Predicting yields of woody biomass in land restoration projects across Australia

Carbon Sequestration Science

Paul, Keryn ; Roxburgh, Stephen

CSIRO

DOI  :  10.25919/5d24e4430726a
RP-ID  :  EP193939

学科分类：地球科学(综合)

澳大利亚|英语

来源: CSIRO Research Publications Repository

【 摘 要 】

Across vast regions of Australia, there are many areas of land where woody vegetation has been cleared or largely supressed post-European settlement, and where land restoration options provide opportunities for sequestration of biomass carbon. These options include either planting native species or facilitating natural regeneration, both of which are simulated at national- and project-scales using the empirical tree yield formula (TYF) of the full carbon accounting model (FullCAM). The spatial input to the TYF that determines site productivity has recently been refined. Further, there has been a 1.6-fold increase in the number of stands measured that provide new calibration data. These developments provided an opportunity to expand the domain of application and capability of the model. We collated 2,340 observations of above-ground biomass (AGB) in planted or regenerating stands across Australia, and applied a novel technique to correct these observations for a baseline biomass (typically 5 Mg DM ha-1) of existing remnant trees or shrubs within these stands (typically 10% of individuals). Statistical analysis of the expanded and corrected dataset identified five categories of stands with statistically-different rates of AGB accumulation, thus informing different TYF calibrations. Yields of AGB were 1.4- to 2.5-times higher in belt plantings of low- and high-stand density ( or1,500 stems per hecatare), and in stands of mesic regions accessing ground or surface water, such as riparian or floodplain zones, than in blocks of plantings or natural regeneration. Yields of AGB from natural regeneration were also 3-times lower when regeneration occurred on land where management for grazing of livestock continued. By accounting for stand age, site productivity and the five stand categories, the overall TYF prediction efficiency was 68% (cf. 45% in the previous calibrations). Accounting for additional variables such as stand density, stand establishment method, species or species mix, and belt width was tested, but considered unwarranted, given the additional resources required to account for these inputs would be substantial, yet they would deliver 10% improvement in prediction efficiencies. As found in the previous calibrations, although predictions were not biased on average, they may nonetheless provide erroneous predictions at a given site. The mean absolute percentage error (MAPE) was 24% across all 2,340 observations of AGB. Given no bias, individual over- and under-predictions will cancel out for carbon accounting at the national-scale, while at the project-scale where carbon offsets are monetarised, discounting for uncertainty may be implemented. In revising the model calibrations, the trade-off between model accuracy and utility was considered.