When you connect to the NBII Metadata Clearinghouse you will be able to search through metadata-based descriptions of biological data sets and information products from many different sources to identify those that meet your particular search criteria.
The NBII Metadata Clearinghouse: http://metadata.nbii.gov/
The NBII Home Page: http://www.nbii.gov/
Powered by Mercury
We collected data in the Hoh (37 plots) and Dungeness (39 plots) watersheds during the summers of 2001 and 2002, as part of the CLIMET (Climate-Landscape Interactions on a Mountain Ecosystem Transect) program (Fagre et al. 2003). We selected plot access points (mostly National Park trails) prior to sampling using vegetation classifications based on Landsat Thematic Mapper imagery (University of Washington Prism project) and topographic maps (A. Hessl, personal communication). We chose exact plot locations in the field, with the goal of sampling a comprehensive range of the vegetation assemblages and physiographic environments of each watershed.
Plots (0.05 ha) ranged in elevation from near sea level to tree line and were located on various aspects and slopes of varying steepness. Plots were located in mature stands with no obvious evidence of disturbance. This "biomonitoring" procedure of sampling a small subset of geographic locations and species is a rapid and cost effective method of estimating trends in productivity as they relate to climatic change (Hessl and Peterson 2004). Within each plot, we noted the location coordinates using a GPS unit and the physical characteristics of elevation, slope, and aspect. We recorded species, diameter at breast height (DBH) and height of all trees greater than 10 cm DBH. We took two increment cores from each of two live individuals representing each 10-cm diameter class for all tree species present on the plot. We collected a total of 532 tree cores in the Hoh watershed and 547 tree cores in the Dungeness watershed.
We crossdated and measured ring widths of all tree cores following the procedures of Stokes and Smiley (1968) and Fritts (1976). We did not include cores that did not crossdate well visually in this analysis. Following crossdating, we measured ring widths for each core to the nearest 0.001 mm using a Velmex measuring system. We remeasured a randomly selected 10-year segment of each core for quality control. If the standard deviation of the difference (absolute value) between measured and remeasured sections was greater than 0.05, we remeasured the entire core. We verified crossdating using the COFECHA program (Holmes 1999).
* Time periods of analysis
We chose the time periods of this analysis for several reasons. We chose to restrict the length of the time period to the past half century both so as not to underestimate total plot productivity through successive losses due to mortality as well as to focus the analysis and conclusions on the recent period of increasing growth that may have been affected by increasing global temperatures and atmospheric CO2 concentrations (e.g., Peterson et al. 1991, Peterson 1994, Innes 1998, McKenzie et al. 2001). In addition, we divided the analysis into two sections according to the two most recent phases of the Pacific Decadal Oscillation (cool, wet PDO = 1947-1976; warm, dry PDO = 1977-2000; Mantua 2002) because tree growth patterns might be affected differently during each phase (Peterson et al. 2002, Keeton et al. 2003).
* Plot-Scale Calculations
We converted ring widths and DBH measurements into annual diameter increments with the following equation, which considers the species-specific increase in bark thickness as a tree grows:
Dt-1 = [Dt - (Bs * Dt)] - Rt / 1 - Bt
Dt-1 = diameter at breast height at year t-1
Dt = diameter at breast height at year t
Bs = bark coefficient for species s (Appendix B)
Rt = 2*ring width during year t
This equation introduces a small amount of error into the growth time series as some bark coefficients are derived outside of the PNW. However, because we are comparing relative growth patterns, this error should not affect interpretation of the results.
We then converted these values into annual basal area increments:
BAI= pi * (Dt/2)^2 -pi * (Dt-1/2)^2.
Unlike ring widths, BAI is not subject to decreasing trends as a tree grows, and therefore no detrending is necessary. In addition, BAI serves as a more accurate measure of tree growth than ring widths (Visser 1995). BAI is also a reasonable proxy for the aboveground net primary productivity (NPP) of a stand (Hessl and Peterson 2004). Complete measures of forest NPP are difficult because they must also include belowground productivity, mortality, litterfall, and grazing-factors that were not quantified in this study. We attempted to minimize the loss of previous years’ BAI due to mortality by restricting the time period of analysis. However, the results from the analysis of the earlier time period (1947-1976) should be interpreted with caution because they may more seriously underestimate plot or forest growth amounts. While we expect that BAI measurements underestimate actual forest NPP, they are an effective means of detecting responses to environmental conditions (Hessl and Peterson 2004).
We assigned uncored individuals in the plot (>30 years) with BAI time series for cored individuals of the same species and size class within that plot or (infrequently) from a plot in the same forest type with similar physical characteristics. The local error inherent in this process is necessary to improve data representation of physical and ecological gradients in each watershed and improve conclusions at larger scales. We summed the tree BAI values to establish total plot BAI time series and calculate total plot basal area. In addition, we calculated a weighted arithmetic average of BAI time series and basal areas for the entire plot. We prewhitened average plot BAI time series by fitting autoregressive models, using the first-minimum Akaike Information Criteria (AIC) to select the model order (number of autocorrelated years). Prewhitening removes autocorrelation trends due to lag effects of growth influences on subsequent years’ growth. Resulting time series are more statistically robust for year-to-year comparisons, but may lack a significant portion of the actual growth amount. Therefore, we report results for multiple growth measures: each plot is represented by total plot basal area and three growth time series, or chronologies, which are the 1) sums of the BAI time series for every individual in the plot and 2) arithmetic averages of the BAI time series for every individual in the plot and 3) the prewhitened plot-average time series.
* Forest-Type Calculations
We partitioned plots into forest types according to the biotic zones and major tree species delineated by Buckingham et al. (1995) in relation to aspect and elevation within the Olympic Peninsula (Figure 4). We expanded the plot measurements to forest types by summing plot BAI and total basal area for all the plots within a forest type, resulting in a single time series of basal area per hectare for each forest type, as well as values of total basal area per ha. We also calculated weighted averages of prewhitened, averaged plot BAI time series within each forest type. We then similarly expanded these forest-type values to obtain overall growth time series for each watershed.
* Correlation Analysis
We calculated mean interseries correlation to assess the degree of similarity among individual BAI time series within each plot as well as among total plot BAI time series within each forest type. Mean interseries correlation calculations present the average of all pair-wise Pearson product-moment correlation coefficients for each BAI series within a plot. Pearson correlation coefficients are a measure of the amount of similarity in year-to-year variation in growth. Similarly, mean "interplot" correlations represent the overall degree of correspondence in growth patterns between plots within a forest type. We also evaluated Pearson correlation coefficients of all possible forest type pairs, as well as between watershed BAI time series to see if growth patterns are statistically similar between certain forest types; this provides information on the scope of growth influences. We also evaluated mean interseries correlation strength among plots grouped by aspect and slope as well as by the non-geographic factors of species, size class, and sample size. Because this correlation analysis encompasses multiple scales, we can examine the correlation strength of growth patterns at these different scales (i.e., plot, forest type, watershed) as an indication of the scale of the dominant growth influences.
* Cluster Analysis
We used agglomerative hierarchical cluster analysis with the single linkage method of creating clusters for the prewhitened plot average BAI data (see Rencher 2002). Cluster analysis compares the average growth amount between successive sets of years. The single linkage method calculates the sum of the annual differences in growth amount between BAI series pairs. For example, if all the differences between series pairs are small, then the overall distance between those two plots on the cluster diagram will also be small. This allows us to assess the amount of variation between plot BAI patterns and to determine the most appropriate groupings of plots according to shared growth variability. Clusters contain plots that have similar magnitudes of difference between paired annual BAI series.
* Sensitivity Analysis
We chose the following measures of tree growth and growth variability to provide a comprehensive view of the relative sensitivity of trees and forests to changes in growth-limiting factors:
1) Mean sensitivity - difference between adjacent BAIs within an individual time series divided by the mean of the two increments, averaged over the entire series. This is a measure of the year-to-year variability in growth. A tree with a high mean sensitivity usually has experienced more growth limitation due to environmental factors, such as climate (Fritts 1976).
2) Standard error - amount of variation in BAI relative to growth amount, independent of sample size.
3) Coefficient of variation - measure of relative variability in BAI, independent of growth amount.
4) Magnitude of growth variability - indication of the possible degree of change. Calculated using prewhitened BAI plot averages to allow comparison among forest types (maximum BAI minus minimum BAI).
5) Autocorrelation - indication of how long (years) a given change in a growth-limiting factor affects tree growth. This information is provided from the prewhitening process.
6) Extreme growth - number of years for which growth exceeds a threshold value (1 and 2 times the standard deviation above and below the mean).
Unless noted above, we calculated these sensitivity measures using individual BAI time series and then averaged the calculated values to obtain relative plot and forest-type sensitivity estimates. All of these measures were considered in making a final estimation and comparison of forest sensitivity. In addition, forest types that grow more and have high growth variability (mean sensitivity and coefficient of variation) may have more dramatic swings in growth and productivity from year to year in response to environmental variability (Hessl and Peterson 2004).