Combining multitemporal remote sensing data and repeated ground observations from the Spanish National Forest Inventory to estimate site index and diameter distribution for uneven-aged management of <i>Juniperus thurifera L.</i> forests

Abstract Spanish juniper, Juniperus thurifera L, forests are typically managed as uneven-aged forests in the Iberian range, which requires information about site index (SI) classes and diameter at breast height (dbh) distributions. The objective of this study is to evaluate and compare the performance of remote sensing models using multitemporal airborne laser scanning (ALS) data, Landsat metrics of state and change along with topographic, geographic and climatic predictors for the prediction of SI classes and diameter distributions. The best model for SI classes was a logistic model that showed an overall accuracy of 83%, a balanced accuracy of 77% and a kappa index of 0.52. The best model for diameter distributions consisted of an ensemble of imputation models for separate diameter classes and had a root mean squared Mahalanobis distance of 0.49, and class-wise coefficients of determination (R2) ranging from 64.95% for the 5–15 cm diameter class, to 89.63% for the 25–35 cm diameter class. Models were used to predict diameter distributions and SI classes, and these predictions were compared to the reference diameter distributions provided by a local silvicultural model. Deviations with respect to a reference silvicultural model were summarized in polygons of the Spanish Forest Map for the study area. These deviations facilitate the identification of areas requiring silvicultural interventions and highlight the value of integrating remote sensing technologies with ground-based inventories with repeated measurements for the adaptive management and conservation of Juniperus thurifera forests.