Skip to main content
Dryad logo

Data from: Light availability predicts mortality probability of conifer saplings in Swiss mountain forests better than radial growth and tree size

Citation

Bianchi, Eva; Bugmann, Harald; Bigler, Christof (2020), Data from: Light availability predicts mortality probability of conifer saplings in Swiss mountain forests better than radial growth and tree size, Dryad, Dataset, https://doi.org/10.5061/dryad.tqjq2bvx4

Abstract

Radial and height growth rates are suitable indicators of impending tree mortality risk of adult trees, but their applicability to saplings remains unknown. We compared radial growth of living and dead saplings of different heights and quantified the effects of light availability, growth and tree size on mortality. We sampled an equal number of living and dead saplings of four coniferous tree species (Pinus cembra, Larix decidua, Picea abies, Abies alba) in nine forests along an elevational gradient of the Swiss Alps. Based on tree-ring widths reconstructed from stem disks at multiple tree heights, we calculated radial growth rates. We observed a divergent pattern in radial growth of living and dead saplings, with reduced growth of dead saplings starting several years prior to death. By matching living and dead saplings of similar ages, we tested whether mortality probabilities of saplings were influenced by light availability, recent growth rates and diameter. Mortality of coniferous saplings in mountain forests was mainly influenced by light availability, with changing effects along the elevational gradient. Recent radial growth rate and tree size were only weakly associated with sapling mortality. Our study establishes the importance of long-term predisposing factors for the mortality probability of conifer saplings in mountain forests, thus extending well-established findings from the adult stage to saplings, which represent a critical stage of forest dynamics.

Methods

Site selection and species sampled

We studied four tree species with different light requirements and shade tolerances, respectively: European larch - Larix decidua (low shade tolerance), Swiss stone pine - Pinus cembra (intermediate shade tolerance), Norway spruce - Picea abies (intermediate to high shade tolerance) and silver fir - Abies alba (very high shade tolerance). Saplings were sampled in nine mountain forests along an elevational gradient in three climatic regions of Switzerland (Southern, Central and Northern Swiss Alps; cf. Appendix 1), each site covering an area of approximately 400 m x 500 m (20 ha). Tree species distribution maps based on data from the first Swiss National Forest Inventory  were used for site selection. The main criteria were: (1) co-occurrence of at least two of the four conifer species; (2) presence of a sufficient number of both living and dead saplings of different heights; (3) natural (as opposed to planted) regeneration of the saplings; and (4) no recent thinning or silvicultural intervention. Two co-occurring tree species were sampled at each site, with the species combination varying along elevation (Table 1). The species P. cembra and L. decidua were sampled at the three highest sites in La Punt and Chamues-ch (2085 and 2052 m a.s.l., both Central Swiss Alps) and Samprou (1900 m a.s.l., Southern Swiss Alps); L. decidua and P. abies at the three intermediate sites in Bergün and Filisur (1585 and 1440 m a.s.l., both Central Swiss Alps) and Predasca (1435 m a.s.l., Southern Swiss Alps); and P. abies and A. alba at the three lowest sites in Campra (1424 m a.s.l., Southern Swiss Alps), Grabs and Entlebuch (1219 and 908 m a.s.l., both Northern Swiss Alps).

Field sampling and data preparation

At each site, an equal number of naturally established saplings were sampled in 2016 across four height classes (10-40 cm, 40-70 cm, 70-100 cm, and 100-130 cm), four species (two co-occurring at each site) and status (dead vs. living), resulting in a perfectly balanced dataset. Only saplings without evidence of browsing or other mechanical damages as major mortality cause were sampled. Dead saplings were sampled only (i) if they had died recently (i.e., in 2016 or 2015), i.e. sampled saplings did not have any green needles left, but still carried brown needles, or (ii) if the stems were very flexible and twigs were moist inside in the case of L. decidua, the only deciduous conifer native to Central Europe. We selected for each dead sapling the nearest living sapling of the same species and similar height (i.e., within the same height subclass of 10 cm height). By sampling saplings in the neighborhood within the same forest stands, our sample consisted of dead and living saplings that were exposed to nearly identical meteorological conditions, which allowed us to exclude death due to climatic influences. For example, 2015 was a particularly warm year in Switzerland (Appendix 1) that could have influenced tree mortality in some regions or specific forest stands. A total of 1’440 saplings were sampled: at each of the nine sites, 160 living and dead saplings were sampled, corresponding to 40 dead and 40 living saplings of each species (i.e., 10 dead and 10 living saplings for each of the four height classes, cf. above). From each sapling, a stem disk was taken as close as possible to the root-shoot boundary. Depending on sapling height, additional stem disks were taken at 40 cm, 70 cm and 100 cm height. A distinct advantage of stem disks compared to increment cores is that disks always contain the pith and thus allow for inferring tree age and cambial age, respectively, in addition to radial growth rate.

For each sapling, the following information was recorded: species, status (dead/alive), height, diameter above the root collar (DRC), and light availability. Light availability was derived from hemispherical photographs using a Canon EOS 70D camera with a Sigma 4.5 mm F2.8 Model EX DC HSM circular fisheye lens. The photographs were taken after sampling the saplings 40 cm above ground at the exact point where each sapling was growing. The photographs were analyzed with the software Hemisfer, version 2.2 to calculate the total (diffuse and direct) effective radiation (W/m2) available for each sapling. In the tree-ring lab, the stem disks were sanded and the ring widths measured with an accuracy of 0.01 mm on a LINTAB measurement table using the TSAP-Win software according to standard dendrochronological procedures. We measured tree ring widths on only one radius, opposite to the widest radius. Due to the low number of tree rings, we could neither visually nor quantitatively cross-date the tree rings of the stem disks. However, we do not assume that partial or missing tree rings occurred frequently because (1) we worked with stem disks that allows more easily to detect partial tree rings unlike increment cores; and (2) the four investigated species rarely show missing rings at the adult stage in these temperate climates, which is probably related to the lack of severe drought.

Based on the measured tree rings, the mean age of the sampled saplings was 21.5 (range: 5 – 63) years for P. cembra, 18.2 (3 – 86) years for L. decidua, 25.4 (5 – 68) years for P. abies and 25.1 (5 – 64) years for A. alba. The mean height was 71.2 (range: 10 – 130) cm for P. cembra, 72.7 (10 – 130)  cm for L. decidua, 71.0 (11 – 130) cm for P. abies, and 72.2 (10 – 130) cm for A. alba. The mean diameter above the root collar (DRC) was 1.9 (range: 0.3 – 6) cm for P. cembra, 1.4 (0.2 – 4.6) cm for L. decidua, 1.9 (0.2 – 5.2) cm for P. abies, and 1.7 (0.1 – 5.0) cm for A. alba.

Statistical analyses

Based on tree-ring data, we analyzed the radial growth rate of the selected radius for stem disks at different heights and calculated height growth rate, respectively, to compare growth patterns of living and dead saplings of different tree species across sites. Each dead sapling was paired retrospectively to a living sapling of the same species and site with tree age as matching variable using the function “pairmatch” of the “optmatch” package, version 0.9-10, in the software R. Tree age is more suitable as a matching variable than diameter or height, because paired saplings with the same or similar age have grown during the same period and under the same climatic conditions. In contrast, paired saplings of the same or similar diameter or height may have substantially different life histories and thus be less comparable. This would for example be the case if a slowly growing living sapling was exposed to different weather conditions than a rapidly growing dead sapling of the same diameter or height.

The difference in light availability, recent radial growth and diameter above the root collar within the matched pairs was statistically tested using a paired t-test, separately for each species but with all height classes pooled together. Because severable observations were available per site, we fitted linear mixed effect models with the function “lme” of the “nlme” package, version 3.1:

∆wij=β+ui+εij                                                                                                                                                                             (1)

ui~N0,σu2, εij~N(0, σ2) i.i.d.

where the response variable Dwij is the difference in light availability, recent radial growth and diameter above the root collar, respectively, of sapling pair j (i.e., difference between a living and a dead paired sapling of the same species) at site i (i.e, one of the nine forests). The fixed effect b is the expected difference between living and dead saplings, ui is the random effect with site as grouping variable, and eij are the errors. To calculate the recent radial growth rate, we considered tree-ring widths at DRC of saplings over the last 5 years only because (1) stem disks from different heights showed similar growth patterns, particularly in L. decidua and P. abies, and (2) on average over all saplings, the growth of living and dead saplings began to diverge approximately 5 years before dying and sampling, respectively. The radial growth rate of the last 5 years (RGR5) was therefore used as explanatory variable in the conditional logistic models (see below) as in other tree mortality studies. We used the same linear mixed effect model (Equation 1) to test for differences within the matched pairs for RGR5 separately for each species and height class.

The matched pairs were then used to fit 18 species- and site-specific conditional logistic regression models with the function “clogit” of the “survival” package, version 2.41‐3. Conditional logistic regression is an extension of logistic regression that offers the advantage of adjusting for non-independent case-control data. The conditional likelihood takes into account the information relative to the matched pairs. The conditional logistic regression model is defined as the log-odds of mortality probability p:

logp1-p=α1+α2g2+…+αkgk+β1x1+…+βnxn                                                                                                                                                                             (2)

where α are the coefficients of the matched pairs variable g, the coefficients β those of the covariates x. The binary variable g encodes living saplings as g = 0 (control) and dead saplings as g = 1 (case). As covariates x, we used (1) effective total radiation, representing light availability; (2) diameter above the root collar (DRC); and (3) mean radial growth rate at DRC of the last five years prior to death or sampling (i.e., radial growth rate, RGR5). A positive coefficient b means that the mortality probability increases with increasing values of the covariate, a negative coefficient means that the mortality probability decreases with increasing values of the covariate.

Usage Notes

This file contains 16 columns (variables) x 54046 rows (excluding the first row, which contains the header with the variable names). The data was used to create the Figures and to fit the survival models. Weather data used for Table 1 and Appendix 1 are property of MeteoSwiss and are not included.

The variables are:

  1. Year

The variable "year" (integer) indicates the year of formation of each tree ring.

  1. IDstemdisk

The variable "IDstemdisk" is a unique identifier of each stemdisk, where the first letter indicates the “alps”, the second letter the “site”, the third letter the “status”, the fourth letter the “species, and the first two numbers indicate a unique number within each site, status, species, and the last number indicate the “stemdisk.position”. E.g. “SCTP20_2” identifies the second stem disk (at position of 40 cm height) of the twentieth sapling of Picea abies dead sampled in Campra in the Southern Swiss Alps.

  1. Tree.ring.width

The variable “tree.ring.width” (unit: 0.01 mm) indicates the width of the tree rings measured on a LINTAB measurement table using the TSAP-Wind software (both of Rinntech, Heidelberg, Germany.

  1. IDsapling

The variable "IDsapling" is a unique identifier of the tree sapling, where the first letter indicates the “alps”, the second letter the “site”, the third letter the “status”, the fourth letter the “species, and the two numbers indicate a unique number within each site, status, species. E.g. “NELA01” identifies the first sapling of Abies alba living sampled in Entlebuch in the Northern Swiss Alps.

  1. Stemdisk.position

The variable “stemdisk.position” indicates 4 categories of sapling heights where stem disks were taken: “1” for stem disks taken at the root-shoot boundary (above root collar), “2” for stem disks taken at 40 cm height, “3” for stem disks taken at 70 cm height, and “4” for stem disks taken at 100 cm height.

  1. Height.class

The variable “height.class” indicates 4 categories of sapling height: “10-40 cm”, “40-70 cm”, “70-100 cm”, and “100-130 cm”.

  1. Height

The variable "height" (unit: cm) corresponds to the sapling height at sampling, which was measured perpendicularly from the soil surface to the apical meristem.

  1. Diameter

The variable "diameter" (unit: cm) corresponds to the sapling diameter above the root collar at sampling, which was measured with a calliper.

  1. Alps

The variable "alps" indicates 3 categories of Swiss Alps: “Southern”, “Central” and “Northern”.

  1. Site

The variable "site" indicates 9 categories of study sites: “La Punt” at 46.57724°/9.90593°, “Chamues-ch” 46.54883°/9.97125°, “Samprou” 46.54446°/8.82848°, “Bergün” 46.63347°/9.7259°, “Filisur” 46.66112°/9.6733°, “Predasca” 46.55417°/8.91808°, “Campra” 46.51982°/8.88549°, “Grabs” 47.19159°/9.36694°, “Entlebuch” 47.02119°/8.09121° (latitude/longitude, WG84 coordinates) in Switzerland.

  1. Status

The variable "status" indicates 2 categories of a sapling state at sampling: "living" and "dead".

  1. Species

The variable "species" indicates 4 categories of tree species sampled: “Pinus cembra”, “Larix decidua”, “Picea abies”, “Abies alba”.

  1. TOTheff

The variable "TOTheff" (unit: W/m^2) indicates the total (sum of direct and diffuse) radiation at the exact position where each sapling was sampled, measured with hemisferical photographs and analysed with the software Hemisfer.

  1. Age.IDstemdisk

The variable “age.IDstemdisk” (unit: yr) indicates the number of tree rings formed in each stemdisk.

  1. Matches

The variable “matches” identifies pairs of dead and living saplings of the same species within the same site paired retrospectively with tree age as matching variable using the function “pairmatch” of the “optmatch” package, version 0.9-10 (Hansen & Klopfer 2006).

  1. RGR5.IDstemdisk

The variable "RGR5.IDstemdisk" (unit: 0.01 mm/yr) indicates the stemdisk-specific radial growth rate of the last five years prior to sampling of living saplings and prior to death of the dead saplings, respectively.

Funding

Swiss National Science Foundation (Advanced Tree MOrtality MOdeling – ATMO^2), Award: Project number 163250

Swiss National Science Foundation (Advanced Tree MOrtality MOdeling – ATMO^2), Award: Project number 163250