Using Blue Intensity from drought-sensitive Pinus sylvestris in Fennoscandia to improve reconstruction of past hydroclimate variability

Abstract

High-resolution hydroclimate proxy records are essential for distinguishing natural hydroclimate variability from possible anthropogenically-forced changes, since instrumental precipitation observations are too short to represent the whole spectrum of natural variability. In Northern Europe, progress in this field has been hampered by a relative lack of long and truly moisture-sensitive proxy records. In this study, we provide the first assessment of the dendroclimatic potential of Blue Intensity (BI) and partial ring-width measurements (latewood and earlywood width series) from a network of cold and drought-prone Pinus sylvestris L. sites in Sweden. Our results show that all tree-ring parameters and sites share a clear and strong sensitivity to warm-season precipitation. The ΔBI parameter, in particular, shows considerable potential for hydroclimate reconstructions, here permitting a cross-validated precipitation reconstruction capable of explaining 56% (1901-2010 period) of regional-scale warm-season high-frequency precipitation variance. Using ΔBI as an alternative to ring-width improves the predictive skill with nearly a 20 percentage points increase in explained variance, reduces signal instability over time as well as allows a broader seasonal window (May-July) to be reconstructed. Additionally, we found that earlywood BI also reflect a positive late winter through early summer temperature signal. These findings emphasize that tree-rings, and in particular wood density parameters such as from BI, are capable of providing fundamental information to advance our understanding of hydroclimate variability in regions with a cool and rather humid climate regime that traditionally has been overlooked in studies of past droughts. Increasing the spatio-temporal coverage of hydroclimate records in northern Europe, and taking full advantage of the opportunities offered by the wood densitometric properties should be considered a research priority.

Keywords: Blue Intensity; Dendroclimatology; Drought sensitivity; Fennoscandia; Hydroclimate; Scots pine; Tree-ring.

Fig. 1

Geographical distribution of the five tree-ring chronology sites in Sweden, together with the outline of the 57–62° N/14–19° E subset of the CRU TS 4.03 product (Harris et al. 2014) and the location of the Stockholm historical weather observatory dataset used in the calibration/validation exercise

Fig. 2

Boxplot summary of the a Rbar statistic and b first-order autocorrelation (AR1) for each tree-ring parameter. The AR1 coefficients are computed over the 1901–2010 modern interval. Also shown is the AR1 of the high-pass filtered warm season precipitation (MJJ pre). For definition of abbreviations, see Sect. 2.1

Fig. 3

Raw non-detrended time series of absorbed blue light in the latewood (MXBI) and earlywood (EWBI) portions of the tree rings, and the derived ΔBI parameter. Site average (grey lines) and network average (green line) are shown

Fig. 4

Biplot of the first two principal components of the PCA performed over the common 1798–2010 period on the multi-parameter collection from the five sites in Sweden. Identified parameter cohorts are highlighted in green. The color of the vectors corresponds to the different parameters (green lines—EWBI; blue—MXBI; pink—LW width; red—TRW; black—EW width). The first two components together represent nearly 50% of the total variation

Fig. 5

Simple linear correlations with CRU TS 3.2 monthly a precipitation and b temperature and the PC1 scores for each tree-ring parameter. Correlations are computed over the 1901–2010 interval using regional (57–62° N and 14–19° E), high-pass filtered, CRU TS 3.2 averages. The numbers in the parenthesis denote the amount of explained variation by the first PC component. Coefficients in the right of the plot are correlations with seasonally averaged climate variables

Fig. 6

Field correlation between selected PC1 composite chronologies and gridded meteorological data from the CRU TS 4.03 product over the 1901–2010 period. a ΔBI and b ring-width (TRW) versus May–July precipitation, c the difference between the correlation fields shown in plots (b) and (a), d EWBI versus February–May temperature. Correlations are reported in color if significant (p < 0.05)

Fig. 7

Moving 31-year window correlation over the 1901–2010 period between selected PC1 composite chronologies and gridded meteorological data from the CRU TS 4.03 product. a Ring-width (TRW) and b ΔBI versus precipitation, EWBI versus c precipitation and d temperature. Precipitation and temperature data have been high-pass filtered and averaged over the region bounded by the latitude/longitude coordinates 57–62° N/14–19° E

Fig. 8

Scaled PC1 composite reconstructions and their target CRU TS 4.03 instrumental data. a Ring width- and b ΔBI-based MJJ precipitation reconstructions, and c comparison between these two reconstructions. d ΔBI-based February–May temperature reconstruction. Correlations between time-series are provided in the bottom of each plot. Note that data have been high-pass filtered and normalized to z-scores over the entire record length

Fig. 9

Time-series of the a ΔBI-based MJJ precipitation (pr) reconstruction and historical meteorological records of sea-level pressure (SLP); b EWBI-based FMAM temperature reconstruction and historical records of temperature (tmp) from the Stockholm meteorological station. Note that SLP explains less than 40% of the variability in summer precipitation in the region (Fig. S3). Data have been high-pass filtered and normalized to z-scores over the entire record length. Correlations between time-series are provided in the bottom of each plot

Fig. 10

Scatter-plot comparisons between warm-season CRU TS 4.03 precipitation and a ΔBI and b ring width. Note that the regionally-averaged precipitation is normally distributed. c Kernel probability density functions of z-scored ring-width and ΔBI individual-site chronologies