The uncertainty in conifer plantation growth prediction from multi-temporal lidar datasets

Abstract

An evaluation of the use of airborne lidar for multi-temporal forest height growth assessment in a temperate mature red pine ( Pinus resinosa Ait.) plantation over a five-year period is presented. The objective was to evaluate the level of uncertainty in lidar-based growth estimates through time so that the optimal repeat interval necessary for statistically meaningful growth measurements could be evaluated. Four airborne lidar datasets displaying similar survey configuration parameters were collected between 2000 and 2005. Coincident with the 2002 and 2005 acquisitions, field mensuration for 126 trees within 19 plots was carried out. Field measurements of stem height were compared to both coincident plot-level laser pulse return (LPR) height percentile metrics and stand level raster canopy height models (CHM). The average plot-level field heights were found to be 23.8 m (standard deviation ( σ) = 0.4 m) for 2002 and 25.0 m ( σ = 0.6 m) for 2005, with an approximate annual growth rate of 0.4 m/yr ( σ = 0.5 m). The standard deviation uncertainty for field height growth estimates over the three year period was 41% at the plot-level ( n = 19) and 92% at the individual tree level ( n = 126). Of the lidar height percentile metrics tested, the 90th (L90), 95th (L95) and maximum (Lmax) LPR distribution heights demonstrated the highest overall correlations with field-measured tree height. While all lidar-based methods, including raster CHM comparison, tended to underestimate the field estimate of growth, Lmax provided the most robust overall direct estimate (0.32 m/yr, σ = 0.37 m). A single factor analysis of variance demonstrated that there was no statistically significant difference between all plot-level field and Lmax mean growth rate estimates ( P = 0.38) and, further, that there was no difference in Lmax growth rate estimates across the examined time intervals ( P = 0.59). A power function relationship between time interval and the standard deviation of error in growth estimate demonstrated that over a one-year period, the growth uncertainty was in the range of 0.3 m (∼ 100% of total growth) reducing to less than 0.1 m (∼ 6% of total growth) after 5 years. Assuming a 10% uncertainty is acceptable for operational or research-based conifer plantation growth estimates, this can be achieved at a three-year time interval.