Carl Boettiger

Ecological forecasting challenges, formalized efforts to empower the ecology community to develop near‐term iterative forecasts of yet‐to‐be‐collected data, are a powerful tool to advance predictive capacities in ecology. The National Ecological Observatory Network (NEON) Ecological Forecasting Challenge, which has processed over 82 million individual forecasts across 81 sites since 2021, exemplifies this approach. However, the cyberinfrastructure (CI) required to support such large‐scale, community‐driven forecasting efforts has not been fully described. Here, we present the design principles, technical implementation, and lessons learned from the CI developed for the NEON Challenge. This serverless, open‐source platform automates key tasks, including target data generation, probabilistic forecast submission‐processing, evaluation using proper scoring rules, and dissemination of results. The CI leverages cloud‐native technologies, including Simple Storage Service (S3)‐based storage, GitHub Actions, and Docker containers, to ensure scalability, reproducibility, and vendor‐agnostic deployment. The CI also includes tools to reduce barriers for contributors, such as R packages for forecast submission, access to weather forecasts as model drivers, and tutorials. A dashboard and searchable catalog enhance transparency and accessibility. Our experience demonstrates that forecasting challenges can serve as focal points for community engagement, education, and synthesis. The modular design of the CI enables adaptation to new observatories and evolving technologies, providing a roadmap for future ecological forecasting efforts. By documenting this infrastructure, we aim to empower researchers to launch and sustain forecasting challenges that advance ecological understanding through prediction.

Provides an R interface to 'rclone' , a command-line program for managing files on cloud storage. 'rclone' supports over 40 cloud storage providers including 'S3'-compatible services ('Amazon S3', 'MinIO', 'Ceph'), 'Google Cloud Storage', 'Azure Blob Storage', and many others. This package downloads and manages the 'rclone' binary automatically and wraps its commands as R functions, returning results as data frames where appropriate.

Ecological forecasting challenges, formalized efforts to empower the ecology community to develop near-term iterative forecasts of yet-to-be-collected data, are a powerful tool to advance predictive capacities in ecology. The NEON Ecological Forecasting Challenge, which has processed over 82 million individual forecasts across 81 sites since 2021, exemplifies this approach. However, the cyberinfrastructure (CI) required to support such large-scale, community-driven forecasting efforts has not been fully described. Here, we present the design principles, technical implementation, and lessons learned from the CI developed for the NEON Challenge. This serverless, open-source platform automates key tasks, including target data generation, probabilistic forecast submission processing, evaluation using proper scoring rules, and dissemination of results. The CI leverages cloud-native technologies, including S3-based storage, GitHub Actions, and Docker containers, to ensure scalability, reproducibility, and vendor-agnostic deployment. The CI also includes tools to reduce barriers for contributors, such as R packages for forecast submission, access to weather forecasts as model drivers, and tutorials. A dashboard and searchable catalog enhance transparency and accessibility. Our experience demonstrates that forecasting challenges can serve as focal points for community engagement, education, and synthesis. The modular design of the CI enables adaptation to new observatories and evolving technologies, providing a roadmap for future ecological forecasting efforts. By documenting this infrastructure, we aim to empower researchers to launch and sustain forecasting challenges that advance ecological understanding through prediction.

Provides 'RDF' storage and 'SPARQL' 1.1 query capabilities by wrapping the 'Oxigraph' graph database library . Supports in-memory and persistent ('RocksDB') storage, multiple 'RDF' serialization formats ('Turtle', 'N-Triples', 'RDF-XML', 'N-Quads', 'TriG'), and full 'SPARQL' 1.1 Query and Update support. Built using the 'extendr' framework for 'Rust'-R bindings.

Detection of wildlife mortality events is critical for timely conservation and natural resource management. We present an open-source, web-based decision support tool that queries, aggregates and summarizes participatory science data from iNaturalist to monitor mortality events worldwide. We demonstrate the effectiveness of this approach using four case studies spanning taxonomic, spatial, and temporal scales. In Canada and the United States, high peaks of bird mortality coincided with zoonotic risk during avian influenza outbreaks. Across Latin America, we detected 75 mortality events of critically endangered species. In California, recorded mammal mortality was associated with human infrastructure, including proximity to roads, and to a lesser extent, the human footprint. Mortality of pumas ( Puma concolor ) was detected across nine countries, highlighting the need for international cooperation to conserve mobile species. Our tool enables resource managers to flag emerging threats and empowers participatory scientists to monitor and integrate mortality records for conservation.

In fishery science, harvest management of size‐structured stochastic populations is a long‐standing and difficult problem. Rectilinear precautionary policies based on biomass and harvesting reference points now represent a standard approach to this problem. While these standard feedback policies are based on analytical or dynamic programming solutions assuming relatively simple ecological dynamics, they are often applied to more complicated ecological settings in the real world. In this paper, we explore the problem of designing harvest control rules for partially observed, age‐structured, spasmodic fish populations using tools from reinforcement learning (RL) and Bayesian optimisation. Our focus is on the case of Walleye fisheries in Alberta, Canada, whose populations display variable recruitment dynamics. We optimised and evaluated policies using several complementary performance metrics representing key trade‐offs in harvest management. The main questions we addressed were: (1) How do standard policies based on reference points perform relative to numerically optimised policies? (2) Can an observation of mean fish weight, in addition to stock biomass, aid in policy decisions?

The NPSdataverse is a suite of R packages developed to create, document, publish, and access data and metadata in open and machine-readable formats. NPSdataverse is modeled off of the tidyverse concept of several packages built with a common goal (Wickham et al., 2019). The NPSdataverse supports Ecological Metadata Language (EML) metadata and .csv data files. Some of the constituent R packages (EML and EMLassemblyline) are general-use and aimed at authoring EML documents. Other R packages (QCkit, EMLeditor, DPchecker and NPSutils) are designed and maintained by the National Park Service (NPS). Although many functions within the NPSdataverse packages are NPS-specific (particularly some API calls), whenever possible the functions are written so that they can also be used by the general public. Scientists conducting permitted research in NPS units can utilize the NPSdataverse to efficiently and consistently meet the data delivery requirements of their permits. Additionally, the packages will be useful for data management plans in a wide variety of grant proposals and for anyone that needs to create open data and machine-readable metadata. The ability to swiftly and easily author, edit, and check Ecological Metadata Language (EML) metadata in a reproducible fashion will be useful for data publication at any number of repositories or data journals. Finally, a scripted interface for downloading NPS data and leveraging metadata while loading it into R or other platforms for subsequent analyses and visualizations will be useful to researchers in the government, academia, and industry as well as the public.

The European Union is committed to achieving ambitious area-based conservation and restoration targets in the upcoming decade. However, there is concern that these targets risk conflicting with socioeconomic needs, particularly for food, timber and bioenergy production. Here we develop an integrated spatial planning approach to identify where restoration, conservation and production allocation could maximize benefits to species conservation and climate mitigation, while acknowledging future land demands of the bio-economy. We show that, while changing production demands risk driving further biodiversity loss by 2030, when these demands are met alongside strategic restoration measures, as outlined by the EU Nature Restoration Regulation, future landscapes could improve the conservation status of populations for more than 20% of species of conservation concern while also increasing terrestrial carbon stocks. Our analysis demonstrates how critical the Nature Restoration Regulation is to achieving biodiversity targets and how integrated planning can align biodiversity policy objectives with future socioeconomic demands.