Selection of sites and trees
The material was collected along a transect from the south of Komi (south taiga sub-zone of boreal forests) to the Arctic spruce timberline. The sampled stands were at similar altitudes. The study stands were grouped into “sub-zones” according to their geographical position in the taiga sub-zones of boreal forests. Totally 261 trees were collected in 5 sub-zones of taiga forests in Komi (Fig. 1, Tab. 2) and 696,929 tree rings were measured.
The sites were selected using GIS datasets of forest management units, old forest inventory maps and satellite images TERRA ASTER (scene size 60x60 km) with a spatial resolution of 15 m. In the procedure for site selection the main aim was to find the most common site types and at the same time exclude possible forest management or any other human impact from the past. Sites with a low productivity index (5 class, according to the classification system for Russian forest productivity) represents 70% of the forest area of Komi Republic (Kozubov & Taskaev 1999). The sites were selected to minimize the differences in exposure, soil properties, topography or vegetation development. To obtain information about changes in site productivity trees of different ages and comparable sites were selected. The trees were randomly sampled on sites of low fertility.
The stands were selected according the following criteria for site conditions:
- spruce or pine dominating species;
- low site index (5 class, according to the Russian forest productivity classification system (Zagreev 1992);
- multistoried mature stands represented by trees of 3-5 different age classes.
In most of the regions in Komi the forest stands are multistoried. Therefore the sample trees were chosen from among trees not dominated by older trees but rather located in openings within the stand. The sample trees were expected to reveal homogeneity in their tree-ring pattern; they showed no obvious signs of near-neighbour competition or forest management. Trees were chosen from different diameter classes, healthy looking with straight, unbroken stems and regularly shaped crown. Mature dominant trees without visible signs of damage were selected as sample trees. The sample trees in the stands were expected to have a common growth trend, which was influenced by a large portion of the climatic effects and other factors which differ among individuals and from site to site. At each site an averaging process, during building chronology, helped to minimize the influence of other factors.
Prior to felling, for visual assessment of the tree ring pattern, a core from each tree was extracted with an increment borer. This allowed exclusion of those trees affected by competition in the past. Siberian spruces and Scots pines were sampled at breast height (1.3 m above the ground). In most cases, discs were cut using a chain saw. If it was difficult to cut discs, cores were extracted from trees with the increment borer from two radii per tree (the first one oriented to the north, the others at 900 -1200 to the first). Geographical coordinates of sample trees were measured using GPS.
Monthly precipitation sums and air temperature means from 5 climate stations (WMO numbers 22996, 23804, 23711, 23412, 23219) were obtained from the Center of Meteorology and Environment Monitoring of Komi Republic (Fig. 1, Tab. 2) and the public archives (Razuvaev et al. 1995 Vose et al. 1992). Climate records were checked for consistency and homogeneity (Razuvaev et al. 1995). Meteorological data were used for identification of dendroclimatic relationships and long-term trends in climate change. To identify long-term trends in climate, the deviations in absolute units from the long-term mean were calculated for the common period. The deviations were then smoothed with a 30-year running mean.
Radial increments were measured to an accuracy of 0.01 mm. During the process of measurement, the raw measurements of tree rings were cross-dated using visual control, by comparing the series graphically. Cross-dating and data quality were assessed using the computer program COFECHA (Grissino-Mayer et al. 1997, Holmes 1999).
To maximize the climatic signals in tree ring series, other factors should be minimized. For example, a typical sample might display exponentially declining growth with age, the classic biological growth curve. Standardizing the sample using a spline curve results in data values that represent a departure from the “expected” value for a given year. This departure from the expected mean value is then used to interpret a proxy environmental signal in the data. The above-mentioned procedure usually is an attempt to remove the growth trends due to normal physiological ageing processes and changes in the surrounding forest community. Therefore individual ring-width series were indexed using spline curves of 60 years with a frequency response of 50%. This approach was selected due to the high amount of variance in the dataset because of using the trees from different age cohorts for chronology building.
The common interval (1954 - 2003) adjusted for order of the pooled autoregressive model was used for analysis of climate-growth relationships. Indices were further prewhitened using Box Jenkins methods of autoregressive and moving average time series modelling (ARMA - Box & Jenkins 1976, Cook 1985, Monserud 1986). The order of the autoregressive-moving mean process was determined by Akaike Information Criteria (Akaike 1974, Cook 1985). Prewhitening allows us to transform autocorrelated series into a series of independent observations by extracting residuals from the modelled process. Chronologies were produced by averaging the annual values of indices. The software ARSTAN (Grissino-Mayer et al. 1997, Holmes 1999) was used to calculate chronologies from tree-ring measurement series by detrending and indexing (standardizing) the series, then applying a robust estimator of the mean value function to remove the effects of endogenous stand disturbances.
Residual ARSTAN chronologies containing high frequency variation were used to examine climate-growth relationships on an interannual basis.
Chronology confidence was determined using statistics of Expressed Population Signal (EPS - Wigley et al. 1984). EPS was computed as a function of mean inter-tree correlation and sample size. In general, it can be assumed that values of EPS equal to or greater than 0.85 are indicators of reliable chronology (Cook & Kairiukstis 1990). Here EPS was computed over the time window of 50 years (1954-2003).
Statistical analysis of growth-climate relationships
Response function analysis (Cook & Kairiukstis 1990) was applied to determine which monthly weather variables significantly affected high frequency variation in radial increment of Scots pine and Siberian spruce. The monthly mean air temperature and monthly sums of precipitation were used to analyse the response of trees to current and previous years. The DENDROCLIM2002 software was applied for the analysis, using bootstrapped confidence intervals to estimate the significance of both correlation and response function coefficients and testing their significance at the 0.05 level. The temporal stability of dendroclimatic relationships was illustrated using correlation coefficients of significant (p