From Pollinators to Portfolios: How the Lantern Project Linked Nature to Credit Risk

[tori_mann]

Discover team Green Finance Lantern's innovative solution- shared by their mentor, Miles Elliott:

*EY France, under the sponsorship of Franck Chevalier, Nathanael Sebbag and Antoine Helouin, elected to enter this year's 2025 SAS Hackathon.  The team, under the operational management of Christophe Beaugendre, identified the topic of linking Biodiversity Risk to Corporate Credit Risk.  In this short paper we will explore what was done, why and what were the key outcomes.  Specific credit goes to the work of AI & Data, ESG and sustainability teams from EY France*

In finance, biodiversity often appears as an abstraction—something beautiful yet detached from balance sheets and credit models.

The Lantern Project set out to change that. Over an intense few weeks during the SAS Hackathon, this team demonstrated how the loss of pollinators—tiny, tireless agents of ecosystem health—can translate into quantifiable financial risk.

As a mentor on this project, I had the privilege of guiding a group of innovators who transformed a complex sustainability concept into a working data-driven prototype. The journey from idea to insight revealed not only technical creativity but also a glimpse of what nature-positive finance might look like in practice.

Defining the Challenge: Making Biodiversity Financially Visible

The team began with a deceptively simple question:
Can biodiversity loss be measured as a financial variable?

More specifically, could we model how variability in pollination—an ecosystem service essential to many crops—affects agricultural production and, ultimately, the creditworthiness of farmers?

Increasingly, financial regulators and institutions are asking how natural capital risks might materialize on balance sheets. Pollinator variability, being both measurable and economically relevant, provided a tangible entry point.

Our goal was to build a proof of concept that would:

1. Quantify the relationship between pollination and farm-level yields;
2. Translate those yield effects into credit risk metrics, such as the probability of default (PD);
3. Present these insights through a SaaS-powered dashboard capable of guiding lending decisions.

Framing the Data Strategy

Early in the process, the team realized that data would make or break the entire model. Biodiversity is inherently spatial—it happens in fields, not spreadsheets. So our data architecture needed to merge geospatial ecological datawith financial data at the level of individual farms.

We began with publicly available sources:

• The French Agricultural Parcel Registry (RPG): a detailed dataset covering roughly 140,000 plots in the PACA (Provence-Alpes-Côte d’Azur) region. Each plot included coordinates and crop types.
• Ecological data from the Dutch PBL Institute, offering land-cover projections from 2020 to 2050 under different sustainability scenarios (SSP1 and SSP5).
• Scientific data on crop pollinator dependence, derived from global studies quantifying how yield responds to pollinator availability.

Using these, the team constructed a data pipeline that could spatially align and harmonize information at the farming plot level.

Each farm was approximated by clustering neighboring plots using a hexagonal grid system, and every plot was assigned a percentage of surrounding land favorable to pollinators. This figure—a measure of pollinator habitat sufficiency—served as the ecological anchor of the model.

To ensure privacy and realism, we then used SAS DataMaker to generate a synthetic portfolio of 3,000 farmers, complete with production and financial attributes like assets, debts, and leverage. The synthetic design allowed us to test relationships safely, while maintaining plausible variability.

Designing the Models: From Ecosystem Service to Default Probability

Once the data foundation was in place, the next step was to create a modeling framework that could translate ecological stress into financial outcomes.

The modeling pipeline followed three logical steps:

1. Yield Estimation:
   Using the pollinator sufficiency metric, we estimated how yield would change under different pollination scenarios. For instance, if the share of pollinator-friendly habitat fell from 27% to 12%, sufficiency—and hence yield—might decline by roughly 60%.
2. Revenue and Profitability Projection:
   These yield changes were then aggregated to the farm level, taking into account each farm’s mix of crops. Farms dominated by pollinator-dependent crops such as apples, melons, or berries showed higher vulnerability than those focused on cereals or oilseeds.
3. Credit Risk Adjustment:
   The financial model linked projected profitability changes to leverage and default risk, adapting concepts from a 2021 European Central Bank climate stress-testing paper. In simple terms, lower yields lead to lower revenue, which raises leverage ratios and increases the probability of default (PD).

Finally, each farmer’s PD under different future scenarios was used to categorize lending decisions:

• Lend normally (low PD and low change in PD)
• Lend with conditions (moderate PD increase)
• Do not lend (high or sharply increasing PD)

This categorization bridged environmental and financial analytics in a single decision framework—a crucial step toward nature-aware lending.

The Role of SAS and AI

All this analysis was orchestrated within the SAS Viya platform, which handled data ingestion, transformation, and modeling at scale. The team also prototyped a large language model (LLM)-powered decision tree to automate recommendations—essentially an AI loan officer that considers both environmental and financial parameters.

An interactive dashboard displayed every farm as a point on the map, color-coded by size and PD impact. Users could drill down into each holding to see:

• Crop composition
• Pollinator habitat metrics (2020 vs. 2050)
• Yield projections and financial indicators
• Lending recommendations

This visual synthesis made the complex interplay between biodiversity and finance immediately tangible for bankers, risk analysts, and policymakers.

Insights and Outcomes

One of the project’s most striking findings was its selective impact.
Out of 140,000 plots analyzed, only about 24,000 (roughly 500 farmers) faced meaningful pollinator habitat decline. However, for those affected, the financial consequences were significant—particularly when high dependency crops dominated their portfolio.

The implication is clear: biodiversity risk is localized but material.
Financial institutions equipped with such insights could tailor credit terms, prioritize resilience investments, or offer transition finance to at-risk farmers.

Beyond the numbers, the Lantern Project proved something more profound: that nature’s value can be expressed in financial terms without diminishing its essence. When biodiversity enters the credit model, sustainability shifts from a moral imperative to a measurable factor in economic resilience.

Reflections on Mentorship and Collaboration

From a mentor’s perspective, what stood out most was the team’s willingness to cross boundaries. Data scientists learned to read ecological maps. Finance analysts debated spatial modeling. Sustainability experts interpreted default probabilities.

This spirit of interdisciplinary fluency turned a theoretical problem into a functioning prototype—one that hints at the next generation of ESG analytics.

Looking Ahead

Lantern is just the beginning. The same architecture could be extended to other ecosystem services—water retention, soil health, or carbon storage—and across other regions.

By embedding natural capital into financial models, we can begin to design credit systems that reward resilience, not just yield. In doing so, we move a step closer to aligning finance with the planet’s true balance sheet.

- Miles Elliott

A huge thank you to our sponsors, Intel and Microsoft, whose support made this Hackathon possible. Their contributions went far beyond sponsorship- bringing expertise, resources, and inspiration that fueled innovation. Together, they created a truly collaborative and transformative experience for all participants.

Green Finance Lantern is just one of many visionary teams from this year’s Hackathon. Find out who the champions are during our award session Dec. 11 on YouTube and LinkedIn!