Landscapes For Sequestering Carbon: a dynamic marginal abatement cost curve approach with Bayesian spatio-temporal modelling

通过改变土地利用和管理将碳封存在陆地生态系统中，是减缓大气中二氧化碳上升的一种手段，而且可以说是扭转这一趋势的唯一经济上可行的手段。为此，大多数国家都在其气候变化承诺中纳入了通过土地利用、土地利用变化和林业（LULUCF）封存碳的目标。例如，苏格兰的造林目标是达到25%的森林覆盖率，并建议在2012-2022年期间创造10万公顷的林地。自2012年以来，英国已经在泥炭地恢复上花费了超过3000万英镑，其目标是隔离碳和保护生物多样性。然而，关于土地利用和土地利用变化的决定是在对土地的竞争需求（例如粮食生产，体育收入等）的背景下做出的，因此经济学进入决策。我们需要知道什么样的政策决定将以最低成本实现碳封存（从而减缓气候变化），以及边际成本如何随着政策选择的增加而变化。例如，在低等级的粗糙草地上造林可能具有成本效益，但在高等级的可耕地上造林则过于昂贵。边际消减成本（MAC）曲线是制定此类决策的既定经济工具。然而，在LULUCF部门，以前只以非常简单的方式应用这些方法，忽略了边际成本的变化、放弃不同土地用途的机会成本和巨大的不确定性。在这里，我们建议基于贝叶斯框架中的时空动态建模，为LULUCF部门开发一种更严格的MAC曲线方法。这建立在之前的工作基础上，该工作开发了一种贝叶斯数据同化方法，将不同的数据源结合起来，对英国过去的土地利用进行时空明确（100米&年）估计。使用马尔可夫链蒙特卡罗方法，我们将有效地探索数千种未来景观的实现，这些景观可能从目前的状态演变而来。该方法具有时空明确性，必须考虑放弃土地利用的机会成本，并包括土地价值的空间变化和边际成本的变化。作为贝叶斯方法，我们建立了MAC曲线的后验概率密度分布，从而量化了相关的不确定性。产出是对哪些土地利用转变将发生、土地利用变化可能发生的地方、将封存多少碳以及以何种代价进行数学上和概率上的严格分析。这将有助于决策者就未来景观如何有助于减缓气候变化做出知情的、基于证据的决策。

Sequestering carbon in terrestrial ecosystems by changing land use and management is one means of slowing the rise in atmospheric carbon dioxide, and arguably the only economically feasible means of reversing the trend. To this end, most nations have included targets within their climate change commitments for sequestering carbon through land use, land-use change and forestry (LULUCF). For example, Scotland has an afforestation target to reach 25 % forest cover, and the creation of 100,000 ha of woodlands in the period 2012-2022 has been recommended. More than £30M has already been spent on peatland restoration in the UK since 2012, with a stated aim of sequestering carbon as well as biodiversity conservation. However, decisions on land use and land-use change are made in the context of competing demands for land (e.g. food production, sporting income, etc.), so economics comes into the decision-making. We need to know what policy decisions will result in sequestration of carbon (thereby mitigating climate change) at least cost, and how the marginal costs change as uptake of policy options increases. For example, afforesting low-grade rough grazing land may be cost-effective, but be prohibitively expensive on high-grade arable land. Marginal abatement cost (MAC) curves are an established economic tool for use in making such decisions. However, in the LULUCF sector, these have been applied in only very simplistic ways previously, ignoring these changes in marginal costs, the opportunity costs of the different land uses foregone and the large uncertainties. Here, we propose to develop a much more rigorous MAC curve approach for the LULUCF sector, based on spatio-temporal dynamic modelling in a Bayesian framework. This builds on previous work, which developed a Bayesian data assimilation approach to combine disparate data sources to make spatio-temporally explicit (100-m & annual) estimates of past land use in the UK. Using a Markov chain Monte Carlo approach, we will effectively explore thousands of realisations of future landscapes which could plausibly evolve from the present-day state. Being spatio-temporally explicit, this approach necessarily accounts for the opportunity costs of the land uses foregone, and includes the spatial variation in land value and the changing marginal costs. As a Bayesian approach, we establish the posterior probability density distribution for the MAC curve, and thereby quantify the associated uncertainty. The output is a mathematically and probabilistically rigorous analysis of which land-use transitions will occur, where land-use change is likely to take place, how much carbon will be sequestered, and at what cost. This will help policy-makers to make informed, evidence-based decisions about how future landscapes can help to mitigate climate change.