eDITH: An R‐package to spatially project eDNA‐based biodiversity across river networks with minimal prior information

Abstract

Ecological and ecosystem monitoring is rapidly shifting towards using environmental DNA (eDNA) data, particularly in aquatic systems. This approach enables a combined coverage of biodiversity across all major organismal groups and the assessment of ecological indices. Yet, most current approaches are not exploiting the full potential of eDNA data, largely interpreting results in a localized perspective. In riverine networks, by explicitly modelling hydrological transport and associated DNA decay, hydrology‐based models enable upscaling eDNA‐based diversity information, providing spatially integrated inference. To capitalize on these unprecedented biodiversity data and translate it into space‐filling biodiversity projections, a streamlined implementation is needed. Here, we introduce the eDITH R‐package, implementing the eDITH model to project biodiversity across riverine networks with minimal prior information. eDITH couples a species distribution model relating a local taxon's eDNA shedding rate in streamwater to environmental covariates, a mass balance expressing the eDNA concentration at a river's cross‐section as a weighted sum of upstream contributions, and an observational model accounting for uncertainties in eDNA measurements. By leveraging on spatially replicated eDNA measurements and minimal hydro‐morphological data, eDITH enables disentangling the various upstream eDNA sources, and produces space‐filling maps of a taxon's spatial distribution at any chosen resolution. eDITH is applicable to both eDNA concentration and metabarcoding data, and to any taxon whose DNA can be retrieved in streamwater. The eDITH package provides user‐friendly functions for single‐run execution and fitting of eDITH to eDNA data with both Bayesian methods (via the BayesianTools package) and non‐linear optimization. An interface to the DHARMa package allows model validation via posterior predictive checks. Necessary preliminary steps such as watershed delineation and hydrological characterization are implemented via the rivnet package. We illustrate eDITH's workflow and functionalities with two case studies from published fish eDNA data. The eDITH package provides a user‐friendly implementation of eDITH, specifically intended for ecologists and conservation biologists. It can be used without previous modelling knowledge but also allows customization for experienced users. Ultimately, eDITH allows upscaling eDNA biodiversity data for any river globally, transforming how state and change in biodiversity in riverine systems can be tracked at high resolution in a highly versatile manner.

Measuring the state of aquatic environments using eDNA—upscaling spatial resolution of biotic indices

“…By using suitable network modelling approaches that explicitly consider transport of eDNA (e.g. eDITH; [51][52][53]), both the point-based (the current method used for biomonitoring) and the spatial projection of the ecological status across a whole river catchment (i.e. the state of a whole ecosystem) can be determined.…”

“…Here, we provide a proof-of-concept of how the combination of eDNA and hydrological modelling allows predictions of the ecological state of a river at the whole river network scale, with a resolution of a few hundred-metre reach lengths. Specifically, we demonstrate the use of eDNA data in combination with the eDITH model [51][52][53]. We explicitly determine the ecological status of a complete river network, based on the occurrence of macroinvertebrate DNA taken from water samples.…”

Environmental DNA: The next chapter

“…At a catchment scale, the eDITH model (Carraro et al, 2020(Carraro et al, , 2021(Carraro et al, , 2023(Carraro et al, , 2024 provides a river network perspective to eDNA transport and allows species distribution maps to be produced across a catchment. Analogous approaches integrating hydrodynamic modelling and eDNA data have also been developed for marine environments (Andruszkiewicz et al, 2019;Fukaya et al, 2021).…”