (-0.027±0.005 W m⁻²; Table ), following the different magnitudes of the changes in concentrations.

While the focus of this research is on the non-CO₂ biogeochemical impacts of land use change, we recognise that CO₂ emissions, surface albedo and other biogeophysical effects, such evapotranspiration and surface roughness, will also have radiative effects. Here, we quantify the impacts of CO₂ and surface albedo for comparison with non-CO₂ atmospheric composition forcings.

The changes in surface albedo are factored into the net simulated ERF_LU, alongside other geophysical impacts of land cover change, such as differences in evapotranspiration driving changes in water vapour. The global mean surface albedo forcing (RF_Alb,surf; Fig. a) calculated following is similar when using the standard and complex chemistry mechanisms at -0.174±0.016 and -0.169±0.012 W m⁻², respectively (Table ). The slight difference in RF_Alb,surf, despite the consistent land cover change, is likely driven by the interactive nature of the chemistry and climate in the model; the calculation of RF_Alb,surf depends partially on cloud cover, which will vary between the experiments. Both of these estimates are within the Intergovernmental Panel on Climate Change (IPCC) land use albedo ERF estimate of -0.2±0.1 W m⁻² . Note that RF_Alb,surf is not calculated at the TOA and therefore may not be directly comparable to other forcings that are calculated at the TOA. The alternative TOA calculations of the albedo forcing (-0.247±0.017 and -0.243±0.019 W m⁻²) remain within the IPCC land use albedo ERF estimate. The global mean surface albedo forcing values are, however, greater for both of the alternative methods as they do not account for the masking effect of clouds, which likely leads to an overestimation of the forcing (Fig. ).

The CO₂ forcing, similarly to that of CH₄, is not included in ERF_LU, due to the prescribed CO₂ concentrations in the experimental set-up. Based on an airborne fraction of 0.5 and the C4MIP land use change carbon emissions (Sect. ), we calculate that atmospheric CO₂ would have increased by 392 Tg CO₂ following the emission of around 214 TgC from land use change between 1850 and 2014. This is equivalent to an increase in atmospheric CO₂ by around 50 ppm. Using the equation from , this change in CO₂ results in a radiative forcing (RF_CO2) of +0.867±0.087 W m⁻².

In this section, we consider the relative impacts from each of the contributions to the land use change forcing. The largest magnitude forcings associated with changes in atmospheric composition, not including CO₂, are driven by IRF_ARI and ΔCRE (Fig. ). These, particularly in combination, are much greater than IRF_O3 or ERF_CH4. However, the global mean ERF_LU, despite not including the negative ERF_CH4, is negative, which is connected to the biogeophysical effects. RF_Alb,surf is of a much greater magnitude than the non-CO₂ biogeochemical forcings. Therefore, RF_Alb,surf drives the negative sign of ERF_LU, while the changes in atmospheric trace gases and aerosols modify the magnitude of the net forcing.

The non-CO₂ forcings in combination, that is ERF_LU + ERF_CH4, offset only around 15 % of the positive CO₂ forcing from PI to PD land use change emissions. Assuming ERF_LU, ERF_CH4 and RF_CO2 combine linearly, the total PI to PD land use change forcing based on our results is around +0.74 W m⁻², or 34 % [25 % to 48 %] of the change in the net anthropogenic ERF from 1850–1900 to 2006–2019 .

Table  compares the land-use driven changes in TOA fluxes, excluding CH₄ and CO₂ impacts, for four model versions: the experimental set-up used here with both the standard and complex chemistry mechanisms, as well UKESM1.0 and UKESM1.1. This allows for an assessment of the impacts of process and parameter updates (Sect. ; ). The net ERF_LU is reduced for both chemistry mechanisms in the new model set-up. The new biosphere-atmosphere processes included in this research reduce ERF_LU from -0.21±0.04 W m⁻² for UKESM1.1 to -0.08±0.05 W m⁻². The impact of new process representation is smaller with the use of the complex chemistry mechanism, for which ERF_LU is -0.12±0.04 W m⁻². While the flux diagnostics for the method of decomposing the forcing were not available for UKESM1.1, a more detailed comparison can be made against UKESM1.0, for which ERF_LU is also higher, with a magnitude of -0.17±0.04 W m⁻².

We find the greatest change between model versions occurs for the clear-sky shortwave fluxes (SWCS′=IRFARI,SW+ΔFclear,clean,SW), which are expected to predominantly be driven by impacts from surface albedo changes and the aerosol-radiation interactions, alongside smaller effects from changes in greenhouse gases. Of all the separated fluxes, the SWCS′ term has the greatest magnitude, regardless of model version. This is also the term that has changed most significantly between UKESM1.0 and the new simulations. SWCS′ is of a much smaller magnitude in the updated model version (-0.19±0.03 and -0.20±0.02 W m⁻² for the standard and complex mechanisms) compared to UKESM1.0 (-0.30±0.02 W m⁻²). Both surface albedo, due to tuning the snow burial of vegetation, and sulphate aerosol representations have been adjusted between UKESM1.0 and UKESM1.1, which may drive some of the differences in SWCS′ . The new process representation included in the model simulations further changes the spatial distribution of BVOC emissions and the formation and growth of aerosols. The difference in the net ERF_LU between the new standard and complex experiment pairs is 0.04 W m⁻². This is less or equal to the uncertainty associated with each ERF_LU, which highlights that the combination of other process updates from UKESM1.1 has had a greater impact on the net ERF_LU than the choice of chemistry mechanism.