Cohort classification methods shape the interpretation of tree growth responses to harvesting: evidence from a natural mixed forest

Cohort classification is fundamental to interpreting tree growth responses to silvicultural interventions, yet whether the choice of classification method shapes these interpretations remains untested in structurally complex natural forests. We compared five cohort classification methods — the IUFRO standard, equal-interval height classification, crown competition height, k-means height clustering, and bivariate DBH–height distribution methods — in broadleaved–Korean pine forests of Northeast China, using monitoring data from 4286 trees across four harvesting intensity gradients over 12 years. The bivariate DBH–height distribution method produced cohorts with the most consistent stand structural relationships, as evidenced by superior G–S (stand basal area–tree density) model fit and the greatest proportion of explained variation in tree relative growth rate. Crucially, the identity of the cohort exhibiting the greatest post-harvest growth response reversed depending on the classification method applied. Under height-based methods, lower-canopy individuals (Cohort III) showed the strongest response. However, this pattern may reflect the mixing of light-demanding canopy saplings and shade-adapted understory shrubs within the same height-defined cohort, rather than a property of lower canopy position per se. Under the bivariate DBH–height distribution method, individuals assigned to the lower-trajectory group (Cohort II) showed the strongest growth responses while those in the upper-trajectory group (Cohort I) showed no significant response regardless of harvesting intensity. Species-level analyses further indicated that Pinus koraiensis and Fraxinus mandshurica differed in the magnitude of their responses within the same structural group, a pattern that was less evident under height-based classification. These results indicate that the choice of cohort classification method leads to different interpretations of tree growth responses to harvesting, and that the species-specific DBH–height grouping performed better than the height-based methods in this dataset by capturing species-specific structural variation that height-based methods conflate with canopy position.