Why a neuroscience lab talks about carbon

Our emission profile

Estimated Total Lab Footprint (2025)

~2.4–4.7 tonnes CO₂e

​

This range reflects methodological uncertainty across categories:

• The lower bound (~2.4 t CO₂e) uses CO₂-only accounting for air travel and the lower estimate for high-performance computing emissions.

• The upper bound (~4.7 t CO₂e) includes aviation’s additional high-altitude climate effects (radiative forcing uplift) and the upper estimate for computing-related electricity use.

​​

The total includes:

• Travel (conferences, invited talks, training events)

• High-performance computing (Snellius, Donders, MPI clusters)

• Digital dissemination (CNSeminars video streaming)

​​

Travel remains the dominant contributor to our footprint.

​​

Digital & Computing

​

In 2025, CNSeminars generated approximately 1,900 hours of watch time. Using published European streaming estimates (~55 g CO₂e per viewer-hour), this corresponds to roughly 0.10 tonnes CO₂e (Carbon Trust, 2020). For comparison, this is substantially lower than the emissions associated with a single long-haul return flight.

​

Our research relies on high-performance computing (HPC) resources, including Snellius (SURF), the Donders HPC cluster, and the MPI cluster. These systems enable large-scale neuroimaging processing, tractography, and computational modelling, but they also consume electricity and therefore contribute to our environmental footprint.

Based on aggregated 2025 scheduler usage (CPU and GPU hours) and standard published energy-intensity models for high-performance computing, we estimate our lab’s annual computing-related emissions to be in the range of:

~0.5–1.2 tonnes CO₂e (2025 estimate)

This range reflects uncertainty in:

• Node-level power consumption

• Data centre power usage effectiveness (PUE)

• National electricity carbon intensity

• GPU vs CPU allocation profiles

The estimate was derived using established computational carbon accounting methodologies (e.g., Green Algorithms framework) and European electricity emission factors. Where possible, we aim to replace model-based assumptions with facility-reported energy data.

​​​

Travel

​

Our estimated 2025 travel emissions for listed work travel (conferences, invited talks, and training events) are estimated at ~1.8 tCO₂e (CO₂-only accounting) or ~3.35 tCO₂e when including aviation’s additional high-altitude climate effects (radiative forcing uplift). Emission factors sourced from the Climatiq emissions database (short-haul flights, rail, and passenger car), with and without aviation radiative forcing uplift.​​​​

​

​

Reduction & Mitigation Strategy

Our primary goal is reduction before compensation.

Travel Choices

Where feasible, we prioritise lower-carbon transport:

• Train over flights for European travel when the journey time is reasonable.

• Cycling and public transport for commuting and local meetings.

• Combining trips (e.g., conference + workshop + collaboration visit) to maximise scientific value per journey.

• Favouring invited talks and workshops that can be delivered online or hybrid when appropriate.

We work with the university travel agency (DGI) to ensure policy-compliant bookings and to maintain oversight of travel-related emissions. Where possible, we select routes and providers with lower environmental impact and avoid unnecessary short-haul flights.

Computing Efficiency

• Avoiding redundant reprocessing of large neuroimaging datasets.

• Sharing derived outputs internally rather than recomputing pipelines.

• Monitoring GPU usage and avoiding over-allocation of resources.

• Favouring institutional infrastructure that increasingly relies on low-carbon electricity.

Digital Practices

• Using Ecosia as our default search engine in the lab, contributing to reforestation initiatives.

• Limiting unnecessary high-resolution streaming and encouraging asynchronous viewing where appropriate.

• Hosting educational content efficiently and reviewing digital footprint annually.

Lab Culture & Everyday Practice

• Encouraging cycling to work where feasible (Nijmegen makes this pleasantly realistic).

• Minimising printing and using digital annotation tools.

• Reusing equipment and extending hardware lifecycles where technically viable.

• Supporting open science practices that reduce duplication of compute-intensive analyses across labs.

• We encourage the use of alternative platforms, such as the German Ecosia web browser (they plant a tree for every search) or the Swiss Euria AI (protects privacy and the planet by repurposing the heat generated by the servers to heat the city of Geneva).

​

​