Welcome to the Environmental Data Science and AI Summit!

Join us at Victoria University of Wellington on 19 August 2024 (Monday) for an engaging day filled with talks, workshops, and networking opportunities centered on the intersection of environmental studies and data science in New Zealand. This summit is ideal for scientists, government officials, students, and community members who are passionate about conservation challenges and seeking a deeper understanding of our environment.

Our speakers include leading experts from academia, Crown Research Institutes, and government.

Attendance is free, but space is limited. To ensure your spot, please register as soon as possible. The registration deadline is Friday, August 9, 2024.

REGISTER HERE

Venue Information

Victoria University of Wellington, Kelburn Campus (Map)

• Lecture room: HMLT205, Hugh Mackenzie Building
• Lunch and coffee break will be served at MY354 Level 3 Foyer, Murphy Building.

DOWNLOAD THE PDF PROGRAMME HERE

Environmental Data Science and AI Summit Speakers

We have an exciting line up of speakers ready to share their expertise:

Dr. Mario Krapp (GNS)

Title: A Bayesian Approach to Reconstructing the Climate of the Past

Abstract: To understand past climates, we are looking into the geological record. However, key information like temperature and rainfall cannot be directly observed. Instead, we rely on proxies such as fossils, biomarkers, stable isotope ratios, pollen abundances, and sediment types. The paleoclimate community has developed empirical methods to interpret these. However, these methods are often challenging to interpret and provide limited insight into the underlying climate drivers. In this presentation, I introduce a Bayesian approach to infer past climate changes from geological records. I will demonstrate the use of Bayesian Inference in the paleoclimate context and discuss how this approach can address complex problems more generally.

Bio: Mario is an environmental data scientist and interested in (almost) everything related to the Earth system. Mario spends his days figuring out how we can we learn from data, how we can better integrate models and data, and how we can make informed decision using the best of both worlds. Examples are future contributions of Antarctica to global sea level rise or past climate sensitivities inferred from the geological record. Mario is currently involved in the Antarctic Science Platform and the MBIE Endeavour programme Te Ao Hurihuri: Te Ao Hou, Our Changing Coast.

Prof. Mengjie Zhang (Victoria University of Wellington / Centre for Data Science and AI)

Title: AIML for Segmentation of Individual Tree Crowns and Salinity Prediction in NZ Oyster Farms

Abstract: This presentation talks about two pieces of works using AIML for image segmentation of individual tree crowns and salinity prediction for oyster farms in NZ, both of which are related for NZ environment. New Zealand has 24% of its 270,000 km2 land covered in forest and actively supports and promotes urban regreening in many of its cities. Sustaining and enhancing biodiversity and healthy living environments are priorities for New Zealand that require careful management of trees in urban areas and forests. This talk will first provide some pieces of work of CDSAI at VUW, collaborating with MWLR, on segmentation of individual tree crowns images for the Wellington City using Convolutional neural networks and variations particular different version of Yolos. Meanwhile, NZ oyster farms provided a sustainable and profitable export for New Zealand, but are sensitive to changes in salinity that can cause significant crop loss if they persist too long. Recent extreme weather events have been leading to increased periods of low salinity, putting the farms at risk. This presentation will then discuss three different methods to assess the viability of salinity prediction systems, including a statistical model, a genetic programming (GP) based symbolic regression model and a convolutional neural network. The results show that both GP and CNN performed better than statistical models to predict salinity, the CNN generalised better on extreme weather conditions, and the GP-learned models are more interpretable. If time allows, some initial work, collaborating with ARC, on climate modelling will be discussed using AI and machine learning.

Dr. Alvaro Orsi (ESR)

Title: Data Science for Environmental Health

Abstract: Environmental health research examines the relationships between environmental factors and human wellbeing. In this presentation, I explore how data science and artificial intelligence (AI) are revolutionising this field, offering unprecedented insights and analytical capabilities. Some examples include the application of remote sensing and AI modelling to characterise on-site wastewater systems, and soil characteristics in polluted areas, and their potential to inform targeted remediation efforts.

I will present an ongoing project to develop an AI-powered digital twin platform that simulates the intricate interplay of environmental, social, and health metrics. This cutting-edge system integrates AI predictive modelling, synthetic population models, and agent-based simulations to create a dynamic virtual representation of environmental health systems. This platform enables scenario testing, risk assessment, and evidence-based decision-making for environmental health management.

Looking ahead, I will explore future directions for integrating data science into environmental health practice, emphasising the importance of interdisciplinary collaboration and the development of environmentally sustainable AI solutions.

Bio: Dr. Alvaro Orsi is a Data Science Lead with extensive experience leveraging AI and advanced analytics to transform insights-driven solutions across scientific and business domains. At the Institute for Environmental Science and Research (ESR), he spearheads AI-powered innovations, orchestrating cutting-edge solutions through technologies such as Generative AI, Digital Twins, Large Population Models, Geospatial data science, and Time series forecasting.

Throughout his career, Dr. Orsi has applied AI and advanced analytics to diverse sectors, including supply chain logistics in primary industries, remote sensing of vegetation, productivity forecasting, and delivering data-driven insights for government agencies. His expertise spans multiple fields, including astrophysics, machine learning, and remote sensing of vegetation.

Before joining ESR, Dr. Orsi was a Principal Research Scientist at PlantTech Research Institute, developing AI and machine learning solutions for New Zealand's primary industry. His background in Computational Astrophysics includes work in Spain, the UK, and his native Chile, contributing to a global perspective that enriches his approach to data science.

Dr. Orsi holds a PhD in Computational Cosmology from Durham University, UK. He has published over 80 peer-reviewed papers across various disciplines, including astrophysics, machine learning, remote sensing, and epidemiological data science. Currently, he serves as a Board member for the AI Researchers Association of New Zealand.

Committed to shaping the future of data science in New Zealand and beyond, Dr. Orsi strives to foster a new era of AI-powered solutions that deliver positive impact for New Zealand's economy, society, and environment.

Prof. Alan Brent (Victoria University of Wellington /Chair in Sustainable Energy Systems)

Title: The Integration of AI and ML into the Operational and Planning Phases of Agrivoltaic Systems

Abstract: The required energy transition to reach a net zero carbon economy by 2050 will see the large uptake of utility-scale solar photovoltaic (PV) electricity generation. Scenarios for Aotearoa New Zealand project in the order of 10 GW – equal to the current generation capacity in the country – over the next two decades. Conventional solar farms have seen the conversion of productive agricultural land with solar arrays and sheep grazing for vegetation management. This is a concern as half of the country’s land area is utilised for agricultural purposes. Globally, agrivoltaic systems have emerged as an innovative solution to address this concern. Optimised systems allow for dual land-usage to both generate electricity and continue with efficient and effective agricultural production. The placement, orientation, and spacing of solar arrays directly impact crop growth and yield, making it crucial to minimize shading while maximizing energy production. Traditional optimisation approaches, such as linear programming, have been beneficial in this context but struggle with the complexity and dimensionality of agrivoltaic systems on a more granular level with more variables to simultaneously account for. The integration of AI and ML into the operational and planning phases of agrivoltaic systems presents a ground-breaking leap forward. These advanced algorithms have the capacity to analyse vast amounts of data, identify intricate patterns, and uncover correlations that traditional methods might overlook. AI and ML provide valuable insights into system performance, leading to data-driven decisions for enhanced efficiency and productivity.

Bio: Alan Brent is a Professor and the inaugural holder of the Chair in Sustainable Energy Systems at Te Wāhanga Ahunui Pūkaha Wellington Faculty of Engineering, Te Herenga Waka Victoria University of Wellington in Aotearoa New Zealand. He holds Bachelor degrees in Engineering (Chemical) and Philosophy (Sustainable Development); Master degrees in Science (Environmental Engineering), Engineering (Technology Management), and Philosophy (Sustainable Development); and a PhD in Engineering Management. He is a Fellow of Engineering New Zealand, and a member of the IEEE Power and Energy Society. More information on his research can be found on his ORCID profile (https://orcid.org/0000-0003-3769-4512), and LinkedIn profile (https://www.linkedin.com/in/alanbrent/).

Dr. Peter Gibson (NIWA)

Title: Opportunities and Challenges for AI in Climate Downscaling

Abstract: Climate models are the main tools used to inform climate change adaptation. They provide quantitative long-term projections of how a range of climate variables may respond to future increases in greenhouse gas concentrations in the atmosphere. However, the coarse spatial resolution of Global Climate Models (GCMs), where each grid cell is on the order of 100-km resolution, means that they are often not up to the task for local and regional-scale projections. Instead, higher resolution regional climate models (RCMs), typically with resolution of 12-25km, are driven from GCMs, including in work carried out by NIWA. However, this task is computationally very expensive to run, which in turn limits the resolution at which the models can be run, and the sampling of extreme events. Here, I will showcase some of the recent work carried out by our group, focused on using generative AI to emulate the traditional downscaling task of RCMs. The future is bright for AI-based RCM emulators, but comprehensive evaluation of the output and careful refinement of the AI training is needed to build trust in the climate projections produced.

Bio: Dr Peter Gibson is a climate scientist at the National Institute of Water and Atmospheric Research (NIWA). His research focuses on extreme events, how they are changing in a warming climate, and how we can best represent them in climate models. This includes using traditional physics-based dynamical models alongside AI/ML models. Peter completed his PhD in climate science at the University of New South Wales, Australia, followed by postdoctoral positions at NASA-JPL and Scripps Institution of Oceanography in California, USA.

Prof. Karin Bryan (University of Auckland/TAIAO)

Title: AI Accelerators for Environmental Change

We are creeping closer and closer to our planetary boundaries, both globally and here in Aotearoa New Zealand. As our environmental impact accelerates, so to must our solutions, not only at the cutting-edge research end, but also at the applied coal face where real differences to environmental outcomes can be made. Data is becoming more and more available, but we struggle to keep pace with applications that can make use of these data streams. This talk will update on recent developments in earth observation and monitoring, showing how such information could be used in combination with recent AI applications developed for other disciplines. However, progress in moving from theory to practice has many impediments, ranging from effective communication to technical challenges.

Bio: Karin Bryan is Professor of coastal environment science at the University of Auckland, in the School of Environment and Institute of Marine Sciences. She researches the effects of climate on coastal and estuarine systems, including increased sedimentation, temperature and sealevel rise, with implications on how our coastal ecosystems function, store carbon and provide services. She uses a variety of techniques, from remote sensing to dynamical modelling to solve problems, with a keen interest in AI innovations through her work in the Taiao Data Platform.

Dr. Dean Meason, (Scion - NZ Forest Research Institute)

Title: Green AI Analysis of Complex Environmental Data for the Forest Flows Programme and its End-User Applications

Abstract: Land-use intensification and climate change are increasing the pressure on water availability, water quality and use around the world and NZ. Managed and unmanaged forests are increasingly vulnerable to an increasing frequency of severe drought and storm events caused by a changing climate. Understanding how water flows through the land, including planted forests is essential to make the best use of water and land while maintaining environmental health. However, it is very difficult to quantify and monitor the growth and water stress on forests, owing to their remoteness, steep terrain and size. The 5-year Forest Flows MBIE Endeavour Research Programme (www.forestflows.nz) is investigating these challenges with the novel integration of various terrestrial and remote sensing technologies in Pinus radiata (D. Don) plantation forests across five watersheds across a range of climatic and physiographic regions. Forest Flows collects real-time, big data from 1,717 terrestrial sensors, with 360,000 observations every 24 hours. It is an example of the increasing complex environment data that is now obtainable by the deployment of ever large number of sensors and sensor networks. This complex data causes issues for researchers and stakeholders as traditional analysis approaches inadequate for the volume and speed of the data, and off-the-shelf machine learning (ML) are not suitable. Scion partnered with TAIAO to develop Green Artificial Intelligence (AI) – an environmental research framework that can analyse complex environmental data from the Forest Flows programme. This presentation will present two novel components of Green AI used by Forest Flows – Kafka Big Data Pipeline and eXplainable AI (XAI). We will demonstrate how these novel approaches (1) found biologically meaningful results from large volumes of raw data, (2) how XAI provided transparency with ML results, (3) ability to make robust predictions beyond the original datasets, and (4) provide useful and timely information for decision makers and stakeholders. We will also outline how Green AI can be applied to other complex environmental data.

Bio: Dean Meason has over 25 years’ experience in forest research and specialises in forestry, forest hydrology, tree ecophysiology, and soil science. His research interests include the spatial and temporal drivers of tree and forest productivity, data fusion of terrestrial and remote sensing, and Green AI. He is a recognised expert in alternative species management and productivity and forest hydrology. Dean is the programme leader of the 5-year, $13.7 million, MBIE Endeavour Forest Flows programme. This programme is investigating the role of planted forests in water storage, use, and release on the regional hydrological landscape and its effects on water quantity and quality to downstream rural and urban water users – and the environmental, cultural, and economic impacts.

Meet the Organising Committee

• Prof. Albert Bifet
• Prof. Karin Bryan
• Dr. Heitor Murilo Gomes (Local Chair, Victoria University of Wellington)