a–d, Supply curves for the 12 core policy mechanisms for M3 Central, with supply curves for all four global outlooks, including sensitivity to agricultural production and adoption behaviour assumptions, presented in Supplementary Figs 1–20. e, Trade-offs between carbon sequestration and biodiversity co-benefits for each policy under M3 Central given a common budget equal to that of the reference policy (U–C–C, AUD$292 billion). The Pareto efficiency frontier (dark dotted line) identifies the maximum carbon and biodiversity co-benefits achievable for the budget and the trade-offs between them. White numbers within marker symbols indicate the total area of reforestation in Mha (CP + EP). aUniform payments with land-use competition and regulation targeting biodiversity were equivalent, with both coded U–CR–B. bDiscriminatory payments with land-use competition and regulation targeting biodiversity were equivalent, with both coded D–CR–B. Colours in the legend relate to all panels.
Policy mechanisms.
Supplementary Table 5 details the payment schemes, land-use policy, targeting strategies, and complementary incentives assessed in this study.
Study area and policy settings.
This analysis was undertaken for the entire intensive agricultural land of Australia—a non-contiguous area of privately owned and managed cleared land stretching across eastern and southern Australia from Queensland to southwest Western Australia (Supplementary Fig. 21). We do not consider potential land-use change in native vegetation, wetland/riparian, or urban areas.
Agricultural policy in Australia has long been focused on increasing agricultural productivity and competitiveness, and has included relatively low levels of public subsidy. Recent climate policy has included both a price on carbon and direct payments, and bipartisan support exists for land sector contribution to climate change mitigation31. Australian biodiversity policy has included many individual incentive schemes for enhancing biodiversity on private land32, 33, 34.
Global outlooks and sensitivity.
We addressed multiple uncertainties using both scenario analysis and sensitivity analysis. To capture the uncertainty in global environmental and economic conditions, we assessed four global outlooks for the period 2013 to 2050 (Supplementary Table 6). Developed through a series of stakeholder interviews and workshops25, outlooks are internally consistent, plausible futures defined by different settings for global action on greenhouse gas emissions abatement, the size of the world economy and human population, and the degree of radiative forcing. Outlooks are benchmarked to the Representative Concentration Pathways35 (RCPs). Integrated assessment was undertaken using the Global Integrated Assessment Model25, 36 to provide projections of key parameters affecting the economic competitiveness of land use including demand for crops and livestock and prices for carbon and oil (Supplementary Fig. 22).
We undertook a sensitivity analysis across variation in agricultural productivity and land-use change adoption behaviour27 (Table 1). Three simple annual increases in agricultural total factor productivity were considered encompassing the range of increases experienced between 1977/1978–2007/08 in Australia37, with the high rate representing a step change increase in productivity. Uncertainty in land-use change adoption behaviour by landholders was captured by three hurdle rates covering the range of variation reported in the land-use change literature38, 39, 40, 41, 42. Thus, most calculations were performed under each global outlook o, adoption hurdle rate h, and agricultural productivity rate u.
Model pre-calculations.
The core analysis involved substantial spatio-temporal, integrated environmental–economic modelling using the Land-Use Trade-Offs model7, 24. The analyses were undertaken using ~1.1-km-resolution raster data (812,383 grid cells) and an annual temporal resolution. Spatio-temporal modelling was conducted with Python43 and NumPy (ref. 44). Model parameters and mathematical notation are summarized in Supplementary Table 7. Bold notation indicates a spatial layer, represented as a vector of grid cells, each with their own individual value for the parameter. Mathematical operations involving spatial layers occurred elementwise (grid cell by grid cell) unless indicated by square brackets, which symbolize operations occurring over all grid cells.
Climate change.
Future climate change estimates were modelled at ~1.8° grid cell resolution using a pattern regression process45 based on the outputs of the MPI-ESM-LR general circulation model (GCM). These change estimates were used to modify high-resolution, ANUCLIM-interpolated spatial layers of annual mean climate (rainfall and temperature) to create future climate layers for each global outlook. Carbon sequestration and dryland agricultural yields were projected under each global outlook on the basis of regressed relationships with annual mean temperature and rainfall7. Estimates of changes in water scarcity were also made on the basis of the climate change modelling and used to modify water prices over time.
Carbon sequestration.
We considered carbon stored in accumulated plant biomass following reforestation as our measure of carbon sequestration in metric tonnes of carbon dioxide equivalent (denoted tCO2), and did not include changes in soil carbon or agricultural emissions. We used 3-PG2-modelled spatial layers of the 20-year carbon sequestration potential (tCO2 ha−1) for mixed environmental carbon plantings and hardwood carbon plantings46 to model carbon sequestration by EP and CP, respectively. This was converted to 100-year carbon accumulation layers and a growth curve was used to model annual carbon sequestration over time7. Carbon sequestration was adjusted for climate change under global outlooks7 and reduced by 20% to account for risk. Carbon sequestration
for both reforestation land uses f in F{CP, EP} under the four global outlooks o was calculated as the average annual climate- and risk-adjusted rate over the 100-year period multiplied by grid cell area a.
Biodiversity co-benefits.
Biodiversity co-benefits accruing from the establishment of EP were estimated from a continuous biodiversity layer7 produced using a generalized dissimilarity model47. The model related plant species compositional turnover between 325,459 site pairs, including over 12,000 species of vascular plants, to environmental layers including 11 downscaled, terrain-adjusted climate metrics and 12 soil and substrate metrics at 0.01° spatial resolution. Compositional turnover was then predicted for each grid cell in response to a change between present-day climate (climatic averages for 1985–2005) and six 2050 climate futures—combinations of two climate scenarios (RCP 4.5 and RCP 8.5) and three GCMs (Can ESM2, MPI ESM2, and MIROC5). The extent to which EP in each grid cell would increase the representation of vascular plant communities requiring similar environmental conditions within a 1,000 km radius was calculated. Higher biodiversity co-benefits were obtained in grid cells that best increase the representation of plant communities under future climate change and have greater landscape connectivity as determined by proximity to remnant habitat and species–area relationships7. The final, single biodiversity priority score layer Bf was a weighted average of the six climate scenario/GCM combinations calculated using the limited degree of confidence approach48 and, hence, is robust to uncertainty in future climate change and remains constant over time.
The generic term biodiversity co-benefits Bsf refers to the sum of the area a of each grid cell multiplied by the biodiversity layer Bf, divided by the sum of this calculation over all grid cells. Bsf was expressed as a percentage of the maximum possible biodiversity co-benefits achievable by reforesting all agricultural land with EP.
There are several sources of uncertainty in the biodiversity co-benefits of reforestation. Although both monoculture plantings and agricultural land may provide ecological benefits49, the benefits of CP relative to EP are uncertain and likely to be low. Hence, we assumed that Bsf = 0 for f = CP. However, in practice, uncertainty also exists in the biodiversity co-benefits from EP. Many habitat resources take decades to develop (for example, tree hollows, fallen boughs, and so on) and many species need to be actively introduced owing to a lack of local populations50. Tree plantings may even have adverse effects on biodiversity if they replace native grasslands and shrublands or diverse farmland assemblages, or where monocultures preclude future biodiversity co-benefits from active or natural reforestation.
Economic returns.
Economic returns to agriculture were calculated on an annual basis in 2010 Australian dollars as profit at full equity (the economic return to land, capital, and management, exclusive of financial debt) using a profit function51, 52, 53, 54. Economic returns were calculated for a set of 23 irrigated and dryland agricultural commodities mapped by ABARES (ref. 55) for the 2005/06 agricultural census year. Yields and commodity prices were sourced from agricultural census data56, fixed and variable cos