Use of multi-temporal LiDAR to quantify stand volume and biomass response to fertilization at larger spatial scales in coastal Douglas-fir forests.

Abstract:

Forest fertilization is becoming increasingly common in the coastal forests of British Columbia as a means to increase wood production and, consequentially, enhance carbon sequestration as a climate change mitigation strategy. Generally, the effects of fertilization are determined through the measurement of sample plots pre- and post-treatment. Due to the labour intensity of installing and measuring these sample plots, the fertilization effects are determined only for a limited portion of the treatment area. In recent years application of remote sensing based enhanced forest inventories have allowed for forest attribute estimation to expand beyond the sample plot to the wider forested area, which offers the potential to examine treatment effects at this scale. The objective of this research is to apply enhanced forest inventory methods using multi-temporal LiDAR data to determine the fertilization effect across a large forested area. In January 2007, a 308 ha area encompassing the research site was fertilized at a rate of 200 kg urea-N ha^-1. LiDAR acquisitions were made in 2004, three years prior to fertilization, as well as in 2008, 2011 and 2016, covering a 524 ha area centred around the treatment zone. A total of 53 paired LiDAR blocks, comprised of four 20m resolution raster cells, were selected on either side of the fertilization boundary for analysis of the effects across several different stand types. Stand types differed in the percentage of Douglas-fir, site index, as well as age. Using non-parametric random forests and Boruta variable selection, a model was developed to estimate total stem volume and total stemwood biomass for each year of LiDAR acquisition using an area-based approach. The models, developed at the plot level, were then applied to the wider rasterized grid and extracted to the paired LiDAR blocks. Preliminary results suggest significant treatment effects for certain stand types, but significant interactions among LiDAR blocks have confounded these findings. Results from this research would validate the use of enhanced forest inventory methods for rapidly expanding the assessment of treatment effects beyond sample plots to the stand, block, or landscape level.