NOTE: This Scoping Study discussion paper was open for community feedback as part of the 2021 National Research Infrastructure Roadmap consultations. Submissions are now closed.

Purpose

The 2016 National Research Infrastructure (NRI) Roadmap identified that investment is needed to establish a National Environmental Prediction System (NEPS) and the 2018 Research Infrastructure Investment Plan1 provided funding for a scoping study on a NEPS which was led by an Expert Working Group2 and with secretariat support provided by the Terrestrial Ecosystem Research Network (TERN).

The scoping study identified that the environment research domain lacks the additional networking and technical linkages necessary to ensure all researchers (and government, industry and wider society) can readily find, review and contribute to improve data and models for prediction. There is no single model that conceptualises complexities of ecosystem function; instead, ensembles of models from different disciplines are required to investigate any complex issue.

The scoping study implementation plan proposed a NEPS capability that builds on existing NCRIS enabled research infrastructure to bring together key elements including a governance hub, synthesis capability, modelling infrastructure and pilot projects.

This paper is intended to be used to test the proposal more broadly with the research sector and stakeholders in the 2021 NRI Roadmap process. The paper reflects the work of the NEPS EWG and the implementation plan received and does not represent the views of the Government.

Background

The 2016 National Research Infrastructure (NRI) Roadmap identified that investment is needed to establish a National Environmental Prediction System (NEPS).

Enhance capability for new infrastructure integrated with eResearch to enable existing and new data with new technologies and modelling to build an Environmental Prediction System for Australia.3

In response to the 2016 Roadmap, the Australian Government released the 2018 Research Infrastructure Investment Plan4 which provided funding for a scoping study on a NEPS. The scoping study commenced in late 2018 and was led by an expert working group (EWG)5, with support provided by TERN.

The objectives of the scoping study were to provide technical assessments and requirement analysis including implementation costs and timeframes to establish and manage a NEPS as national research infrastructure to suit researchers as well as public policy decision-makers. The expert working group and TERN (Australia’s land ecosystem observatory) have provided the government with an implementation plan. This has been developed in consultation with key experts and stakeholders, including the existing National Collaborative Research Infrastructure Strategy (NCRIS) network.

Context and need for NEPS

Our knowledge, models, tools and standards are becoming insufficient for the challenges of the future. A national prediction capability will help us learn, explore the limits of possibility, understand how ecosystems are changing and understand future uncertainties and risks. It will inform choice and better decisions, inform adaptive planning, effective interventions and understanding of cumulative impacts. It will give credibility to decisions and provide a basis for action and learning (CSIRO 2019).6

Australia has diverse terrestrial, aquatic and marine ecosystems that deliver public benefit when managed appropriately, and huge costs when impaired. Much of Australia’s natural capital has already been degraded and is subject to on-going stressors such as continuing development, invasive plants and pests and climate change-driven weather extremes such as drought and fire. According to the NEPS implementation plan, Australia will lose more of its precious ecosystems unless it manages them more sensitively and improves their resilience. Current management of these ecosystems involves assessment, regulation, conservation and restoration, guided by State and Commonwealth legislation and agency functions and considerable investments. In the face of climate change, related land use changes, and large-scale development projects such as ports, mines, roads and so on there is an increasing need for management practices to take into account predictions. These predictions need to identify the magnitude and other dimensions of identified change on the environment given a particular event or action, in comparison with the situation without the event or action, as depicted in Figure 1.

NEPS prediction

Figure 1 Environmental management increasingly includes prediction of change impact.

Impact prediction then leads to an evaluation of the significance of impacts based on factors such as level of public concern; scientific and professional judgement; measure of disturbance to ecological systems; and impacts on social values and quality of life.

Environmental predictions generate significant public benefits. For example, the value of seasonal climate forecasts to agriculture is estimated at more than $1500 million per annum7. The integration of environmental observations with predictive modelling would ultimately lead to improved environmental management and knowledge and would benefit Australia.

The Final Report of the Review of the EPBC Act (Samuel, 2020)8 stresses the need for improvements in the information used in environmental assessment, stating “… governments need the capability to model the environment, including the probability of outcomes from proposals, drawing on predictive modelling capabilities” and “to do this well, investment is required to improve knowledge of how ecosystems operate and to develop the capability to model them.”

According to the implementation plan, sustaining natural assets requires on-going monitoring, assessment and fore-sighting of changes in environmental conditions and Australia is missing a point of focus for environmental prediction; a brand that is known and respected for quality, independence and authority. There is currently no platform that brings together the following dispersed expertise: data and models to support rich transdisciplinary integration and foster new capabilities; short-term forecasting; and longer-term identification of risks. The environment research domain lacks the additional networking and technical linkages that are necessary to ensure that all researchers (and government, industry and wider society) can readily find, review and contribute to improve data and models for prediction. There is no single model that conceptualises complexities of ecosystem function; instead, ensembles of models from different disciplines are required to investigate any complex issue. This creates the need for a modelling infrastructure that can harness domain expertise, models and data assets from diverse environmental modelling communities. Figure 2, depicts some of the sub-optimal researcher community impacts arising from a lack of predictive research capability in ecosystem science.

Neotic slides

Figure 2 Effects of significant deficits in research infrastructure for environmental prediction

Australia has strong foundations in ecosystem science, motivated by the need to understand and protect our natural environment. However, our abilities to simulate ecosystem functions and predict how they might respond to future disturbances remain quite immature. There is a need to engage scientists of diverse abilities using innovative digital approaches. NEPS is proposed as a mechanism to realise this ambition.

NEPS Overview

The NEPS implementation plan is a ten-year investment proposal to develop and operate a National Environmental Prediction System, and due to vulnerability, with a focus on terrestrial natural and managed ecosystems in its start-up phase. The goal of a NEPS NRI is to facilitate and accelerate the proliferation and creation of knowledge about terrestrial ecosystem function and condition in relation to a range of future scenarios associated with the environment. According to the implementation plan, Australia needs to improve the national research infrastructure that is required for enabling ecologists, computer scientists, environmental modellers and statisticians to co-develop the next generation data sets and models. Existing NCRIS facilities and government agencies provide strong foundations for NEPS (see Figure 3) but there are large gaps limiting the effectiveness of the community of practice involved in predicting ecosystem function.

The project is designed so that it can be scaled up to incorporate freshwater, agricultural, coastal and marine ecosystems. The proposal presents a roadmap for building the necessary technical and social infrastructures that are needed, as well as models for governance and research utilisation and aims to increase Australia’s environmental prediction capability through leveraging and coordinating existing research hubs and facilities.

Figure 3 NEPS will be integrated with relevant existing national research and operational infrastructures.

NEPS aims to provide organisational infrastructure to aid the synchronisation of environmental research for predictive purposes. The support for NEPS would produce the necessary transformation in transdisciplinary environmental research by enabling ecologists, computer scientists, environmental modellers and statisticians to co-develop data sets and models required to underpin improved management of Australia’s terrestrial ecosystems, and ultimately, other environmental domains. NEPS will be a source of technical expertise related to models, data and prediction and will co-invest in development activities with NRI partners to address interoperability challenges.

Research Infrastructure and NEPS Proposed Solutions

No nation has a fully connected infrastructure that addresses the set of challenges that NEPS aims to handle, although several centres and institutes are exploring opportunities around ecological and environmental prediction. Transformation of ecosystem prediction into a rigorous, scientific endeavour has only recently been made possible through:

  • observing capacity of operational satellites from which Australia acquires data streams
  • the continental surveillance programs of TERN and IMOS
  • cloud computing and data storage
  • high-performance computing for modelling of complex, dynamic systems
  • machine learning and AI for mining of large quantities of data.

Much of Australia’s existing NRI is considered world class. Over more than a decade, Australia has delivered a mature, strong, agile, efficient, and accessible national research infrastructure ecosystem able to facilitate collaborations and accelerate research across multiple areas of importance to Australia’s future, and is well connected internationally. This NCRIS-enabled research infrastructure provides the foundations for a national environmental prediction system. However, existing capabilities are not designed to support environmental prediction, including delivery into new scales and incorporating new processes, such as dynamic evolution. Figure 4 is a modular description of the gaps in NRI to be filled in by NEPS.

NEPS diagram

Figure 4 Demonstration of the gap in national research infrastructure to be filled by NEPS (brown)

The existing NCRIS capabilities that could provide resources to NEPS, either as data or data services and analytics, do not provide environmental prediction research infrastructure. For example, the mission of the Australian Research Data Commons, ARDC, is to accelerate research and innovation by driving excellence in the creation, analysis and retention of high-quality data assets. As such, ARDC’s purpose is to provide Australian researchers with competitive advantage through data. ALA, TERN and BPA each collect and aggregate data that links to terrestrial ecosystems. None, however, have the social infrastructure needed to build trust in models developed from the data or to oversee authentication of models. ALA and TERN have developed tools and standards to bring data into their repositories and to display and visualise data and derived data products, yet their roles stop short of building models that become readily interoperable as components in more complex meta-models to represent dynamics of coupled systems at multiple scales from local to planetary. There is no infrastructure currently with the mandate necessary for delivery of trusted models and predictions to coordinate standards and interoperability between domains and across the environment. Furthermore, none of the NRIs have a role in assisting with the integration of big data for use in ensembles of models – a service which NEPS would provide. The ACCESS climate model NRI is aimed at providing modelling capability for researchers in weather and climate science, whereas NEPS aims to provide modelling and synthesis support to ecosystem scientists and perhaps ultimately, across the full suite of social, ecological, environmental, and geophysical sciences.

The NEPS Implementation plan proposes the guiding principles for access to the research infrastructure available to all Australian researchers as: researcher-centric, assists research, has low barriers to entry and is efficient and enables the majority of researchers and research-intensive institutions to collaborate.

Key Elements of NEPS

The NEPS implementation plan is recommending solutions that address the four elements of services needed for a functioning NEPS – governance hub, synthesis capability, modelling infrastructure and pilot projects.

  • A Governance Hub responsible for developing and managing partnership agreements and policies, including principles and procedures for curating and managing access to data sets and models. This component has oversight of activities such as user management, authorisation, infrastructure status and security.
  • A Synthesis Capability to enable all NEPS stakeholders to come together as a transdisciplinary community of practice and encourage new collaboration across the prediction field where science and policy meet. Through its Synthesis Capability, NEPS will invite participation from researchers and institutions to participate in Pilot Projects that involve stakeholders in their co-design. Synthesis Capability is an interface between researchers, government and industry to close the gaps between them and to optimise use and integration of research.
  • A loosely coupled Modelling Infrastructure to harness domain expertise, models and data assets from diverse environmental modelling communities, and to enable robust integration of different models and data-model fusion to enhance skilful predictions. It will be a standards-based, FAIR-aligned framework to enable scientists to create next generation modelling technology for coupled earth and environment systems. The Modelling infrastructure will enable scientists to leverage each other’s expertise and work in ways not currently possible and address research issues related to the many types of model integration. NEPS support of the proposed research will come in the form of technical skills related to developing accessibility standards so that model code can be easily discovered and retrieved from persistent Internet locales; developing standards for comprehensive documentation to enable models to be widely used and replicated; identifying and adopting technology standards to facilitate reusability; developing and adopting standards for APIs and common ontologies so that models can be more readily interconnected in an open model ecosystem; creating cyberinfrastructure to help modelling scientists implement technologies to make their models reusable and interoperable; creating educational programs and professional incentives to encourage modellers to adopt such standards; data wrangling; benchmarking; competitive funding for Pilot Projects; and strategic scientific and infrastructure leadership.
  • A number of Pilot Projects, to galvanise collaboration between researchers and end-users and demonstrate the value of environmental prediction, forging a bottom-up alliance of scientific communities that share the goals of understanding complex earth and environment systems and enabling decision-making for sustainable management of the environment for the future. Through these Pilot Projects, NEPS will initially be drawing upon terrestrial ecosystem function and condition data and expertise to support the interdisciplinary research field of ecosystem science. In particular, Pilot Projects will provide an incentive for transdisciplinary research and complex problem solving through digital infrastructure, filling knowledge gaps, supporting decision making and improving decision support tools.

The implementation plan also offers a solution that uses the governance hub to assist the existing NRI facilities to enhance their data management, discovery and delivery capabilities in support of research into environmental prediction. This will take place within virtual laboratories and NEPS will be a source of technical expertise related to models, data and prediction. There will be a combined effort from existing NRI facilities required for enhanced delivery of data, including time series, spatial and sensor data, for the operation of researcher models and decision tools.

The proposal suggests an existing NCRIS facility could host NEPS in the first instance as this would allow cost effective access to its existing governance, communication and management services.

1 https://www.dese.gov.au/national-research-infrastructure/resources/2018-research-infrastructure-investment-plan

2 The Chair of the Panel was Professor Rob Vertessy and the members were: Dr Andrea Hinwood, Dr Adam Lewis, Dr Phil McFadden AO, Mr Warwick McDonald and Dr Steve Morton. The Terrestrial Ecosystem Research Network (TERN), an NCRIS-funded initiative headquartered at University of Queensland, provided support to the Expert Panel.

3 https://2021nriroadmap.dese.gov.au/wp-content/uploads/2021/01/2016-National-Research-Infrastructure-Roadmap.pdf

4 https://www.dese.gov.au/national-research-infrastructure/resources/2018-research-infrastructure-investment-plan

5 The Chair of the Panel was Professor Rob Vertessy and the members were: Dr Andrea Hinwood, Dr Adam Lewis, Dr Phil McFadden AO, Mr Warwick McDonald and Dr Steve Morton. The Terrestrial Ecosystem Research Network (TERN), an NCRIS-funded initiative headquartered at University of Queensland, provided support to the Expert Panel.

6 Pirzl R, Doerr V, Lemon D, McDonald W, Ahmad M, Grigg N, Fletcher C, Mokany K, Schandl H, O’Connell D, Palmer R, and Woodward E (2019) Environmental Prediction Symposium Synthesis. CSIRO Land and Water, Australia.

7 Centre for International Economics (CIE), 2014. Analysis of the Benefits of Improved Seasonal Climate Forecasting for Agriculture. Managing Climate Variability R&D Program, CIE, Canberra.

8https://epbcactreview.environment.gov.au/resources/final-report