Determining the success of forest restoration projects requires methodologies that can quantify forest conditions at landscape-scales while simultaneously providing fine-scale metrics (e.g., tree and group attributes). These needs are especially important in southwestern ponderosa pine (Pinus ponderosa) forests where anthropogenic impacts have increased tree densities and reduced diversity, heterogeneity, and ecosystem function. Individual tree detection (ITD) algorithms, as applied to lidar, have displayed the versatility needed to take on this challenge. Several algorithms exist, each with their own nuances. Forest structure and composition also influence the accuracy of ITD methods. This research aims to: 1) evaluate ITD methods and parameters for estimating basal area and tree density across a gradient of forest structure following restoration treatments and 2) quantify fine-scale metrics combining the most accurate ITD methodology with canopy height models (CHMs) to assess structural differences among treatments. A 20-year landscape-scale experiment near Flagstaff Arizona provided stands with varied treatment intensity (i.e., thinning and repeated burning) and pre-treatment conditions (i.e., blocks differed by abundance of old trees). We stem-mapped fifty 0.04-ha plots and compared them to lidar-derived metrics to assess performance of ITD algorithms and parameters. We then used the most effective ITD algorithm to quantify structural differences between treatments.
Results/Conclusions
When looking across treatments, all ITD methods varied in their ability to predict tree metrics. We found applications of the default Li et al. algorithm best explained variation in tree height (R2 = 0.80) and tree density (R2 = 0.57). Other methods explained less of the variance in relation to height (R2 ranging 0.66 to 0.27) which would become problematic when predicting DBH and calculating basal area from tree heights. Using the Li et al. algorithm, we found mean basal area per tree increased with treatment intensity but total basal area per treatment decreased. ITD also allowed us to measure tree density per patch which decreased with treatment intensity. From the CHM, we saw an increase in canopy cover from 60% (heavily thinned) to 91% (control) while the number of patches per treatment decreased from 240 to 22 patches, respectively. All treatments had a higher number of smaller patches compared to the control which suggests restoration was successful. Overall, increasing treatment intensity results in more open conditions (i.e., more patches of larger trees covering less of the landscape). This pattern provides insight in how shifts in degraded forests towards desired conditions may provide increased resilience, diversity, heterogeneity and ecosystem function.