Evaluation of modelling approaches in predicting forest volume and stand age for small-scale plantation forests in New Zealand with RapidEye and LiDAR

Abstract

In New Zealand, 30% of plantation forests are small-scale (<1000 ha) and knowledge of these forests, especially those less than 100 ha, is limited. These forests are expected to comprise more than 40% of the total harvest volume by 2020, so it is critical to understand the small-scale forest resource in order to plan effectively for marketing, harvesting, logistics and transport capacity. A remote sensing solution to small-scale forest description is necessary because conducting a comprehensive ground-based survey of those patchy forests is impractical. However, the utility of remote sensing prediction techniques for application in small-scale forests is unknown. This research evaluated two parametric models (multiple linear regression and seemingly unrelated regression) and two non-parametric models (k-Nearest Neighbour and Random Forest) models to predict stand variables (mean top height, basal area, volume and stand age) using model inputs including RapidEye-derived metrics and LiDAR-derived metrics. LiDAR-derived metrics were better at predicting all forest stand variables relative to RapidEye metrics. Combining LiDAR metrics with RapidEye metrics did not improve variable prediction results (on average 0.2% reduction in RMSE). Non-parametric models and parametric models performed similarly. Of all approaches tested in this study, multiple linear regression (MLR) using LiDAR-derived metrics was deemed to be the best performing modelling approach for predicting stand variables for small-scale plantation forests in New Zealand. MLR predicted mean top height (MTH) with a root-mean-square-error (RMSE) of 1.81 m, basal area (BA) with an RMSE of 9.92 m2 ha−1, stand volume with an RMSE of 94.93 m3 ha−1 and age with an RMSE of 2.17 years.