Fig. 4

Class-wise error metrics (bias, root mean square error RMSE, mean absolute error MAE) for maximum likelihood (ML) and the studied PRM methods. \(PRM({\overline{D}D}_{\mathrm{g}})\): PRM system of Eqs. (8, 9) applied to the field values of \(\overline{D}\) and \({D}_{\mathrm{g}}\). \(PRM({\widehat{\overline{D}}\widehat{D}}_{\mathrm{g}})\): PRM system of Eqs. (8, 9) used with values of \(\widehat{\overline{D}}\) and \({\widehat{D}}_{\mathrm{g}}\) obtained from age and site index. \(PRM({{D}_{\mathrm{o}}D}_{\mathrm{g}})\) and \(PRM({{\widehat{D}}_{\mathrm{o}}\widehat{D}}_{\mathrm{g}})\): PRM system of Eqs. (8, 10) using the field values of \(N\), \({D}_{\mathrm{o}}\) and \({D}_{\mathrm{g}}\) and their predicted values \(\widehat{N}\), \({\widehat{D}}_{\mathrm{o}}\) and \({\widehat{D}}_{\mathrm{g}}\), resp. N: stand density.\({D}_{g}\): quadratic mean diameter. \({D}_{o}\): dominant diameter (mean diameter of the 100 thickest trees ha⁻¹)