Peter Steward

Peter Steward is a climate adaptation scientist who builds the evidence and data infrastructure that food and land systems decisions rest on. Based in Nairobi with the Alliance of Bioversity International and CIAT, he works where continent-scale climate risk analysis meets the evidence from thousands of field trials.

He is technical lead for the Africa Agriculture Adaptation Atlas, an open, harmonised data infrastructure covering the African continent on a consistent sub-national grid, and writes the pipeline behind it: downscaled CMIP6 and observational climate data turned into crop- and livestock-specific hazard risk surfaces, intersected with exposure and published as open, cloud-native data. The aim is practical — moving climate risk from a barrier into a basis for informed investment. He represents the Alliance on the leadership team of the CGIAR Climate Data Hub, responsible for its metadata standard, data catalogue and a system-wide map of climate data assets across eleven CGIAR centres, so that evidence scattered across the system becomes findable and usable.

Alongside this, Peter manages data extraction and meta-analysis for Evidence for Resilient Agriculture (ERA), synthesising outcomes from thousands of agronomic and livestock experiments across Africa and Latin America. The ERA database was published in Scientific Data in 2024.

Much of his work addresses the climate finance gap: a geospatial methodology for the World Bank to screen and prioritise nature-based solutions across rural landscapes; climate rationales that give Green Climate Fund proposals a defensible evidence base; and a free Monte Carlo uncertainty calculator, built with the Global Methane Hub, that lets national teams report livestock emissions with credible confidence intervals under the Paris Agreement's transparency framework. With AIMS Rwanda and the McKnight Foundation he leads an open-source Python Climate Toolkit that lets agroecology projects attach measured climate context to their field records.

The users are consistent: Green Climate Fund proposal teams, the World Bank, AfDB and IFAD, national meteorological services and ministries, and regional partners including RCMRD, AGNES, ANAPRI, ASARECA and ATI Ethiopia. This work is judged by whether those users can defend a methodological choice in front of a reviewer, so Peter is deliberate about what the evidence will and will not support.

He trained as an ecologist, with a PhD from the University of Leeds on ecosystem services in small-scale tropical agriculture, followed by field research in Malawi on conservation agriculture under climate stress and in South Africa on pollination in apple orchards. He works primarily in R and Python, building on STAC, cloud-optimised geospatial formats and reproducible pipelines, and treats AI coding agents as research instruments, run inside audited loops with deterministic checks and traceable provenance.