Harnessing Ensemble Models for Robust Near-Term Population Forecasts under Environmental Change
Lead Research Organisation:
University of Sheffield
Abstract
Being able to forecast how wildlife populations change in the near future is becoming more urgent than ever. As climate change, habitat loss, and other pressures intensify, ecosystems are facing new and unexpected challenges. From setting sustainable harvest limits to protecting endangered species, conservation managers and policymakers rely on forecasts to make timely, effective decisions. Yet the reality is that reliable forecasts are often missing when they’re needed most.
Traditionally, ecologists have made long-term predictions about how species will change over decades or even centuries. These long-term forecasts are crucial for understanding broader trends, but they are often too uncertain and too slow to be useful for real-time decision-making. One major problem is that many forecasting methods assume the relationships observed in historical data will remain unchanged into the future. In today’s world of rapid climate shifts and unpredictable environmental events, this is rarely true.
Near-term iterative ecological forecasting (NTEF) offers a promising solution. Instead of projecting far into the future, NTEF focuses on making short-term predictions — days to a few years ahead — and then regularly updating those forecasts as new data come in. This “forecast-update-repeat” cycle makes NTEF uniquely suited for today’s changing world. It enables forecasts to remain current as conditions change, providing managers with up-to-date, decision-relevant information when they need it most.
Despite its promise, NTEF is still in its early days. So far, ecologists have limited understanding of which modelling approaches work best, how to deal with uncertainty, or how to adapt forecasts when conditions change. Other fields, like weather forecasting and epidemiology, have overcome similar challenges by using ensemble forecasting—combining multiple different models into a single, more reliable prediction. But in ecology, ensemble forecasting for populations remains underdeveloped.
This project will deliver a major leap forward by combining the strengths of NTEF and ensemble approaches. We will develop and rigorously test new methods that show us how to make short-term forecasts more accurate and useful in the face of uncertainty. By blending insights from different models, our ensemble forecasts will capture a wider range of possibilities and help managers act with greater confidence.
To achieve this, we will connect three strands of work. First, we will conduct computer simulations to test how species' life cycles, environmental variability, and data quality affect the performance of near-term ensemble forecasts. Second, we will validate these ideas through laboratory experiments with live populations under controlled conditions, testing how well near-term, updated forecasts handle real-life unpredictability. Finally, we will apply what we learn to a unique long-term study of wild sheep, demonstrating how near-term ensemble forecasting can inform real conservation decisions in the wild.
Beyond advancing science, this project will deliver real-world impact by co-developing open, user-friendly forecasting tools and training resources with leading conservation partners, ensuring they are practical and ready for adoption by researchers, conservation organisations, and policymakers. By pioneering near-term, iterative population forecasting, we will equip future decision-makers to stay ahead of change and take timely, effective action for nature in an unpredictable world.
Traditionally, ecologists have made long-term predictions about how species will change over decades or even centuries. These long-term forecasts are crucial for understanding broader trends, but they are often too uncertain and too slow to be useful for real-time decision-making. One major problem is that many forecasting methods assume the relationships observed in historical data will remain unchanged into the future. In today’s world of rapid climate shifts and unpredictable environmental events, this is rarely true.
Near-term iterative ecological forecasting (NTEF) offers a promising solution. Instead of projecting far into the future, NTEF focuses on making short-term predictions — days to a few years ahead — and then regularly updating those forecasts as new data come in. This “forecast-update-repeat” cycle makes NTEF uniquely suited for today’s changing world. It enables forecasts to remain current as conditions change, providing managers with up-to-date, decision-relevant information when they need it most.
Despite its promise, NTEF is still in its early days. So far, ecologists have limited understanding of which modelling approaches work best, how to deal with uncertainty, or how to adapt forecasts when conditions change. Other fields, like weather forecasting and epidemiology, have overcome similar challenges by using ensemble forecasting—combining multiple different models into a single, more reliable prediction. But in ecology, ensemble forecasting for populations remains underdeveloped.
This project will deliver a major leap forward by combining the strengths of NTEF and ensemble approaches. We will develop and rigorously test new methods that show us how to make short-term forecasts more accurate and useful in the face of uncertainty. By blending insights from different models, our ensemble forecasts will capture a wider range of possibilities and help managers act with greater confidence.
To achieve this, we will connect three strands of work. First, we will conduct computer simulations to test how species' life cycles, environmental variability, and data quality affect the performance of near-term ensemble forecasts. Second, we will validate these ideas through laboratory experiments with live populations under controlled conditions, testing how well near-term, updated forecasts handle real-life unpredictability. Finally, we will apply what we learn to a unique long-term study of wild sheep, demonstrating how near-term ensemble forecasting can inform real conservation decisions in the wild.
Beyond advancing science, this project will deliver real-world impact by co-developing open, user-friendly forecasting tools and training resources with leading conservation partners, ensuring they are practical and ready for adoption by researchers, conservation organisations, and policymakers. By pioneering near-term, iterative population forecasting, we will equip future decision-makers to stay ahead of change and take timely, effective action for nature in an unpredictable world.