Wildfires are rapidly emerging as one of the most critical climate risk factors of the 21st century. Their impacts extend beyond immediate destruction—affecting biodiversity, carbon cycles, land-use systems, and long-term ecosystem resilience. A funded PhD position (75%) in Computer Science / Environmental Data Science is now open at the Karlsruhe Institute of Technology (KIT), Germany. The project focuses on advancing probabilistic deep learning methods for spatiotemporal wildfire forecasting within the international C4LaNd research training group.

This position is designed for candidates seeking to combine state-of-the-art artificial intelligence with high-impact climate science applications. As climate risks intensify globally, integrating deep learning with Earth system modelling represents one of the most promising frontiers in environmental science. This PhD offers the opportunity to shape that frontier through both theoretical AI advancement and real-world climate impact modelling.

Institutional Framework and International Collaboration

The doctoral project is embedded in C4LaNd, an interdisciplinary research training group uniting natural sciences, social sciences, economics, and engineering to address sustainable land use under climate change.

Participating institutions include:

• Karlsruhe Institute of Technology (KIT)
• University of Melbourne
• University of Hohenheim

The programme offers:

• A dual PhD degree (KIT & University of Melbourne)
• One-year funded research stay in Melbourne
• International supervision framework

Primary location: Karlsruhe, Germany.

Research Environment

The project is based in the AI in Climate and Environmental Sciences group at KIT’s Institute of Theoretical Informatics (ITI).

Primary supervisor: TT-Prof. Peer Nowack
Collaborators at KIT include experts in atmospheric environmental research.
Melbourne co-advisor: Dr. Benjamin Henley.

The research integrates AI method development with Earth system modelling communities.

Scientific Focus: Probabilistic AI for Wildfire Risk

Wildfire risks are inherently multidimensional and uncertain. Accurate prediction requires models capable of representing:

• Spatiotemporal dynamics
• Extreme event probabilities
• Interactions between climate and land use
• Impacts on biodiversity and vegetation systems

Core Objective

Develop advanced probabilistic deep learning models for spatiotemporal forecasting of wildfire risks, using both Earth observation data (e.g., satellite datasets) and computer simulations.

Methodological Innovation

A central part of the project is advancing AI algorithms themselves.

Research includes:

1. Probabilistic Deep Learning

• Designing uncertainty-aware forecasting models
• Developing next-generation generative approaches
• Exploring diffusion-model-based architectures

2. Spatiotemporal Modelling

• Handling high-dimensional geophysical datasets
• Learning from Earth observation data
• Integrating physical constraints into AI systems

3. Climate and Land-Use Risk Assessment

• Modelling combined effects of climate change and land-use change
• Assessing wildfire impacts on biodiversity
• Supporting climate adaptation and mitigation research

Integration into Earth System Models

Your wildfire model will contribute to major ecosystem modelling frameworks, including:

• LPJ-GUESS
• WOW

This integration will improve the representation of wildfires in future dynamic vegetation and climate simulations—linking AI innovations directly to Earth system science.

Research Responsibilities

• Advance probabilistic deep learning for spatiotemporal modelling
• Apply methods to wildfire risk assessment under future scenarios
• Collaborate with LPJ-GUESS and WOW modelling communities
• Conduct literature reviews and data integration
• Publish in leading AI and climate journals
• Present at international conferences

Candidate Profile

Applicants should hold:

• Master’s degree in Computer Science, Mathematics, Physics, Data Science, or related field

Required qualifications:

• Strong programming skills (ideally Python and C++)
• Solid mathematical foundation
• Fluency in English
• Motivation for interdisciplinary research

You must be willing to complete a one-year research stay at the University of Melbourne.

Employment Conditions

• Start date: November 2026
• Contract duration: 3.5 years
• Employment level: 75% (TV-L E13 salary scale)
• Application deadline: 17 May 2026
• Vacancy number: 1085/2026

Host institution: Institute of Theoretical Informatics (ITI), KIT.

Application Requirements

Applications must include:

• Letter of motivation
• CV
• Bachelor’s and Master’s transcripts (including preliminary transcript for Master’s if applicable)
• Contact details of two academic referees
• APPLY NOW
• Application deadline: 17 May 2026

Scientific inquiries:
TT-Prof. Peer Nowack
Email: peer.nowack@kit.edu

Why This PhD Is High-Impact

This project sits at the intersection of:

• Artificial Intelligence
• Climate risk modelling
• Earth system science
• Biodiversity research

It offers rare exposure to both methodological AI innovation and applied climate risk assessment.

Career pathways after completion include:

• Academic AI research
• Climate modelling institutions
• Environmental risk analytics
• Earth observation agencies
• Advanced AI labs focusing on scientific machine learning

Given increasing global investment in climate resilience, probabilistic AI forecasting skills are in exceptionally high demand.

Frequently Asked Questions (FAQ)

Is this position fully funded?

Yes. The PhD is funded at 75% employment under Germany’s TV-L E13 salary scale for 3.5 years.

Is this more AI-focused or climate-focused?

Primarily AI-focused, with methodological innovation at its core. Climate and wildfire risk serve as the high-impact application domain.

Do I need prior wildfire research experience?

No. Strong quantitative and programming skills are more important than prior domain-specific experience.

What are diffusion models?

Diffusion models are advanced generative AI architectures that iteratively learn probability distributions. They have recently achieved state-of-the-art results in complex data modelling tasks.

Will I receive two doctoral degrees?

Yes. The cotutelle model allows for a dual PhD degree from KIT and the University of Melbourne.

Is knowledge of German required?

No. The programme operates in English.

What makes an applicant competitive?

Strong mathematical background, demonstrated experience in machine learning, and interest in probabilistic modelling for scientific applications.