Detecting growth releases of mature retention trees in response to small-scale gap disturbances of known dates in natural-disturbance-based silvicultural systems in Maine.

• Rate of disturbance detection within two-years of a known date was 64.1% or better. • Tsuga canadensis and Thuja occidentalis had detection probabilities of ∼80% • High pre-disturbance growth rate reduced detection rates consistently. • Time-lapses before the peak growth rate ranged from 5.4 to 7.2 years on average. Disturbances play a fundamental role in shaping forest communities. It is therefore important to accurately quantify their frequency and magnitude in forest ecosystems. Tree-ring series are commonly used for relatively accurate detection of past disturbances. Currently, several dendroecological techniques exist that enable the detection of disturbances in tree-ring series. In this study, increment cores (n = 312) were collected from trees that were deliberately retained in harvested gaps of two natural-disturbance-based silvicultural systems in the Acadian Forest of Maine, USA. These systems were designed to approximate the gap-based natural disturbance regime of the region. Since the specific harvest dates and severities were known, we compared the efficacy of three predominant disturbance detection techniques in accurately detecting disturbance within two years of the harvest date across five common, yet ecologically distinct species -- Acer rubrum , Picea rubens , Pinus strobus , Thuja occidentalis , and Tsuga canadensis -- and the length of elapsed time prior to when maximum growth responses were experienced. We tested the growth-averaging method, the boundary-line method, and the combination of the growth-averaging and boundary-line methods (Splechtna method) using the TRADER R package. We found little variation among detection techniques in their -- relatively high success -- rate of detecting disturbance within two years of the harvest date (growth-averaging method: 69.9%, Splechtna method: 68.6%, and boundary-line method: 64.1%). We most frequently found negative correlations with pre-treatment growth rates and detection rates among detection techniques. Our estimates of disturbance severity did not influence the probability of detecting the year-of-harvest. Of the species tested, Tsuga canadensis (73.7% − 85.5%) and Thuja occidentalis (78.4% − 83.7%) had the greatest detection rates across the techniques while Pinus strobus was consistently the lowest (51%). Time-lags – defined as the number of years post-harvest before a tree reaches its peak growth rate – were also analyzed. Pre-disturbance basal area increment and crown ratio were marginally negatively correlated with the length of time-lags in response to harvest and Thuja occidentalis (7.2 ± 0.4) and Tsuga canadensis (6.8 ± 0.3) had significantly longer time-lags than Acer rubrum (5.4 ± 0.3). Overall, all techniques demonstrated relatively high detection rates (>60%), given the range of sizes and ages of the species tested and reduced two-year time window for detection. Absent from all models were variables that measured disturbance severity, which suggests tree responses to disturbance were more sensitive to tree-level attributes and local changes to the growing environment. [ABSTRACT FROM AUTHOR]