Generation and application of river network analogues for use in ecology and evolution

Abstract

1. Several key processes in freshwater ecology and evolution are governed by the connectivity inherent to dendritic river networks. These networks have extensively been analyzed from a geomorphological and hydrological viewpoint, yet network structures classically used in modelling have only been partially representative of the structure of real river basins, and have often failed to capture well known scaling features of real river networks. Pioneering work has identified optimal channel networks (OCNs) as spanning trees that reproduce all scaling features characteristic of real, natural stream networks worldwide. While these networks have been used to generate landscapes for studies on metapopulations, biodiversity and epidemiology, their generation has not been generally accessible.2. Given the increasing interest in dendritic riverine networks by ecologists and evolutionary biologists, we here present a method to generate OCNs and, to facilitate its application, we also provide the R-package OCNet. Owing to the random search process that generates OCNs, multiple network replicas spanning the same surface can be built, allowing one to perform computational experiments whose results do not depend on the particular shape of a single river network. The OCN construct also enables the generation of elevational gradients derived from the optimal network configuration, which can constitute three-dimensional landscapes for spatial studies in both terrestrial and freshwater realms. Moreover, the OCNet package provides functions that aggregate the OCN into an arbitrary number of nodes, calculate several metrics and descriptors of river networks, and draw relevant features of the network.3. We describe the main functionalities of the package and present how it can be integrated into other R-packages commonly used in spatial ecology. Moreover, we exemplify the generation of OCNs and discuss an application to a metapopulation model for an invasive riverine species. 4. In conclusion, OCNet provides a powerful tool to generate and use realistic river network analogues for various applications. It thereby allows the design of spatially realistic studies in increasingly impacted ecosystems, and enhances our knowledge on spatial processes in freshwater ecology in general.

Dispersal evolution in currents: spatial sorting promotes philopatry in upstream patches

“…Application of our model on more complex landscapes, or more highly dimensional habitats such as are experienced by terrestrial, wind‐dispersed organisms (2D) or marine species (3D), would provide valuable insight into the interaction of the effects of currents and system complexity on dispersal strategies, genetic variability (Morrissey and De Kerckhove 2009, Paz‐Vinas et al 2015) and community structure (Fagan 2002, Tonkin et al 2014). Running dispersal evolution models within suites of simulated artificial river networks (Carraro et al 2020) would provide an effective means for generating predictions on how different river characteristics are likely to exert different selective forces. Introducing temporal variability in current strength would allow us to investigate the resilience of populations to occasional disturbances (Bonte et al 2012, Travis et al 2012) such as storm surges, flooding events or local patch extinction due to pollution events.…”

How to design optimal eDNA sampling strategies for biomonitoring in river networks

“…The virtual river network used in the following simulations is an Optimal Channel Network (OCN) built via the R‐package OCNet (Carraro, Altermatt, et al, 2020; Carraro, Bertuzzo, et al, 2020). OCNs are idealized constructs that reproduce the topological and scaling features of real river networks and are therefore suitable for simulation studies on various ecological and ecohydrological issues (Rinaldo et al., 2014; Rodriguez‐Iturbe et al., 1992).…”

“…Representation of the optimal channel network used in this study, spanning a square of side 20 km. The aggregation of the OCN at the AG level (see Carraro, Bertuzzo, et al, 2020) identifies 200 nodes, which are here displayed at the downstream end of the corresponding river reaches. Red indicates nodes whose drainage area is lower than the median drainage area across the 200 nodes (referred to in the text as upstream nodes); black identifies nodes with drainage area higher than the median (downstream nodes)…”

“…A threshold area of 461 pixels () is used to distinguish the fraction of lattice pixels that effectively constitute the river network. At this so‐called RN (river network) aggregation level, the network is constituted by 4,468 nodes (each corresponding to a 50‐m pixel with drainage area ≥ A T ; see Carraro, Bertuzzo, et al (2020) for details on aggregation levels of an OCN). Subsequently, such network is further coarsened into the AG (aggregated) level, where nodes represent sources and confluences of the network at the RN level, while edges follow the drainage directions previously identified.…”

“…Here, we fill this gap by making use of computer simulations in order to assess different scenarios of eDNA release, transport, and detection. We test these scenarios based on the application of the hydrology‐based eDNA transport model of Carraro, Hartikainen, et al (2018) using synthetic analogues of river networks (so‐called Optimal Channel Networks, see Carraro, Bertuzzo, et al (2020) and Rinaldo, Rigon, Banavar, Maritan, and Rodriguez‐Iturbe (2014)). This allows testing realistic spatial scenarios in a generic setting that is not constrained to the particular shape of a real river network.…”