Valuing uncertainty part I: the impact of uncertainty in GHG accounting

Background: It has become increasingly evident in the literature that a correlation needs to be made between uncertainty in GHG emissions estimates and the value of emissions. That is, emissions with larger uncertainty are less desirable than those with smaller uncertainty. In fact, concrete advances in trade and reduction agreements depend on finding a set of methodologies for dealing with uncertainty that is acceptable to all parties. Results: Here, we assume that a cost, or value, can be assigned to changes in GHG emissions. As this cost can be assigned to emissions (or sequestrations), then so must a cost be assigned to the associated uncertainty. Standard methods from the actuarial sciences provide an approach to this valuation and we apply these same ideas to dealing to GHG accounting. Conclusion: This framework will allow us to address issues related to agreement structures and motivations for reducing uncertainty, and will enable objective comparisons between options. Marland, E., J. Cantrell, K. Kiser, G. Marland, and K. Shirley, 2014. Valuing uncertainty part I: the impact of uncertainty in GHG accounting, Carbon Management 5(1) 35-42. ISSN: 1758-3004 Uncertainty With an evolving political environment of negotiated commitments, legal restrictions or other measures to limit emissions of GHGs, and of markets to trade in emissions permits, there is a growing need to accurately evaluate carbon stocks and f lows [1]. Our concerns here are: ƒ Finding a general accounting framework that makes sense and is generally acceptable; ƒ Properly accounting for how much carbon is released to the atmosphere; ƒ Correctly valuing carbon released to the atmosphere; ƒ Understanding the implications of carbon accounting frameworks and potential management strategies (for biofuels, forestry, land-use change, industrial processes, and so on). In all cases it is important to deal with how much carbon is released to (or sequestered from) the atmosphere as CO 2 , when it is released, and the uncertainties in evaluating both the quantity and time of release. When different carbon flows have substantial differences in the uncertainty of their measurement or estimation, it is important to deal proactively with this uncertainty. The time of carbon release is of particular concern currently in considerations of land-use change and biofuels where emissions and sequestration occur within the same spatial system but not necessarily contemporaneously; but time is relevant for lifecycle analyses generally where emissions and sequestration cover the interval from production to end-of-life management. When uncertain quantities are distributed over time, the uncertainty gains another key dimension. It makes sense that a party should not receive credit for sequestration unless it is reasonably certain that it has happened and it makes sense that emissions should not be charged until it is reasonably certain that they have happened. Our focus in this article is on dealing with the uncertainty in CO 2 emissions estimates, which is followed by a second article in this issue of Carbon Management that shows the effect on accounting for the time value of emissions [2]. ƒ Anthropogenic sources of CO2 emissions Human activities that can lead to significant quantities of CO 2 emissions include: coal, petroleum, natural gas, biofuel and bioproducts consumption; natural-gas flaring; industrial processes such as cement Valuing uncertainty part I: the impact of uncertainty in GHG accounting Eric Marland*1, Jenna Cantrell1, Kimberly Kiser1, Gregg Marland2 & Kevin Shirley1 Background: It has become increasingly evident in the literature that a correlation needs to be made between uncertainty in GHG emissions estimates and the value of emissions. That is, emissions with larger uncertainty are less desirable than those with smaller uncertainty. In fact, concrete advances in trade and reduction agreements depend on finding a set of methodologies for dealing with uncertainty that is acceptable to all parties. Results: Here, we assume that a cost, or value, can be assigned to changes in GHG emissions. As this cost can be assigned to emissions (or sequestrations), then so must a cost be assigned to the associated uncertainty. Standard methods from the actuarial sciences provide an approach to this valuation and we apply these same ideas to dealing to GHG accounting. Conclusion: This framework will allow us to address issues related to agreement structures and motivations for reducing uncertainty, and will enable objective comparisons between options. manufacture; land-cover and landuse change; and human and domestic-animal respiration. Our ability to measure these anthropogenic emissions, their time profiles, and their long-term (and short-term) implications varies widely, and yet we need to deal with them within a consistent framework. Globally, we now have formal agreements in place, pending, or under discussion that dictate specific reductions in some types of emissions based on current or prior levels. For the time being, no restrictions on CO 2 emissions related to human respiration or to agricultural food products and their trade (e.g., domestic-animal respiration) have been imposed and we do not consider these further in this article. For all of the other categories emissions are measured or estimated in some way, and emissions limits and/or trading of emissions permits among sources or categories is often permitted without general consideration of the uncertainty in measurement or the time of emissions. Some categories of emissions or emissions reductions are excluded from some systems when it is judged that uncertainties are so large as to preclude fungibility [101]. Emissions commitments, carbon taxes, cap-and-trade systems, emissions offsets, carbon footprints, mitigation and adaptation strategies are all now generally discussed, and sometimes implemented, without considering independent monitoring and verification or the uncertainty or timing of emissions estimates [3]. The issues of uncertainty and time are especially acute currently for consideration of carbon flows related to biomass energy and land-use change, because these are areas where measurement of carbon flows is particularly difficult and subject to uncertainty, and because carbon flows occur in both directions (both to and from the atmosphere), and yet these two directions of flow are often not contemporaneous, so that net flows within a typical accounting period (e.g., a year) may not reflect the net flows integrated over longer times such as a forest rotation period ([4–7]). This article is focused on the treatment of uncertainty for carbon flows, particularly when there is a need to compare or trade carbon flows with significantly different levels of uncertainty. ƒ Uncertainty in emissions estimates The IPCC volumes on methods for national GHG emissions inventories provide extensive and valuable discussions for estimating uncertainty [8], but the challenge is how to deal with this uncertainty in emissions commitments or markets. Several edited volumes have collected a variety of papers that begin to confront this challenge [9,10]. In basic scientific inquiry there is generally a desire to reduce uncertainty, but in climate change and emission inventory estimates there are some very important additional issues that confound that objective. We understand that reducing the level of uncertainty will help create better estimates of emissions, and help guide better science and more accurate decision-making. In the name of science then, we can agree that more accurate data is desirable. With current policies it is not clear, however, that it is to everyone’s benefit to reduce uncertainty. It also is not yet clear what will be done with this knowledge of uncertainty. We argue here that as it is accepted that carbon emissions have some value, or cost, the question of uncertainty takes on another role. Recognizing the value of carbon emissions, perhaps in a quest to mitigate or adapt to the increasing carbon in the atmosphere, brings greater importance to the level of uncertainty in emissions estimates. Uncertainty, and differences in the levels of uncertainty, raise important questions about the cost of emissions, whether or not emissions commitments have been met, and how emissions permits can be purchased or traded. ƒ Uncertain ambiguity We now have formal agreements in place that require specific reductions in emissions based on current or prior levels. In order to fulfill those obligations, careful measurements or estimates must be made at some reference point and at ongoing periodic points in time to demonstrate that those reductions have truly taken place. We should have some understanding of the confidence conveyed in our estimates. Although the choice of confidence interval can be chosen depending on one’s risk tolerance, in this article we have chosen to represent that the uncertainty conveyed captures the correct value with 95% confidence. To introduce some of the primary issues surrounding uncertainty in agreements to limit emissions, we present three simplified illustrations. Does carbon sequestration offset carbon emissions? In the first illustration, we suppose that one party will release 100 tons CO 2 -e of emissions and wishes to compensate by trading with a party that will sequester 100 tons CO 2 -e (Figure 1). Suppose further, however, that the emissions have a level of uncertainty estimated to be ±5% and the sequestration has an uncertainty of ±10%. Are the emissions and sequestration equivalent? What is the difference in value between the emissions Key terms Accounting: In this context, accounting refers to utilizing an inventory of quantities and attribute those quantities to various parties according to some set of rules. For carbon accounting, this may refer to allocating the costs of adaptation or mitigation to various parties according to their activities in carbon emissions or sequestration. Uncertainty: Value that defines the accuracy level of a reported value. This can be due to measurement error, lack of available data, modeling assumptions or future estimation. Anthropogenic: ‘Of human source’ or ‘caused by human activity’. Anthropogenic emissions are humancaused emissions from power plants, automobile emissions, and industrial processes. and the sequestration? What is the value of reducing uncertainty relative to the value of the emissions and sequestration? Have carbon emissions commitments been met? As a second illustration, suppose that a country had 100 tons CO 2 -e of emissions in 1990 and has agreed to reduce those emissions by 10% by the year 2020 (Figure 2). This means that the emissions in 2020 need to be 90 tons CO 2 -e or less. This sounds relatively simple, until we include uncertainty. An uncertainty level of 20% at 95% confidence is not out of the realm of possibility (large, but useful for this demonstration). Instead of emissions of 100 tons CO 2 -e of emissions in 1990, the country now has emissions of 100 tons, ±20 tons (with 95% confidence). That gives a range of possible reference emissions that extends between 80 and 120 tons at the 95% confidence level (Figure 2). How much does the country have to reduce their emissions by 2020 and how is it calculated? If we are concerned only at the central tendency, the requirement is still 90 tons, but if we are interested at the 95% confidence level, then a 10% reduction from the 120 (100 ± 20) is 108. So the requirement could be met if the upper bound of the 95% confidence level is reduced from 120 to 108 tons, and this can be demonstrated if emissions can be calculated to be 100 tons CO 2 -e of emissions with an uncertainty level of ±8% (100 ± 8). That would mean that the 10% reduction at the 95% confidence level would be met (we are 95% confident that the correct value lies between 92 and 108) by reducing the uncertainty between 1990 and 2020, even though no real change might have occurred. Note that Figure 2 assumes that the 1990 and 2020 values are estimated independently. In fact, it may sometimes be possible to estimate the change in emissions over time (trend uncertainty) independently of the two end values and with less uncertainty than the difference between the two end values. Consider alternatively for Figure 2, that the country reports 90 tons CO 2 -e emissions in 2020 but the uncertainty has increased from ±20% in 1990 to ±40% in 2020. Then, the reported central number has indeed dropped by 10%, but the top of the 95% confidence interval has actually increased by 6 tons (90 ± 36 tons). This means that the true value might have actually increased. In fact, even with the same uncertainty in 2020 (90 tons ±20%), we cannot be sure that the emissions have dropped since the 95% confidence intervals overlap. How much carbon tax is owed? Finally, on a different scale, we might look at a third illustration that is relevant to a single power plant that is required to pay a fee in proportion to its emission levels (Figure 3). The plant could report that it released 30 tons of CO 2 -e with 5% uncertainty, or it could report 28 tons with 12.5% uncertainty. The second quantity covers the same upper range of emissions values, but with larger uncertainty. What is the result? The plant has good reason to maintain the higher level of uncertainty because it pays for only 28 tons released rather than 30. There are additional possibilities for problematic results if the uncertainty is asymmetric [11]. In all of these cases, uncertainty creates a problem in the calculations. In the first case, we see the need to quantify uncertainty for the purposes of comparison. In the second case, we see that methods are needed to make the use of uncertainty in agreements (at every scale) uniform. In the last case, we see the possibility for a motivation to keep uncertainty levels high. In this article, we reinforce the notion that we need to develop a consistent methodology for dealing with uncertainty and show an approach that effectively deals with all of the issues outlined above. We do this by borrowing ideas from the insurance industry, revealing how their treatment of uncertainty translates into a reliable but flexible methodology that deals with uncertainty but also motivates reducing uncertainty. We recognize that some consequences of the approach may still be intuitively negative, but we believe that the methods are clear and consistent and provide a basis for useful policy. 60 80 100 120 140 A