Retrieving Chl-a Is Not Retrieving Phytoplankton Biomass

Applied remote-sensing papers often retrieve one measurable variable and then use it to discuss a related environmental object or state. Chlorophyll-a (Chl-a), for example, is frequently used to discuss phytoplankton biomass or bloom condition; total suspended solids (TSS/SPM) to infer turbidity or particulate transport; leaf area index (LAI) to characterize canopy structure or vegetation condition; and above-ground biomass (AGB) to discuss terrestrial carbon storage or ecosystem function.

The full inferential chain from remote-sensing observation to environmental interpretation can be written as:

X → R ≈ M → O

X is the observed signal: radiance, reflectance, remote-sensing reflectance (Rrs), fluorescence-related signal, or another quantity recorded by a sensor. R is the retrieval result, an estimate derived by an algorithm from that observation. M is the reference variable, the variable actually used for model training, parameter calibration, and accuracy validation. O is the environmental object or state of interest: the ecological object, environmental process, or spatial pattern that the study ultimately wants to explain.

The chain therefore contains two different operations. X→R≈M is the remote-sensing retrieval. M→O is the environmental interpretation.

Consider Chl-a. A satellite does not directly observe the ecological object “phytoplankton biomass.” It records optical or radiometric properties of the water (X), and an algorithm produces a Chl-a retrieval (R). Model training, calibration, and validation are referenced to Chl-a concentration measured by a laboratory or field method (M), not to phytoplankton biomass itself (O). When the retrieved Chl-a is then used to discuss phytoplankton biomass, bloom status, or trophic condition, the analysis has moved from retrieval into environmental interpretation.

This pattern is common in the literature. Ferreira et al. (Nature Communications, 2024), studying the West Antarctic Peninsula, used a regionally calibrated Chl-a algorithm and then interpreted the retrieved Chl-a in terms of changes in phytoplankton biomass and bloom phenology. The directly retrieved variable remained Chl-a; the biomass and phenology claims depended on interpreting Chl-a as a biomass proxy. A global study of coastal phytoplankton blooms similarly described Chl-a as a proxy for phytoplankton biomass while explicitly noting that satellite-detected blooms mainly represent high surface biomass and cannot directly identify toxicity or species composition (Dai et al., Nature, 2023).

The distinction between retrieval and environmental interpretation matters because a representation relation sits between them.

The core strength of remote sensing is its ability to take a reference variable M that is otherwise local, discrete, and dependent on field or laboratory measurement, and extend it into a spatially continuous, comparable, repeatable regional estimate. What the satellite product directly provides is a regionalized estimate of M—not an unmediated observation of environmental object O. O enters the discussion because M is known, or assumed, to represent it in some way.

That representation can be strong or weak, robust or conditional on region, scale, physiology, or ecological context. But it must be stated. It should not be silently skipped.

Many remote-sensing studies blur the actual path from R to O. The path is R≈M and then M→O. It is a composite chain, not one direct retrieval. In that sense, R→O is a proxy of a proxy: the retrieval first approximates the reference variable, and the reference variable is then used to represent the environmental object.

Once the Proxy-of-a-Proxy structure is made explicit, several risks become visible. Retrieval error propagates along the interpretive chain and acquires stronger meaning at the object level; an error that is acceptable for M may no longer be acceptable for O. More importantly, the M–O relation is never unconditional. It depends on region, scale, and ecological context. Outside those conditions, an interpretation that appears natural can lose its evidential basis.

Any remote-sensing study that ultimately makes a claim about O should therefore state the representation relation it relies on. What exactly is M? Under what conditions is M associated with O? What are the limits of that relation? These are not optional qualifications. They are part of the validity conditions of the result.

References

• Ferreira, A., Mendes, C. R. B., Costa, R. R., Brotas, V., Tavano, V. M., Guerreiro, C. V., Secchi, E. R., et al. Climate change is associated with higher phytoplankton biomass and longer blooms in the West Antarctic Peninsula. Nature Communications, 15, 6536 (2024).
• Dai, Y., Yang, S., Zhao, D., Hu, C., Xu, W., Anderson, D. M., Li, Y., Song, X.-P., Boyce, D. G., Gibson, L., Zheng, C., & Feng, L. Coastal phytoplankton blooms expand and intensify in the 21st century. Nature, 615, 280–284 (2023).