The national inventory of carbon stock in New Zealand’s pre-1990 planted forest using a LiDAR incomplete-transect approach

Abstract

New Zealand has implemented its Land Use and Carbon Analysis System (LUCAS) to meet its reporting commitments to the United Nations Framework Convention on Climate Change (UNFCCC) on carbon stock changes related to forests. A double sample design using airborne Light Detection and Ranging (LiDAR) and ground plots has been used for its recent national inventory of carbon stocks in planted forests that were established prior to 1990, using the experience gained with LiDAR from an earlier inventory of the post-1989 planted forests. For the pre-1990 inventory, ground-based, 0.06hectare circular permanent sample plots were installed on an 8×8km grid, with LiDAR flown in North–South transects across only those forests that contain a ground plot. LiDAR metrics were extracted for up to four 0.06ha circular cells spaced at 1km intervals either side of the ground plot and another cell over the ground location. Effectively, LiDAR data were collected on an incomplete 8×1km grid, where individual forests have an unequal probability of being included in the sample. This paper describes the inventory design, those features of the plot measurement protocols that differ from the earlier inventory and the methods to calculate means and variances.Carbon stocks in each of the above- and below-ground live biomass, dead wood and litter pools were calculated for each ground plot at the time of measurement and predicted into the past and future using the Forest Carbon Predictor system. Field measurements of woody debris and litter are laborious and error prone when there has been a silvicultural operation and decay of residues. Given standing tree measurements and a description of past silviculture, the carbon stocks in the woody debris and litter pools are most efficiently estimated by the system rather than measured directly. The LiDAR data were used with regression estimators that explained 80% of the variance in total carbon, where the height percentile P40ht was the best variable. The combination of 686 LiDAR-only plots with a further 191 LiDAR/ground measured plots reduced the confidence interval to 11.1% of the mean total carbon stock per hectare from 14.1% using ground plots only. It had the advantage of being easier and faster to implement than an inventory with the increased number of ground plots required for an equivalent level of precision. A standard double sampling procedure with LiDAR plots on the complete 8×1km grid would have required lengthy flying and been too costly.