Comments on Soimakallio et al. (2022) “Closing an open balance: The impact of increased tree harvest on forest carbon”

Soimakallio et al. (2022) reviewed literature to draw leveraged conclusions on how forest carbon balances are affected when “tree harvest rates are increased compared to a reference.” They condensed a large set of forest development scenarios into a ratio of changes in C stocks and harvests (H), that is, ΔC/ΔH calculated between scenarios. When both the numerator and denominator were expressed in tonnes of C, the division produced a dimensionless carbon balance indicator, CBI(T), as a function of time T. The resulting values were instructed to be interpreted as follows: “A positive CBI(T) value means that the forest carbon balance is reduced (i.e., less carbon is stored in the forest) when the harvest rate is increased. A CBI(T) value of one implies that the forest carbon stock is reduced by exactly the amount of carbon that is harvested,” and reasoned how the CBI should differ from one for different time horizons. This hypothesized behavior of the CBI may have oversteered the analyses, leading to the following choices, but altogether neglecting the question of whether all the contexts studied could be characterized by a simple indicator alone:


Comments on Soimakallio et al. (2022) "
Closing an open balance: The impact of increased tree harvest on forest carbon" Soimakallio et al. (2022) reviewed literature to draw leveraged conclusions on how forest carbon balances are affected when "tree harvest rates are increased compared to a reference." They condensed a large set of forest development scenarios into a ratio of changes in C stocks and harvests (H), that is, ΔC/ΔH calculated between scenarios. When both the numerator and denominator were expressed in tonnes of C, the division produced a dimensionless carbon balance indicator, CBI(T), as a function of time T. The resulting values were instructed to be interpreted as follows: "A positive CBI(T) value means that the forest carbon balance is reduced (i.e., less carbon is stored in the forest) when the harvest rate is increased. A CBI(T) value of one implies that the forest carbon stock is reduced by exactly the amount of carbon that is harvested," and reasoned how the CBI should differ from one for different time horizons.
This hypothesized behavior of the CBI may have oversteered the analyses, leading to the following choices, but altogether neglecting the question of whether all the contexts studied could be characterized by a simple indicator alone: 1. Pooling and equal analysis of studies with very different forest and forestry conditions, including variation from forest holding to national and continental scales. 2. Turning all scenarios as increased harvests regardless of the original scenario description. 3. Filtering of the resulting CBIs for extreme values, considered as outliers explained by factors other than differences in harvesting rates.
Contrary to #1, the scales considered essentially affect the ability to reconcile carbon stocks and harvests (e.g., Pohjanmies 2018). Contrary to #2, the reviewed studies often compared business-as-usual (BAU)-tointensive and BAU-to-extensive types of scenarios, that is, increasing and decreasing harvests. The direction importantly reflects both the transition from BAU to an uncertain counterfactual and the different allocation of harvests when forest use is intensified or abandoned. The CBI equation does not constrain the calculation of ΔC/ΔH with their original signs, and it could be argued that the resulting CBIs and all influencing factors should be (meta-)analyzed in this way instead of choices #2 and 3.
It is reasonable to assume that CBIs will differ depending on the scale considered and its initial forest state, how harvests are allocated to that forest structure, and how close the scenarios are to, for example, maximum sustained yield. It is somewhat difficult to justify the production of generic as opposed to scenario-specific carbon balance information, as a scenario is usually associated with the underlying factors mentioned above. Sufficient metadata should allow analyses of variation in these factors (Gurevitch et al. 2018), but the current supplementary data are not sufficiently consistent to even recalculate ΔC/ΔH with their original signs, which was communicated in more detail personally with the first author.
Consider how informative a CBI value, or the discussed range of ~0.34 to 1.57 tC/tC (Fehrenbach et al. 2021), is as such, without information that the lower value was actually caused by a series of storms and drought that reduced both C and H in Germany. The likely existence of similar or opposite (+ΔC/+ΔH) cases argues against the use of filtered average CBIs. If none of the studies analyzed a "normal forest," which a reference for fully regulated forestry, the discussion of the structure deviating from this ideal is misleading. Nevertheless, this discussion may indicate that the CBI reasoning of the authors is limited to a uniform age class distribution. Given that a similar dimensionless indicator (opposite numerator and denominator) may have led to erroneous conclusions when establishing principles for EU-LULUCF forest reference levels, a similar test as in Vauhkonen et al. (2021) with other generic | 537 LETTER TO THE EDITOR descriptive forms of a random distribution than just uniform may be motivated to challenge thinking.
In summary, the conclusions of Soimakallio et al. (2022) are based on a carbon balance indicator that is not transparent and loses information when averaged over multiple factors. While potentially useful for comparing fossil fuel displacement by wood in alternative harvest scenarios of a given forest resource, the indicator should not be used to compare scenarios based on different forest and forestry conditions. The results should be subjected to a proper meta-analysis with respect to these factors, which is not possible with the data presented.