How to Extract Climate Variability from Tree-Rings

Before the field trips have the permission of the owners, in case of a conservation area, or the corresponding authorities. It is very important that some personnel representing the authority participate in the field work to avoid any problem.

1. Sampling strategy

1. Determining the study area
   1. Select the most appropriate sampling area based on climatic information and forest composition (forests can be highly heterogeneous; Figure 1A,B).
   2. Check that the sampling site shows evident annual climatic seasonality and inter-annual climatic variation including dry/wet or cold/hot season during the year. Inspect the climatic records from nearby meteorological stations to determine the annual climatic seasonality and inter-annual climatic variation.
   3. Ensure that moderate to high climatic inter-annual variability is present so that the trees in the study site show enough year-to-year ring width variation to cross-date samples among trees.
   4. Carry out field trips in the area of interest to identify potential sites with the species of interest (Figure 1B)
   5. Use some of the recommended tools such as cartography, drones, and satellite images to explore a larger forested area and detect more potential sampling areas. Verify the information from these sources in the field.
   6. Gather information from complementary sources such as the regional stakeholders which include forest service providers, forest producers, rural communities, and small landowners. Choose the best study sites and the most suitable individuals to fulfill the objective based on the information obtained from both sources.
   7. Select areas where the longest-lived individuals of the species of interest are observed (Figure 2A,B). Observe standing dead trees, fallen trees, and stumps. Old dead samples are very important since they allow the chronology to be extended back in time (Figure 2A,C,D,E)
   8. Register the location of individuals with the characteristics mentioned above using a GPS.
2. Considerations for selecting the best tree
   1. Once the best site has been located, select the trees to be appropriately sampled. Trees located in shallow and rocky soils and steep slopes are more sensitive to climatic variability. Use these ecophysiographic characteristics to determine the limiting factors that trees most likely will record (Figure 2A).
      NOTE: Avoid taking tree samples in places of high competition; in these high-density locations, trees will have a strong forest stand dynamics signal and a reduced climatic one.
   2. Record the site information in a field format. Collect geographic and ecological information on the area, such as coordinates, elevation, the slope of the terrain, location's name, vegetation type, dominant species, and current land use.
   3. Record information of the sampled trees, such as diameter, height, presence of damage, if it is located near or on a stream, on a steep slope, or ravine.
      NOTE: The above information will be handy when analyzing the samples to corroborate and better interpret the study results. Since trees or samples might be exposed to damage or the site's conditions where the trees grow can alter the annual variations in growth. This metadata will help explain variations in growth independent of climatic factors, giving the elements to consider or eliminate noisy samples, always considering highlighting the climatic signal.
   4. Give a code for each sample based on the site name and the sample number, which comprises the first three letters of the site name, tree number, and sample number. For example, the first sample taken in this site will have the following code: RMI01A, which corresponds to the site Río Miravalles (RMI), tree number one (01), and first sample (A).
      NOTE: The term sample in this case refers to an increment core or a piece of a cross section taken from one tree.
   5. Perform selective sampling as done in most of the dendroclimatic studies by selecting individuals with specific phenotypic characteristics and growing in specific environmental conditions to address the objectives of the research. Perform a non-selective sampling if the goal is to project climate effects on tree growth and to integrate tree size and stand dynamics.
   6. Select trees with a long-lived appearance, on some occasions with dry top, dieback, twisted stem (i.e., spiral shape), and dropping branches (Figure 3A-C). Long-lived individuals will extend environmental records further back in time.
   7. Identify long-lived trees by observing highly compact, narrower rings during their most recent years which are by consequence difficult to observe and cross-date. Identify younger trees growing during these same periods as they register wider and more conspicuous rings, facilitating the dating of older ones.
   8. Consider sampling between 10%-20% of young individuals among the sampled trees within the sampling strategy.
   9. Ensure that the trees have a solid trunk to obtain the longest possible increment core. In addition, avoid rotten areas because they can cause sectioned samples and loss of internal rings and may get the increment borer stuck.
   10. Ensure that the selected trees are not hollow. Gently tap the tree with a plastic hammer and listen to the resonance of the wood. If the resonance is strong or deep, it means that the tree might be hollow. If the sound is dry, there is a low probability that the tree is hollow.
       NOTE: This step is important because the increment borer may get stuck in hollow trees, making it difficult to extract the increment borer, and potentially the sample might not be of good quality.
   11. Even when the above is considered and no rot is detected, pay close attention to the following. When introducing the increment borer, apply certain degree of force to penetrate the tree. When this required force changes, the borer becomes softer, at this point stop and draw the sample.
       CAUTION: If force is continued to drive the increment borer in, the decaying wood mixed with resin will collect in the increment borer barrel and form a plug that is difficult to remove with the extractor. When this happens, do not use a knife or some steel material to free the wooden plug from the increment borer (this could damage the cutting part, rendering it useless).
   12. The edge of the increment borer is very sensitive to metal rust, use a lubricant and a piece of wood to press the plug and release the increment borer cylinder. The wood does not cause any damage to the edge of the increment borer.
   13. When working with resinous trees or with large quantities of sap, clean the borer often with oil. Use lubricants or ethanol, for cleaning the resin residues that adhere to the metal.
3. Sample collection (collecting increment cores)
   1. Collect the samples with the Pressler increment borer (Figure 4A), a precision tool designed to extract a small core from a living tree without significant damage⁶. Use any of the available increment borers which come in different lengths (100-1000 mm), diameters (4, 5, 10, 12 mm), and threads (2 and 3; Figure 4B,C). As with any wood-cutting tool, keep the borer sharp and clean; unsharpened borers might lead to twisted and broken cores.
   2. Select the right borer depending on the tree species to be sampled. For most woods, use a three-threaded borer of any length or diameter for sampling. For hardwood species, use a two-threaded borer of small diameter and a short length for a slower penetration, less friction and stress on the wood, and a lower probability of being broken during the sampling process.
   3. In species that show a high frequency of false rings or interannual density fluctuation (IADF) and or micro-rings, use the 12 mm diameter borer instead of the 5 mm. This allows to extract a wider sample surface for better visualization of the difficult rings and facilitates the identification of these problems (Figure 4B). Do not attempt to use longer borers in hardwoods since there is the risk of breaking it in during the sampling process.
   4. In order to take a wood sample, orient the increment borer by aiming it toward the center of the tree, 90° (perpendicular) to the axis of the trunk.
   5. Push the increment borer into the tree and turn the handle clockwise, simultaneously. This part is essential since the lack of pressure during the initial penetration of the borer bit might cause irregular or broken cores. Once the bit has penetrated completely, relax the pressure and turn the handle until the desired depth is reached (Figure 5A).
   6. Obtain at least two samples per individual to ensure good sample quality. If trees are growing on a slope, take samples parallel to the slope's contour (Figure 4A) to avoid the reaction wood produced by the trees⁷.
      NOTE: In conifers, trees produce reaction wood in the form of wide rings down the slope to keep the tree upright and it is called compression wood. In angiosperms (broadleaf trees) wide rings are produced up the slope and call it tension wood. Reaction wood must be considered to find the center of the tree and avoid non-climatic influences on tree-ring widths.
   7. When the borer has been rotated deep enough to the tree's center (Figure 5B), insert the extractor into the borer and push it towards the center of the tree (Figure 5C).
   8. When the extractor is inserted to its full length, turn the borer slightly counterclockwise to break the connection between the sample and the tree (Figure 5C). Then, remove the extractor carrying the core (Figure 5D,E) and finish the extraction by removing the borer from the tree turning it counterclockwise (Figure 5F).
   9. After taking the sample, pay close attention that the tree forms a seal of resins or sap exudation followed by secondary growth. Under special conditions, the injury might be the pathway for the entry of pathogens that could damage the tree⁶.
   10. When working in restricted areas, for instance, protected natural areas and national parks, consider taking extra measures to protect the sampled trees. Cover the minor injury made by the increment borer with Campeche wax or beeswax.
   11. In case problems occur during fieldwork, such as having wood stuck inside the borer, a broken tip, or having the increment borer stuck on a tree refer to steps 1.2.9.-1.2.12. In addition, take more than one increment borer to the field.
       NOTE: Remember that there is no golden rule to extract the core. Avoid irregularities and attempt to get the best information needed (Figure 4). For further information about taking care of the increment core and sampling, consult Maeglin⁹ and Phipps¹⁰ papers freely available on-line.
   12. Handle the cores with care as they are brittle. Store each sample immediately after extraction. For samples with 5 mm or thinner diameter, place them in plastic straws with perforations or paper straws for better ventilation and to avoid fungal growth (Figure 6A). For samples with a 12 mm diameter, wrap them in newspaper or any other type of paper (Figure 6B).
   13. During fieldwork and transportation to the laboratory, protect the samples and store the samples on a solid plastic tube with plastic caps.
   14. In places where dead trees or stumps are found, extract cross-sections using a chainsaw. This allows samples from both small trees and large trees (Figure 6C).
       NOTE: The objective of this type of sample is to extend the period of the chronology and to help detect missing rings not found on the cores. The missing rings are or will be evident if the whole circumference of the tree is exposed⁶.
   15. For samples taken with the chainsaw and with a certain degree of wood decomposition, it is possible to lose sample fragments. Wrap the samples in plastic to avoid this (Figure 6D,E).

Figure 1: Temperate mixed-conifer forest. (A) Mixed-conifer forest of Pinus montezumae, Pinus arizonica, and Pinus ayacahuite. (B) Mixed-conifer forest of Pseudotsuga menziesii, Pinus arizonica, and Pinus ayacahuite. Please click here to view a larger version of this figure.

Figure 2: Site selection. (A) Forested areas with limiting conditions (shallow, dry soil and a steep slope) with a high probability of finding long-lived individuals. (B) Long-lived individuals are essential for dendroclimatic studies. (C, D, E) Locating and selecting deadwood (stumps, fallen trees, and wood with a certain degree of deterioration) that allows the chronology to be extended in time. Please click here to view a larger version of this figure.

Figure 3: Selection of the best tree specimens. (A) Tree with a dead canopy top and thick branches, characteristic of long-lived individuals, and (B, C) images of trees with twisted stems and branches, that is, in a spiral shape, indicative of long-lived individuals. Please click here to view a larger version of this figure.

Figure 4: Tools used for sample collection. (A) Increment borer (Pressler), the tool to extract dendrochronological samples. (B) A 12 mm diameter borer, recommended for cases where more material is needed to define the tree rings, allowing the extraction of a larger sample volume, which improves the visualization of intricate rings, and facilitates the identification of growth problems. (C) A 5 mm diameter borer used in most cases. This type of borer is used for core sampling. Please click here to view a larger version of this figure.

Figure 5: Sample collection process. (A) Orient the drill pointing to the center of the trunk, positioned at a 90° angle, perpendicular to the axis of the trunk, simultaneously push the borer towards the tree and turn clockwise. (B) When the borer has been inserted 1 inch deep, keep turning clockwise to reach the center of the trunk, the extractor is inserted into the inner cylinder of the borer. (C) When the extractor is inserted to its full length, rotate the borer one turn counterclockwise to break the connection between the sample and the tree. (D, E) Wood sample extraction. (F) The borer is removed from the trunk by turning counterclockwise. Please click here to view a larger version of this figure.

Figure 6: Techniques to protect wood samples. Because the samples can be fragile, each sample must be stored properly after being collected. (A) The samples taken with the 5 mm diameter borer are placed in plastic straws with perforations or paper straws. The perforations allow better ventilation and prevents fungal growth. (B) The 12 mm diameter specimens are firmer. These samples are wrapped in newspaper or other paper type or manila envelopes. (C) When collecting cross-sections with a chainsaw (D, E), they should be wrapped in plastic to provide further support and avoid fragments being lost during transport. Please click here to view a larger version of this figure.

2. Sample preparation in the laboratory

1. Follow the standard procedures indicated by Stokes and Smiley⁶ for preparation and dating of samples in the laboratory.
2. Let the samples dry in shade so that the loss of moisture from the wood gradually minimizes wood deformations (Figure 7A). After the cores have lost enough moisture, mount them on wooden mounts or rails, fix with glue (Figure 7B) and fasten them with tape or thread (Figure 7C,D).
3. Pay attention to the orientation of the wooden cores when placing them on the mounts. Fix the cores such as the xylem cells of the wood, which are oriented perpendicular to the plane, are observed and surfaced (Figure 7E). This orientation allows clear visualization of the wood anatomy of the tree rings.
4. Sand and polish the samples using sandpaper of different grits, ranging between 120 to 1200 grit. In cross-sections that can show significant surface irregularities, follow one of two possible options.
   1. Option 1: Work with an electric brush and later sand the sample. Option 2: Begin the sanding process with a coarser sandpaper grit, in the range of 30 and gradually increase the grit to 1200. This will allow the growth rings to be seen and differentiated more easily (Figure 7F,G).
5. Polish the entire upper part of the sample (Figure 7E). Polish a minimum of 30% and a maximum of 50% of the sample part opposite to the section glued to the wooden rack. This will allow to have enough portion of wood for later polishing processes with the objective of greater clarity of the rings, erase points and marks that are placed during the dating process.

Figure 7: Preparation of sample. (A) Drying samples in the shade ensures that the loss of moisture is gradual to minimize the deformation of the wood (twisted cores). (B) Example of how to mount samples on a wooden rack, fixed with glue, and (C, D) show how they are attached to the trim with tape or thin rope. (E) Indicates the correct position of the wood fibers, which must be oriented perpendicular to the growth rings. This orientation will allow clear visualization of the anatomy of the growth rings. (F, G) It is an example of the quality of sanding and polishing using sandpaper grits from 120 to 1200. This procedure allows to visualize and differentiate the growth rings. Please click here to view a larger version of this figure.

3. Tree-ring dating

1. After the samples have been polished, analyze each core under a stereoscope at 10x to 15x magnification. Consider a stereoscope that allows observing and comparing several growth rings at the same time.
2. Once the researcher has a good idea of what a growth ring is, depending on the species used, count the growth rings of each sample. This step will provide a tree age approximation. Additionally, recognize the type of variations that might be encountered during the cross-dating process (Figure 8A). For tree ring counting, start from the inner ring (center of the tree) to the outer ring (bark).
3. Make small marks on the sample to go back and revisit the sample knowing the place in time. Place a tiny dot for each decade, two dots for every fifty-year segment, and three dots for every hundred years (Figure 8A).
4. Use other types of marks to highlight rings that have particular characteristics. For example, when a micro ring with only a tiny part of the growth band is evident, use two parallel points to mark them. When there is a suspicion or certainty of the absence of a ring, use two alternate points or a circle to mark it, and when a false ring is identified, use a diagonal line to indicate that it is a single ring.
   NOTE: For more details on the counting technique, follow the standard procedures indicated by Stokes and Smiley⁶.
5. Once the tree rings are counted, use growth graphs or skeleton plots to compare the temporal patterns and variability between wide and narrow rings. This graphical part allows comparing several samples simultaneously and determining common and synchronized growth patterns (Figure 8B). This technique allows detecting growth discrepancies that could have been mistakenly marked when counting the rings.
   NOTE: Please see reference⁶ and the link below for more details on making a skeleton plot: https://www.ltrr.arizona.edu/skeletonplot/plotting.htm.
6. In samples of young living trees, where the date of the last outer ring (besides the bark) is known, perform a preliminary tree ring dating directly on the sample. For example, if the sample was collected from a forest in northern hemisphere in December 2021, which is the end of the growing season and tree ring formation is completed, the date of the last fully formed ring will be 2021. Using this, count the rings from the outer part (bark) to the center of the sample.
7. The samples from older trees show periods of narrower rings, generally at the outermost part of the core. Generate a skeleton plot for these cores (Figure 8C) to compare their growth pattern with a known well-dated sample or with a previous regional ring-width master chronology (Figure 8D).
8. To compare the sample look for the synchrony between thin and wide rings between different trees (Figure 8A,B). The sample is considered dated when successful match is found based on the cross-dating technique.
9. In samples where growth synchrony patterns are not clear, due to differences in growth variability, absent rings, or false rings, detect the problem by reviewing ring by ring between the samples and compare it with perfectly dated samples. Use climate records from nearby stations to verify suspicious missing rings, since this kind of ring anomaly occurs in years with extreme dry or cold conditions.
10. After potential problems are identified, correct the count and test if the synchrony is achieved.
11. After all living trees are cross-dated, develop an average growth chart commonly called Master Chronology (Figure 8D), which is the average of all dated growth plots and indicates the growth pattern of the site on a time-domain⁶. It is useful as a dating tool for more samples that need to be cross dated, like dead trees with unknown date of death (Figure 8C).

4. Measuring the tree-ring

1. Once the samples have been dated, measure tree ring widths. Measure total ring width as well as interannual band – earlywood and latewood- widths if possible. Use a measuring system with a precision of 0.001 mm¹¹ to perform these measurements (Figure 8E).
2. Measure the growth rings and partial rings one by one by sliding the measuring system stage and observing the sample through a stereoscope with a cross-linked eyepiece. Depending on the measuring system, begin the measurement with the innermost ring to the outer ring (Figure 8F).
3. If a mechanical measurement system is not available, in that case, use a scanner to take high-resolution images and perform tree-ring measurements using a specialized software such as CooRecorder or R measuring from CRAN.

Figure 8: Cross-dating and tree-ring measurement. (A) Shows ring count and growth pattern comparison between two samples. (B) An example of how the growth variability of both samples is reflected in paper graphs (skeleton plot). This type of graph allows comparisons between the growths of many samples simultaneously (cross-dating) and is an essential techniques for achieving the correct dating. The marks at the top of the graph 0, 50, 60, etc., indicate the number of rings counted in sample shown in A. (C) Skeleton plot of a dead wood sample dated to the exact year using the master chronology. (D) Example of a master chronology, average of correctly dated living trees. (E) A measurement system with a precision of 0.001 mm was used to measure each of the annual growths. (F) Schematic showing the annual growth in Pinus lumholtzii and the three different band portions of an annual ring (total ring, earlywood, and latewood). Please click here to view a larger version of this figure.

5. Verification of cross-dating

1. Once the ring widths have been measured, test them for their dating accuracy and quality. Use the software COFECHA¹² (https://ltrr.arizona.edu/research/software) and dplR¹³ for the statistical verification of cross-dating. Identify not significant correlations (< 0.3281; p > 0.01) between segments of the ring series amongst a master chronology built using the same samples in COFECHA analysis software.
2. Look for flags in the software which identify segment correlation values that are not significant, making it easy to identify the potential discrepancies between the segments of any sample with the overall master chronology (the average of all standardized values of each sample analyzed).
3. The discrepancies may be related to errors due to measuring or ring identification attributed to specific microsite conditions for selected trees, which is not synchronized with the overall variability among the rest of the samples. Verify these with observations and notes taken in the field and decide whether to conserve or eliminate this sample.
4. For more details on the interpretation of the COFECHA statistics, see Speer⁷.

6. Chronology development

1. Detrend or standardize the tree-rings measurements to remove all the non-climatic information (noise), such as the age, tree geometry, stand dynamics and disturbance effects as described.
   1. Fit a mathematical equation to the sample data-negative exponential (Figure 9A), straight line (Figure 9C), or cubic splines- depending on the criteria and temporal trends found on the samples (Figure 9A). Then divide each measured ring width by its fitted or expected value.
   2. Average the standardized values of individual trees together into a mean-value function and adjust for differential growth rates due to differing tree ages and differences in the overall growth rate. This will generate a standardized time series with a relatively constant variance and a mean equal to one⁸ (Figure 9B,D)
2. There is not a perfect recipe to standardize tree ring width series; perform a graphical inspection of all the ring-width measurements to identify the embedded trends in the samples before applying any detrending method.
3. Use any statistical platform to achieve standardization. Software like ARSTAN or dplR are especially for this type of analysis^13,18 and are freely available at https://ltrr.arizona.edu/research/software and as an R package from CRAN.
4. Perform autocorrelation removal using the statistical procedure called autoregressive moving average modelling (ARMA modelling) which is automatically applied in the two programs already mentioned. This is required to study the effect of interannual climatic variability on the tree rings.
5. Once the tree-ring measurements have been standardized and autocorrelation has been removed, develop the site chronology (Figure 10A). The tree ring site chronology is the average of all standardized series using a robust bi-weighted mean that, unlike the arithmetic mean, attenuates the influence of atypical years (outliers).
6. Assess the quality and the climate reconstruction potential using the three key statistical indicators from the site chronology generated by ARSTAN or dplR, namely the expressed population signal (EPS), the mean sensitivity (MS), and the intercorrelation between series (ISC).
7. Estimate the degree of similarity between the different samples used in the chronology and a hypothetical chronology with an infinite number of samples by using the EPS (Figure 10B). A value greater than 0.85 is considered acceptable and suggests that the chronology has a sufficient number of samples to express the common signal of a given site¹⁹.
8. Measure the relative variability between ring widths by using the MS. The value of the mean sensitivity ranges from zero to two, with zero values meaning there is no difference between two adjacent rings and two meaning that a ring has a zero value next to a ring where the value is greater than zero³. A mean sensitivity value greater than 0.3 indicates sufficient interannual variation and potential for climatic reconstruction.
   NOTE: The mean sensitivity can be interpreted as a metric of the potential relationship between tree growth and climate.
9. Calculate the mean Pearson correlation coefficient of each sample against its master chronology produced from all the other time series at the site using the ISC. This statistic indicates the common signal of tree growth among trees.

Figure 9: Examples of detrending and standardization procedures of tree-ring width measurements (RW), from measurements to indices. The standardization to a ring width index (RWI) is calculated, so the mean is around one and has a homogeneous variance. (A) Ring width series RW indicates the exponential decrease in growth due to the age effect, the detrending curve of best fit is applied, and in this example, we use a negative exponential curve (red color). (C) This is a second example of a straight line (red color). (B, D) Normalized indices (RWI) are generated after dividing the value of the curve by the RW series. This division eliminates the trends fitted with the curve, maximizing the climatic signal (time series in gray color) and a 20 year smoothing spline (red color) to observe low-frequency events such as droughts and wet periods. Please click here to view a larger version of this figure.

7. Monthly correlation analysis

1. Perform correlation analysis to identify the relationship between annual tree growth (site chronology) and monthly climatic variables (precipitation, temperature, evaporation, relative humidity, among others). Use this analysis to assesses the period to reconstruct the tree ring based on the highest correlations values between seasonal climatic variables and the site chronology.
2. Carry out the correlation analysis with the previous and current year monthly climatic records (Figure 11A). Use one of the several programs available to run this type of analysis, see references ^3,20,13.
   NOTE: The analysis at a monthly to seasonal resolution indicates the degree of association, using Pearson correlation coefficients, between climatic variables and the tree ring site chronology.
3. Use the climate records from the nearest meteorological stations or the nearest grid point from a gridded dataset for performing the correlation analysis. Additionally, assess the records' quality and use the best available option.
4. When performing correlation analysis on multiple meteorological stations, explore each regional record one by one before compiling one regional climate record. Use only those stations that display the highest correlation values with the tree-ring chronology.
5. Compare the regional average with the best available gridded dataset to assess and avoid losing climate variability when combining multiple meteorological stations.

8. Simple linear regression model and reconstruction of the climatic variable

1. Once the seasonal period that shows the strongest climate-growth is identified (Figure 11B and Figure 12A), perform a simple or multiple linear regression analysis to build up the reconstruction model (Figure 12B).
2. Perform this procedure on an extensive range of monthly combinations to obtain the best reconstruction model (the one with the highest explanatory power, adjusted R² value). In this analysis, consider the tree-ring chronology index as the independent variable, and the precipitation for a seasonal accumulated monthly period as dependent variable.
3. After the regression model has been generated, apply it to the chronology in the common period of the observed data.
4. Subsequently, divide the common period of observed and reconstructed data into two periods each one containing half the data used in the entire common regression model to statistically validate the model and perform a calibration and verification test.
5. Determine the following statistical variables to verify the statistical predictive power and uncertainty of the regression model (see Discussion for a detailed description): Correlation coefficient (r), adjusted R², reduction of error (RE), signs test, paired sample t-test, standard error of estimate (SE), root-mean-square error of validation (RMSEv), and Durbin-Watson test.
6. Once the regression model has been statistically validated, use it to reconstruct the climatic variable of response using the tree ring chronology.
7. Finally, to provide additional reliability and certainty to any climate reconstruction, verify the reconstruction with historical documented records or other dendroclimatic reconstructions from nearby locations.