Columbia Climate School Fuels Impactful Research with FY27 Seed Grants 

September 15, 2026

Columbia University's Climate School Office of Research is proud to announce the selection of four outstanding research teams to receive support through the FY27 Research Seed Fund Program. This year's internal seed grant initiative offered awards ranging from $20,000 to $100,000, aimed at propelling interdisciplinary research that generates significant societal impact and advances the understanding of critical climate, Earth, and society challenges. The competition accepted 49 proposals, and 8 reviewers generously lent their time and expertise to help select this year's awardees.

"We are thrilled to support these four teams as they pursue bold, interdisciplinary ideas that push into new territory," said Janice Savage, Associate Dean of Research Administration at the Columbia Climate School. "This was an incredibly competitive round — we received so many strong, creative proposals from across the Climate School, and narrowing the field down was genuinely difficult. We're grateful to everyone who applied, and we're excited to see the insights this year's awardees generate." 

FY2027 SEED FUND AWARDEES

INTEGRATING MOLECULAR AND MACHINE LEARNING TECHNIQUES TO RESOLVE VIRAL CONTROL ON CARBON CYCLING
PI: Annika Gomez | Co-PI: Sonya Dyhrman

Accurately quantifying biologically-driven carbon fluxes in the ocean is an essential tool in understanding and addressing climate change. Every year phytoplankton in the surface ocean fix roughly 65 billion metric tons of CO2 into organic carbon, moving carbon between the atmosphere and the ocean. The subsequent fate of this carbon is shaped by the ubiquitous presence of phytoplankton viruses, which may either enhance long-term storage of organic carbon or release of CO2 back into the atmosphere. Currently, our understanding of how viruses redirect newly fixed carbon in the ocean is limited by the fact that the specific host identity is unknown for the vast majority of phytoplankton viruses. Without a known association with phytoplankton hosts, the impact of viruses on marine carbon cycling and sequestration remains unquantifiable.

With this project, the team will implement molecular techniques to connect viruses with their hosts in samples collected in the North Atlantic Ocean, building an accurate, unbiased host-virus interaction database. Using this new database, the team will apply a machine learning approach to predict phytoplankton host-virus interactions from field genomic data. The resulting model will be used to infer host-virus interactions in publicly available datasets, offering an unprecedented opportunity to rapidly advance parameterization of viral dynamics into biogeochemical models. The generalizable conclusions from this work and its further application will provide critical data needed to constrain the microbial processes that shape the fate of carbon, and predict the success of marine carbon dioxide removal strategies that use microbial processes.

ACCELERATING PHYSICS-BASED LAVA FLOW MODELS TOWARDS MACHINE LEARNING EMULATION AND REAL-TIME SIMULATION
PI: Einat Lev | Co-PI: Dhruv Balwada

When a volcano erupts, decisions have to be made fast: which communities to evacuate, which flight paths to close, where ash and lava are headed next. The computer models that answer those questions are detailed and physically realistic, but they can take hours or days to run — time a crisis rarely allows. This project is making several of those models dramatically faster, through a combination of code optimization and GPU computing. The gains matter in two ways. In an emergency, a model that runs in minutes rather than days can actually inform the response. Over the longer term, speed makes it practical to run thousands of scenarios instead of a handful, producing hazard assessments that account for the full range of ways an eruption might unfold.

Speed also opens a further door. A large library of simulation results can be used to train machine-learning emulators — compact models that learn the behavior of the full simulation and reproduce it in seconds. The accelerated codes and emulators will be shared through a cloud-based computational portal, making them available to volcano observatories, civil protection agencies, and researchers anywhere in the world. The project pairs volcanologist Einat Lev with oceanographer Dhruv Balwada. Ocean and climate modelers have spent years accelerating their codes and building emulators; volcanology has been slower to adopt these methods. The collaboration is a deliberate transfer of expertise between two fields that share more computational ground than their subject matter suggests.

A KNOWLEDGE, ATTITUDES, AND PRACTICES (KAP)-INFORMED FLOOD RISK COMMUNICATION PILOT IN DOUALA
PI: Pavithra Priyadarshini Selvakumar | Co-PI: Lovees Ahfembombi Lueong

Douala, Cameroon's economic capital, faces escalating urban flood risks driven by rapid unplanned urbanization, poor drainage infrastructure, low-lying coastal geography, and increasingly intense rainfall. In Bonaberi, one of Douala's largest informal settlements, flooding recurs throughout the rainy season from April to October, causing fatalities, displacement, livelihood disruption, infrastructure damage, and waterborne disease outbreaks. This project proposes a nine-month seed pilot that combines a Knowledge, Attitudes, and Practices (KAP) study with community workshops to understand how flood risk information currently reaches residents in Bonaberi and how it can be improved.

The KAP study will provide a simple baseline on what residents know about flood risks, which information sources they trust, how they understand warning messages, and what actions they take before, during, and after flooding. The team will then use community workshops to discuss these findings with residents and trusted stakeholders, including women's groups, faith leaders, health workers, community organizers, municipal actors, digital content creators, and local information providers. These workshops will help identify where communication is breaking down, which messengers are trusted, and what types of messages are most likely to be understood and acted upon. During these workshops, the team will develop and test a small set of practical communication materials, such as sample flood warning messages, preparedness guidance, and trusted messenger pathways. The goal is to generate clear, community-grounded evidence on what works, what does not, and what would be needed for a larger proposal.

CONFLICT, FERTILIZER SUPPLY CHAINS, AND AGRICULTURAL VULNERABILITY: QUANTIFYING CROP PRODUCTION RISKS FROM FERTILIZER DISRUPTIONS
PI: Meijian Yang | Co-PI: Jonas Jägermeyr Additional Co-PIs: Michael Puma, Walter Baethgen, Jeffrey Schlegelmilch, Jyoti Singh

Food production around the world depends on stable access to key agricultural inputs, especially fertilizer. However, conflicts and geopolitical crises can disrupt fertilizer trade, delay shipments, raise prices, and create uncertainty for farmers. These shocks can change fertilizer use, increase production costs, reduce crop yields, and heighten food-security risks, especially in import-dependent or climate-stressed regions.

This project will examine how conflict-related fertilizer disruptions can affect agricultural production and food security using past and ongoing geopolitical disruptions — including the Russo-Ukrainian war-related Black Sea trade disruptions and the closure of the Strait of Hormuz — to understand how fertilizer supply shocks move through trade systems, input markets, farmer decisions, and crop production. It will develop a prototype dataset to identify countries and crop systems most exposed to fertilizer supply risks, and use crop modeling tools to estimate how changes in fertilizer availability, price, application rate, and timing may affect major crops. The analysis will include global-scale and selected regional examples to show why the same fertilizer shock may have different impacts across countries, crops, soils, climates, and farming systems.

Through the seed project, the team expects to produce preliminary datasets, case studies, crop-impact simulations, and maps identifying areas most vulnerable to fertilizer-related agricultural risk. It will support future early-warning systems and targeted strategies that help governments, humanitarian organizations, and agricultural planners anticipate fertilizer-related production risks, identify vulnerable regions before crises deepen, and design timely interventions to strengthen food security under geopolitical uncertainty.