Spatial networks with an underlying physical structure are ubiquitous throughout nature. These physical networks comprise many disparate systems, but share underlying concepts that enable similar morphological and topological analysis. My thesis develops a set of methods for digitising and analysing 2D networks from photographic imagery. I start by reviewing network theory and physical networks, providing examples from nature. In Chapter 2, I develop and test a novel, flexible pipeline for converting binary maps of physical networks into a graph-theoretical representation. I apply this system to 2D projections of tree root systems to derive a suite of functional traits. In Chapter 3 I focus on spider webs and the generation of imagery required for the techniques in Chapter 2. I propose modifications to standard fieldwork techniques that use ultraviolet (UV) light and fluorescence to increase contrast between a foreground structure (i.e. the spider’s web) and the background. This method significantly increases contrast with the background relative to visible-light approaches and aids in sampling when the background cannot be excluded through other means. I also develop a proof-of-concept machine-learning method for creating binary maps of spider webs from these images. In Chapter 4, I take images generated using the techniques in Chapter 3 and create network representations using the pipeline from Chapter 2. I process these spatial networks to derive established and novel traits of spiders’ webs, and perform a set of Monte Carlo simulations to assess optimal prey size. This set of novel computational and empirical methods provides those wishing to study the topology and morphology of physical network structures with an easily adaptable set of tools. The fast, simple, and flexible tooling provides the opportunity to develop large databases of network-based functional traits for a variety of systems where previously it was logistically impractical due to laborious and time-consuming methods.