%0 Research article
%T Strip-road tree identification and stand structure estimation using harvester data
%A Tarvainen, Riku
%A Riekki, Kirsi
%A Ovaskainen, Heikki
%A Kärhä, Kalle
%A Malinen, Jukka
%D 2026
%J Silva Fennica
%V 60
%N 3
%R doi:10.14214/sf.26021
%U https://silvafennica.fi/article/26021
%X Harvester data are a potentially valuable source of detailed information on forest inventory in first-thinning stands. In first-thinning operations performed using modern cut-to-length harvesters, all the trees within the strip roads are removed. The resulting systematic strip road network covers approximately 15–20% of the stand area. We investigated how strip-road trees can be identified from harvested production data to quantify within-stand variation and evaluate how strip-road sample placement affects the representativeness of the sample. Three different strip-road tree identification methods were evaluated: (1) a boom-angle and -length method; (2) a buffer method using reconstructed strip-road centrelines in a global coordinate system; and (3) a boom-angle method. The accuracy of the resulting diameter distributions, stem counts and basal-area estimates was assessed using Reynold’s error index and root mean square error metrics against detailed field reference data. In addition, theoretical offset strip roads were established and systematically shifted across each measurement area to quantify within-stand variation and sampling sensitivity. The most accurate stem count estimates were obtained using the boom-angle and -length method and the buffer method, whereas the boom-angle sector method performed poorly. Offset strip-road tree samples revealed substantial variation in stem count and basal area in some stands, indicating that strip-road alignment and spatial heterogeneity can significantly influence the representativeness of the sample. Our findings show that, in general, harvester-based strip-road sampling is a feasible and cost-efficient approach for estimating stand-level forest variables in first thinnings if tree identification methods and spatial biases are carefully considered.