Integration of tree-ring data, Landsat time series, and ALS-derived topographic variables to quantify growth declines in black spruce

Forest ecosystems and timber products are expected to play a determining role in climate change mitigation and adaptation. As the frequency and severity of natural disturbances increase, remote sensing technologies prove crucial for detecting and assessing the impact of natural disturbance on forest condition and productivity. Although satellite-based remote sensing is commonly used for measuring the extent and severity of stand-replacing disturbances such as fire and harvest, capturing the effects of subtler, non-stand replacing disturbances (NSR) remains challenging. Studies based on the analysis of tree rings have yielded insights on the impact of such disturbances at a local scale, yet tools for estimating the compound impact of NSR disturbances on broader scales are lacking. The objective of this study was to use Landsat time series spectral reflectance information in conjunction with tree-ring data to generate spatially explicit estimates of growth declines attributable to NSR disturbances at a 30-meter resolution across black spruce-dominated stands in two managed boreal forests. Basal area increment (BAI) calculated from 1545 increment cores collected in 52 plots across the two study sites were used to assess the growth decline in each plot due to NSR disturbance events from drought and insect defoliation. Plots that experienced a severe growth decline were identified using a threshold (i.e., a decrease in BAI ≥ 23.7% between two consecutive 11-year periods) established after examining the historical variability in growth rates from the tree-ring data. Subsequently, a logistic regression model was developed to predict the probability of a plot sustaining a severe growth decline, using predictors from Landsat time series and topographic variables derived from airborne laser scanning data, achieving an accuracy of 79.2% after five-fold cross-validation. In a subsequent modelling step, plots that experienced severe declines in BAI were used to construct a linear model, predicting the magnitude of the decline in BAI attributed to NSR disturbances using predictors from Landsat time series, and resulting in a model with an R2 = 0.70 when using five-fold cross-validation. Model predictions were then applied to black spruce-dominated stands at both study sites to estimate the impact of disturbances on forest productivity. Depending on the study site, between 22.6% and 57.6% of the analysis area had a predicted probability of severe growth decline > 50%. Within the most affected areas, the median decrease in annual BAI predicted using the OLS model was 95.3 and 64.4 mm² yr−1, respectively, for the two study sites. This research demonstrates the utility of combining tree ring and Landsat data to assess the impact of non-stand-replacing disturbances on forest growth at the scale of a forest management unit that in turn could inform salvage and/or silvicultural interventions that enhance the resistance and resilience of vulnerable stands.