the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 23 Jul 2026
Consistent and Harmonized Forest Statistics for Forest-use Intensity Analyses Across Europe (2000–2023)
Abstract. Forests play a crucial role in Europe as carbon sinks, biodiversity reservoirs, and sources of renewable raw materials. Reliable Europe-wide statistics on forest biomass stocks, forest area, and harvest at national and subnational scales are essential for monitoring biomass dynamics and supporting sustainable forest management. However, the availability of such data remains incomplete, and differences in definitions, methodologies, and reporting standards introduce substantial uncertainties that hinder robust interpretation. Here, we present EuFor, an open-access database of key forest indicators for 38 European countries, derived from National Forest Inventories (NFIs), census statistics, and datasets from international organizations. It is openly available at https://doi.org/10.5281/zenodo.20815146. EuFor consists of two subsets: (i) EuFor-reported, a compilation of primary statistical data for 1990–2023, and (ii) EuFor-harmonized, a stock-flow consistent annual time series of forest area, roundwood volumes of growing stock and harvest for 2000–2023 across 21 national and 193 subnational units. EuFor shows near-perfect agreement with an authoritative reference dataset at the European aggregate level. At the country level, however, discrepancies vary substantially. Agreement is highest for forest area (mean relative deviation: 3.2 %; mean Pearson correlation: r = 0.743), followed by growing stock (7.4 %; r = 0.872), and harvest (14.5 %; r = 0.654). We further derived four indicators of forest-use intensity. Data uncertainties propagated into these estimates, leaving the direction of change ambiguous in 20 % of assessments. Moreover, the indicators often revealed contrasting spatial patterns and trends, highlighting both substantial uncertainty in trend interpretation and the multidimensional nature of forest-use intensity. By providing open-access, harmonized forest statistics together with an explicit assessment of uncertainty, EuFor offers a valuable resource for future analyses of forest biomass dynamics, carbon balances, forest-use intensity, and sustainable forest management across Europe.
- Preprint
(2352 KB) - Metadata XML
-
Supplement
(1791 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on essd-2026-540', Anonymous Referee #1, 16 Aug 2026
-
RC2: 'Comment on essd-2026-540', Anonymous Referee #2, 01 Sep 2026
Comments to the Author
This manuscript presents the EuFor-reported and EuFor-harmonized datasets, which compile and harmonize forest area, growing stock, harvest, and increment statistics for 38 European countries over 2000–2023, with additional historical observations where available. The authors integrate heterogeneous national forest inventories, forest censuses, State of Europe’s Forests statistics, FAOSTAT, and spatial datasets to produce annual national and subnational time series. The dataset addresses an important need for spatially and temporally consistent forest statistics and has clear potential for forest-use intensity, carbon-budget, and land-use analyses. The manuscript is well structured, and the effort involved in collecting and documenting these fragmented sources is substantial. However, issues remain regarding quality control, the reliability of reconstructed data, validation independence, methodological transparency, and uncertainty characterization.
General comments:
- The quality-control procedures require strengthening. Much of the source information appears to have been manually transcribed or translated from heterogeneous reports, yet the manuscript mainly describes basic cleaning and visual inspection. A systematic QC workflow should include duplicate detection, range and sign checks, treatment of missing and zero values, verification of unit conversions, detection of temporal discontinuities, national–subnational consistency checks, and stock–harvest–increment plausibility checks. The manuscript should clearly describe these procedures and summarize their outcomes.
- A substantial proportion of the harmonized dataset is reconstructed rather than directly observed. Approximately one third of the area and harvest series is interpolated or extrapolated, while most growing-stock values are generated using the CRAFT model; additional records involve allocation between forest-use classes. The current calculation flags provide useful provenance but are insufficient to communicate reliability. The authors should provide additional indicators, such as reconstruction method, extrapolation length, temporal distance from the nearest observation, number of observations supporting model calibration, and relevant model-fit diagnostics.
- The manuscript currently lacks sufficiently independent validation of the harmonized and reconstructed data. The comparison with State of Europe’s Forests (SoEF) cannot be considered fully independent because SoEF data or related information are used at several stages of the harmonization, including area adjustment, growing-stock expansion, and harvest normalization. Moreover, the underlying national sources may overlap. The current agreement is therefore partly expected by construction and should be described as an external consistency assessment rather than independent validation. The authors should conduct out-of-sample validation by withholding a subset of reported observations and evaluating how accurately the interpolation, extrapolation, and modelling procedures reconstruct them. Appropriate metrics, including bias, MAE, and RMSE, should be reported. Where genuinely independent sources are available, additional comparisons would further strengthen the credibility of the dataset.
- The CRAFT model is central to the dataset because it generates most harmonized growing-stock values, but its implementation and evaluation require clearer documentation. The authors should report the parameter-estimation procedure, parameter constraints and initialization, treatment of regions with sparse observations, natural-loss assumptions, convergence criteria, and regional model-performance statistics. Temporal or spatial holdout validation should be provided.
- Several harmonization choices require clearer justification and sensitivity assessment. These include the calculation of forest-area harmonization factors when multiple overlapping years exist, treatment of regions without matching observations, assumption that the factors remain constant over time, selection of one source series primarily on the basis of completeness, and unweighted averaging of EuFor-estimate and SoEF-normalized harvest series. Some reported area factors are relatively extreme, suggesting possible differences in definitions or spatial boundaries. The authors should examine these cases and evaluate the sensitivity of the final dataset to alternative harmonization choices.
- The uncertainty analysis currently represents cross-source variation rather than the full uncertainty of the harmonized estimates. It does not appear to propagate uncertainty associated with source observations, conversion factors, spatial mapping, interpolation, extrapolation, CRAFT parameters, or manual classification. The term “uncertainty envelope” should therefore be qualified or replaced with “cross-source sensitivity range.”
Specific comments:
- Page 18, line 477: The periods 2000–2005 and 2018–2023 each contain six years when both endpoints are included. They should not be described as “five-year averages.”
- Page 24, line 614: A decline from 311 to 183 corresponds to a reduction of approximately 41%, not 59%. Please correct the percentage.
- Page 24, line 623: Total European harvest is reported as approximately 2.3–2.8 million m³ yr⁻¹. These values appear to represent area-normalized harvest intensity, with units of m³ ha⁻¹ yr⁻¹, rather than total harvest, which is several hundred million m³ yr⁻¹. Please correct the description, values, and units.
Citation: https://doi.org/10.5194/essd-2026-540-RC2 -
RC3: 'Comment on essd-2026-540', Anonymous Referee #3, 06 Sep 2026
Dear authors,
This manuscript presents the EuFor dataset, which compile forest area, growing stock, harvest, and increment statistics from national forest inventories, forest censuses, SoEF, FAOSTAT, and other sources. The dataset addresses an important need for spatially and temporally consistent forest statistics. The separation between reported and harmonized subsets is useful, and the effort required to collect information from fragmented national sources is substantial. However, some aspects should be improved. Specific comments are as follows.
- Although the target definitions and processing steps are documented, the manuscript does not yet demonstrate whether estimates obtained through different harmonization pathways have a comparable basis. I suggest the authors examine whether the resulting values differ systematically according to the principal data sources or processing methods, and clarify the extent to which EuFor-harmonized improves comparability across countries and regions, and identify any remaining limitations associated with differences in data sources and processing methods. Any remaining limitations for comparing absolute values across countries and regions should be stated explicitly in section 3.3.
- The SoEF dataset contributes to the construction of parts of EuFor and is subsequently used in the data-agreement assessment. The degree of independence therefore varies among variables and countries. Please identify where EuFor and SoEF share input data or harmonization information, reconsider the description of SoEF as an “independent international forest statistics dataset”. The results would be more accurately presented as data agreement rather than independent validation. Furthermore, the assessment could be supplemented with suitable observations or datasets not used in constructing EuFor, including withheld source observations. If no sufficiently independent reference is available, this limitation should be acknowledged more explicitly.
- The manuscript states that the SoEF comparison covers only the years for which matched observations are available rather than the complete 2000-2023 EuFor-harmonized record. However, the corresponding agreement statements in the abstract and conclusions do not carry these temporal qualifications. I suggest the authors report the temporal coverage and number of matched observations more clearly and state that the most recent years have not been assessed through the SoEF comparison. If independent observations are available for later years, please consider supplementing the comparison.
- The uncertainty-envelope analysis assess how differences between datasets propagate into forest-use intensity estimates, but the interpretation remains limited. Please clarify which input variables and data combinations are mainly responsible for contradictory trends, and discuss why these cases occur. Relating the results to the data agreement and processing methods would make the analysis more useful for identifying countries and indicators for which the estimated changes are robust or should be interpreted cautiously. A brief case-focused discussion would be sufficient.
- Figures 4 and 6 use the same visual treatment for estimates with substantially different levels of support from source observations. Although this information is available through the calculation flags, a concise spatial or regional summary would improve interpretation of the maps. For example, the Supplement could report the latest available observation and the prevalence of interpolation, extrapolation, CRAFT modelling, or spatial allocation for each variable and region. This would help users identify areas where the mapped patterns depend more strongly on reconstructed values.
Overall, the EuFor dataset has the potential to become a valuable resource for analyses of European forest dynamics and forest-use intensity. Addressing the points above would improve the traceability of the processing workflow, clarify the reliability of the dataset, and provide users with more appropriate guidance for using the dataset.
Citation: https://doi.org/10.5194/essd-2026-540-RC3
Data sets
EuFor database: Consistent and Harmonized Forest Statistics for Forest-Use Intensity Analyses Across Europe (2000–2023) Florian Weidinger et al. https://doi.org/10.5281/zenodo.20815146
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 249 | 81 | 101 | 431 | 177 | 113 | 84 |
- HTML: 249
- PDF: 81
- XML: 101
- Total: 431
- Supplement: 177
- BibTeX: 113
- EndNote: 84
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Dear authors,
I have reviewed the study “Consistent and Harmonized Forest Statistics for Forest-use Intensity Analyses Across Europe (2000–2023)”. This manuscript introduces the EuFor database, an open-access compilation of European forest statistics. Using this database, the authors compute four indicators of forest-use intensity (harvest per area, per growing stock, per net annual increment, and per potential NPP) and systematically assess how data uncertainties affect trend interpretation. My comments are as follows.
1 Section 2.2.3 uses deviation factors based on the JRC dataset to adjust EuFor-reported forest areas to the “total forest” definition. However, the JRC data themselves are harmonised products with inherent uncertainties. The manuscript does not discuss how uncertainties in the JRC reference data propagate into EuFor-harmonised area estimates, I recommend the authors add more contents on it.
2 In Section 2.3.2, NPPpot is kept constant at 2018 values for 2019-2020 and at 2020 levels for 2021-2023. Given that NPPpot is climate-sensitive, this constant extrapolation may introduce systematic bias, please add more contents on it.
3 Section 2.2.2 describes a selection process where one “data stream” per country/indicator is chosen among multiple candidates (different NUTS+ levels, forest classes, sources). Although two prioritisation criteria are given, the actual decisions for each country are not documented. I request that a supplementary table be added, listing for each country which candidate streams were considered and the reasons for selecting or rejecting them.
4 The “flag_calc” column is mentioned as a tool for assessing data provenance (lines 820828), but no aggregated reliability metric is provided.
5 Although the paper primarily emphasises the construction and utility of this database, I would like the authors to explicitly summarise its innovativeness at the methodological level.