Geostatistical prediction of height/diameter models

Height/diameter models are used in forest inventories and in stand projection systems mainly because height measurements are expensive and time-consuming. Mixed models have been frequently employed in H/D modelling assuming that model parameters can vary among stands. Using mixed models, the prediction of the random effects requires at least one additional measurement in the stand. In the present paper, we advocate the idea that empirical predictors for the random stand effects of a mixed model can be obtained using geostatistics when spatial correlation is present. Using the proposed methodology we obtained site-specific height/diameter curves without additional measurements. The case study was carried out in pure maritime pine stands of central Spain. Height/diameter measurements were available from the Second Spanish National Forest Inventory. Final predictions of random stand effects were obtained by kriging. Cokriging was also used for random stand effect prediction, using as a secondary attribute the stand density.