Tree Core Analysis with X-ray Computed Tomography

1. Core sampling

1. Sample the tree with a Pressler borer. Consult references on how to core a tree manually³⁹. This toolchain is presented for 5.15 mm cores.
2. Put the tree cores unglued into 6 mm paper straws or in any other receptor that does not require glue. Do not glue the tree cores on a wooden support. If the cores were already mounted, unmount them with scalpel/saw or solvent depending on the glue type.
3. Avoid wrapping them in plastic tubes, try to dry them first to avoid mold growth and fungal decay.
4. Use pencil to write on the paper straws, since the extraction could fade marker writings.
   NOTE: This toolchain is also suited for 4 mm cores. Already mounted and sanded cores can be used, but these need to be unmounted first. For archeological or remnant wood, a core-sized sample needs to be made (by sawing/reducing the size) for scanning.

2. Core preparation before scanning

1. Perform an ethanol/toluene or ethanol or acetone⁴⁰ Soxhlet extraction for at least 24 h (Figure 4A,B). This is necessary to ensure a resin-free wood volume to have an accurate density calculation (resin attenuates the signal, Figure 4C^{13, 41}).
   1. Make a solution of 1000 mL of ethanol with 427 mL of toluene (according to e.g., ASTM D 1107 – 96) or pure ethanol or acetone. Many universities have a lab glass workshop that can make an adapted version of the Soxhlet apparatus for tree cores (Figure 4B).
   2. Fill the round-bottomed flask with the solvent, and put the samples within the straws in a stainless-steel support in the Soxhlet apparatus with cooler on a laboratory heater. Allow for at least 6 siphonings per h.
      NOTE: A closed water circuit is preferably used for cooling (or any other modern cooling system). In this case, a water pump circulates the water, stored in a 100 L basin with chlorine tablets (to avoid algae development), at 120 L/h (but this will depend on the size of the Soxhlet apparatus as well as the applied heat by the heat element). An air cooler is used to cool the water in the 100 L basin. To save on solvent, glass beads can be used to fill-up the void volume of the Soxhlet apparatus.
   3. Check the temperature and number of siphonings regularly. Dry the cores under a chemical hood (check lab policy for this) to remove the solvent residue or put a drying oven under the chemical hood.
2. Dry the cores for 24 h in a drying oven (103.5 °C) under exhaust ventilation and perform a hot water bath (90 °C) or hot water Soxhlet extraction for 24 h as described above.
   1. Keep the samples in the paper straws in the steel sample holders. Dry again in a dry oven for 24 h at 103.5 °C, then condition to 60% relative humidity (RH) before scanning.
      NOTE: The 60% value is chosen in this case because these are approximately the ambient average relative humidity conditions in the X-ray CT scanning room. Oven dry samples can also be used. The most important factor is that all samples are scanned under the same conditions.
   2. In case the water extraction degenerates the paper straws, put the tree cores back into new 6 mm paper straws²¹. Both 5.15 mm and 4 mm cores can be put into 6 mm paper straws.
   3. Make sure the entire paper straw is filled with wood. Wrap both straw-ends up and cut the ends. This enables easy insertion of the straws in the cylindrical holder.
   4. Make sure the cambium (bark) side is clearly indicated on the core because the CoreProcessor and RingIndicator assume that the bark side is down. Remove bits and pieces as well as cores with cracks as these are difficult to work with.

Figure 4. Workflow for extracting and drying cores prior to scanning. (A) Increment cores in paper straws are first put in a hot water bath and subsequently in a hot ethanol-toluene mixture with a Soxhlet apparatus for 24 h then dried, put in a hot water bath for 24 h, then dried again at 103.5 °C, then conditioned before scanning. (B) Images of the adapted Soxhlet set-up at Ghent University. Multiple adapted Soxhlet apparatuses are serially attached. Note the aluminum foil and insulation tubes to keep the solvent or solvent mixture warm enough for extraction. (C) Example of the surface of an increment core from Pinus longaeva before and after extraction. Resins and other extractives mask the true density signal, due to attenuation of the X-rays. Please click here to view a larger version of this figure.

3. Core scanning

1. Select the proper sample holder type (Figure 5) as described below, depending on the research purpose as described in Figure 1.
   1. For inter-ring scale purposes, use holder type1 – cardboard and perforated floral foam.
   2. For the MXD data use holder type 2 – a plastic sample holder that holds 6 cores of 1-15 cm length and has a 3 cm diameter.
   3. Use holder type 3 – for dendro-anatomy purposes that consists of simply tying paper straws together.
2. Load the tree cores in paper straws into the sample holder and fill in the spreadsheet template that can be downloaded on https://dendrochronomics.ugent.be/#software. Each number corresponds to a given position in the sample holder.
3. Scanning in an X-ray µCT facility: consult an X-ray CT expert for the proper settings and scanning protocol. For the scanner used in this protocol, see³⁷.
4. Reconstruction of the scanned images: consult an X-ray CT expert for the proper settings of the reconstruction, or the settings that come with the CT scanner (software package of the X-ray CT scanner). For reconstruction parameters for this protocol, see³⁷.

Figure 5. Sample holder types and resolution. Design of the three main holder types, that correspond to (A) inter-ring scale, (B) tree ring scale, and (C) anatomical scale. Due to the resulting size, the number of samples decreases with increasing resolution requirement. The corresponding 3D renderings for (D) inter-ring scale (holder 1), (E) ring width and MXD (holder type 2), and (F) anatomical parameters (holder type 3). Scale bar = 5 mm. Please click here to view a larger version of this figure.

4. Obtain TRW data, density data and anatomical data

1. Getting started with the software.
   1. To obtain density values (holder type 1 and 2), install the following three software packages: (i) CoreProcessor, (ii) RingIndicator and (iii) CoreComparison toolboxes from
   2. Since these toolboxes were made in MATLAB, install the correct MATLAB compiler runtime (MCR), available on: http://nl.mathworks.com/products/compiler/mcr/. The version needed currently is MATLAB2022B. The MCR is free, just as the 3 software packages.
   3. Next to the folder of the reconstructed volumes, make a new empty folder for the MAT files (a file format from MATLAB that stores variables, etc.), e.g., matfiles_useful_name_sample_holder, this is necessary to store temporary parameters.
   4. Make sure the spreadsheet is in the folder of the reconstructed volumes, in order to fill in the samples in a later step.
2. Preprocessing the core volumes with the CoreProcessor.
   NOTE: For high-resolution anatomy data, following processing steps are not necessary (Figure 1), once the reconstructed volumes are available, refer to the ImageJ guidelines for 3D images⁴². Segmentation can be done through the Weka plugin⁴³ for indicating vessels on individual slices for instance¹², or in software plugins such as ROXAS⁴⁴.
   1. Prepare the core extraction: Prepare the automated extraction of the increment cores (depending on inter-ring scale Figure 6A or ring scale Figure 6B) as described below.
      1. Select the folder with reconstructed files (16-bit cross-sectional TIFF slices) and the spreadsheet file (XrayForm.xlsx).
      2. Select the empty folder from 4.1.3. This is the folder where CoreProcessor will create a MAT file that will be used in 4.2.2. for mass core extraction.
      3. Select the newly created empty MAT file folder or, if not created, select any other folder. The volume is loaded, and a longitudinal cross-section of the cylinder appears, check a few horizontal slices to ensure if all cores and references were correctly mounted into the cylinder.
      4. Click from bottom to top to select slices to evaluate this. These cross-sections will be used for the next step, so click across the entire length. Indicate the dark (air) and white references (Figure 6), as many times as indicated in the spreadsheet file. Double click after each selection.
      5. For sample holder 1, inspect the multiscreen with 9 cross-sections, generated through the reference material to check whether the density calculation was performed on the wood (and not on the paper straw or other non-wood material). This is essential for correct conversion to actual wood density values.
      6. In a new pop-up window, select every core separately by drawing a circle/ellipse around it. Perform labelling based on the information from the spreadsheet file, check whether the name is correct. Respect the order in which the template is filled. All info is stored in a MAT file, that will be used during the next step for extracting all cores at once.
   2. Click on Mass Core Extraction to carry out core extraction for all the cores in a given cylinder (that were indicated in 4.2.1.2), given that the metadata and spreadsheet files are located in the same folder.
      NOTE: When using Mass core extraction, all cores will be extracted according to the labelling of the user (that are stored in the mat file folder). As a result of this operation, a new folder named Extracted will be created which contains all the separate cores. When multiple sample holders are prepared in the first step, and all the MAT files are stored in the same folder, all MAT files in this folder will be processed.
   3. In the CoreProcessor toolbox, click on Manual Tg Correction, then select the Extracted folder to ensure proper orientation of the transversal and radial plane of every core volume, similar to mounting physical samples in a sample holder before sanding keeping the transversal plane in view (Figure 7B).
      1. Some cross-sections are displayed so that the user can see the grain direction. In the last plot (bottom right), inspect the slice that is presented and draw a line to indicate the grain direction. Double click to automatically rotate the core.
      2. Then a screen appears, here crop the volume of the core (Figure 7C), making sure that the rectangle encompasses the entire core diameter. This is necessary because the central point of the rectangle should be as close as possible to the core axis (for correct alignment in the RingIndicator module).
         NOTE: If scanning at high resolution, select only material within the core at this stage and use the converter to switch to sample holder type 2. This is certainly helpful for ring-porous species for example.
   4. Perform the following optional steps in case of misaligned, flipped, or broken cores or incompatible image formats.
      1. Core tilt correction (only for sample holder 1): To automatically correct for core tilt (e.g., when the core is slightly inclined in the cylinder; Figure 7A). This is important for later steps. Select the Extracted folder created in the previous step.
         NOTE: This is normally not needed if you have a sample holder type 2 or type 3 (Figure 5) or when all samples and sample holders are put perfectly upright.
      2. Flip volumes: In case the sample was not put with the bark side down (the pith-bark direction is not left to right), change the direction by flipping cores. The function asks for a folder and will create a new folder.
      3. Glue cores: Stitch large parts of individual cores together virtually. The idea is not to stitch perfectly, but to make sure that, if needed, single cores are in the same multipage TIFF volume. Use the broken core indication (see step 4.3.4.9) to overcome the spaces in between the core pieces.
      4. Converter: If the volumes have been manipulated in another software package and are saved outside the toolbox (e.g., ImageJ), perform this step to get the header info correct. Use this step for 16-bit multipage TIFF volumes coming from another CT scanner as well and for using the RingIndicator toolbox.
         ​NOTE: All steps from here are for structure direction corrections and finally, densitometry and tree ring analysis.
3. Correcting the angles and indicating rings in RingIndicator.
   NOTE: Following steps need to be taken chronologically in the RingIndicator module: (i) visualization of the transversal and radial plane, (ii) structure correction, (iii) creating the density profile, and (iv) indicating the rings manually or semi-automatically.
   1. Perform visualization of transversal and radial plane as described below.
      1. To get familiar with the GUI, open the toolbox and select a multipage TIFF File (one that was created after extraction and after tg corrected, so from the folder name Tg corrected).
      2. Select the Half Thickness of Slice Image to display an average image of the central X slices, with X ranging from 1-10 with 1 being no averaging (i.e. only display the central slice), and 10 being an average image of the 10 central slices.
         NOTE: Averaging can be beneficial to highlight ring borders better, and suppress noise, on the other hand it can also make ring borders look less distinct, especially when the fiber angle is substantial. Please tune this value according to the needs. It is important to mention that this is merely for displaying, it does not impact the underlying volume, nor does it impact the eventual calculations/results.
      3. Inspect the graphical user interface (GUI). The GUI contains a top figure with the transversal plane of the core and a bottom figure with the radial plane of the core, and a menu (Figure 8).
      4. To load a new core volume, search for the Volume > Load Volume function to select a new multipage TIFF file in the directory. Now the user is ready to indicate the structure direction to subsequently indicate the actual tree rings.
   2. Structure correction to facilitate automatic tree ring indications.
      1. Manually click on the image to insert green bars for structure correction. Start with the pith (or oldest ring available) and end with the most recent ring. Along the core correct for structural deviations on both the transversal and the radial plane.
      2. Make sure to indicate the ring and fiber angle properly, because all subsequent indications will take an interpolated angle between the respective indications.
      3. Insert a green bar on one plane, it will automatically generate one on the other plane. After inserting the bar, move the nodes (white squares) at the end of the bar to change the angle. Use the middle node to change the bar position. Adjust the size of the nodes for ease of use.
      4. After every step, make sure to press Data > Export and Rings > Export Rings, to ensure that ring and fiber indications are written to the proper .txt files. Throughout all corrections, make sure to always perform these two actions.
   3. Calculating the density profile.
      1. Use the Densitometry > Densitometry plot to calculate the density profile. Create and plot a density profile by selecting Overlay Plotting > Plot Density Profile, that can serve as an overlay on the core planes.
      2. For MXD calculation, indicate a percentage (recommended is 20%, e.g., 20% means that only 20% of the highest density values are included), to exclude lower values from resin ducts, earlywood etc. For mean density values use 100%. For minimum density also use 100%.
   4. Perform manual and/or automatic indication of ring boundaries using one left mouse click per ring as described below.
      NOTE: Insert the same green bars as in step 4.3.2.1., this time the user should do this now ring per ring (Figure 9).
      1. Use Overlay Plotting > Plot Density Profile, to visualize the density profile on the image. It is best to indicate the ring boundary also in conjunction with this profile.
         NOTE: Open as many instances from the toolboxes as needed, in combination with the CoreComparison toolbox (see further).
      2. For automatic indications (recommended for conifers for instance), use Auto > Max/Min/Inflection detection. This function will automatically indicate tree ring boundaries based on the densitometry profile created in the previous step.For automatic indications (works best for boreal conifers or species with clear ring boundaries defined by the density profile), use Auto > Max/Min/Inflection detection. This function will automatically indicate tree ring boundaries based on the densitometry profile created in the previous step.
      3. Depending on the species (diffuse-porous/ring-porous or coniferous), use the maximum/minimum, or curve inflection (right after the density peak) as the tree ring boundary criterion. The first parameter is a threshold for the peak function, the second is a smoother function.
      4. After performing the automated ring indications, perform automatic shifting (with window size of 1 pixel and user-defined iterations) , while making sure that no ring indications are plotted on each other due to this shift and re-run the densitometry plot.
      5. Change the felling date to the year at which the increment cores were taken on living trees or any other date that is suitable. The default is set at 1900.
      6. In Overlay Plotting, select Plot Rings and the years are displayed. Also select the plane on which to plot rings and/or the density profile.
      7. Once there are two cores fully indicated (e.g. from the same tree), do pairwise comparisons of core ring width curves by opening the CoreComparison toolbox. Open one RingIndicator instance per core and edit the indications if necessary. Export them again and arry out this iterative and interactive approach to obtain good quality data. In the next step (step 4.4.1.), do this for all cores to export both the TRW data and the corresponding density values.
      8. Check double-ring positions in the Rings > Check Rings option; this will check if any indications have a position outside the images as well as rings that are overlapping, e.g., in case if a double click has happened twice on the same position.
      9. For broken cores that have bits and pieces, start at the pith, and demarcate anomalies (air due to broken cores, bright spots due to resin pockets, contamination, and decayed parts) as if they were rings to delete them afterwards. End with the tree ring boundary of the last formed ring and export the data. Try to avoid broken cores during sampling (see step 1.1. and³⁹).
         NOTE: Both missing as well as broken rings can be indicated, see Figure 10. Two cases are defined, one where the core breaks in the middle of a ring, and one where the core breaks on a ring boundary. For the first one (Figure 10A), indicate the gap, note the number of that gap followed by number 1 and close the brackets. Then the program will sum the two parts of the ring to obtain a correct TRW value. In the second case, the gap number is indicated followed by number 2, and the program will omit this section (Figure 10B).
      10. By indicating rings and generating density profiles, several .txt files are created and stored in the same folder as the multipage TIFF files, and the most important ones are listed in Table 1. Keep these .txt files in the same folder as the TIFF images, in order to further treat them in the CoreComparison module (step 5.4).
          NOTE: For calculating density values in a different way based on the raw density profile (for instance latewood width vs earlywood width⁴⁵ or other metrics not supported by the CoreComparison and CoreProcessor toolboxes), use the zpos_corr and density_corr text files and use these for such calculations.
4. For comparing TRW series and exporting the data via CoreComparison , follow the steps below.
   NOTE: When using this toolbox, two windows are opened simultaneously, the CoreComparison and the PatternMatching module. The CoreComparison deals with cross dating and exporting of TRW data, whereas the PatternMatching module can be used for exporting density profiles and density values (MXD, minimum density (MND), mean density, and 4 quartile values). The main goal of CoreComparison is to generate the TRW values and density values, and to have a rough estimate of cross dating quality, with a coupling to RindIndicator in case of missing rings or wrong indications.
   1. Perform the following steps for cross dating or matching the ring width pattern with other series.
      1. Once the RingIndicator section is completed, select the text files to compare, and a screen will open with ring widths, together with cross dating or statistical parameters such as Gleichlaüfigkeit (GLK)⁴⁶ and or Spearman correlation between individual series (Figure 11).
      2. To make changes in the ring indications (for instance during cross dating), re-open RingIndicator, import data, export the modified data and rings, and use the Refresh button in the CoreComparison module (see top left-hand side) to see the changes.
      3. Open multiple RingIndicator instances to make changes to different tree-ring series and use multiple screen monitors to do this.
      4. In the GLK and CORR menu (Figure 11), perform a series of actions related to GLK, such as thresholding the figures, exporting the GLK values to a spreadsheet and ranking them.
   2. Obtain TRW values from CoreComparison as described below.
      1. Use the plotting and exporting function in CoreComparison to change the offset between the curves (a larger value result in a smaller offset). Visualize the TRW data in a plot by selecting Plotting and Exporting > Ring width.
      2. Export the TRW data by clicking Plotting and Exporting > Export RW Data. Export ring widths in spreadsheet format or in Tucson format.
   3. Obtain the density profiles from the PatternMatching module as described below.
      1. To obtain the density profiles in spreadsheet, go to the PatternMatching module, and after loading in the profiles, in the menu 기타 select Plotting > Export Corrected Density Profiles.
      2. To batch process profiles or to assess the data using different methods, directly use the .txt files from Table 1 (see step 4.3.4.11).
   4. Obtain MXD, MND, and other density values from the PatternMatching module as described below.
      1. For obtaining the mean density, MXD, MND and quartile data per tree ring, use Other Plotting > Export Clustered Data. A spreadsheet will be generated and can be found in the same folder as the TIFF files.
      2. The tabs in the spreadsheet are detailed in Table 2. Export these tree ring variables to Rstudio or other software for further analysis.

Figure 6. Obtaining wood density values from grey values. Conversion of the grey values of the image to actual wood density estimates, through calibration with a reference material. (A) Sample holder 1, at 60 µm resolution, showing the air reference, white reference, and the core. (B) Sample holder, at 20 µm, air, reference and core are shown as well. Please click here to view a larger version of this figure.

Figure 7. Pre-processing steps of the 3D core volume. Pre-processing steps of the 3D core volume before visualization and indicating the tree rings. (A) Tilt (axial) correction, (B) tangential correction, with the grey bars indicating the fiber direction, (C) the cropping of the core volume to obtain a volume with only woody voxels. Please click here to view a larger version of this figure.

Figure 8. Screenshot of a core. Screenshot of a core (tangential width approximately 3 mm, resolution µm) opened in the RingIndicator toolbox. Top pane is the transversal view, bottom pane is the radial view. Green bars show the structure indication (in this case on the ring boundary), red line is the density profile. The yellow numbers indicate calendar year, and the blue numbers the generic ring number (counting from the first indication). Please click here to view a larger version of this figure.

Figure 9. Ring indication and density profile visualization in the RingIndicator module. Green lines are indications, red line is the wood density profile. Please click here to view a larger version of this figure.

Figure 10. How to deal with cracks in the cores. (A) When a crack occurs in the middle of the ring, the generic ring number (starting from the pith) will be labeled as a gap, and TRW will be calculated by summing the widths of the two ring parts. Parameters are 3 (number of the ring) and 1 (case 1: crack in the middle of a ring)³¹. (B) When a crack occurs at a ring boundary, the gap will be omitted from the TRW calculation. Parameters are 3 (number of the ring) and 2 (case 2: crack at the ring boundary)³¹. Please click here to view a larger version of this figure.

Figure 11. Screenshot of the Core Comparison module. Screenshot of the Core Comparison module, where ring width series are distributed vertically for easy visual comparison. The top pane shows the ring width of 4 cores from Fagus sylvatica, the bottom left pane shows the Gleichlaüfigkeit (GLK) and the bottom right pane the Spearman correlation coefficient. Please click here to view a larger version of this figure.

Table 1. Text files generated by RingIndicator. List of text files that are generated after indicating rings and exporting the density profile in RingIndicator. These should remain in the same folder as the .tiff file in order to be opened in the CoreComparison module.

Table 2. Spreadsheet details. Tree-ring variables that are exported into a spreadsheet named export clustered data.