Testing different Earlywood/Latewood delimitations for the establishment of Blue Intensity data: A case study based on Alpine Picea abies samples

Abstract

For dendroclimatological Blue Intensity (BI) studies based on earlywood (EW) or latewood (LW) information, a demarcation between the two is necessary, which can be difficult to establish for species where the transition is subtle. Often, a percental value k is used that calculates an EW/LW boundary value for each tree ring individually based on the difference between maximum and minimum absorption. Several laboratories and authors have used different values for k (e.g. k = 30 % or k = 50 %), while wood anatomical and visual studies suggest that k is on the order of 80 %. Here, we test how different settings of k, and thus different definitions of the EW and LW proportions of a tree ring, influence the dendroclimatic potential of derived time series. To this end, we correlate instrumental temperature measurements with tree ring chronologies that are based on EW and LW information (e.g. EW absorption (EWBI), LW absorption (LWBI)), where the EW/LW proportion is varied by setting different values for k. The tree ring samples utilized are 30 cores of spruce (Picea abies) trees from a high-elevated site (ca. 1700 m a.s.l.) in the northern Alps, Austria. Overall, we achieve high correlations between temperature data and our tree ring chronologies. Regarding the stability of the climate signal under different k values, the results show that absorption intensity based parameters (ΔBI, EWBI, LWBI) are only mildly influenced by different settings of k, while width based parameters (EW width, LW width) show a larger dependence on k. LW width, for instance, was stronger correlated with temperature, the smaller the LW was chosen (and thus the higher k was set). Based on our results and the wood anatomical definition of the EW/LW boundary, we suggest that k = 80 % may be a good choice for future studies. However, since this is only a case study from one site, careful screening of the respective data set regarding an appropriate k value must accompany each dendroclimatological study.

Introduction

Tree rings in conifers of the mid and high latitudes are usually characterized by a transition of large-lumen and thin-walled cells in the first section (the so-called earlywood) to narrow-lumen and thick-walled cells in the second and last part of a ring (latewood) (Schweingruber, 1988). Earlywood (EW) cells form in the beginning of the growth season and ensure efficient transport of water and nutrients, but are prone to cavitation (Pratt et al., 2007). Latewood (LW) cells appear distinctly darker, have higher densities and provide more mechanic stability, but are less efficient in water transport (Sperry et al., 2006; Björklund et al., 2017). As a response to the local and regional climate, a treeʾs allocation of resources to the formation of either EW or LW cells, and the properties of those so formed cells, vary between and within years resulting in different characteristics of each annual increment and the cells therein (Schweingruber, 1988; Björklund et al., 2017). Dendroclimatologists measure those variations, summarize them in different so-called parameters (e.g. ring width) and thus draw conclusions about the climate of the past. Owing to the different times of formation, EW and LW comprise different signals which can be used to reconstruct different climatic variables of different seasons (Schweingruber, 1988; Wood and Smith, 2015).

A wood property exhibiting climate induced inter- and intra-annual variability, is density. If there is one growth limiting factor (e.g. temperature at the altitudinal tree line), it determines the density characteristics of each annually formed tree ring to a high degree (Schweingruber, 1996). Therefore, wood density is a good climate proxy and has successfully been applied to a wide range of paleoclimatic studies (e.g. Cleaveland, 1986; Briffa et al., 1988; Chen et al., 2012). However, the establishment of density values is time consuming and the costs for acquiring required X-ray measurement devices are considerable (McCarroll et al., 2002). This has impeded establishment of high-quality tree ring based climate reconstruction where access to such a device was not available.

The development of the novel Blue Intensity (BI) method (McCarroll et al., 2002), which can serve as a surrogate for density measurements, thus means methodological progress for many tree ring researchers worldwide. BI works on the principle that the absorption of light in the blue spectrum of a tree ring has been shown to be highly correlated with X-ray wood density measurements, and thus is similarly an attractive proxy for temperature (McCarroll et al., 2002; Campbell et al., 2011; Rydval et al., 2014; Björklund et al., 2019). However, it has also been reported that the signals in BI and density are not entirely the same (Kaczka et al., 2018). To date, it is still not entirely clear, what exactly the BI values represent (Björklund et al., 2019). Nonetheless, BI has recently been applied successfully to create long chronologies (Fuentes et al., 2018), improve archaeological dating (Wilson et al., 2017), investigate drought periods (Babst et al., 2016) and address the growth divergence problem (Buras et al., 2018). A growing body of methodological literature deals with continuously refining and developing the method (e.g. Kaczka et al., 2018).

The intra-annual variations of density/BI in tree rings allow the establishment of different parameter time series, e.g. mean densities of EW or LW. Due to the gradual transition from EW to LW in most coniferous species (e.g. Picea abies), however, a defined delimitation is a prerequisite for the establishment of EW or LW based time series. Different approaches exist to delimit the two. First, the EW/LW boundary is set where the double cell wall thickness is greater than the radial lumen diameter (Mork, 1928). This method can only be used on wood anatomical micro sections and is therefore time-consuming. Second, one fixed density/BI threshold (usually between 0.40 and 0.55 g/cm³ for density (Koubaaa et al., 2002)) is defined which considers those parts of the tree ring with higher values than the threshold as LW and the ones with lower values as EW (Schweingruber, 1983; Cleaveland, 1986). This can be problematic if inter-annual density/BI fluctuations within a tree ring series are high (Björklund et al., 2017; Samusevich et al., 2020). And third, a percental value k is set which calculates an EW/LW boundary threshold T for each individual ring based on the difference between maximum density/BI (MXBI; acronyms in this paper always referring to the BI parameters) and minimum density/BI (MNBI) of each ring according to the formula (Schweingruber, 1983):

$$T=MNBI+\left(\frac{k}{100}*\left(MXBI-MNBI\right)\right)$$

This ‘floating threshold’ approach is a more dynamic method that is now commonly implemented in standard analysis programs such as CooRecorder (Cybis Electronics, Sweden), WinDendro™ (Regent Instruments, Canada) and LignoVision™ (Rinntech, Germany). Given the subtle transition of EW to LW, all approaches are arbitrary. Nonetheless, when undertaking a dendroclimatic study that involves a parameter based on EW or LW information, it is desirable to choose a k value that can extract as much information as possible. Parameters of interest with dependence on the EW/LW boundary, that have tradition in both density and BI studies, are: Earlywood width (EWW), Earlywood absorption/density (EWBI), Latewood width (LWW) and Latewood absorption/density (LWBI). Additionally, one can calculate ΔBI as introduced by Björklund et al. (2014) as the difference between MXBI and EWBI for each tree ring. This parameter is widely used alongside its refinement, adjusted ΔBI (Björklund et al., 2015). Originally, ΔBI was proposed for Scots pine (Pinus sylvestris) with the goal of mitigating the influence of a color transition occurring between heartwood and sapwood in that species and which, if not considered, would bias the obtained BI values. Even though the investigated species in this study, spruce (Picea abies), does not feature such a marked transition, the parameter was calculated as it, too, is potentially influenced by changes in the definition of the EW/LW boundary.

The previously mentioned parameters MXBI and MNBI, the average absorption/density of the whole tree ring (TRBI) and the total ring width (TRW) are also standard tree rings parameters computed automatically by most software programs. They do not change with different k values, but still were calculated in this study for comparison.

Different laboratories and authors have used different k values to obtain dendroclimatic parameter time series, e.g. k = 30 % (Österreicher et al., 2015) and k = 50 % (Björklund et al., 2017), but the reasoning behind this choice is not explained. At the same time, k = 80 % has been shown to align best with visual and wood anatomical definitions (Samusevich et al., 2020). We hypothesize that the k value has a large influence on the climatic signal of obtained time series, which thus impacts the potential of tree ring studies to derive climate information. Indeed, previous studies have shown that different methodological approaches in the preparation and analysis of samples can impact the climatic signal of derived parameter time series (Björklund et al., 2019). Therefore, the goal of this study is to test different k values on the EW/LW boundary depended tree ring parameters and correlate the obtained series with instrumental climate data. Thus, we aim to explore whether and how different definitions of EW and LW, based on the k value, influence the retained climatic signal of a range of dendroclimatic parameter time series.