Stochastic forest growth simulation: incorporating growth prediction uncertainty with wind and fire damage into carbon sequestration estimates and discounted cash flow analysis.
Degree GrantorUniversity of Canterbury
Degree NameMaster of Forestry Science
Uncertainty in forest productivity prediction and the variable reduction from wind and fire damage makes predictions of forest Net Present Values (NPVs) and carbon sequestration uncertain, creating a distribution of possible future outcomes for new forestry investments. To quantify the distribution of NPV and carbon sequestered at harvest a stochastic modelling system was used incorporating uncertainty around the expected site productivity of radiata pine (Pinus radiata D. Don) (Palmer, Hock et al. 2009) with the influence of catastrophic wind and fire damage (Moore, Manley et al. 2011, Anderson, Doherty et al. 2008). By combining uncertainty in forest productivity estimation with that of volume reduction from catastrophic damage this system captured a complete picture of potential forest growth outcomes for forest volume and carbon sequestration.
The stochastic forest modelling system used repeated runs of the 300 Index growth model (Kimberley, West et al. 2005) inside the Atlas Forecaster forest modelling software (Snook 2010). Site Index and 300 Index values were used to calibrate the model to site productivity. They were generated using the R statistics package (R Core Team 2012) with means, variances and correlation based on the sampled uncertainty of spatially modelled productivity estimates referenced against measured PSP data. The generated productivity indices were used to construct Forecaster project files which were used to initiate model runs. Simulated stem volume and stocking from forest growth simulation were used to calculate the likelihood of wind damage (Moore and Quine, 2000) for catastrophic damage simulation. The simulated average probability of damage was calibrated to match the surveyed regional average rates of wind and fire occurrence in Moore, Manley et al. (2011) and Anderson, S. A. J., et al. (2008) respectively. Growth simulations selected for damage were modified to represent salvage harvest operations and re-run with an earlier harvest matched to the year of the simulated event. Damaged simulations were modified with a higher proportion of logging waste and NPV calculations using increased logging costs. The NPVs and total carbon sequestered for the collated final simulations were used to display the likelihood of each simulated outcome.
Results were compared across differing levels of catastrophic damage on high and low productivity sites. The addition of wind and fire damage to NPV estimates lowered the mean NPV especially for high productivity sites. However, the stochastic NPV estimates contained large variation and no statistically significant results were achieved, even with high numbers of repeated simulations. A reduction in total carbon sequestered prior to harvest due to wind and fire of between 3 and 5% was found for the case study forest in the Central North Island of New Zealand. However, there was a large amount of variation in the simulations around these means and the observed reduction was not statistically significant. Confidence intervals for estimates of carbon sequestration without wind and fire damage were between 5% and 10% of the mean, and the addition of wind and fire damage increased variance and confidence intervals to 20% of the mean.
The results and conclusions show the significance of uncertainty for new forest investments on previously un-forested land both in terms of NPV and carbon sequestration. The trend across all scenarios with catastrophic damage is a tailed distribution with the main proportion of outcomes centred on the expected deterministically calculated value, for both carbon sequestered and NPVs. On average NPVs were reduced by $600/ha through the inclusion of average rates of catastrophic damage. However, there was high variation in the result and the predicted reduction is not statistically significant.