A new satellite-driven modeling framework developed by CBIOMES researchers and others enables global-scale estimation of photosynthesis–irradiance (P-I) parameters for marine phytoplankton, bypassing the need for ship-based incubations.
Reporting by Helen Hill for CBIOMES
The new study led by Gregory Britten (WHOI) with co-authors Bror Jönsson (UNH), Mick Follows (MIT), and Gemma Kulk (PML), Heather Bouman (Oxford), and Shubha Sathyendranath (PML), introduces a breakthrough approach for predicting how marine photosynthesis responds to sunlight and temperature – using only satellite observations. Published in Limnology and Oceanography Letters, the research offers a powerful tool to help estimate primary production across the global ocean, a process that drives Earth’s carbon cycle and sustains marine food webs.
Phytoplankton, microscopic plants drifting in the ocean, perform approximately half of Earth’s photosynthesis. Their growth depends on light and nutrients, and scientists often describe this relationship using photosynthesis – irradiance (P-I) curves, which quantify how photosynthetic rates change with light intensity and in turn depend on local ecosystem conditions. Traditionally, measuring these curves requires ship-based incubations and specialized sensors – a costly and logistically challenging task that limits coverage to small regions.
Britten’s team sought to overcome this limitation by leveraging satellite remote sensing. Modern satellites provide global measurements of ocean color, which can be used to infer chlorophyll concentration, light penetration, and temperature. The researchers developed a machine learning model that links these satellite-derived properties to P-I parameters, enabling predictions of photosynthetic efficiency without direct sampling. “Our goal was to bridge the gap between local measurements and global-scale observations,” said Britten. “By using satellite data, we can estimate photosynthetic responses across the entire ocean in near real time.”
The study compiled a large dataset of in situ P-I measurements from diverse ocean regions, paired with coincident satellite observations. Using machine learning techniques, including random forests and random effects regression, the team trained models to predict key P-I parameters – such as the initial slope of the curve (α) and the light-saturated photosynthetic rate (Pmax) – from satellite-observed variables including chlorophyll concentration, sea surface light intensity, and sea surface temperature.
The results were promising. The satellite-based predictions captured regional and seasonal variability in P-I relationships, outperforming traditional empirical algorithms and achieving >70% when predicting P-I relationships that the model was not trained on. For example, the model correctly identified higher photosynthetic efficiency in nutrient-rich high-latitude waters and lower efficiency in oligotrophic subtropical gyres. These patterns are critical for estimating primary production and understanding how ocean ecosystems respond to climate change.
Importantly, the approach also quantified uncertainty, allowing researchers to assess confidence in predictions and identify regions where additional measurements are needed. This feature makes the method suitable for integration into global biogeochemical models and Earth system simulations. “Primary production is a cornerstone of ocean ecology and climate,” Britten explained. “Improving our ability to estimate it from space will enhance carbon cycle forecasts and help us understand how marine ecosystems adapt to environmental change.”
The implications extend beyond academic research. Accurate, large-scale estimates of photosynthesis can inform fisheries management, carbon sequestration strategies, and assessments of ocean health. As climate-driven shifts in temperature and stratification alter light and nutrient regimes, tools like this will be essential for tracking ecosystem responses.
Looking ahead, the team plans to refine the model by incorporating additional satellite products, such as hyperspectral ocean color and data from upcoming missions like NASA’s PACE. They also aim to couple the predictions with dynamic models of phytoplankton physiology, creating a more mechanistic framework for global primary production estimates.
Publication
Gregory L. Britten, Bror Jönsson, Gemma Kulk, Heather A. Bouman, Michael J. Follows, Shubha Sathyendranath (2025), Predicting photosynthesis–irradiance relationships from satellite remote-sensing observations, Limnology and Oceanography Letters, doi: 10.1002/lol2.70062