After almost five years of work, we have a new paper out in New Phytologist on venation network architecture. You can read the article at the journal or directly download a PDF.
The paper is part of the international and multi-institution BALI project, with fieldwork carried out in Sabah, Malaysia primarily in 2015.
Sabine Both led the field campaign, with major contributions from Unding Jami (head of tree climbing team) and a large team of Malaysian, Peruvian, and British researchers (all acknowledged in the manuscript). The Peruvian team was a consequence of the earlier and similar CHAMBASA project in Peru – several of the best people were recruited to share their expertise, leading to a fun reunion and a very multilingual field experience.
The study focuses on understanding leaf venation network architecture at different spatial scales – the small veins up to the big veins. It develops a conceptual framework for describing network architecture across these scales, using an adaptation of the hierarchical loop decomposition (HLD) methods that were published a few years ago, and is able to generate multi-scale statistics, which are geometric descriptors of network architecture that are functions of vein size. These statistics can then be compared across species, or across spatial scales, providing a nice way to unite prior studies of network architecture at single scales, and bringing HLD concepts more closely to biologically relevant descriptors.
We are able to generate multi-scale statistics for hundreds of species, leveraging both months of intensive fieldwork by dozens of people, months more of lab work analyzing the resulting leaves, and also some big recent developments in deep learning-based image analysis that we have pioneered (with a companion paper on this subject accepted and coming out shortly).
The study demonstrates that multi-scale statistics provide a powerful way to quantify variation in network architecture. We show that variation among species and scales is very high, and also coordinated with functional axes like mechanical strength and defense chemistry. I think that this approach will help clarify the ways in which network architecture has evolved, and influences leaf functioning.
The big limitation of the study is that the leave analyzed span spatial scales only up to 1 cm in length – so we are missing many of the major veins in this work. In ongoing studies we are extending the methodology to whole leaves, and are excited to start sharing results in the coming months.
Looking back on these photos from the 2015 campaign, it is a pleasure to remember familiar faces and shared experiences in the forest. And it is equally hard to imagine being able to do something similar in today’s world. I’m grateful for all the international contributions to this project, and hope the day will come where more work like this can happen again.
