Quantifying natural disturbances using a large-scale dendrochronological reconstruction to guide forest management

Estimates of historical disturbance patterns are essential to guide forest management aimed at ensuring the sustainability of ecosystem functions and biodiversity. However, quantitative estimates of various disturbance characteristics required in management applications are rare in longer-term historical studies. Thus, our objectives were to: (1) quantify past disturbance severity, patch size, and stand proportion disturbed, and (2) test for temporal and sub-regional differences in these characteristics. We developed a comprehensive dendrochronological method to evaluate an approximately two-century-long disturbance record in the remaining Central and Eastern European primary mountain spruce forests, where wind and bark beetles are the predominant disturbance agents. We used an unprecedented large-scale nested design dataset of 541 plots located within 44 stands and 6 sub-regions. To quantify individual disturbance events, we used tree-ring proxies, which were aggregated at plot and stand levels by smoothing and detecting peaks in their distributions. The spatial aggregation of disturbance events was used to estimate patch sizes. Data exhibited continuous gradients from low- to high-severity and small- to large-size disturbance events. In addition to the importance of small disturbance events, moderate-scale (25-75% of the stand disturbed, >10 ha patch size) and moderate-severity (25-75% of canopy disturbed) events were also common. Moderate disturbances represented more than 50% of the total disturbed area and their rotation periods ranged from one to several hundred years, which is within the lifespan of local tree species. Disturbance severities differed among sub-regions, whereas the stand proportion disturbed varied significantly over time. This indicates partially independent variations among disturbance characteristics. Our quantitative estimates of disturbance severity, patch size, stand proportion disturbed, and associated rotation periods provide rigorous baseline data for future ecological research, decisions within biodiversity conservation, and silviculture intended to maintain native biodiversity and ecosystem functions. These results highlight a need for sufficiently large and adequately connected networks of strict reserves, more complex silvicultural treatments that emulate the natural disturbance spectrum in harvest rotation times, sizes, and intensities, and higher levels of tree and structural legacy retention.
Methods
We used a plot network established in the remaining primary mountain spruce forests of Central and Eastern Europe (www.remoteforests.org). Data span large geographical gradient covering the Carpathian Mountains of southern and northern Romania, Ukraine, southern and northern Slovakia, the Bohemian Forest in the Czech Republic, and the Harz Mountains in Germany. We used a hierarchical sample design of 541 plots nested within 44 stands (4-63 plots·stand-1; median: 11) nested within 6 sub-regions (4-13 stands·sub-region-1; median: 6). The “West” sub-region was comprised of the Harz and Bohemian Forest mountain ranges to avoid any sub-region with a single stand.

We used a comprehensive dendrochronological method to reconstruct past disturbances based on analyses of increment cores extracted on each study plot. To quantify individual disturbance events, their severities, and stand proportions disturbed, we used tree-ring proxies (two types of radial growth patterns were assumed to indicate past disturbance: release from suppression and rapid early growth rate), which were aggregated at plot and stand levels by smoothing and detecting peaks in their distributions. The spatial aggregation of disturbance events was used to estimate patch sizes. The data collection and analyses are described in detail in the related article published in Ecological Applications. The supplementary material of the article also contains the R code which were used to obtain and process the data [besides the patch size data, which were obtained by analysis in ArcGIS 10.5 (ESRI, Redlands, United States, http://www.esri.com)].