AI for Climate 
R&D Roadmap

Consultation paper

May 2025



Citation

Naughtin C, Evans D, Mason C & Trinh K, 2025. AI for 
Climate R&D Roadmap: Consultation paper. CSIRO. 

Copyright

© Commonwealth Scientific and Industrial Research 
Organisation 2025. To the extent permitted by law, all rights 
are reserved and no part of this publication covered by 
copyright may be reproduced or copied in any form or by 
any means except with the written permission of CSIRO.

Important disclaimer

CSIRO advises that the information contained in this 
publication comprises general statements based on 
scientific research. The reader is advised and needs to 
be aware that such information may be incomplete or 
unable to be used in any specific situation. No reliance 
or actions must therefore be made on that information 
without seeking prior expert professional, scientific 
and technical advice. To the extent permitted by 
law, CSIRO (including its employees and consultants) 
excludes all liability to any person for any consequences, 
including but not limited to all losses, damages, costs, 
expenses and any other compensation, arising directly 
or indirectly from using this publication (in part or in 
whole) and any information or material contained in it.

CSIRO is committed to providing web-accessible content 
wherever possible. If you are having difficulties with 
accessing this document, please contact csiro.au/contact.

Acknowledgement of Country

CSIRO acknowledges the Traditional Owners of the 
lands, seas and waters, of the area that we live and work 
on across Australia. We acknowledge all Aboriginal 
and Torres Strait Islander peoples and their continuing 
connection to their culture and pay our respects 
to Elders past and present. CSIRO is committed to 
reconciliation and recognises that Aboriginal and Torres 
Strait Islander peoples have made and will continue 
to make extraordinary contributions to all aspects of 
Australian life including culture, economy and science.

Acknowledgements

The authors would like to acknowledge the contributions 
of the various individuals who shared their expertise and 
experience as part of this roadmap development process. 
This includes government, industry, academia and the 
not-for-profit sector representatives who participated 
in expert interviews. These insights were instrumental 
in shaping the direction of this research, identifying 
candidate artificial intelligence (AI) use cases for climate 
mitigation and adaptation and brainstorming potential 
models for cross-sectoral collaboration. We also thank 
Mahesh Prakash, Moana Simpson, Leah Moss, Cara Stitzlein, 
Brent Henderson and Nandini Ramesh for their helpful 
guidance and feedback in developing this report and 
Shanae Burns for assisting in the report production.

Images: all photographs Adobe Stock except pages 5 and 16 
(Shutterstock), page 8 (SmartSat CRC) and page 25 (CSIRO).

Currency conversions

All dollar value figures denote Australian 
dollars unless otherwise specified.



Abbreviations

ABARES

Australian Bureau of Agricultural and 
Resource Economics and Sciences

ACCESS

Artificial Intelligence for Changing Climate 
and Environmental Sustainability 

ACT

Australian Capital Territory

ADViCE

Artificial Intelligence for Decarbonisation’s 
Virtual Centre of Excellence

AI

artificial intelligence

AI4ER

AI for the Study of Environmental Risks

CCTV

closed-circuit television 

ICIP

Indigenous Cultural and Intellectual Property

IPCC AR6

Intergovernmental Panel on Climate 
Change Sixth Assessment Report

LLM

large language model

ML

machine learning

NASA

National Aeronautics and 
Space Administration

NRT

National Science Foundation 
Research Traineeship

NSF

National Science Foundation 

NSW

New South Wales

R&D

research and development

SAMMI

Seqwater’s Autonomous Motorised 
Monitoring Instrument

SustAI

UKRI AI Centre for Doctoral 
Training in AI for Sustainability

UK

United Kingdom

UKRI

UK Research and Innovation

UN

United Nations

US

United States





Contents

Introduction...................................................................................................................................................1Scope of this paper.............................................................................................................................................................1Defining artificial intelligence...........................................................................................................................................2Current and emerging AI use cases for climate action...........................................................3Foundational climate and weather modelling.................................................................................................................3Critical infrastructure.........................................................................................................................................................4Emergency management...................................................................................................................................................7Agriculture..........................................................................................................................................................................9Energy................................................................................................................................................................................12Water .................................................................................................................................................................................14Healthcare ..............................................................................................................................................................................................16Other cross-domain applications....................................................................................................................................17Responsible AI-for-climate applications.......................................................................................19Ensuring climate-related AI is safe and responsible ...................................................................................................................19The energy and emissions footprint of AI......................................................................................................................20The water footprint of AI.................................................................................................................................................20Recognising and respecting Indigenous Cultural and Intellectual Property...............................................................20Pathways to accelerate AI‑for-climate R&D................................................................................21Bringing together multidisciplinary teams of technical and domain experts.............................................................21Supporting research collaboration and innovation across sectors..............................................................................22Setting the strategic focus with a clear pathway to impact..........................................................................................23Promoting trusted, safe and responsible AI...................................................................................................................24References..................................................................................................................................................26

Introduction

Australia is among many countries actively responding to 
the global challenge of reducing greenhouse gas emissions 
and adapting to a changing climate. Despite progress 
towards net zero, recent forecasts suggest that the 
world is not on track to meet the Paris Agreement [12]. 
A continuation of current policies is projected to result 
in global warming of 3.1 °C by 2100 and the world would 
need to reduce greenhouse gas emissions by 30% to limit 
global warming to 2 °C [12]. These impacts are already 
being felt in Australia, with annual costs associated 
with extreme weather events expected to increase 
from $38 billion in 2020 to $73 billion by 2060 [13].

With its superior processing speed, capacity to handle 
large and diverse datasets and powerful predictive 
capabilities, there is growing interest in leveraging 
artificial intelligence (AI) to tackle the climate crisis. 
While AI applications have been developed to solve or 
support various climate mitigation and adaptation needs, 
many of these developments have been opportunistic 
and lack strong collaboration between researchers, 
industry, government and not-for-profit sectors. 

Many leaders are also not clear on how AI can support 
their climate action. In 2022, Boston Consulting Group 
found that 87% of global business leaders working 
in climate or AI felt that AI would be a helpful tool in 
responding to climate change, but less than half (43%) 
had clarity on how to use AI in their climate change 
efforts [14]. Where AI is being applied as part of climate 
efforts, these applications are skewed towards mitigation, 
with less focus on how AI can support organisations 
and communities in adapting to a changing climate [15]. 
Leveraging the full potential of AI to accelerate climate 
action requires working across sectors to align future AI 
directions with current and emerging market needs.

With AI and climate, we are dealing 
with a problem space that is very 
broad and a piece of technology 
that can mean many things.

– Expert interviewee

Scope of this paper

This report serves as a discussion starter on where and 
how Australia can leverage AI to accelerate climate action. 
It draws on insights from industry, government and 
not‑for-profit representatives and AI experts to provide a 
market-led view of future AI research and development 
(R&D) opportunities for climate mitigation and adaptation. 
While AI can support climate and weather forecasting, 
this paper focuses on downstream applications of AI 
concerning climate challenges facing critical infrastructure, 
emergency management, agriculture, energy, water and 
health care, noting that this is not an exhaustive list of areas 
where AI can be applied as part of climate responses.

Specifically, this report:

• provides an overview of existing domain‑specific 
AI applications for climate mitigation and 
adaptation and untapped opportunities 
where AI could be applied in the future
• identifies cross-cutting AI applications that 
could support responses to common climate 
challenges impacting multiple sectors
• highlights the potential risks posed by the increasing 
use of AI in tackling the climate crisis and considerations 
for ensuring AI is used safely and responsibly
• seeks feedback on potential pathways for 
Australia to accelerate the role of AI in climate 
adaptation and mitigation responses, considering 
learning from other existing initiatives.


Feedback on this paper will help inform future 
AI‑for‑climate efforts across industry, government, 
academic and not-for-profit organisations, including CSIRO. 
These include mechanisms for engaging stakeholders across 
the Australian R&D ecosystem, building awareness and 
capabilities, coordinating R&D efforts and resources, and 
promoting the safe and responsible use of AI. This paper 
does not seek to cover all potential needs and issues 
relating to AI, but rather to build a shared understanding 
of the types of climate problems with which AI can assist 
and how industry, government, academia and not-for-profit 
sectors can work together towards these opportunities. 



Defining artificial intelligence

While there is no universally accepted definition 
for AI, in this paper we define it as a collection 
of computer science and interrelated statistical 
approaches and technologies that can be used to 
learn from data, solve problems and perform tasks 
autonomously to achieve defined objectives [4]. 

The following table provides examples of some 
of the AI technologies referred to in this report 
and a high-level snapshot of how they can be 
deployed for a range of functions and the specific 
ways in which they can support climate-related 
challenges, which are expanded on in this paper.

EXAMPLE AI TECHNOLOGIES

GENERAL APPLICATIONS

CLIMATE-SPECIFIC APPLICATIONS

Machine learning (ML)

Building computer systems 
that can learn from data to find 
relationships and patterns, and 
make predictions from training 
data without explicit instructions

Diagnosing melanomas: ML can 
help clinicians identify malignant 
skin lesions from images, enabling 
early clinical intervention

Weather forecasting: ML models 
can provide faster and more 
granular weather forecasts and test 
a range of climate scenarios

Computer vision

Enabling computers to analyse and 
extract information from images

Protecting a national icon: 
Computer vision, combined with 
other AI technologies, has been 
applied to the Great Barrier Reef to 
detect crown-of-thorns starfish – a 
key contributor to coral bleaching

Detecting infrastructure faults: 
Computer vision can be used to 
detect maintenance issues early 
(e.g. pipeline leaks) so that they 
can be repaired, strengthening 
resilience to future extreme 
weather events

Robotics and autonomous systems

An interdisciplinary field of AI that 
designs and develops systems that 
can interact with their environment

Industrial robotics: Autonomous 
haulage trucks managed by remote 
operators are used on mining 
sites, improving worker safety and 
productivity 

Disaster evacuation: Driverless 
vehicles, combined with other 
AI technologies, can help 
evacuate affected individuals 
while protecting the safety of 
frontline responders

Human language technologies

Advanced models that are 
trained on large volumes of 
human language to generate 
human‑like outputs 

Virtual assistants: Human 
language technologies underpin 
virtual assistants like Siri, Alexa 
and Google Assistant, enabling 
them to interact conversationally 
with users and streamline everyday 
administrative tasks

Assessing damage: Natural 
language processing models can be 
used to analyse social media data 
following a natural hazard event to 
direct resources to individuals or 
properties that require support







Current and emerging AI 
use cases for climate action

There is a wealth of examples of where AI has already 
been applied to improve foundational climate and 
weather modelling capabilities and address climate 
mitigation and adaptation challenges in specific 
domains. This section highlights examples of how AI 
can be applied in weather and climate forecasting and 
specific application domains at different scales, focusing 
on critical infrastructure, emergency management, 
agriculture, energy, water and health care. We also 
build on these existing applications to identify future 
opportunities to use AI to further mitigate or solve climate 
challenges emerging in one or more of these domains.

Foundational climate 
and weather modelling

How can AI improve climate 
and weather forecasting?

Given the capacity of AI to be applied to diverse datasets 
and provide inferences at finer scales and under 
multiple scenarios, it can help improve foundational 
climate and weather modelling capabilities. A host of 
AI-enabled tools have been developed to improve the 
speed, accuracy, and spatial and temporal resolution 
of weather forecasts [16–19]. For example, Microsoft’s 
Aurora uses AI to forecast a range of weather variables 
10 days in advance [16, 20] and Google’s GenCast 
probabilistic weather model can generate 15-day weather 
forecasts at a 0.25° latitude–longitude resolution [19].

AI models can also expand the capabilities of traditional 
climate models, which are based on the fundamental 
laws of physics to simulate the Earth’s climate system. 
While physics-based climate models can accommodate 
poorer data environments and extrapolate events that go 
beyond historical data, they are underpinned by complex 
equations and can be computationally intensive. AI can be 
applied to additional datasets to model climate processes 
that are currently not well predicted by physics‑based 
model s, thereby revealing how to improve the parameters 
within these traditional models (e.g. [21–23]). 

This capability has proven useful in a range of 
climate‑related applications that draw on climate variables 
(e.g. temperature, weather patterns, ocean currents) and 
human activities (e.g. emissions, land use and population 
patterns) of varying spatiotemporal scales [15]. Examples 
include Microsoft’s ClimaX, a novel foundational model that 
uses diverse datasets collected at different spatiotemporal 
scales to produce climate and weather projections [24]. 
Because AI-based approaches are limited to the data on 
which they are trained, they are unable to extrapolate to 
unknown future climate conditions [25]. Hybrid models 
that combine the accuracy, precision and efficiency 
of AI with physics-based simulation approaches 
may therefore offer an optimal approach [25, 26]. 



Future opportunities to expand foundational 
AI capabilities for climate action

Emerging AI developments in multimodal foundational 
models are opening new avenues for climate and weather 
forecasting and associated downstream applications. 
Foundational AI models provide key efficiency benefits 
in that they can be trained on unlabelled data and then 
fine‑tuned using a smaller amount of labelled data for 
specific applications [27–29]. Technology company IBM and 
the National Aeronautical and Space Administration (NASA) 
have joined forces to develop a multimodal AI foundational 
model for climate and weather prediction, which aims to 
overcome the limitations of current AI climate models [29]. 
Specifically, the model will have the capacity to predict 
extreme events that differ from average conditions and to 
continuously update as climate conditions and data evolve.

Experts also acknowledged the opportunity to develop a 
geospatial AI foundational model for Earth observation 
that could be used for a range of industry and community 
applications. In addition, NASA and IBM Research 
have developed the first open-source geospatial AI 
foundational model trained on satellite data collected 
by NASA [27, 28]. These capabilities can make it easier 
to deploy industry-specific AI tools for monitoring 
natural hazards, predicting crop yields and more [27, 28]. 
Importantly, the realisation of these foundational AI model 
opportunities will rest on access to high-performance 
computing infrastructure, which has typically been a 
challenge for Australian companies and researchers [30].

The foundational AI model developments could support 
climate resilience and capacity building in developing 
nations in the Pacific. Small Pacific Island nations are 
among the most vulnerable to the impacts of climate 
change [31] and the Australian Government has committed 
to supporting the Pacific Island region in strengthening 
its climate and disaster resilience [32]. Foundational 
models could be leveraged to reduce the computational 
cost and complexity of downscaling climate models 
[26], making this information more accessible, timely 
and cost-effective for decision-makers in the Pacific.

AI could also be used to support the management 
of data sources that underpin climate modelling. 
There are currently large volumes of data that are poorly 
structured or documented, limiting the capacity to use 
these data in climate-related AI applications. AI could act 
as a ‘data custodian’, assisting in extracting metadata 
and documenting information on climate datasets and 
querying these datasets. This capability could enhance and 
simplify the process of data discovery and facilitate greater 
opportunities for integration between research projects.

Critical infrastructure

How can AI help adapt critical infrastructure 
to a changing climate?

Critical infrastructure includes a broad range of systems, 
assets and supply chains that are central to the social and 
economic wellbeing, sovereignty and security of Australia 
(e.g. transportation, energy, telecommunication) [33]. 
Protecting Australia’s critical infrastructure from natural 
disasters – along with cyberattacks and global supply 
chain disruptions – is a key priority for government and 
industry in continuing to deliver essential services [33]. 

AI can pre-emptively detect and predict infrastructure 
vulnerability to maintenance issues, helping to reduce 
the costs of or need for repair after a natural hazard 
event. Unmanned aerial vehicles or sensors can 
collect data on infrastructure performance remotely, 
which are analysed by computer vision algorithms to 
detect damage to structural components (e.g. cracks 
in concrete, leaks in pipelines) [34, 35]. For example, 
Sydney-based company VAPAR analyses closed-circuit 
television (CCTV) footage using AI and machine learning 
(ML) to assess the condition of wastewater pipes, 
reducing the cost of infrastructure management [36].

Other applications use AI to rapidly assess damage 
to critical infrastructure during or after a disaster 
event, improving on previous time-intensive processes 
[37, 38]. Using deep neural network and ML approaches, 
images collected via satellites or social media users can 
be compared pre- and post-disaster to diagnose the 
level of damage [37, 39]. Researchers have also used 
AI with social media or satellite data to detect areas 
affected by power and communications outages in real 
time [40, 41]. These insights can help decision-makers 
direct resources, including emergency management. 

Bayesian neural network and neural network-based causal 
approaches have been used to assess the resilience of 
critical infrastructure to a range of interconnected factors 
and the potential cascading impacts of a natural hazard 
event across multiple assets [42, 43]. National Hazards 
Research Australia plans to explore these approaches in 
the Australian context [44]. Understanding the impacts 
of natural hazards on interconnected infrastructure 
networks could better pre-empt the effects of natural 
hazards across multiple systems and coordinate 
investment for climate mitigation and adaptation.



How can AI help decarbonise 
critical infrastructure?

AI can measure and monitor the carbon footprint 
associated with critical infrastructure to identify 
opportunities to reduce emissions. Google’s Environmental 
Insights Explorer can help city planners in the United 
States (US) assess the emissions generated by buildings 
and transport through aerial measures of tree canopy 
coverage and available rooftop solar installations [45]. 
This approach could be extended by using AI to assess 
the emissions profile of individual assets or to identify 
planning areas that would benefit from decarbonisation 
interventions (e.g. planning more green cover).

There are emerging use cases for AI to track and audit 
direct and indirect emissions (scope 1 and 2 emissions, 
respectively) across industrial supply chains (scope 3 
emissions). Existing commercial applications use AI to 
combine multiple data streams, match emissions to specific 
products and business activities, and predict the impact 
of different interventions [46–48]. AI applications, such 
as ClimateBERT, have also been developed to analyse 
climate risk disclosures from company reports to assess 
compliance between climate-related reporting and 
demonstrable climate actions [49, 50]. This capability 
is increasingly relevant to many Australian companies 
that will be required to monitor and report scope 1, 2 
and 3 emissions under new mandatory climate-related 
financial disclosure reporting requirements [51]. 

AI has been deployed to help decarbonise heavy-emitting 
industries like transportation. In aviation, Google 
Research, in collaboration with American Airlines and 
Breakthrough Energy, has used AI to predict contrails 
during flight – a leading source of aviation emissions [52]. 
By optimising aircraft routes using AI, they were able to 
reduce contrails by 54% over 70 flights [53]. AI has also 
been used to identify more fuel- and carbon-efficient 
road transport routes. Examples include Google Maps’ 
Project Green Light, which uses AI to identify routes that 
get users to their destination using fewer emissions [54] 
and the United Parcel Service, which has introduced similar 
capabilities to support more fuel-efficient logistics [55].

Future climate-related use cases for AI 
in critical infrastructure

While there are emerging use cases using AI to assess the 
co-dependence between critical infrastructure [42, 43], 
industry and government experts consulted in this research 
noted further opportunities to use AI to understand the 
second- and third-order impacts of natural hazards 
across critical infrastructure. With this information, 
decision‑makers could make strategic investments that 
limit the flow-on impacts of natural hazards across systems. 
These use cases will need to consider regulations related 
to data sharing, concerns about commercially sensitive 
information and differing stakeholder interests.

The transition to renewables and investments to 
improve the resilience of critical infrastructure can 
be costly. AI could be used to mine existing datasets 
(e.g. government approvals, company announcements) 
to identify opportunities to coordinate infrastructure 
investments and minimise duplication. For example, if 
the regional energy provider is building resilience into 
their network, other infrastructure owners in the region 
may not need to invest in backup power generators. 

New mandatory climate-related financial disclosure 
reporting requirements for Australian companies [51] 
will likely bring greater scrutiny to the estimates 
associated with emissions. While commercial offerings 
are available, industry representatives noted difficulties 
in evaluating the quality of these commercial ‘black 
box’ solutions. They also seek independent guidance 
regarding standards and best practices for modelling 
emissions across industrial supply chains. There are 
opportunities to use AI to improve the monitoring 
and reporting of emissions and other climate‑related 
risks and assess the impact of such disclosures.

Emergency management

How can AI help emergency management 
adapt to a changing climate?

Early prediction of natural hazards can help emergency 
management agencies better direct resources. AI can 
detect and predict natural hazards better than 
traditional methods by learning from patterns in historical 
meteorological and satellite data [56–59]. Google Research 
has used AI as part of Flood Hub, which provides a 5-day 
prediction for extreme riverine events [60, 61] and FireSat, 
which can detect bushfires that are 5 square metres in 
area [62]. Emerging applications are also using satellites 
with onboard AI to detect bushfire smoke (see Case study: 
Detecting bushfires from space) and AI with unmanned 
aerial vehicle imagery to detect hazards at a finer spatial 
and temporal resolution than satellite data [63–70]. 



CASE STUDY 

Detecting bushfires from space

Australian scientists are working to 
provide faster and more reliable bushfire 
detection as part of the Kanyini mission 
[3]. This mission is led by the University 
of South Australia and supported by the 
Cooperative Research Centre for Smart 
Satellite Technologies and Analytics.

Researchers are using a satellite equipped with a 
hyperspectral imager that captures reflected light from 
Earth across various wavelengths to create detailed 
surface maps for monitoring bushfires, water quality 
and land management [9]. A computationally efficient 
(lightweight) AI model processes these satellite data 
onboard and differentiates smoke from cloud. 

This approach allows for the early detection of fire 
smoke before the fire generates significant heat, 
resulting in quicker alerts to ground crews and 
response times. The performance of the Kanyini system 
marks a substantial improvement over conventional 
approaches, detecting smoke 500 times faster than 
traditional ground-based methods [3]. The system 
is scheduled for full implementation in 2025. 

True colour image of Kangaroo 
Island taken by the Kanyini 
satellite on 2 February 2025.

©2025 SmartSat CRC. This image 
has been used with the express 
permission of SmartSat CRC.



Social media is also increasingly used in AI applications 
to streamline rescue and evacuation operations. Here, 
natural language processing models are used to analyse 
social media posts during a disaster event to assist rescue 
teams in detecting and prioritising requests for help [71]. 
CSIRO has worked with emergency services agencies in 
Queensland to use social media data and other datasets to 
forecast call centre requests during an emergency event. 
AI has also been used with social media data to provide 
a fine‑grained temporal and geospatial representation of 
affected individuals and areas during and after the event [72]. 

AI can support more sustainable workloads for emergency 
management personnel working in call centres. 
These personnel are often required to process, interpret 
and respond to information from many sources during 
an emergency. The RapidSOS system used in emergency 
call centres in the US uses AI to generate insights and 
alerts based on data collected from a range of devices, 
informing operators’ decisions about crisis response [73]. 
AI-based recommendation systems can reduce staff 
information overload and decrease perceived workloads 
and heart rates, indicating potential decreases in stress 
levels [74, 75]. However, the evidence is mixed on how these 
systems impact operators’ situational awareness [74, 75].

Future climate-related use cases for AI 
in emergency management

AI could be used to combine various data sources 
(e.g. social media, prior interactions with emergency 
or government assistance, mobile phone usage) to 
provide tailored emergency information to individuals 
and households based on their risk profile [76]. 
This information could be delivered via AI-enabled 
chatbots to help citizens prepare during high-risk 
periods or take the appropriate safety measures during 
a natural hazard [76]. To ensure AI applications are safe 
and responsible, developers need to consider the ethical, 
data security and privacy issues associated with using 
AI in these emergency management contexts [76].

With the prevalence of cybercrime reports on the rise 
in Australia [77], there is a growing need to ensure that 
emergency alert systems are secure and resilient to 
cybersecurity risks. For instance, there is the risk that 
cyber actors could gain access to these systems and issue 
fake alerts to the public. AI could help to detect false 
or malicious emergency alerts. Increasing reliance 
on AI-generated predictions and alerts in emergency 
management creates a need for systems, such as an 
immutable record, that allow users to establish the 
provenance and quality of forecasts and information.

Agriculture

How can AI help the agricultural sector 
adapt to a changing climate?

AI tools can provide climate insights to farmers to help 
them optimise farm operations and strengthen climate 
resilience in the face of more severe and frequent 
extreme weather events. In collaboration with the 
Bureau of Meteorology, CSIRO has developed My Climate 
View – a digital tool that provides Australian farmers 
with insights into current climate trends relevant to 
their specific region and commodity [78]. The Australian 
Bureau of Agricultural and Resource Economics and 
Sciences (ABARES) has also developed ‘farmpredict’ 
which is a microsimulation model that uses ML to make 
climate‑informed predictions on a range of business inputs 
and outputs (see Case study: ABARES farmpredict model).

Precision agriculture technologies can help farmers tailor 
their operations to crop and environmental conditions. 
AI can optimise and automate precision agriculture 
systems based on current climate conditions [79], helping 
farmers adapt to changing and variable climate conditions 
while adopting more sustainable farming practices. 
Existing applications use AI to draw on data about climate, 
soil, crop and weather conditions to automate irrigation 
decisions [80] and reduce herbicide use through targeted 
weed and pest spraying [81, 82]. Future developments in 
wireless sensors and AI algorithms could enable sensors 
to process information and make on-farm decisions 
autonomously without relying on external servers [83].

AI can accelerate the identification of plant varieties 
that are more responsive or resilient to emerging 
climate conditions [84]. By analysing large and 
complex datasets generated by genomics, phenomics 
and other ‘omics’ technologies, AI can simulate plant 
responses to environmental stressors and identify 
genetic markers associated with desirable traits 
[84]. While existing AI applications have focused on 
single‑omics data, emerging AI methods have the 
potential to integrate data from multiple sources [84].



CASE STUDY 

ABARES farmpredict model

ABARES farmpredict is a microsimulation 
model designed to analyse Australian 
broadacre farming businesses [2]. 
The model simulates physical and 
financial outcomes for Australian farming 
businesses under prevailing climate 
conditions and commodity prices. 

It does so by drawing on a range of climate variables, 
including rainfall and temperature, to assess their 
impact on crop yields, livestock management and 
overall productivity. This approach enables a detailed 
understanding of how environmental conditions 
influence farming decisions and outcomes.

The simulation component of farmpredict builds on 
these statistical insights to generate scenario-based 
forecasts that simulate farm performance under different 
conditions [10]. The model forecasts production for crops, 
livestock and inventory holdings, and combines these 
with input and output prices to estimate financial results. 
The simulation provides insights into revenue, costs and 
stock changes, helping farmers make informed decisions 
and adapt to environmental and economic changes. 



How can AI help decarbonise 
the agricultural sector?

AI approaches are being increasingly used to measure 
and predict the carbon sequestration potential of land 
and oceans. ML approaches have been used to integrate 
satellite data and other data sources to measure soil carbon 
stocks [85] or to continuously monitor forest cover and 
its carbon impact, as illustrated by the Pachama platform 
[86]. In addition, CSIRO is working with the Australian 
Government and Google Australia to use ML to map 
seagrass ecosystems in the Indo-Pacific and Australia. 
Seagrass is a critical long-term carbon sink that can store 
35 times more carbon than tropical rainforests [87, 88]. 

Alternative proteins, such as plant-based proteins, are 
increasingly in demand and can significantly reduce the 
reliance on more carbon-intensive animal proteins [89]. 
AI can utilise vast amounts of scientific data to accelerate 
innovations in alternative proteins, including optimising 
the formulations needed for growing cells or identifying 
candidate bioactive ingredients [90–92]. Examples of 
companies using these applications include: Vivici, which 
has partnered with Ginkgo Bioworks in using AI to select 
promising strains for dairy protein production [93]; and 
Shiru, which has leveraged AlphaFold [94] to predict 
protein structure and to identify and produce proteins 
that can replace traditional food ingredients [95].

Future climate-related use cases for AI 
in agriculture 

Autonomous field management enabled by AI reflects 
the next evolution of precision agriculture, with emerging 
AI developments providing the building blocks towards 
fully autonomous farming systems [96, 97]. Here, AI 
can bring together large and more diverse datasets for 
predictive analysis [98], enhance pest and disease detection 
[99] and enable higher levels of autonomy in tasks such 
as spraying, planting and harvesting [100]. While fully 
autonomous farming systems are yet to be realised due 
to technical and non-technical reasons (e.g. legislative, 
economic) [101], they could provide avenues for farmers 
to streamline decisions and processes that reduce 
environmental impacts and maximise economic returns.

AI could be used to improve the granularity of emissions 
monitoring to the individual-property level, helping 
to identify and potentially incentivise farmers who 
are engaging in low-emissions farming practices. 
Farmers undertaking carbon sequestration activities are 
required to use very complex sensors and software and 
often need to rely on external consultants for support. 
Satellites equipped with spectrometers can measure 
changes in methane and nitrogen levels in the atmosphere, 
which could be integrated and analysed using AI to 
monitor emissions associated with farming practices.

Methane emissions produced from enteric fermentation 
in livestock (i.e. the natural process through which 
microbes break down food as part of the digestive 
process in ruminant animals, producing methane as 
a by-product) account for 70% of greenhouse gas 
emissions in Australian agriculture [102]. AI can help 
the agricultural sector reduce the amount of methane 
produced through enteric fermentation. This could 
include using AI in the discovery of new livestock feeds 
and feed supplements that inhibit methane production 
[102] or to identify low-methane breeding traits [103].



Energy

How can AI help the energy sector adapt 
to a changing climate?

Like other types of critical infrastructure, AI can 
support the maintenance and optimisation of energy 
infrastructure. Rather than relying on fixed schedules 
or reactive responses, deep learning models can bring 
together historical performance data, weather data and 
sensor data to predict potential failures based on weather 
conditions [104–106]. Researchers from the Argonne 
National Laboratory in the US have developed AI models for 
predicting maintenance issues in energy infrastructure and 
found these tools helped to reduce maintenance costs by 
up to 56% and unnecessary crew visits by up to 66% [107]. 

How can AI help decarbonise the 
energy sector?

One of the most common use cases for AI in the 
energy sector is to optimise energy distribution [108]. 
AI is already used to manage fluctuations across the 
energy grid, including controlling grid networks and 
forecasting supply and demand [109]. AI can support 
the consolidation of information across complex and 
diverse energy networks, which often include a range 
of energy sources and organisations. This information is 
being used to identify ways to optimise energy use and 
energy-related emissions as part of Alphabet’s Tapestry 
project (see Case study: Alphabet’s Tapestry project). 

With the growing adoption of renewable energy 
technologies, the energy grid is becoming more 
decentralised and variable [108]. AI can better support 
the management of renewable energy sources by 
improving supply predictions and aligning decisions with 
forecasted energy needs and weather conditions [108, 109]. 
For example, AI could be used to automatically adjust solar 
panels and wind turbines to maximise energy production 
or identify the optimal location for these systems based 
on weather conditions [109]. Google’s DeepMind has 
developed an AI algorithm that can predict wind power 
output up to 36 hours in advance, which has increased 
the value of its wind energy by approximately 20% [110]. 

Future climate-related use cases for AI 
in energy

As the uptake of distributed energy systems increases 
(e.g. residential solar photovoltaic units, electric vehicles, 
battery storage devices), the volume of energy-related 
data will grow too. There are opportunities to use AI to 
further leverage the value generated through these 
data to optimise energy grid management and provide 
real-time insights that can be used to detect vulnerabilities 
or threats in the system (e.g. due to misconfiguration 
or malware introduced by malicious actors). 

Lithium-ion batteries used in electric vehicles and other 
energy storage solutions are critical to the renewable 
energy transition. Lithium is a finite resource, however, 
creating the need to identify other candidate metals 
that could be used in batteries. AI can play a role in 
accelerating the materials discovery and design process 
for new energy storage solutions. Emerging examples 
have used AI combined with physics-based models 
and high-performance computing to identify other 
suitable candidates for battery applications [111–113]. 

Just as AI can be used to optimise the distribution of 
energy across a network, this technology could also 
be used to manage energy supply to electric vehicles 
based on optimal conditions. For example, aligning 
electric vehicle charging times to periods where energy 
demand is low or when renewable energy sources are 
high. These AI models could underpin dynamic pricing 
models for electric vehicle charging to incentivise 
charging during off-peak (cheaper) periods [114].



CASE STUDY

Alphabet’s Tapestry project

Alphabet’s Tapestry project aims 
to improve the management of 
the electricity grid, shifting from 
a one-way flow of electricity from 
fossil fuel plants to consumers to 
a more distributed network [1]. 

Currently, electricity grids lack cohesive management 
and visibility, with information fragmented across 
various organisations. Tapestry seeks to address these 
challenges by increasing grid visibility to promote a 
greener and more reliable system for stakeholders.

Tapestry is developing AI-powered tools to modernise grid 
management and planning [1]. They include GridAware, 
which accelerates and automates asset inspections, 
and the Grid Planning Tool, which simplifies complex 
simulations of various scenarios (e.g. the impact of 
low wind on power usage during peak demand). 

Tapestry and CSIRO are collaborating to develop advanced 
smart inverters to improve the integration of renewable 
energy into the grid [11]. Traditional inverters are often 
inefficient and costly and can cause grid instability. 
The Tapestry and CSIRO smart inverters will feature 
grid‑forming intelligence, allowing them to actively 
manage energy flow, communicate with other grid devices 
and respond dynamically to real-time conditions.



Water

How can AI help the water sector adapt 
to a changing climate?

AI can help anticipate future water demand and optimise 
water sources, particularly under a more variable climate 
with more extremes (droughts and floods). AI algorithms 
can model complex, non-linear hydrological processes 
and integrate data from diverse sources, such as ground 
gauges, remote sensors and Internet of Things devices 
[115–119]. CSIRO has developed WaterWise, a cloud-based 
platform that uses data from soil-based sensors and ML 
to predict future water needs in real time [120]. If this 
technology was rolled out across Australia’s four most 
water-intensive crops (cotton, sugarcane, tomatoes and 
almonds), it could generate $1 billion in value by 2030 [120].

Traditional irrigation systems in urban environments 
operate on pre-set timers. Several research initiatives 
have explored how AI combined with sensor data and 
cloud computing can improve the responsiveness of 
urban irrigation systems. Researchers from Central 
Queensland University used AI to automate irrigation 
decisions using data from sensors installed in parklands 
in Cairns, which helped save 583 litres of water per 
square metre each year [121]. The New South Wales 
(NSW) Government similarly used ML to optimise water 
efficiency in parks based on weather conditions [122].

Natural hazards can pose water quality risks through the 
growth of harmful bacteria and pathogens at higher water 
temperatures or agricultural run-off from floodings. AI can 
support the capacity to monitor water quality risks under 
changing climate conditions. For example, Seqwater and 
the Queensland University of Technology have developed 
the SAMMI (Seqwater’s Autonomous Motorised Monitoring 
Instrument) robot, which uses AI to autonomously collect 
water samples from difficult-to-reach locations (see Case 
study: SAMMI by Seqwater). Researchers from Los Alamos 
National Laboratory are also investigating the use of 
AI to forecast harmful algal blooms from water sample 
data, weather conditions and satellite imagery [123]. 

How can AI help decarbonise the 
water sector?

AI can be used to optimise the operation of water 
treatment plants, reducing greenhouse gas emissions. 
An example is the DARROW project, a European research 
initiative, which uses AI on data collected from sensors 
that monitor water quality and treatment processes 
[124–126]. Such models will provide insights that can 
be used to inform plant operations (e.g. appropriate 
chemical treatment levels), reduce energy consumption, 
increase biogas production and recover valuable 
resources such as phosphorus and nitrogen.

Future climate-related use cases 
for AI in water

Deciding on which water source to draw from (e.g. dam 
water, desalination, treated recycled water) is currently 
a manual process. AI could support the optimisation 
of water supply decisions, taking into account current 
and forecasted water supply, capacity, pressure within 
the system and energy use associated with distributing 
water. AI could also play a role in increasing industrial 
use of treated recycled water where appropriate, 
reducing unnecessary use of freshwater sources.

Data centres are a growing source of water demand 
given that they rely upon water-based cooling systems 
to manage excess heat generated by servers. There are 
emerging circular models for repurposing this heated 
water (see Responsible AI-for-climate-related applications) 
and AI could also play a role in coordinating water 
flows as part of circular solutions for data centres, 
optimising water use between data centres and other 
domains that can repurpose the heated water. 

Changes in climate conditions can impact water quantity 
and quality, but current modelling approaches do not 
integrate climate projections with long-term water 
forecasting. AI could be used to integrate climate and 
water forecasting capabilities to inform water planning 
decisions during periods of extreme shortage (droughts) 
or abundance (flooding). WaterNSW has integrated 
climate projections with demand forecasting through 
its NSW and ACT (Australian Capital Territory) Regional 
Climate Modelling (NARCliM) project as an emerging 
example of what this capability could look like [127]. 

While there are promising capabilities for detecting 
algal blooms in water samples, similar capabilities are 
needed to monitor the proliferation of deleterious 
bacteria in water sources [128]. AI tools could be utilised 
to efficiently monitor and detect unsafe levels of 
harmful bacteria in water. Researchers have trained 
neural networks to detect specific types of bacteria 
in water samples with greater precision and speed 
than traditional methods [129]. Further development 
of these methods and their application in routine 
water monitoring workflows could reduce the risk of 
exposing the population to harmful water sources. 



CASE STUDY

SAMMI by Seqwater

The SAMMI (Seqwater’s Autonomous 
Motorised Monitoring Instrument) robot 
developed by Seqwater in collaboration 
with the Queensland University of 
Technology is improving water quality 
monitoring in South East Queensland 
[5, 6]. SAMMI is a solar‑powered, 
self‑driving robot designed to operate 
autonomously in waterways. 

Since its launch in 2019, SAMMI has demonstrated 
its ability to collect water samples, measure quality 
parameters and create sonar maps of reservoirs. 
These robots enable more frequent measurements 
of water quality and wider spatial coverage than 
traditional methods, including hard-to-reach areas. 

New versions of SAMMI extend the capabilities of the 
initial version with multiple AI-based enhancements 
[6]. For example, SAMMI 2 can autonomously identify 
and spray weeds that are detrimental to water quality, 
such as water hyacinth and salvinia. Future versions 
will focus on collision-avoidance technology to 
navigate around moving objects and debris.



Healthcare 

How can AI help the healthcare sector adapt 
to a changing climate?

A growing prevalence of extreme heat events heightens the 
risk of heat-related morbidity and mortality, particularly 
among vulnerable populations (i.e. those aged 65 years 
and over, young children or people with existing chronic 
conditions) [130, 131]. Researchers have used AI to predict 
climate-related health risks to help the healthcare 
sector prepare for surges in demand [132–135], including 
heatstroke prediction [133, 135] and forecasting mortality 
risks related to weather-sensitive cardiovascular diseases 
[134]. Other modelling approaches have quantified 
heat health risks across geographical areas to help 
decision-makers better prepare for extreme heat across 
Australia (e.g. the Heat-Health Risk Index [136, 137]). 

AI can also be used to monitor for workplace heat stress, 
which can lead to work accidents, lapses in concentration, 
fatigue and poor decision-making [138–140]. AI can 
analyse physiological data collected from body sensors 
(e.g. heart rate, temperature and humidity inside clothes, 
chemical markers in sweat) to detect signs of heat stress 
[141, 142]. These signals enable managers to intervene 
early, encouraging workers to take breaks or work 
indoors when they start showing signs of heat stress [141]. 
Australian-based start-up EMU Systems has developed an 
environmental monitoring system to predict heat stress in 
athletes, with broader application for workers in mining, 
manufacturing, construction and agriculture [143]. 

Climate change is also contributing to heightened infectious 
disease risks and the emergence of vector-borne infections 
in new locations [144]. Epidemic intelligence systems 
enabled by AI can be used to detect early warning signals 
of disease outbreaks, mapping these signals spatially, 
simulating the impact of different interventions and 
detecting sources of misinformation [145]. These capabilities 
have been demonstrated through platforms such as 
EPIWATCH, BlueDot and the Global Biosurveillance 
Portal, which use natural language processing to 
analyse social media data and news reports combined 
with location data to identify and categorise potential 
disease outbreaks at their precise locations [145–147]. 

AI can be used to monitor air quality, helping 
decision‑makers identify and respond to emerging health 
risks early. Air pollution and greenhouse gas emissions 
are often driven by common factors (e.g. burning of fossil 
fuels via coal-fired power plants), meaning that addressing 
one of these challenges (reducing emissions) can 
co‑benefit the other (improved health outcomes) [148–151]. 
BreezoMeter is a platform that uses ML to analyse data 
from various sources (e.g. air quality monitoring stations, 
satellite and meteorological data) to create high-resolution 
real‑time heat maps of pollution and pollen levels [152]. 



Future climate-related use cases of AI 
in healthcare

A changing climate poses significant risks to populations 
living in regions that are predicted to exceed the ‘human 
climate niche’ (i.e. the climate conditions favourable 
to sustaining human life and activity) [153]. To support 
proactive responses to these climate risks, AI could 
be used to monitor and identify locations that are 
trending towards exceeding the human climate 
niche, informing community relocation responses. 
Existing modelling approaches can quantify how 
global populations might be impacted under differing 
degrees of global warming [154], which could be 
used as the basis for such early warning systems.

With the increasing digitisation of health records, there 
are opportunities to expand the use of AI to predict 
climate-related impacts on the health system [155]. 
Example use cases highlighted by the experts we consulted 
in developing this roadmap paper include using AI to 
integrate health system data with climate modelling to 
provide more precise estimates of the impacts of extreme 
weather events on healthcare services. There could 
also be opportunities to factor healthcare-related costs 
of climate change into existing economic modelling 
approaches [155] and strengthen the case for investment 
into climate adaptation measures in healthcare [156, 157].

Global supply chains account for approximately 75% of 
the Australian health system’s carbon footprint [158]. 
Australia has recently jointly committed with the US 
and United Kingdom (UK) to decarbonise healthcare 
supply chains [158] and the Australian Government is 
developing a health system decarbonisation roadmap 
(due for release in 2025), which will cover scope 1, 2 and 3 
emissions and patient travel [159]. Just as AI has been used 
to monitor and manage emissions across supply chains 
in other industries (see Critical Infrastructure), AI could be 
used to bring together diverse datasets and quantify 
supply chain emissions in the healthcare sector.

Other cross-domain applications

AI can translate and tailor climate insights for a variety 
of users. Emerging examples, such as ChatClimate, are 
leveraging the conversational capabilities of generative 
AI tools to personalise climate insights to specific 
queries (see Case study: ChatClimate). Another platform, 
This Climate Does Not Exist, has used AI to generate images 
that illustrate plausible future climate scenarios tailored to 
an individual’s location to raise awareness and encourage 
climate action [160]. These capabilities demonstrate how AI 
could be used as a ‘data concierge’ to help decision-makers 
understand climate risks and the actions they can take.

Methane modelling is central to decarbonisation efforts 
across a range of domains, particularly agriculture, energy 
and water – the largest anthropogenic sources of methane 
emissions [161]. The Global Methane Pledge has been 
established to accelerate global efforts to reduce methane 
emissions by at least 30% from 2020 levels by 2030 to limit 
global warming to 1.5 °C [162]. AI has been used to improve 
the timeliness, accuracy and granularity of methane 
emissions monitoring, including MethaneSAT [163], 
GHGSat’s SPECTRA platform [164] and Climate TRACE [165]. 
Climate TRACE arguably provides the most comprehensive 
global inventory of greenhouse gas emissions, including 
methane, covering over 350 million assets worldwide [165].



CASE STUDY

ChatClimate

ChatClimate leverages advanced natural 
language processing capabilities to 
provide timely access to reliable and 
credible information on climate change 
impacts [7, 8]. This conversational 
AI is underpinned by an LLM that 
was trained on data from the 
Intergovernmental Panel on Climate 
Change Sixth Assessment Report 
(IPCC AR6) to produce tailored answers 
to questions relating to climate change. 

The performance of this model has been tested against 
other current state-of-the-art LLMs (e.g. GPT-4) and a 
hybrid ChatClimate approach that utilises IPCC AR6 data as 
well as in-house GPT-4 knowledge to test the value of this 
potential use case [7]. When evaluated by a team of authors 
from the IPCC AR6 team, the hybrid ChatClimate solution 
was found to produce the most accurate responses [7]. 

These results demonstrate the value of a tailored 
LLM solution that considers current, domain-specific 
data [7]. While these tools will not negate the 
need to translate climate information in complex 
decision-making contexts, they could support 
knowledge-sharing in non-expert contexts.



Responsible AI-for-climate 
applications

AI offers great promise in accelerating climate mitigation 
and adaptation efforts, but there are potential risks 
that need to be managed. The design, development 
and implementation of climate-related AI applications 
must be done responsibly so that they do not add to, or 
exacerbate, existing social, economic or environmental 
challenges. This section highlights key considerations 
around energy and water use, AI ethics and Indigenous 
Cultural and Intellectual Property (ICIP) rights.

Ensuring climate-related AI is safe 
and responsible 

Australia, along with other national governments, has 
developed ethical guidelines for the safe, secure and 
responsible development and deployment of AI [166], 
and these need to be considered in climate-related AI 
applications. These include ensuring that AI systems:

• benefit human, societal and environmental wellbeing
• align with human values and rights, including 
diversity and the autonomy of individuals
• are inclusive, accessible and do not unfairly discriminate 
against individuals, communities or groups
• respect the privacy, protection and security of data
• are reliable, accurate and reproducible, operating 
in line with their intended purpose
• are transparent so that people understand when they 
are engaging with, or impacted by, an AI system
• can be contested when they significantly impact an 
individual, community, group or environment
• are accountable by those individuals or organisations 
who are responsible for the outcomes of the AI system.


Some of the use cases presented in this paper highlight 
specific responsible AI considerations. For example, in the 
case of emergency management, respect for the privacy, 
protection and security of people’s data collected from 
social media, government services and other sources 
needs to be considered in the development of AI-enabled 
evacuation alert systems that can personalise warnings to 
the public. Providing citizens with the option to opt out of 
these services could be one way to manage this concern.

The inclusivity and accessibility of AI tools also need 
to be considered. Given that social media data can 
be influenced by digital access and socioeconomic 
characteristics, applications that rely primarily on 
these data may be inadvertently biased towards certain 
community segments [72]. The broader concentration of 
AI developments and investments in the Global North – 
the world’s developed and wealthier nations [167] – could 
lead to geographical biases in AI models and applications 
and capability gaps in less-developed Global South 
nations. Targeted efforts, such as dedicated AI-for-climate 
funding opportunities for the Global South, have been 
introduced to address these geographical gaps [167].

The establishment of safe, secure and responsible design 
and deployment of climate-related AI systems is critical 
in building trust in these applications. The spread of 
misinformation has previously hindered the capacity 
for climate science to drive climate action [168], so 
emerging AI applications that aim to communicate 
climate risks or predictions (see, for example, Case study: 
ChatClimate) must be technically robust, transparent 
and reliable [169]. Research focused on advancing our 
understanding of best practices for fostering trust in, 
and the trustworthiness of, climate-related AI could 
support the translation of responsible AI principles into 
practice and the evaluation of trust in such tools [170].



The energy and emissions 
footprint of AI

The data centres and transmission networks that underpin 
AI are energy intensive [171]. The Electric Power Research 
Institute estimates that data centres could consume up to 
9.1% of electricity in the US by 2030 (compared with 4% in 
2023) depending on the pace of future developments in AI 
applications and energy efficiency measures [172]. If the size 
of these models and computing demand continue to grow 
at the current rate, the International Energy Agency predicts 
that AI-driven energy demands associated with data centres 
could exceed the existing energy efficiency gains [171]. 

The emissions associated with data centres will depend 
on access to clean energy sources. For instance, the carbon 
intensity associated with electricity generation is much 
less in Norway (30 gCO2 equivalents per kilowatt-hour) 
than in Australia (549 gCO2) [173]. Some of the world’s 
leading technology companies – including Microsoft, 
Google and Amazon – are exploring nuclear energy as 
a potential energy source, either by purchasing nuclear 
power or partnering with a nuclear energy provider [174].

Data centre energy needs are highly influenced by 
server and infrastructure requirements. A Google-led 
analysis estimated that its cloud data centre infrastructure 
was 1.4–2 times more energy efficient than traditional 
data centres and its hardware specifically designed for 
ML training and inference was 2–5 times more efficient 
than generic systems [175]. Importantly, these analyses 
consider only the emissions associated with operating 
data centres, but the embodied carbon associated with 
constructing data centres and server hardware is another 
key consideration in understanding AI’s emissions footprint.

The water footprint of AI

Data centres generate heat, with water-based systems 
being the most common approach for keeping 
servers and infrastructure at the optimal temperature. 
Like energy, future expansions in AI could place 
unsustainable pressure on water reserves if not 
well managed. There are emerging circular energy 
models where heated water from data centres has been 
repurposed for aquaculture production [176] and to 
heat residential homes [177, 178] and public swimming 
pools [179]. The viability of circular models associated 
with data centres will depend on proximity to partners 
who can repurpose water and other resources. 

Recognising and respecting 
Indigenous Cultural and 
Intellectual Property

Researchers, developers and other partners need to 
consider ICIP rights when engaging and working 
with First Nations p eople on climate-related AI 
applications. ICIP rights reflect the rights of First Nations 
people to the tangible and intangible aspects of their 
cultural heritage. The importance of First Nations data 
sovereignty and ICIP rights is acknowledged in the safe 
and responsible development and deployment of AI 
in the Australian Government’s proposed mandatory 
guardrails for AI in high-risk settings [180].

There are existing examples where ICIP has been 
misappropriately used without consent or attribution. 
These include generative AI tools that have been 
trained on First Nations artworks without permission 
and used to create inauthentic AI-generated First 
Nations art [180, 181]. These practices can pose risks to 
ICIP and increase the spread of misinformation about 
First Nations people when false or inappropriate data 
sources are used to generate AI outputs [182].

To protect ICIP in climate-related AI applications, First 
Nations people should be involved in the AI design and 
development to identify where ICIP is used inappropriately 
or misused and to provide oversight of the information used 
to train the AI system [183]. CSIRO has previously partnered 
with rangers and Traditional Owners in the Kakadu National 
Park to use AI and First Nations knowledge to support 
the management of para grass – an invasive weed that 
impacts wetlands [184]. CSIRO is also partnering with First 
Nations communities in developing digital capabilities 
under the Climate Services for Agriculture initiative.



Pathways to accelerate 
AI‑for-climate R&D

In addition to identifying where AI could be applied 
in response to climate challenges, this paper serves 
as a discussion starter on how the Australian R&D 
ecosystem can leverage AI to accelerate climate action. 
This section presents a series of options, drawing 
on expert insights and learnings from related local 
and international examples. These suggestions are 
intended to stimulate conversations about how industry, 
government, academia and not-for-profit sectors can 
work together towards AI-for-climate opportunities.

Bringing together multidisciplinary 
teams of technical and 
domain experts

AI-for-climate research requires multidisciplinary 
teams with expertise and capabilities in climate science, 
AI and the application domain. ‘Knowledge brokers’ 
are also valued team members who can assist with 
relationship building, cross-disciplinary knowledge 
sharing and feedback loops between researchers 
and end users. Embedding these multidisciplinary 
capabilities in education and training pathways for 
emerging scientists and engineers could further help 
build capacity for future AI-for-climate applications.

Experts consulted in developing this paper emphasised 
that it is insufficient for AI experts to only consult domain 
experts at the start of the R&D process. The connection 
between the AI and domain experts needs to be embedded 
across multiple stages so that the AI experts understand 
the nuances of the data and the domain experts can 
understand the assumptions and constraints underpinning 
the AI model. Through this process, domain experts 
can also improve their AI literacy around the types of 
problems for which AI is best suited and its limitations.

You can’t throw things over and expect 
the AI researchers to understand the 
intricacies of the dataset… You have to 
engage in the design process so that the 
people who are designing [the model] 
understand what the needs are and the 
requirements. Also, the knowledge transfer 
has to happen the other way around. If you 
are going to consume the model, you need 
to understand the assumptions made in [its] 
architecture and design so that you know 
where it’ll work and where it won’t work.

– Expert interviewee 

WHAT THIS COULD LOOK LIKE

EXISTING EXAMPLES FROM WHICH WE CAN LEARN

Engage a network of Australian 
and international climate, AI 
and domain experts to come 
together (e.g. via a regular 
annual or bi-annual forum) to 
share knowledge and learnings 
on the intersection of AI and 
climate action

CSIRO hosted an AI for Climate Symposium in 2024 that brought together over 120 Australian and 
international AI and climate experts across industry, government and academia [185]. It included 
presentations on current AI use cases for climate mitigation and adaptation challenges and a 
‘design sprint’ that encouraged participants to develop novel AI solutions to an emerging climate 
challenge. This event raised awareness around the need for stronger cross-sector collaboration in 
AI-for-climate opportunities in an Australian context.

The Climate Change AI Summer Schools bring together a global network of industry, government 
and academic participants working at the intersection of climate change and ML [186]. Running since 
2022, the Summer School is an annual, hybrid event held over seven weeks and attracting up to 10,000 
participants from over 140 countries. It includes lectures on use cases for AI and ML in addressing 
climate change and hands-on tutorials on applying AI and ML to solve climate-related problems.

Support programs for early 
and mid-career researchers 
(e.g. PhDs, fellowships and 
other professional training 
programs) that foster the next 
generation of interdisciplinary 
experts bridging linkages 
between AI and climate in 
different domains

UK Research and Innovation (UKRI) has funded a series of Centres for Doctoral Training that support 
PhD positions at the intersection between AI and climate, environment and sustainability domains. 
These include Intelligent Earth: UKRI AI Centre for Doctoral Training in AI for the Environment 
hosted by the University of Oxford [187]; AI for the Study of Environmental Risks (AI4ER): UK 
Centre for Doctoral Training hosted by Cambridge University [188]; and the UKRI AI Centre for 
Doctoral Training in AI for Sustainability (SustAI) hosted by the University of Southampton [189]. 

In the US, Morgan State University has established the National Science Foundation (NSF) Research 
Traineeship (NRT) program on Artificial Intelligence for Changing Climate and Environmental 
Sustainability (ACCESS) [190]. The program aims to provide PhD students with interdisciplinary 
training in environmental science, water quality management, climate science, and AI and ML to 
support the application of AI solutions for climate action.







Supporting research collaboration 
and innovation across sectors

Partnerships and the coordination of R&D activities 
are critical when using AI to support climate mitigation 
and adaptation, given the global nature of climate 
change and the scale and urgency of climate action 
required. This type of research requires a diverse 
combination of resources and expertise that expands the 
boundaries of any one organisation, sector or region.

For example, as highlighted in the Critical infrastructure 
section, natural hazard risks impacting water, energy, 
transport, telecommunications and healthcare 
infrastructure are highly connected. When one node 
breaks down, these impacts can flow through to 
other parts of the network. By joining forces across 
organisations, sectors and disciplines, there could be 
opportunities to maximise and scale mitigation and 
adaptation efforts, increase the use of shared tools 
and approaches, and prevent duplicated efforts .

What needs to happen is collaborations. 
AI for science requires collaborations 
of different groups providing different 
expertise in [skills] and resources and 
trying to solve a problem together. 

– Expert interviewee

WHAT THIS COULD LOOK LIKE

EXISTING EXAMPLES FROM WHICH WE CAN LEARN

Establish a dedicated 
AI‑for‑climate grand challenge 
or innovation hub for 
researchers, innovators and 
end users to collaborate, 
share resources and develop 
novel AI‑enabled solutions in 
response to emerging climate 
challenges impacting Australia 
and Pacific Island nations

The AI Innovation Grand Challenge was held in 2024 as part of the United Nations (UN) Framework 
Convention on Climate Change’s Technology Mechanism Initiative on Artificial Intelligence for 
Climate Action [191]. This grand challenge was designed to bring together teams of students, 
entrepreneurs, academics and non-government organisations to pitch an AI solution for climate 
mitigation and adaptation. All AI applications developed through this challenge were open source, 
with a strong focus on building capability in developing countries.

The Artificial Intelligence for Decarbonisation’s Virtual Centre of Excellence (ADViCE) initiative – 
funded by the UK Government and delivered by Digital Catapult, Energy Systems Catapult and 
The Alan Turing Institute – is focused on the development of AI applications that support the 
transition to net zero [192]. This virtual centre brings decarbonisation stakeholders across sectors 
together to collaborate, disseminate information, and identify priority decarbonisation challenges.

Other grand challenges and innovation hubs have focused on specific sectors or 
geographies. These include AI for Climate Resilience in Rural Areas, which aims to identify 
AI‑driven solutions that respond to climate challenges facing rural communities in Asia, 
Africa, Latin America, the Caribbean and beyond [193]. AI for Good, convened by the UN’s 
International Telecommunication Union, the Government of Switzerland and other UN agency 
partners, hosted a series of AI/ML Solutions for Climate Change initiatives in 2023 focused on 
water management, food and agriculture and 5G network energy consumption [194]. Finally, the 
USD$100 million Bezos Earth Fund’s AI for Climate and Nature Grand Challenge aims to create and 
scale AI solutions and incentivise cross-sectoral partnerships in the areas of sustainable proteins, 
energy grid optimisation and biodiversity conservation [195]. 

Climate Change AI has systematically analysed 215 past and ongoing grand challenges, competitive 
grants, ‘hackathons’, incubators and accelerators, and innovation hubs globally in climate, AI and 
AI-for-climate domains [167]. While these initiatives have grown exponentially since 2019, most are 
concentrated in the Global North. Successful initiatives tend to involve heterogeneous participant 
cohorts, position AI innovations as a vehicle for achieving a climate or nature objective (but not the 
primary focus), encourage scalable solutions and build in evaluation measures to assess the impact 
of projects.







Setting the strategic focus with 
a clear pathway to impact

A large share of current AI-for-climate developments 
have been opportunistic and lack a clear path to impact 
or, where relevant, a path to market. There could be 
opportunities to develop a set of sector-specific or 
national priorities or challenges that consider what is 
technically feasible, how AI could have maximum impact 
and where there are ‘low-hanging fruit’ opportunities.

It is important to acknowledge that some climate 
challenges that need to be solved may have a strong public 
interest but limited commercial imperative (e.g. emissions 
or deforestation monitoring platforms). Such AI use 
cases could present a role for government and the 
not-for-profit sector in supporting the development 
of AI-for-climate applications where there are 
insufficient commercial incentives, ensuring that these 
developments create opportunities for shared benefits.

I think there is a lot to be said for programs 
that are thoughtfully put together and 
bring together the smartest heads to solve 
a specific problem and then figure out 
how you can get that solution a pathway 
to market where it can have impact.

– Expert interviewee

WHAT THIS COULD LOOK LIKE

EXISTING EXAMPLES FROM WHICH WE CAN LEARN

Define a set of priority 
sectors, national challenges 
or use cases for AI for climate 
mitigation and adaptation to 
coordinate resources across 
public, private, academic and 
not-for-profit sectors

The ADViCE initiative was defined around four high-emitting sectors, comprising agriculture, 
the built environment, energy and manufacturing [196]. A set of AI-for-decarbonisation 
grand challenges was then identified under each of these sectors (i.e. unlocking domestic 
decarbonisation, enabling net zero infrastructure, maximising flexibility in energy networks, 
decarbonising manufacturing inputs, improving manufacturing process efficiency, optimising 
soil management and minimising methane in agriculture [196]). Identifying these priority areas is 
designed to support stakeholders in connecting with relevant partners and capabilities, improve 
access to data needed to develop AI solutions and stimulate investment opportunities [196].

Identify pathways and 
partnerships to support the 
development and maintenance 
of AI-for-climate solutions 
that support public interest, 
non‑commercial datasets, 
models and applications

The Lacuna Fund aims to address the data gaps that commonly exist for low- and middle‑income 
countries, which can lead to biased or harmful outcomes for marginalised populations when 
these data are used in ML [197]. Funded by public-sector agencies, private organisations 
and philanthropies, the Lacuna Fund has a specific focus on supporting the creation and 
maintenance of datasets and ML models to help low- and middle-income communities respond 
to climate change.







Promoting trusted, safe 
and responsible AI

In ensuring that AI-for-climate applications are safe and 
responsible, there is a need to create trust with users 
so that organisations and decision‑makers can be 
confident using the insights generated by such 
tools [198]. Experts consulted in developing this 
roadmap raised concerns about relying on AI systems 
that have not been validated or do not report on 
the performance of the model. They cited instances 
where commercial providers claimed that their tools 
could provide predictions at a level of granularity 
that is not possible with state-of-the-art offerings.

One way to strengthen trust in AI-for-climate tools is 
through transparency. Transparency can be achieved 
through sharing datasets and code, acknowledging 
the sources of training data, validating and evaluating 
the impacts of the model, and publishing on the model 
performance. Experts we spoke to were supportive of the 
sharing of climate modelling, datasets and AI tools so that 
they could be tested and improved via collaborations across 
organisational and national boundaries. Shared resources 
could include benchmarked datasets, open‑source 
code and results from tests and evaluation exercises.

WHAT THIS COULD LOOK LIKE

EXISTING EXAMPLES FROM WHICH WE CAN LEARN

Create a governance 
framework that can be used to 
evaluate alignment between 
AI‑for‑climate applications with 
Australia’s AI ethical principles 
and guardrails 

Building on Australia’s AI Ethics Principles [166], several resources have been developed to support 
Australian organisations in aligning their development and use of AI to these principles.

The Gradient Institute, in collaboration with the National Artificial Intelligence Centre and CSIRO, 
has developed a practical set of guidelines for Implementing Australia’s AI Ethics Principles [199]. 
These guidelines are designed to raise awareness around Australia’s AI Ethics Principles among 
senior leaders, system owners and AI developers. They also provide practical resources for ensuring 
adherence to these principles.

The Australian Government has released a Voluntary AI Safety Standard [198], which specifies 
10 voluntary guardrails for Australian organisations to ensure safe and responsible AI use. 
This standard aligns with emerging international standards and practices around safe and 
responsible AI use.

Federal, state and territory governments in Australia have jointly established a National Framework 
for the Assurance of Artificial Intelligence in Government [200]. This framework is designed 
to support government agencies in applying Australia’s AI Ethics Principles when designing, 
developing and implementing AI safely and responsibly.

Encourage and incentivise 
open access to climate‑related 
datasets, models and 
tools developed across 
research, private, public and 
not‑for‑profit organisations

Climate TRACE, mentioned in Other cross-domain applications, is a source of timely and granular 
global emissions information and an example of an open and accessible climate-related data 
asset [165]. Building on initial funding from Google in 2019 to monitor power plant emissions, 
this platform has been expanded with other collaborators and partners to now cover up to 75% of 
global asset-level emissions. All data and methodology documentation housed on this platform are 
free and publicly available to maximise the extent to which these resources can be used to drive 
climate action.

To help accelerate progress towards addressing gaps in quality and accessible data needed in 
AI-for-climate applications, Climate Change AI has established the CCAI Dataset Wishlist [201]. 
This platform allows stakeholders across sectors to identify and classify critical data gaps that are 
limiting their ability to use AI in response to a climate mitigation or adaptation challenge. Users can 
categorise the current state of available data (e.g. private data that need to be released or public 
data that need structure).







Participants from the design sprint 
at CSIRO’s AI for Climate Symposium 
held on 16–17 October 2024 at the 
State Library of Victoria, Australia



References

Note: For brevity, minimal reference details are provided. However, most reports can be readily accessed by copying 
the information provided into a search engine. All online publications were accessible as of 21 March 2025.

1. Tapestry, 2025. Home page. Available from: https://x.company/projects/
tapestry/

2. Australian Bureau of Agricultural and Resource Economics and Sciences, 
2024. ABARES farmpredict model. Available from: https://www.
agriculture.gov.au/abares/research-topics/climate/drought/farmpredict

3. UniSA, 2024. Fighting fires from space in record time: How AI could 
prevent a repeat of Australia’s devastating wildfires. University of South 
Australia, 5 June.

4. Hajkowicz S, Karimi S, Wark T, Chen C, et al., 2019. Artificial intelligence: 
Solving problems, growing the economy and improving our quality of 
life. CSIRO.

5. Water Source , 2019. SAMMI the robot sets sail to safeguard Seqwater’s 
water supply. Australian Water Association, 15 May.

6. Seqwater, 2022. Artificial intelligence and robotics to revolutionise water 
quality monitoring. Australian Water Association, 9 May.

7. Vaghefi SA, Stammbach D, Muccione V, Bingler J, et al., 2023. 
ChatClimate: Grounding conversational AI in climate science. 
Communications Earth & Environment, 4(1):480.

8. ChatClimate, 2025. Home page. Available from: https://www.chatclimate.ai/

9. Lu S, Jones E, Zhao L, Sun Y, et al., 2024. Onboard AI for fire smoke 
detection using hyperspectral imagery: An emulation for the upcoming 
Kanyini Hyperscout-2 Mission. IEEE Journal of Selected Topics in Applied 
Earth Observations and Remote Sensing 17:9629–9640.

10. Hughes N, Soh WY, Lawson K & Lu M, 2022. Improving the performance 
of micro-simulation models with machine learning: The case of 
Australian farms. Economic Modelling, 115:105957.

11. Tapestry, 2025. Case study: Tapestry and Australia’s Commonwealth 
Scientific and Industrial Research Organisation (CSIRO) – prototyping 
advanced inverters to power a sustainable future. Available from: 
https://x.company/case-study/tapestry-csiro/ 

12. UNEP, 2024. No more hot air ... please! With a massive gap between 
rhetoric and reality, countries draft new climate commitments. Emissions 
Gap Report 2024. United Nations Environment Programme.

13. Deloitte Access Economics, 2021. Special report: Update to the economic 
costs of natural disasters in Australia. Australian Business Roundtable for 
Disaster Resilience & Safer Communities.

14. Maher H, Meinecke H, Gromier E, Garcia-Novelli M, et al., 2022. AI is 
essential for solving the climate crisis. Boston Consulting Group, 7 July.

15. Dannouni A , Deutscher SA, Dezzaz G, Elman A, et al., 2023. Accelerating 
climate action with AI. Boston Consulting Group.

16. Wong C, 2024. Superfast Microsoft AI is first to predict air pollution for 
the whole world. Nature, News, 4 June.

17. Huawei, 2023 . Huawei Cloud Pangu-Weather Model now available on 
European weather agency website, 3 August.

18. Nvidia, 2024. FourCastNet. Nvidia.

19. Price I, Sanchez-Gonzalez A, Alet F, Andersson TR, et al., 2025. 
Probabilistic weather forecasting with machine learning. Nature, 
637(8044):84–90.

20. Lam R, Sanchez-Gonzalez A, Willson M, Wirnsberger P, et al., 2023. 
Learning skillful medium-range global weather forecasting. Science, 
382(6677):1416–1421.

21. Climate Modeling Alliance, 2025. Home page. Available from: https://
clima.caltech.edu/

22. European Centre for Medium-Range Weather Forecasts, 2025. Artificial 
Intelligence in DestinE: The explainer. Available from: https://destine.
ecmwf.int/artificial-intelligence-in-destine-the-explainer/

23. Kitsios V, O’Kane TJ & Newth D, 2023. A machine learning approach to 
rapidly project climate responses under a multitude of net-zero emission 
pathways. Communications Earth & Environment, 4(1):355.

24. Nguyen T, Brandstetter J, Kapoor A, Gupta JK, et al., 2023. ClimaX: A 
foundation model for weather and climate. arXiv:2301.10343.

25. Kochkov D, Yuval J, Langmore I, Norgaard P, et al., 2024. Neural general 
circulation models for weather and climate. Nature, 632(8027):1060–1066.

26. Saha A & Ravela S, 2024. Statistical-physical adversarial learning from 
data and models for downscaling rainfall extremes. Journal of Advances 
in Modeling Earth Systems, 16:e2023MS003860.

27. IBM, 2023. IBM and NASA open source largest geospatial AI foundation 
model on Hugging Face. IBM, 3 August.

28. NASA, 2023. NASA and IBM openly release geospatial AI foundation 
model for NASA earth observation data. National Aeronautical and Space 
Administration, 3 August.

29. IBM, 2023. IBM and NASA are building an AI foundation model for 
weather and climate. IBM, 29 November.

30. Hajkowicz SA, 2024. Artificial intelligence foundation models: Industry 
enablement, productivity growth, policy levers and sovereign capability 
considerations for Australia. CSIRO.

31. UNDP, 2024. Making our future: New directions for human development 
in Asia and the Pacific. United Nations Development Programme.

32. Department of Foreign Affairs and Trade, 2024. The Pacific: Building a 
stronger and more united Pacific family. Available from: https://www.
dfat.gov.au/geo/pacific

33. Department of Home Affairs, 2023. Critical infrastructure resilience 
strategy. Australian Government.

34. Spencer BF, Hoskere V & Narazaki Y, 2019. Advances in computer vision-
based civil infrastructure inspection and monitoring. Engineering, 
5(2):199–222.

35. Yang J, Mostaghimi H, Hugo R & Park SS, et al., 2022. Pipeline leak and 
volume rate detections through artificial intelligence and vibration 
analysis. Measurement 187:110368.

36. Hajkowicz S, Bratanova A, Schleiger E & Naughtin C, 2023. Australia’s 
artificial intelligence ecosystem: Catalysing an AI industry. CSIRO.

37. Alam F, Imran M & Ofli F, 2017. Image4act: Online social media image 
processing for disaster response. In: Proceedings of the 2017 IEEE/ACM 
International Conference on Advances in Social Networks Analysis and 
Mining, p.601–604.

38. Presa-Reyes M & Chen S-C, 2020. Assessing building damage by learning 
the deep feature correspondence of before and after aerial images. 
In: 2020 IEEE Conference on Multimedia Information Processing and 
Retrieval (MIPR), p.43–48.

39. Sit MA, Koylu C & Demir I, 2019. Identifying disaster-related tweets and 
their semantic, spatial and temporal context using deep learning, natural 
language processing and spatial analysis: A case study of Hurricane Irma. 
International Journal of Digital Earth, 12(11):1205–1229.

40. Paul U, Ermakov A, Nekrasov M, Adarsh V, et al, 2020. #Outage: 
Detecting power and communication outages from social networks. In: 
Proceedings of The Web Conference 2020. Association for Computer 
Machinery, p.1819–1829.

41. Cui H, Qiu S, Want Y, Zhang Y, et al., 2023. Disaster-caused power outage 
detection at night using VIIRS DNB images. Remote Sensing, 15(3):640–640.

42. Sriram LMK, Ulak MB, Ozguven EE & Arghandeh R, 2019. Multi-network 
vulnerability causal model for infrastructure co-resilience. IEEE Access, 
7:35344–35358.

43. Sun J, Bathgate K & Zhang Z, 2024. Bayesian network-based resilience 
assessment of interdependent infrastructure systems under optimal 
resource allocation strategies. Resilient Cities and Structures, 3(2):46–56.

44. National Hazards Research Australia, 2023. Modelling impacts of natural 
hazards on interconnected infrastructure networks. National Hazards 
Research Australia.

45. Google, 2025. Environmental Insights Explorer. Available from: https://
insights.sustainability.google/

46. CO2 AI, 2025. Home page. Available from: https://www.co2ai.com/

47. Toustone, 2025, Home page. Available from: https://www.toustone.com/.

48. Avarni, 2025. Home page. Available from: https://www.avarni.co/

49. Bingler JA, Kraus M, Leippold M & Webersinke N, 2022. Cheap talk and 
cherry-picking: What ClimateBert has to say on corporate climate risk 
disclosures. Finance Research Letters, 47:102776.



50. Bingler JA, Kraus M, Leippold M & Webersinke N, 2024. How cheap talk in 
climate disclosures relates to climate initiatives, corporate emissions, and 
reputation risk. Journal of Banking & Finance, 164:107191.

51. Treasury, 2024, Treasury Laws Amendment Bill 2024: Climate-related 
Financial Disclosure. Australian Government.

52. Lee DS, Fahey DW, Skowron A, Allen MR, et al., 2021.The contribution 
of global aviation to anthropogenic climate forcing for 2000 to 2018. 
Atmospheric Environment, 244:117834.

53. Google Research, 2024. How AI is helping airlines mitigate the climate 
impact of contrails, 8 August.

54. Google, 2023. New ways we’re helping reduce transportation and energy 
emissions, 10 October.

55. UPS, 2020. UPS to enhance ORION with continuous delivery route 
optimization, 29 January.

56. Chen R, Zhang W & Wang X, 2020. Machine learning in tropical cyclone 
forecast modeling: A review. Atmosphere (Basel), 11(7):676.

57. Subrahmanyam KV, Ramsenthil C, Imran AG, Chakravorty A, et al., 2021. 
Prediction of heavy rainfall days over a peninsular Indian station using 
the machine learning algorithms. Journal of Earth System Science, 
130(4):240.

58. Yuan C & Moayedi C 2020. Evaluation and comparison of the advanced 
metaheuristic and conventional machine learning methods for the 
prediction of landslide occurrence. Engineering with Computers, 
36(4):1801–1811.

59. Jaafari A, Zenner EK, Panahi M & Shahabi H, 2019. Hybrid artificial 
intelligence models based on a neuro-fuzzy system and metaheuristic 
optimization algorithms for spatial prediction of wildfire probability. 
Agricultural and Forest Meteorology, 266–267:198–207.

60. Google Research, 2025. Flood forecasting. Available from: https://sites.
research.google/gr/floodforecasting/

61. Nearing G, Cohen D, Dube V, Gauch M, et al., 2024. Global prediction of 
extreme floods in ungauged watersheds. Nature, 627(8004):559–563.

62. Google Research, 2025. FireSat: How we’re using AI to create a 
breakthrough in wildfire detection. Available from: https://sites.research.
google/gr/wildfires/firesat/

63. Boroujeni SPH, Razi A, Khoshdel S, Afghah F, et al., 2024. A 
comprehensive survey of research towards AI-enabled unmanned aerial 
systems in pre-, active-, and post-wildfire management. Information 
Fusion, 108:102369.

64. Kim S, Lee W, Park Y-S, Lee H-W, et al., 2016 Forest fire monitoring 
system based on aerial image. In: 2016 3rd International Conference on 
Information and Communication Technologies for Disaster Management 
(ICT-DM). IEEE, p.1–6.

65. Bouguettaya A, Zarzour H, Taberkit AM & Kechida A, 2022. A review 
on early wildfire detection from unmanned aerial vehicles using 
deep learning-based computer vision algorithms. Signal Processing, 
190:108309.

66. Lee W, Kim S, Lee Y-T, Lee H-W, et al., 2017. Deep neural networks 
for wildfire detection with unmanned aerial vehicle. In: 2017 IEEE 
International Conference on Consumer Electronics (ICCE), p.252–253.

67. Gebrehiwot A, Hashemi-Beni L, Thompson G, Kordjamshidi P, et al., 2019. 
Deep convolutional neural network for flood extent mapping using 
unmanned aerial vehicles data. Sensors, 19(7):1486.

68. Ichim L & Popescu D, 2020. Segmentation of vegetation and flood from 
aerial images based on decision fusion of neural networks. Remote 
Sensing, 12(15):2490.

69. Ogunjinmi PD, Park S-S, Kim B & Lee D-E, 2022. Rapid post-earthquake 
structural damage assessment using convolutional neural networks and 
transfer learning. Sensors, 22(9):3471.

70. Takhtkeshha N, Mohammadzadeh A & Salehi B, 2022. A rapid self-
supervised deep-learning-based method for post-earthquake damage 
detection using UAV data (Case study: Sarpol-e zahab, Iran). Remote 
Sensing, 15(1):123.

71. Zhou B, Zoua L, Mostafavib A, Lina B, et al., 2022. VictimFinder: 
Harvesting rescue requests in disaster response from social media with 
BERT. Computers, Environment and Urban Systems, 95:101824.

72. Fan C, Wu F & Mostafavi A, 2020. A hybrid machine learning pipeline 
for automated mapping of events and locations from social media in 
disasters. IEEE Access, 8:10478–10490.

73. RapidSOS, 2025. Home page. Available from: https://rapidsos.com/

74. Abbas AN, Amazub CW, Mietkiewicz J, Briwa H, et al., 2024. Analyzing 
operator states and the impact of AI-enhanced decision support in 
control rooms: A human-in-the-loop specialized reinforcement learning 
framework for intervention strategies. International Journal of Human–
Computer Interaction, 20241–35.

75. Mietkiewicz J, Abbas AN, Amazu CW, Baldissone G, et al., 2024. 
Enhancing control room operator decision making. Processes, 12(2):328.

76. Natural Hazards Research Australia, 2024. Be ahead of ready. Natural 
Hazards Research Australia.

77. Australian Signals Directorate, 2024 ASD cyber threat report 2023–2024. 
Australian Government.

78. CSIRO, 2024. My Climate View: Tailored insights for farmers. Available 
from: https://www.csiro.au/en/about/challenges-missions/drought-
resilience/mission-progress/climate-services-for-agriculture

79. Navarro-Hellín H, Martínez-del-Rincon J, Domingo-Miguel R, Soto-Valles 
F, et al., 2016. A decision support system for managing irrigation in 
agriculture. Computers and Electronics in Agriculture, 124:121–131.

80. Sitharthan R, Rajesh M, Vimal S, Saravana Kumar E, et al., 2023. A novel 
autonomous irrigation system for smart agriculture using AI and 6G 
enabled IoT network. Microprocessors and Microsystems 101:104905.

81. Blue River Technology, 2025. Home page. Available from: https://www.
bluerivertechnology.com/

82. James Cook University, 2023. Weed sprayer a game-changer for farmers, 
5 September.

83. Qazi S, Khawaja BA & Farooq QU, 2022. IoT-equipped and AI-enabled 
next generation smart agriculture: A critical review, current challenges 
and future trends. IEEE Access, 10:21219–21235.

84. Harfouche AL, Jacobson DA, Kainer D, Romero JC, et al., 2019. 
Accelerating climate resilient plant breeding by applying next-generation 
artificial intelligence. Trends in Biotechnology, 37(11):1217–1235.

85. Yuzugullu O, Farjaoui N, Don A & Liebisch F, 2024. Satellite-based soil 
organic carbon mapping on European soils using available datasets and 
support sampling. Science of Remote Sensing, 9:100118.

86. Pachama, 2025. Home page. Available from: https://pachama.com/

87. Macreadie PI, Trevathan-Tackett SM, Skilbeck CG, Sanderman J, et al., 
2015. Losses and recovery of organic carbon from a seagrass ecosystem 
following disturbance. Proceedings of the Royal Society B: Biological 
Sciences, 282(1817):20151537.

88. McLeod E, Chmura GL, Bouillon S, Salm R, et al., 2011. A blueprint for 
blue carbon: toward an improved understanding of the role of vegetated 
coastal habitats in sequestering CO2. Frontiers in Ecology and the 
Environment, 9(10):552–560.

89. Wynn K, Sebastian, B, 2019. Growth opportunities for Australian food 
and agribusiness – Economic analysis and market sizing. CSIRO.

90. Nikkhah A, Rohani A, Zarei M, Kulkarni A, et al., 2023. Toward sustainable 
culture media: Using artificial intelligence to optimize reduced-serum 
formulations for cultivated meat. Science of The Total Environment, 
894:164988.

91. Alasi SO, Sanusi MS, Sunmonu MO, Odewole MM, et al., 2024. Exploring 
recent developments in novel technologies and AI integration for plant-
based protein functionality: A review. Journal of Agriculture and Food 
Research,15:101036.

92. Doherty A, Wall A, Khaldi N & Kussmann M, 2021. Artificial intelligence 
in functional food ingredient discovery and characterisation: A focus on 
bioactive plant and food peptides. Frontiers in Genetics, 12:768979.

93. Vivici, 2023.Vivici selects Ginkgo Bioworks to extend its range of novel 
dairy proteins, 27 November. 

94. Jumper J, Evans R, Pritzel A, Green T, et al., 2021. Highly accurate protein 
structure prediction with AlphaFold. Nature, 596(7873):583–589.

95. Shiru, 2021. Shiru is breaking ground in leveraging AlphaFold, 3 
September.

96. Gackstetter D, von Bloh M, Hannus V, Meyer ST, et al., 2023. Autonomous 
field management – An enabler of sustainable future in agriculture. 
Agricultural Systems, 206:103607.

97. Saleem MH, Potgieter J & Arif KM, 2021. Automation in agriculture 
by machine and deep learning techniques: A review of recent 
developments. Precision Agriculture 22(6):2053–2091.

98. Sarker IH, 2021. Data science and analytics: An overview from data-driven 
smart computing, decision-making and applications perspective. SN 
Computer Science, 2:377.



99. Batz P, Will T, Thiel S, Ziesche TM, et al., 2023. From identification to 
forecasting: the potential of image recognition and artificial intelligence 
for aphid pest monitoring. Frontiers in Plant Science, 14:1150748.

100. Shaikh TA, Rasool T & Lone FR, 2022. Towards leveraging the role of 
machine learning and artificial intelligence in precision agriculture and 
smart farming. Computers and Electronics in Agriculture, 198:107119.

101. Hansen BD, Leonard E, Mitchell C, Easton J, et al., 2023. Current status 
of and future opportunities for digital agriculture in Australia. Crop and 
Pasture Science, 74(6):524-537.

102. Department of Climate Change, Energy, the Environment and Water, 
2023. Livestock Emissions Framework for Feed Technologies. Australian 
Government.

103. Nickel R, 2023. The climate-friendly cows bred to belch less methane. 
Reuters, 9 August.

104. Onwusinkwue S, Osasona F, Ahmad I, Anyanwu A, et al., Artificial 
intelligence (AI) in renewable energy: A review of predictive maintenance 
and energy optimization. World Journal of Advanced Research and 
Reviews, 21(1):2487–2799.

105. Zhao W, Equsquiza M, Valero C, Valentín D, et al., 2020. On the use of 
artificial neural networks for condition monitoring of pump-turbines with 
extended operation. Measurement, 163:107952.

106. Nasser A & Al-Khazraji H, 2022. A hybrid of convolutional neural 
network and long short-term memory network approach to predictive 
maintenance. International Journal of Electrical and Computer 
Engineering 12(1):721.

107. Thompson L, 2024. Revolutionizing energy grid maintenance: How 
artificial intelligence is transforming the future. Argonne National 
Laboratory, 28 May.

108. Rozite V, Miller J & Oh S, 2023. Why AI and energy are the new power 
couple. International Energy Agency, 2 November.

109. Gailhofer P, Herold A, Schemmel JP, Scherf C-F, et al., 2021. The role of 
artificial intelligence in the European Green Deal. European Parliament.

110. Elkin C & Witherspoon S, 2019. Machine learning can boost the value of 
wind energy. Google Deep Mind, 26 February.

111. Zheng L, Zhang S, Huang H, Liu R, et al., 2023. Artificial intelligence-
driven rechargeable batteries in multiple fields of development 
and application towards energy storage. Journal of Energy Storage 
73:108926.

112. Sendek AD, Ransom B, Cubuk ED, Pellouchoud LA, et al., 2022. Machine 
learning modeling for accelerated battery materials design in the small 
data regime. Advanced Energy Materials, 12(31):2200553.

113. Aionics, 2025. Home page. Available from: https://aionics.io/

114. Golsefidi AH, Hüttel FB, Peled I, Samaranayake S, et al., 2023. A joint 
machine learning and optimization approach for incremental expansion 
of electric vehicle charging infrastructure. Transportation Research Part 
A: Policy and Practice, 178:103863.

115. Nourani V, Baghanama AH, Adamowskib J & Kisic, 2014. Applications 
of hybrid wavelet–artificial intelligence models in hydrology: A review. 
Journal of Hydrology, 514:358–377.

116. Nourani V, Komasi M & Alami MT, 2012. Hybrid wavelet–genetic 
programming approach to optimize ANN modeling of rainfall–runoff 
process. Journal of Hydrologic Engineering, 17(6):724–741.

117. Rajaee T, Mirbagheri SA, Zounemat-Kermani M & Nourani V, 2009. Daily 
suspended sediment concentration simulation using ANN and neuro-
fuzzy models. Science of The Total Environment, 407(17):4916–4927.

118. Pagendam D, Janardhanan S, Dabrowski J & MacKinlay D, 2023. A log-
additive neural model for spatio-temporal prediction of groundwater 
levels. Spatial Statistics, 55:100740.

119. Singh KP, Basant N & Gupta S, 2011. Support vector machines in water 
quality management. Analytica Chimica Acta, 703(2):152–162.

120. CSIRO, 2025. WaterWise: helping farmers reduce the water footprint 
of high value crops. Available from: https://www.csiro.au/en/research/
natural-environment/water/waterwise

121. ARDC, 2021. Leading Australia to data-driven research impact. Australian 
Research Data Commons.

122. NSW Government, 2024. Smart Irrigation Management for Parks and 
Cool Towns (SIMPact). Available from:https://www.nsw.gov.au/business-
and-economy/smart-places/case-studies/smart-irrigation-management-
for-parks-and-cool-towns-simpact

123. Marrone BL, Banerjee S, Talapatra A, Gonzalez-Esquer CR, et al., 2023. 
Toward a predictive understanding of cyanobacterial harmful algal 
blooms through AI integration of physical, chemical, and biological data. 
ACS ES&T Water, 4(3):844–858.

124. Smart Water Magazine, 2023. Unlocking the potential of wastewater 
using AI – the DARROW project introduces itself. Smart Water Magazine, 
29 August.

125. McGovan J, 2024. The role of AI in building wastewater treatment plants 
for the future. Recover Web, 16 May.

126. DARROW, 2025. The DARROW project. Available from: https://www.
wastewater.ai/the-project/

127. AdaptNSW, 2025 Securing water for Sydney’s future. Available from: 
https://www.climatechange.environment.nsw.gov.au/stories-and-case-
studies/securing-water-sydneys-future

128. Dupke S, Buchholz U, Fastner J, Förster C, et al., 2023. Impact of climate 
change on waterborne infections and intoxications. Journal of Health 
Monitoring, 8(Suppl 3):62.

129. Nehal SA, Roy D, Devi M & Srinivas T, 2020. Highly sensitive lab-on-chip 
with deep learning AI for detection of bacteria in water. International 
Journal of Information Technology, 12(2):495–501.

130. Ebi KL, Capon A, Berry P, Broderick C, et al., 2021. Hot weather and heat 
extremes: health risks. The Lancet, 398(10301):698–708.

131. Falchetta G, De Cian E, Wing IS & Carr D, 2024. Global projections of heat 
exposure of older adults. Nature Communications, 15(1):3678.

132. Jiang S, Warren JL, Scovronick N, Moss SE, et al., 2021. Using logic 
regression to characterize extreme heat exposures and their health 
associations: a time-series study of emergency department visits in 
Atlanta. BMC Medical Research Methodology, 21(1):87.

133. Wang Y, Song Q, Du Y, Wang J, et al., 2019. A random forest model to 
predict heatstroke occurrence for heatwave in China. Science of The Total 
Environment, 650:3048–3053.

134. Ohashi Y, Ihara T, Oka K, Takane Y, et al., 2023. Machine learning 
analysis and risk prediction of weather-sensitive mortality related to 
cardiovascular disease during summer in Tokyo, Japan. Scientific Reports, 
13(1):17020.

135. Ogata S, Takegami M, Ozaki T, Nakashima T, et al., 2021. Heatstroke 
predictions by machine learning, weather information, and an 
all-population registry for 12-hour heatstroke alerts. Nature 
Communications, 12(1):4575.

136. Department of Climate Change, Energy, the Environment and Water, 
2024. New heat maps will help communities adapt to climate impacts. 
Australian Government, 12 April.

137. Australian Climate Service, 2024. Heat and our health: A glance at the 
changing heat risk to human health in Australia. Australian Climate 
Service, 15 April.

138. Tawatsupa B, Yiengprugsawan, V, Kjellstrom T, Berecki-Gisolf J, et al.,2013. 
Association between heat stress and occupational injury among Thai 
workers: findings of the Thai Cohort Study. Industrial health, 51(1):34-46.

139. Tamm M, Jakobson A, Havik M, Burk A, et al., 2014. The compression 
of perceived time in a hot environment depends on physiological and 
psychological factors. Quarterly Journal of Experimental Psychology, 
67(1):197–208.

140. Morabito M, Cecchi L, Crisci A, Modesti PA, et al., 2006. Relationship 
between work-related accidents and hot weather conditions in Tuscany 
(central Italy). Industrial health, 44(3):458–464.

141. Kalasin S, Sangnuang P & Surareungchai W, 2021. Satellite-based sensor 
for environmental heat-stress sweat creatinine monitoring: The remote 
artificial intelligence-assisted epidermal wearable sensing for health 
evaluation. ACS Biomaterials Science and Engineering, 7(1):322–334.

142. Togo H & Hirata A, 2021. Novel health risk alert system for occupational 
safety in hot environments. IEEE Pulse, 12: 24–27.

143. Lynch G, 2024. EMU Systems. ON Program Alumni, CSIRO, 14 May.

144. Baker RE, Mahmud AS, Miller IF, Rajeev M, et al., 2022. Infectious disease 
in an era of global change. Nature Reviews Microbiology, 20(4):193–205.

145. MacIntyre CR, Chen X, Kunasekaran M, Quigley A, et al., 2023. 
Artificial intelligence in public health: The potential of epidemic 
early warning systems. Journal of International Medical Research, 
51(3):03000605231159335.

146. O’Brien J, 2022. EpiWatch, the AI-driven early warning system for 
epidemics. InnovationAus.com, 14 November.



147. Kirby Institute, 2025, EPIWATCH: Prevent the next pandemic with 
Epidemic Intelligence. Available from: https://www.kirby.unsw.edu.au/
research/projects/epiwatch

148. Kinney PL, 2018. Interactions of climate change, air pollution, and human 
health. Current Environmental Health Reports, 5(1):179–186.

149. Orru H, Ebi KL & Forsberg B, 2017. The interplay of climate change and air 
pollution on health. Current Environmental Health Reports, 4(4):504–513.

150. Harvard School of Public Health, 2023. Allergies are getting worse with 
climate change. Harvard School of Public Health, 11 April.

151. AIHW, 2020. Australian bushfires 2019–20: Exploring the short-term 
health impacts. Australian Institute of Health and Welfare.

152. Copernicus Atmosphere Monitoring Service, 2025. BreezoMeter: 
Information on air quality and pollen. Available from: https://
atmosphere.copernicus.eu/breezometer-information-air-quality-and-
pollen

153. Xu C, Kohler TA, Lenton TM & Scheffer M, 2020. Future of the human 
climate niche. Proceedings of the National Academy of Sciences, 
117(21):11355.

154. Lenton TM, Xu C, Abrams JF, Ghadiali A, et al., 2023. Quantifying the 
human cost of global warming. Nature Sustainability, 6(10):1237–1247.

155. RACP, 2024. Modelling the health impacts of climate change. Royal 
Australasian College of Physicians.

156. Alcayna T, O’Donnell D & Chandaria S, 2023. How much bilateral and 
multilateral climate adaptation finance is targeting the health sector? 
A scoping review of official development assistance data between 2009–
2019. PLOS Global Public Health, 3(6):e0001493.

157. Khan FM, Gupta R & Sekhri S, 2021. A convolutional neural network 
approach for detection of E. coli bacteria in water. Environmental Science 
and Pollution Research, 28(43):60786.

158. Department of Health and Aged Care, 2024. Australia joins US and UK 
statement on decarbonising healthcare. Australian Government, 23 April.

159. Department of Health and Aged Care, 2023. National Health and Climate 
Strategy. Australian Government.

160. This Climate Does Not Exist, 2025. Home page. Available from: https://
thisclimatedoesnotexist.com/home

161. UNEP, 2022. Global methane assessment: 2030 baseline report. United 
National Environment Programme.

162. Global Methane Pledge, 2024. Home page. Available from: https://www.
globalmethanepledge.org/.

163. MethaneSat, 2025. Home page. Available from: https://www.methanesat.
org/.

164. GHGSat, 2024. Home page. Available from: https://www.ghgsat.com/en/.

165. Climate TRACE, 2023. Climate TRACE unveils open emissions database of 
more than 352 million assets. Climate TRACE, 2 December.

166. Department of Industry, Science and Resources, 2025. Australia’s AI Ethics 
Principles. Available from: https://www.industry.gov.au/publications/
australias-artificial-intelligence-ethics-principles/australias-ai-ethics-
principles

167. Yaakoubi Y, Donti PL, Kaack LH, Rolnick D, et al., 2024. Grand Challenge 
Initiatives in AI for Climate & Nature: Landscape assessment and 
recommendations. Climate Change AI.

168. Oreskes N & Conway EM, 2010. Merchants of doubt: How a handful of 
scientists obscured the truth on issues From tobacco smoke to global 
warming. Bloomsbury Press.

169. Maußner C, Oberascher M, Autengruber A, Kahl A, et al., 2025. 
Explainable artificial intelligence for reliable water demand forecasting to 
increase trust in predictions. Water Research, 268:122779.

170. Bostrom A, Demuth JL, Wirz CD, Cains MG, et al., 2024. Trust and 
trustworthy artificial intelligence: A research agenda for AI in the 
environmental sciences. Risk Analysis, 44(6):1498–1513.

171. International Energy Agency, 2023. Data centres and data transmission 
networks. Available from: https://www.iea.org/energy-system/buildings/
data-centres-and-data-transmission-networks

172. EPRI, 2024. Powering intelligence: Analyzing artificial intelligence and 
data center energy consumption. Electric Power Research Institute.

173. Ember and Energy Institute, 2024. Carbon intensity of electricity 
generation. Our World in Data.

174. Mazhar M, 2024. Microsoft, Google and Amazon turn to nuclear energy 
to fuel the AI boom. CBC Radio, 31 October.

175. Patterson P, Gonzalez J, Le Q, Liang C, et al., 2021. Carbon emissions and 
large neural network training. arXiv:2104.10350.

176. Green Mountain, 2021. Land-based trout farm will use data center waste 
heat. Green Mountain, 28 June.

177. Judge P, 2022. Stack shares heat from Oslo data center. Data Centre 
Dynamics, 21 September.

178. Wretborn S, 2018. DigiPlex data center to heat 10,000 Stockholm 
households. Stockholm Data Parks, 12 March.

179. Kleinman Z, 2023. Tiny data centre used to heat public swimming pool. 
BBC, 14 March.

180. Department of Industry, Science and Resources, 2024. Safe and 
responsible AI in Australia: Proposals paper for introducing mandatory 
guardrails for AI in high-risk settings. Australian Government.

181. Worrell T, 2024. AI affects everyone – including Indigenous people. It’s 
time we have a say in how it’s built. The Conversation, 11 October.

182. Wilson C, 2024. AI is producing ‘fake’ Indigenous art trained on real 
artists’ work without permission. Crikey, 19 January.

183. Fitch E, McKenzie C, Janke T & Shul A, 2024. The new frontier: Artificial 
intelligence, copyright and Indigenous culture. Terri Janke and Company, 
23 November.

184. Cranney K, 2019. Magpie geese return with help from ethical AI and 
Indigenous knowledge. CSIRO, 20 November.

185. CSIRO, 2024. AI for Climate Symposium. Available from: https://wp.csiro.
au/ai4c/

186. Climate Change AI, 2024. Climate Change AI Summer School 2024. 
Available from: https://www.climatechange.ai/events/summer_
school2024

187. University of Oxford, 2025. Intelligent Earth: UKRI AI Centre for Doctoral 
Training in AI for the Environment. Available from: https://intelligent-
earth.ox.ac.uk/home

188. University of Cambridge, 2025. AI for the study of Environmental Risks 
(AI4ER): UKRI Centre for Doctoral Training. Available from: https://ai4er-
cdt.esc.cam.ac.uk/

189. University of Southampton, 2025. SustAI: The UKRI AI Centre for Doctoral 
Training in AI for Sustainability. Available from: https://sustai.info/

190. Morgan State University, 2025. NSF NRT: Artificial Intelligence for Climate 
Change and Environmental Sustainability. Available from: https://www.
morgan.edu/biology/nsf-nrt-access

191. Enterprise Neurosystem, 2025. AI Innovation Grand Challenge. Available 
from: https://enter.innovationgrandchallenge.ai/

192. Digital Catapult, 2025. AI for Decarbonisation’s Virtual Centre of 
Excellence (ADViCE). Available from: https://www.digicatapult.org.uk/
programmes/programme/advice/

193. Asian Development Bank, 2023 . AI for Climate Resilience in Rural 
Areas. Available from: https://challenges.adb.org/en/challenges/m4d-
challenge-2023

194. AI for Good, 2025. AI/ML solutions for climate change. Available from: 
https://aiforgood.itu.int/about-us/aiml-solutions-for-climate-change/

195. Bezos Earth Fund, 202 5. AI for Climate and Nature Grand Challenge. 
Available from: https://aiforclimateandnature.org/

196. Haben S & Young S, 2023. ADViCE: AI for Decarbonisation Challenges. 
Alan Turing Institute.

197. Lacuna Fund, 2025. Home page. Available from: https://lacunafund.org/.

198. Department of Industry, Science and Resources, 2024. Voluntary AI Safety 
Standard. Australian Government.

199. Reid A, O’Callaghan S & Lu Y, 2023 Implementing Australia’s AI Ethics 
Principles: A selection of responsible AI practices and resources. Gradient 
Institute & CSIRO.

200. Department of Finance, 2024. National framework for the assurance of 
artificial intelligence in government. Australian Government.

201. Climate Change AI, 2025. Data gaps (beta). Available from: https://www.
climatechange.ai/dev/datagaps/info



For further information

CSIRO Data 61

Dr Claire Naughtin
+61 7 3833 5658
claire.naughtin@data61.csiro.au

CSIRO Data 61

Dr Mahesh Prakash
+61 3 9545 8010
mahesh.prakash@data61.csiro.au

As Australia’s national science agency, 
CSIRO is solving the greatest 
challenges through innovative 
science and technology.

CSIRO. Creating a better future 
for everyone.

Contact us

1300 363 400
+61 3 9545 2176
csiro.au/contact
csiro.au