Land-use change matrix with IPCC land-use change nomenclature (Bucki <em>et al</em> 2012) IFL: intact forest land, NIFL: non-intact forest land, OL: other land (= non-forest land)

Table 1. Land-use change matrix with IPCC land-use change nomenclature (Bucki et al 2012) IFL: intact forest land, NIFL: non-intact forest land, OL: other land (= non-forest land). The nine land-use change areas associated with the matrix were coupled to a bookkeeping carbon model to compute emissions due to land-use change in Panama over the 1992–2000 period. Results are for (a) areas (103 ha) and (b) associated emissions (106 MgCO2 ha−1) computed for the REL and (c) corresponding REDD+ activities. Values in between parenthesis are the sensitivity for each matrix element. The two values of the key parameters selected to compute the total uncertainty were Emi = Maxj{ΔEmij}/REL: time interval (8/10 yrs), IFL stocks (168.6/204.6 MgC ha−1), IFL increment (−0.16/0.75 MgC ha−1 yr−1), NIFL stocks (120.0/102.3 MgC ha−1), NIFL increment (4.43/2.37 MgC ha−1 yr−1); for areas: edge width (1000/250 m), resolution (250/50 m), fallows (= forests/other lands). Abstract The United Nations Framework Convention on Climate Change (UNFCCC) defined the technical and financial modalities of policy approaches and incentives to reduce emissions from deforestation and forest degradation in developing countries (REDD+). Substantial technical challenges hinder precise and accurate estimation of forest-related emissions and removals, as well as the setting and assessment of reference levels. These challenges could limit country participation in REDD+, especially if REDD+ emission reductions were to meet quality standards required to serve as compliance grade offsets for developed countries' emissions. Using Panama as a case study, we tested the matrix approach proposed by Bucki et al (2012 Environ. Res. Lett. 7 024005) to perform sensitivity and uncertainty analysis distinguishing between 'modelling sources' of uncertainty, which refers to model-specific parameters and assumptions, and 'recurring sources' of uncertainty, which refers to random and systematic errors in emission factors and activity data. The sensitivity analysis estimated differences in the resulting fluxes ranging from 4.2% to 262.2% of the reference emission level. The classification of fallows and the carbon stock increment or carbon accumulation of intact forest lands were the two key parameters showing the largest sensitivity. The highest error propagated using Monte Carlo simulations was caused by modelling sources of uncertainty, which calls for special attention to ensure consistency in REDD+ reporting which is essential for securing environmental integrity. Due to the role of these modelling sources of uncertainty, the adoption of strict rules for estimation and reporting would favour comparability of emission reductions between countries. We believe that a reduction of the bias in emission factors will arise, among other things, from a globally concerted effort to improve allometric equations for tropical forests. Public access to datasets and methodology used to evaluate reference level and emission reductions would strengthen the credibility of the system by promoting accountability and transparency. To secure conservativeness and deal with uncertainty, we consider the need for further research using real data available to developing countries to test the applicability of conservative discounts including the trend uncertainty and other possible options that would allow real incentives and stimulate improvements over time. Finally, we argue that REDD+ result-based actions assessed on the basis of a dashboard of performance indicators, not only in 'tonnes CO2 equ. per year' might provide a more holistic approach, at least until better accuracy and certainty of forest carbon stocks emission and removal estimates to support a REDD+ policy can be reached.表1. 采用政府间气候变化专门委员会（IPCC）土地利用变化命名法的土地利用变化矩阵（Bucki等，2012）。其中IFL：原生林（intact forest land），NIFL：非原生林（non-intact forest land），OL：其他土地（即非森林土地，other land）。该矩阵关联的9类土地利用变化区域被耦合至簿记碳模型，以计算1992-2000年间巴拿马因土地利用变化引发的碳排放。结果分为(a) 区域面积（单位：10³ 公顷）、(b) 对应碳排放量（单位：10⁶ MgCO₂·ha⁻¹），该结果基于参考排放水平（REL）计算得到，以及(c) 相应的减少毁林和森林退化所致排放以及森林保护、可持续管理和增强森林碳储量的行动（REDD+）活动。括号内的数值为每个矩阵元素的敏感性系数。用于计算总不确定性的两组关键参数取值为：排放因子Emi = Maxⱼ{ΔEmᵢⱼ}/REL；时间间隔（8/10年）；原生林碳储量（168.6/204.6 MgC·ha⁻¹）；原生林碳增量（-0.16/0.75 MgC·ha⁻¹·yr⁻¹）；非原生林碳储量（120.0/102.3 MgC·ha⁻¹）；非原生林碳增量（4.43/2.37 MgC·ha⁻¹·yr⁻¹）；区域面积相关参数：边缘宽度（1000/250 m）、空间分辨率（250/50 m）、休耕地（=森林/其他土地类型）。 摘要 联合国气候变化框架公约（UNFCCC）确立了针对发展中国家减少毁林和森林退化所致排放以及森林保护、可持续管理和增强森林碳储量的行动（REDD+）的政策方法与激励机制的技术和财务模式。当前仍存在诸多技术挑战，阻碍了与森林相关的碳排放与碳汇清除量的精准估算，以及参考水平的设定与评估。这些挑战可能会限制各国参与REDD+行动，尤其是当REDD+的减排量需要达到发达国家排放合规抵偿所需的质量标准时。本研究以巴拿马为案例，测试了Bucki等（2012，Environ. Res. Lett. 7 024005）提出的矩阵方法，开展敏感性与不确定性分析，区分了“建模不确定性来源”（指模型专属参数与假设）与“常规不确定性来源”（指排放因子与活动数据中的随机与系统误差）。敏感性分析结果显示，由此产生的通量差异范围为参考排放水平的4.2%至262.2%。休耕地分类以及原生林的碳储量增量/碳积累量是敏感性最高的两个关键参数。通过蒙特卡洛模拟传播得到的最大误差源于建模不确定性来源，这要求我们必须格外关注以确保REDD+报告的一致性，而这对于保障环境完整性至关重要。鉴于这类建模不确定性来源的影响，采用严格的估算与报告规则将有助于提升各国减排量的可比性。我们认为，除其他因素外，通过全球协同努力改进热带森林异速生长方程，可降低排放因子的偏差。向公众开放用于评估参考水平与减排量的数据集与方法，将通过提升问责制与透明度增强该体系的可信度。为保障结果的保守性并应对不确定性，我们认为发展中国家需要利用现有真实数据开展进一步研究，以测试保守折扣法的适用性，该方法包含趋势不确定性及其他可提供实际激励并随时间推动改进的可选方案。最后，我们主张，基于绩效指标仪表盘评估的REDD+基于结果的行动，不应仅局限于“每年二氧化碳当量吨数”，这或许能提供更全面的途径，至少在能够获得更精准、确定的森林碳储量排放与清除量估算以支撑REDD+政策之前是如此。