Facing the modelling complexity
The FRL is an outcome of complex modelling exercise performed individually by each country. Based on our assessment, the main difficulty faced by countries was ensuring the consistency between the FRL and GHG inventories, or with other estimates (e.g. national GHG emission projectionsFootnote 6) (see Fig. 2 and Additional file 1: Table S5). Compared to the FMRL under the Kyoto Protocol, where the amount of harvest was implemented as an exogenous driver within the modelling framework (including economy or policy assumptions), the FRL concept is more difficult to implement because in principle it requires forest models to simulate management practices, age dynamics and the resulting harvest endogenously (cf. ; see also Fig. 1). The majority of countries adopted advanced forest ecosystem models, while some needed to develop ad hoc FRL models to comply with the requirements of the LULUCF Regulation (see Additional file 1: Table S10).
Additional challenges derived from country capacities, in terms of data, know-how, and resources availability. In some cases, the limited availability of more detailed data on forest management practices, ecosystem health and economics, might affect the advancements in forest modelling (e.g. [31,32,33]). Driven by international commitments and an increasing interest on climate change mitigation and adaptation policies, the overall capacity for forest modelling has constantly increased in recent years (e.g. ). For the purposes of the FRL, the majority of countries adopted empirical models mostly based on data and information from NFIs (see Additional file 1: Table S10). Despite NFIs are conceived as the most reliable information source for forest state and management [16, 35], usually over periods (inventory cycles, every 5–10 years), such information is not always comparable between subsequent periods (see e.g. ). Moreover, not all countries have full matching between inventory cycles and the period 2000–2009 (e.g. Poland; see Additional file 1: Table S10).
Based on our assessment, several countries faced difficulties in collecting reliable information to adequately represent the forest management practices in the period 2000–2009. To overcome this issue, countries adopted very different approaches for quantifying the impact of forest management practices, i.e. harvest intensity, and used aggregated data or ancillary information along with NFIs (see Additional file 1: Table S10).
Other major challenges are linked to the consideration of age-related forest characteristics, including the simulation of their dynamics. In our analysis, we refer to “age” as an explicit model parameter, but we recognise that other dynamic parameters (e.g. DBH, biomass density) might be also used to adequately simulate the development of forest stands. Indeed, some countries (e.g. Germany, Italy, Portugal) used age-related proxies, such as biomass densities, volume classes or area-based increment (see Additional file 1: Table S10). Individual choices of the best proxies for age dynamics were likely driven by biophysical circumstances, data availability and parameters in statistics at country scale. The use of age “as is” can be meaningless in complex structures, such as for example, uneven-aged, or multi-layered stands in Mediterranean forests (e.g. ).
These findings reveal that, while countries demonstrated huge efforts in data collection and elaboration, a further improvement of data on forest management (on practices, target species, rotation length or tree cutting characteristics, harvesting rates) and characteristics (age structure, area, increment, health status, soil conditions, regeneration, etc.) would enable more robust comparison between past and future management practices, and ultimately support the decision-making process. Further harmonization of NFIs, i.e. common definition of key parameters and data processing procedures, may be an effective solution to improve comparability of forest indicators and estimates among countries . Several attempts to harmonizing NFI data have been made so far, such as for example, those concerning the assumptions and definitions of stem volume , and of the area restrictions to forest management . An improvement of NFIs should also aim at a more holistic knowledge of forests, as forest data is used for other purposes than wood resources, including climate, energy and biodiversity in the context of current policy settings. The use of remote sensing techniques, if duly combined with ground plots, will increasingly complement country statistics in providing timely spatial and temporal patterns on forest management [40, 41]. Additional efforts can be oriented to improving the robustness of national forest statistics and implicitly their reporting within the EU frameworks (e.g. EUROSTAT) or at a broader scale (e.g. FAOSTAT, Forest Europe) (e.g. [34, 42, 43]). Joint efforts aimed at assessing, comparing and enhancing forestry models in Europe can be conveyed into a common platform for sharing experiences, ideas and main findings (e.g. community of practice on forest management decision support systems; see Footnote 6).
Ensuring consistency and comparability with historical estimates
The LULUCF Regulation requires ensuring consistency between the FRL-related simulations and GHG inventories. The reason is twofold: the accounting will be based on the GHG inventories and historical estimates presented by GHG inventories are subject to accurate and robust review process. Medium and low fulfilment in ensuring consistency with GHG inventories can be partly explained by only limited model adequacy (e.g. about pools and gases) (see Fig. 3 and Additional file 1: Table S8). The main challenges are linked to the difficulties in transitioning from simplified methods used in GHG inventories (i.e. few strata) to an increased modelling complexity for simulating the impact of past harvest and age structure development, as required by the LULUCF Regulation. This is particularly the case of modelling living biomass, for which countries further developed their modelling capacity through adopting specific modelling tools and collecting/refining detailed country-specific data (see Additional file 1: Table S10; see also the approaches used in the GHG inventoriesFootnote 7).
From our assessment, the majority of countries put efforts in modelling living biomass and HWP carbon pools, and only partly deadwood, and often omitted the CO2 and non-CO2 emissions linked to biomass burning (i.e. controlled burning and wildfires), thus triggering an obvious inconsistency with the GHG inventories (see Additional file 1: Table S11). For HWP, all countries used the “production approach” following the IPCC guidelines and as required by the LULUCF Regulation (see also ), so consistency with the GHG inventory was not a concern.
The omission, notably of non-CO2 emissions, from biomass burning (prescribed and wildfires) may be due to the fact that they were considered negligible in the reference period 2000–2009, by Northern countries , or are to be included later using the background level for applying the natural disturbance provision, particularly concerning the fire-prone countries (e.g. Greece). Depending on the model used, countries faced difficulties in incorporating the deadwood pool (mandatory for the LULUCF Regulation), likely because of the lack of reliable data (some GHG inventories lack estimations for this pool and instead assume the pool to be in balance) (cf. ). In addition, many countries did not incorporate other carbon pools such as litter and soil (see Additional file 1: Table S11). This performance outcome is closely linked to the adopted modelling framework (from simplified to full carbon models), and associated data requirements. On the one hand, empirical models running exclusively aboveground biomass growth (see Additional file 1: Table S10), which are robust in simulating stand productivity, are often not able to represent carbon and nutrient cycles in other C pools, below-ground processes, and the impact of environmental disturbances (see e.g. ). On the other hand, widely tested models, i.e. through years of application for forestry operations and for scientific purposes at national scale or in international contexts, were used by some countries, including EFISCEN Space by Netherlands; CBM by Czech Republic, Ireland and Poland for living biomass; and Yasso by Austria, Finland, Germany and Latvia for soils (see Additional file 1: Table S10). However, the use of an advanced modelling tool providing full carbon simulations (i.e. comprising living biomass, dead organic matter and soil) made it difficult to ensure a consistent representation of all carbon fluxes as reported in the GHG inventories (e.g. Poland and Czech Republic). This is also due to different data processing and aggregation to national scale, and models’ capacity to represent the disturbances and management practices, compared to the simplified assumptions as in GHG inventories. Ensuring consistency with other information sources (i.e. time series in GHG inventory) requires additional efforts for model calibration and validation in order to improve model robustness and reduce uncertainty, such as e.g., adequate representativeness of forest diversity, accuracy of allometric equations, spatial extrapolation of local data, and conversions from standing volume to entire carbon stocks . To our knowledge, six countries showed an inconsistency in the model output . Three of them (Greece, France and Finland) smoothed this discrepancy by adopting an ex-post calibration, while for the remaining (Cyprus, Bulgaria, Germany), the EC put forward a correction of the FRL value because of a detected inconsistency of model outputs with GHG inventory estimates [8, 20].
The improvement of comparability between FRL and GHG inventories would require a further development of forest ecosystem models to feed both GHG inventory data and projections towards robustly incorporating both reliable input data and representation of the effects of management and environmental disturbances on stand development and growth (e.g. forest landscape models; ). Based on our findings, the FRL exercise resulted in an increased availability of updated data and previously disclosed information on forest management within the NFAPs, particularly on harvest . These data may facilitate for example, the effectiveness of the EU-level forest initiatives (e.g. the Forest Information System for Europe—FISE,Footnote 8 the EU forest observatory, the ThinkForestFootnote 9 platform) in providing timely evidence-based support to current EU policies also beyond climate . Advances in modelling approaches and data quality may also improve the reporting of GHG emissions and removals for forest land under the UNFCCC, and foster the comparability of estimates within the LULUCF sector .
FRLs as a tool for understanding the mitigation potential of EU forests
The FRL represents the projected evolution of the forest sink (including HWP) for the period 2021–2025, with the assumption of continuing the 2000–2009 management practices and without external influences from policy and market development. This way, the FRL is a benchmark for measuring the climate impact of management changes in forestry—but it is important to note that the FRL is not a projection of probable or preferable development of the carbon sink for the period 2021–2025 (Fig. 1). The trend of the total EU forest carbon sink under the FRL (− 18% in 2021–2025 relative to 2000–2009) can be largely attributed to (i) the impact of increased harvest rates (+ 16%; see ) driven by the evolution of the age class distribution; and (ii) the effects of forest aging on reduced increment ([48, 49]).
The link between the age class distribution and the evolution of harvest within the period 2021–2025 is evident where even-aged forests are predominant, and harvest is mostly provided through clear-cuts. In most of these cases, the overall shape of the age class distribution confirms that the increasing amount of harvest reported within the period 2021–2025 is mostly due to the expected evolution of the age structure . In other cases, however, where an irregular or an uneven-aged structure is predominant, and harvest is mostly provided through thinnings or single-tree selection systems, age structure does not play a key role. This is, for example, the case of Spain and Greece, where most of the forest area is classified as uneven-aged. In other cases, the effect of exceptional natural disturbances affecting some countries within the period 2000–2009 (i.e. Germany or Austria) or during the most recent years (such as in case of Czech Republic) might have altered the age class distribution. In these cases, salvage logging activities—which do not have a direct relation with the age class distribution—may prevail on ordinary management practices carried out within the period 2000–2009. The current FRL design tried to balance the impact of all these factors—certainly having different roles due to country-specific circumstances—and, at the same time, factored out possible expectations due to policy and economic assumptions, allowed under the Kyoto Protocol .
Our analysis suggests that the projected carbon sink in living biomass decreases more than proportionally (− 26%) compared to the increasing amount of harvest (+14%) projected in the FRLs (Fig. 4). Since this sink is the difference between net increment and harvest, when most of the increment is harvested, then a relatively small increase in harvest causes a significant drop in the sink. For example, if the increment is 100 tC, the harvest 80 tC and the sink is 20 tC, a 10% increase in harvest (88 tC) causes a 40% drop in the sink (from 20 tC to 12 tC, assuming a constant increment). This projected trend in age-related increase in harvest calls for additional efforts in order to reverse the current declining sink and align the forest sector with the mitigation expected in 2030. On the one hand, an urgent increase in net increment would be required , e.g. through new forest area or improved forest management practices (thinning etc.). On the other hand, a climate-smarter use of any extra age-related harvest becomes even more important, i.e. using this extra wood in long-lasting products with high material substitution benefits may partially compensate the impact of the declining forest sink .