Dendrochronology and extreme pointer years in the tree-ring record (AD 1951–2011) of polar willow from southwestern Spitsbergen (Svalbard, Norway)

Complete individuals of Salix polaris, including their root and branch systems, were collected during the short Arctic summer seasons in 2007, 2008, 2011 and 2012. The total of 45 samples of polar willow from 10 sites were collected from the flat raised marine terraces at the elevation of 12–28 m a.s.l. in the vicinity of the Polish Polar Station in Hornsund (Fig. 1). Each individual was documented in a digital photograph. Each sampling point was situated on relatively homogeneous surface, that does not show clear evidence of periglacial activity (e.g. soli-fluction lobes). The sites were flat and well-drained, never in local depressions, channels or on steep edges. Each collected individual was cut with pruning shears and placed in string bags. To protect the samples, the bags were filled with a mixture of ethanol and glycol.

The collected samples were sectioned with GSL 1 sledge microtome. The 15–20 μm cross-sections were taken along each individual from four to seven different places (depending on the length of the root and wooden branches) in order to develop their chronology (Fig. 6). This method was described in detail in previous research (e.g. Bär et al., 2006; Hallinger et al., 2010; Owczarek et al., 2014a; Myers-Smith et al., 2015b). In the laboratory, microtome cross-sections were prepared from the whole stem diameter of the selected segments. After staining with 1% solutions of safranin and astra blue dyes, digital photographs of the micro-sections were taken for tree-ring analyses. The width of rings distinguished in the collected individuals ranges from relatively wide 0.8 mm, to extremely narrow of less than 0.01 mm (Fig. 5b). Partially or completely missing growth rings are very common in the analysed specimens.

The Salix polaris measurements followed the standard dendrochronological procedures (Cook and Kairiukstis, 1990 and Speer, 2010). The ring widths were measured along two or three radii using the OSM 3.65 and PAST4 software (VIAS, 2005). The TSAP software was used for visual cross-dating of growth width series, and the COFECHA software for the statistical quality control of cross-dating (Holmes, 1983). Locally absent or wedging rings were examined by applying serial sectioning for each individual (Fig. 6) (Kolishchuk, 1990).

In general, the raw data fluctuated around a horizontal straight line, without a distinct age trend. Therefore the measurements were transformed into chronology by averaging. The ARSTAN (ver.41d) standardisation software was used to calculate the chronology and its parameters (Cook, 1985). In order to compare the resulting chronology with other chronologies and to assess its reliability and relevance for the dendroclimatological research, the appropriate descriptive statistics, including mean sensitivity (MS), mean series inter-correlation (Rbar) and expressed population signal (EPS), were calculated (formulas after Wigley et al. (1984) and Speer (2010)). In the standardised dendrochronological data, the extreme years were distinguished using the threshold value of ±1 standard deviation. In the next step the distinguished negative years were compared with climatic variables. The meteorological data from the Hornsund station, covering the period 1978–2011, were obtained from Marsz and Styszyńska (2013) and Hornsund Database (2014). The climate variables investigated in this study included the monthly data: mean air temperature (°C), mean maximum diurnal air temperature (°C), mean minimum diurnal air temperature (°C), monthly absolute maxima of air temperature (°C), monthly absolute minima of air temperature (°C), number of days with positive mean diurnal air temperature, sunshine duration (hours), percentage of relative sunshine duration, total monthly precipitation (mm), monthly mean thickness of snow cover (cm) and number of days with snow cover. In addition we used daily air temperature and precipitation data for the selected years.

Finally, 21 out of the 45 samples were included in the chronology and became the basis for further analyses. The remaining sequences were removed due to a large number of missing rings, eccentricity, presence of scars and clearly visible reaction wood. The oldest individuals of Salix polaris used in the chronology construction were 62 years old, while the mean plant age was 38. After truncation up to minimum three series, the constructed chronology covers the years 1951–2011 (Table 1, Fig. 7). The annual increments are in general very narrow and vary between 18 μm/yr to 564 μm/yr, with the mean approx. 137 μm/yr. In the course of the sample curves of the dwarf shrub no age trend is observed, as is in the case of trees. In general, from the beginning of the 1980s an increase of the mean and maximum growth ring width is observed (Fig. 8). Increase in temperature of 1.1°C per decade and 43 mm of annual precipitation per decade is accompanied by positive trend in growth ring width (12 and 35 μm for mean and maximum growth ring width, respectively) (Table 2). The reduced growth ring width over the last five years is a result of unequal samples replication in comparison to the previous three decades. In the detailed course of the chronology curve a decline can be observed in the 1980s (Fig. 7). The periods of increased growth rate can be noticed in the 1970s and 1990s.

The developed chronology shows very high inter-annual variations expressed by quite a high value of the mean sensitivity (0.56) (Table 1). The Expressed Population Signal values were slightly below the 0.85 threshold (Wigley et al., 1984), and their mean value for the whole chronology was 0.82 (this value decreased in the oldest part of the chronology). This means that in order to reduce the reconstruction uncertainty in the future it will be necessary to increase the sample size (number of plants), especially of the oldest individuals.

Consistent extreme years (narrow or wide ring widths) could be distinguished on many sections and in the resulting chronology. We found 13 extreme years in the Hornsund chronology, of which nine were negative: 1958, 1967, 1975, 1980, 1981, 1989, 1999, 2006 and 2010, and four were positive: 1963, 1966, 1976 and 1997. Positive years are usually the effect of regeneration after unfavourable conditions in the previous years. Negative years are much more important for dendroclimatological and ecological analyses, indicating unfavourable environmental conditions. Interpretation of extremely low growth or even a missing ring on the basis of existing meteorological data for the Hornsund station showed that there is no a single factor influencing growth of polar willow.

Detailed interpretation of the extreme years showed that very narrow rings occur when a few unfavourable climatic factors before or during the year of ring formation act together (Table 3). Despite the fact that there were not enough observational data for the years 1958, 1967 and 1975, we can trace the course of the weather in subsequent extreme negative years since 1978. In the year 1980 low annual air temperature of the previous year and the second largest absolute minimum of the air temperature were accompanied by a number of days with positive mean diurnal air temperature during winter. Such thermal fluctuations could have damaged the plant tissue and previous year's shoots. Another threat for the dwarf shrubs in 1980 was very low precipitation that spring (May-June). Similar meteorological situations, i.e. the occurrence of a number of days with positive temperatures in winter, were observed in 2006. That year had up to 13 days with positive temperature in January and the highest absolute maximum of air temperature in January (4.9°C) recorded for Hornsund. Despite the high precipitation in January 2006 (321% of the norm for the reference period, 1978–2011), the above average temperatures led to fast snow melt and consequently very low mean thickness of snow cover in January. The lack of any substantial snow cover along with the lowest recorded absolute minimum of air temperature (–35.9°C) that same month, could have caused damage to the dwarf shrubs, and thus explain the very narrow ring produced later that year. Additionally, the lowest recorded absolute minima of air temperature in May and June of that same year (2006) and the lack of snow cover at that time would have caused a water shortage for plants and thus also explain the formation of a very narrow ring. In 2010, a rapid thaw took place in January, an effect of nine days with positive temperature. Additional occurrence of liquid precipitation resulted in only 26 days with snow cover. After episodes of snow melt, a rapid temperature drop took place (Fig. 9). Lack of a deep and stable snow cover which isolates and protects Salix shrubs against frost damage resulted in a very narrow growth ring formation that year.

The reason for extremely narrow growth of polar willow could have also been very low precipitation at the beginning of the growing season accompanied by a lack of snow cover, which severely restricted water availability to the shrubs and resulted in the formation of a very narrow ring. Low amount of spring precipitation is rather typical for this area but exceptional years with values, which are significantly below the average, can be distinguished. The growing periods for Salix communities are determined based on the start of daily temperature T > 0°C. Although Kappen (1993) reported, that vascular plants can be active under snow, no photosynthesis is observed if the ground is frozen. Such weather conditions occurred, for example, in 1989, when there was a low annual air temperature the previous year, followed record lowest precipitation in May 1989 and a lack of snow cover the next month. Also, a similar weather pattern occurred in 1999, with low (snow) precipitation during that winter and spring, and in fact the lowest annual snow-cover thickness of the whole instrumental record (1978–2011). Development of narrow rings in 1989 and 1999 as an effect of physiological drought connected with low winter precipitation can be linked to more complex weather conditions. Other factors like predominance of anticyclonic circulation at the beginning of the 1989 vegetation period (Niedźwiedź, 2006) and relatively high June temperature during these years (respectively 2.3°C and 2.6°C mean monthly air temperature when the average for the period 1978–2011 is 1.9°C) led to increase of evaporation and soil desiccation.

In addition, most of the extreme years were characterised by very low (above average for the reference period) sunshine duration in summer (Table 3). Sunshine is not the growth limiting factor at this site, but its values are connected to temperature, which logically drives the tree growth during the polar summer.