Examining the transferability of height–diameter model calibration strategies across studies

Abstract

Height–diameter (H–D) models are fundamental tools for predicting the relationship between tree H–D at breast height, for numerous applications in forestry. Increasingly, studies develop H–D models that can be calibrated to achieve a high level of precision with only a few observations. Different calibration methods and strategies are employed and compared in these studies, often disregarding the data used to develop the models and the H–D function used. In this study, we examined the transferability of optimal calibration strategies across studies, conducting a literature review and an empirical study. We compared the performance of six H–D functions and different calibration methods when using the same calibration strategies and dataset. Based on our literature review, we found that the most commonly employed calibration strategy is random-effects calibration and that the most common variable used to develop generalized H–D models is dominant height. We observed that different calibration methods can lead to varying results due to their different emphases on various aspects of the data and their individual limitations. Moreover, when the same dataset is used for calibration, different H–D functions may exhibit various performances. However, we found high percentages of agreement for the Curtis, Schumacher, and Wykoff H–D functions across all three calibration methods and low agreement between all functions and the Power H–D function. These observations underscore the need to consider all relevant factors, including the H–D function used, when selecting an H–D function and calibration strategy to ensure optimal transferability of the model. Our study provides insights that can improve the accuracy of H–D models, which are essential for predicting forest growth and structure in the context of changing environmental conditions.