The Wood Image Analysis and Dataset (WIAD): Open‐access visual analysis tools to advance the ecological data revolution

Abstract

1. Ecological data are collected and shared at an increasingly rapid pace, but it is often shared in inconsistent and untraceable processed forms. Images of wood contain a wealth of information such as colours and textures but are most commonly reduced to ring-width measurements before they can be shared in various common file formats. Archiving digital images of wood samples in libraries, which have been developed for ecological analysis and are publicly available, remains the exception.2. We developed the Wood Image Analysis and Dataset (WIAD), an open-source application including a web interface to integrate basic visual analysis of wood samples, such as increment cores, thin sections or X-ray films, basic data processing, and archiving of the images and derived data to facilitate transparency and reproducibility in studies using visual characteristics of wood.3. WIAD provides user-friendly tools to manipulate images of wood samples, mark and measure wood characteristics such as growth increments, density fluctuations, early-and latewood widths and fire scars, and to visualise, process and archive images, metadata, and the derived data. 4. WIAD constitutes a step towards the reproducible automation of tree-ring analysis while establishing an open-source foundation to create improved communitydeveloped repositories which would enable novel ecological studies harnessing the wealth of existing visual data.

Automation of tree-ring detection and measurements using deep learning

“…The CNN's outputs include JSON files of all detections as well as position (.POS) files that are compatible with Coo Recorder (Larsson 2003). The POS files make the CNN results easily editable through Coo recorder and the recent development of new, open source image analysis programs such as WIAD (Rademacher et al 2021)…”

“…The CNN’s outputs include JSON files of all detections as well as position (.POS) files that are compatible with Coo Recorder (Larsson 2003). The POS files make the CNN results easily editable through Coo recorder and the recent development of new, open source image analysis programs such as WIAD (Rademacher et al 2021) or Dendro Elevator (https://dendro.elevator.umn.edu/) provide additional ways to display both images and annotations.This software combined could provide a freeware open source image analysis pipeline.…”

“…Also, large datasets or meta-analyses may be biased when multiple people contribute to measurements. Analysis on digital images of tree-ring samples is becoming increasingly common, as are developments in both image acquisition systems (Griffin et al, 2021; Levanič, 2007) and programs for analysis (Crawford, 2021; Rademacher et al, 2021)). Additionally, some forms of dendrochronological analysis that depend on high resolution images such as blue intensity (Rydval et al, 2014) or quantitative wood anatomy (von Arx & Carrer, 2014) require digital images.…”

“…Also, large datasets or meta-analyses may be biased when multiple people contribute to measurements. Analysis on digital images of tree-ring samples is becoming increasingly common, as are developments in both image acquisition systems (Griffin et al, 2021;Levanič, 2007) and programs for analysis (Crawford, 2021;Rademacher et al, 2021)).…”

“…CNNs are widely used across many fields from facial recognition (Yang & Yin, 2017) to human health (Kido et al, 2018), finance (Gudelek et al, 2017), genomics (Washburn et al, 2019) and ecology (Fricker et al, 2019). Increased use of imaged increment cores (Griffin et al, 2021;Larsson, 2003;Rademacher et al, 2021) provides an opportunity to automate detection and subsequent measurement of annual ring widths. Previous attempts at identifying ring boundaries have had variable and sometimes impressive AUTOMATED TREE-RING DETECTION AND MEASUREMENT success (Conner et al, 1998;Fabijańska et al, 2017;Fabijańska & Danek, 2018;Martinez-Garcia et al, 2021;Resente et al, 2021) but the only deployable method is limited to three-dimensional images (Martinez-Garcia et al, 2021).…”