Development of a site form equation for predicting and mapping site quality. A case study of unmanaged beech forests in the Cantabrian range (NW Spain)

Abstract

Site form (SF), expressed as the dominant height of a stand at a reference dominant diameter, is used less often than site index (SI) to estimate forest site quality. However, it has the advantage that it is age-independent and can therefore be applied in a wider set of situations in forestry practice. Like SI, elaboration of SF has traditionally required stem analysis or repeated measurements in permanent plots; however, Airborne Laser Scanning (ALS) can nowadays be used to generate site quality practical maps, thereby simplifying the method. The aims of this study were to fit a dynamic equation to stem analysis data in order estimate SF in natural beech forests in NW Spain, as well as to compare the performance of SF for site quality estimation and to analyze the possibility of using ALS data for site quality prediction in these forests. The Algebraic Difference Approach formulation of the Bertalanffy-Richards model provided the best results and defined four curves for dominant heights of 5, 10, 15 and 20 m at a reference dominant diameter of 20 cm. A significant relationship between SF and SI was observed, and we therefore believe that SF is a good approach for site quality estimation. These results can be used directly at inventory plot level for establishing site quality classes without the need to know the stand age, either in terms of SF or translated into SI. On the other hand, ALS data allowed estimation of both dominant height and dominant diameter, although better results were obtained with the former. Applying the SF dynamic model to both data sets enabled prediction of SF at 25x25 m/pixel (and SI using the SI-SF relationship). The overall accuracy of the relationship between the observed SI and that predicted from ALS metrics yielded a coefficient of determination of 0.456 without bias, heteroscedasticity or absence of normality. The results of this unbiased raster model were considered rather good, as predictions were obtained for a pixel size of 25x25 m (0.0625 ha). Prediction of mean SI value for one hectare would thus be necessary to average the values obtained in 16 pixels with an expected error compensation. Use of the raster model based on ALS metrics will enable site quality estimation for current beech stands at high spatial resolution without the need for fieldwork, providing very valuable information for forest managers and researchers.