Manually reconstructing clonal plant networks in the field
1) First select appropriate candidate plants, take a reference picture before cutting of biomass (Figure 1a).
2) Cut off all shoots at the base (collect plant material if interested) and place a labelled pin in each shoot (Figure 1b).
3) Deploy the calibration frame (which holds a spatial scale for translating pixels to centimetres) as level as possible. Take a photo from an appropriate distance (position the camera parallel to measurement frame to prevent measurement inaccuracies) so all four corners are visible from the image. If you are unable to shoot the whole frame in one image, than take 2 or 4 images from the separate corners that can later be stitched together (test.png in Supplementary file includes an example of the latter).
4) Either draw the spatial shoot organization in the grid with the labels or use a field proof tablet or computer to assign labels to each individual shoot in the picture (Figure 1c).
5) Start digging out the plant and note the nearest connection/neighbour of each shoot in the clonal network (Figure 1d). This step requires some practising and experience. Please be careful when excavating the plants, depending on the species the rhizomes can disintegrate rapidly. When parts of the plants have all been noted down, you can remove them from the clonal plant to avoid confusion. Use your fingers when branching nodes (where shoots and rhizomes sprout from) are closely spaced apart. They may appear to be coming from the same point, but sometimes you can still feel a tiny piece of rhizome connecting the nodes.
6) After the plant has been excavated and all shoots in the network are connected, remove all labelled pins and replace the sand on the plant; it is likely to sprout again!
7) Computer work: Load the images into the computer in Photoshop or equivalent software. Straighten the image into a square gird using a perspective crop tool (Figure 1c).
8) Use Image J or an equivalent program to set the scale of the photos (use the scale on the images to translate pixels to centimetres). Using Image J, you can now measure the length between connected shoots or measure the reorientation angle.
9) An example of a fully labelled image including the connections found in the field is included in the Supplementary file (connected.jpg).
Reconstructing clonal networks: the automated approach
Field methods:
This method was validated for two dune grass species (Ammophila arenariaand Ammophila breviligulata) only (see associated publication).
Step 1-3 need to be performed as previously described, with the exception of the labelled pins. This method requires coloured pins that can be extinguished from still images, but they don’t need to be labelled. After the image is taken, the plants need to be excavated to make sure all shoots are connected. However, the precise inter-shoot connections do not need to be noted down. Instead note down which clusters of shoots are connected to one another.
Computer methods
· Repeat step 7 from the manual approach and straighten the image.
· To make sure all pins are recognized in the Matlab script it is recommended to highlight all individual pins in Photoshop. Use different layers for each individual plant in the image and store the image as a .png image of a fixed size: for example 3600 x 3600 pixels. To easily recognize the pins, use the same colours (for example the shoots of plant 1 in the image get red pins (R-G-B :255-0-0) and the shoots of plant 2 get blue pins (R-G-B:0-0-255). The custom-made Matlab script (clonal_plant_analyses.m) uses the function ..._pins.m to recognize pins, functions for red (red_pins.m) and blue (blue_pins.m) are included. To make sure the script (clonal_plant_analyses.m) can recognize the pins it is recommended to create dots of sufficient size (for example 15 pixels).
· Open the readme.text from the ZIP package in Supplementary file 1. First use the script on the test.png image to understand the code. Read the annotations and change the plant parameter to whichever colour of pins you want to read (red = 1, blue = 2) in from your image. Note: you can change the colour of the pins yourself by creating new functions in matlab. Just copy the …_pins.m function and change the values in the RGB channels.
· The code saves the spatial coordinates for all shoots of an individual plant in a csv file (xy_plant…csv). The unit of these coordinates is in pixels, it and can be translated to a spatial dimension (e.g. cm or mm) using the calibration scale from your images. These spatial coordinates can be used for connecting the shoots (set the connect parameter in the clonal_plant_analyses.m to 1) or to perform other spatial statistics (see the associated publication for examples).
· The script also connects all shoots within a clonal individual using two connecting algorithms: Nearest Neighbour search (NN) and Traveling Salesman (TS) (set connect to 1). These methods create a singular path through the clonal nodes (no branching probability is included), which results in a possible step size distribution which can be used as a proxy for clonal expansion strategy. The first connecting algorithm tested is based on the Travelling Salesman principle (TS). The Travelling Salesman in a classical NP (non-deterministic polynomial time) problem from computer science that deals with computing the shortest possible route given N number of cities in which every city has to be visited once 5. In our case we used a numerical approach that connected all shoots N in the clonal network in an open circuit until the total route length did not shorten anymore for N times (see ts_solve.m). The second algorithm searches for the nearest neighboring shoots consecutively until all shoots N are connected. The algorithm was iterated N times, starting at every individual shoot, and the route with the shortest total length was selected (see clonal_plant_analysis.m section: estimate pin (shoot) connections using NN). The results from these connecting algorhitms are stored in the associated csv files (distancesNN_plantx.csv or distancesTS_plantx.csv).