When Everything Is a Heat Wave and Every Change Is Climate

Charles Rotter

A recent paper in Science Advances bears the authoritative title “A major heat wave in the North Atlantic had widespread and lasting impacts on marine life.” The phrasing is careful but decisive. A single physical event is asserted, and its ecological consequences are presented as both extensive and enduring. For readers accustomed to climate-ecology literature, the implication is clear: an extreme climatic anomaly occurred, ecosystems responded, and the link between the two is sufficiently established to justify confidence.

The paper positions itself as both synthesis and analysis. It compiles physical temperature records, surveys a wide range of ecological indicators, and proposes that the year 2003 marks a pivotal moment—a thermal disruption that reorganized the North Atlantic marine ecosystem. The ambition is not modest. The authors are not merely cataloging observations; they are advancing a causal narrative.

Early on, the paper states its central premise plainly:

“Here, we test whether this sudden and stark rise in temperature classifies as an MHW, examine the physical drivers behind it, and investigate its impacts on marine ecosystems.”

This sentence does a great deal of work. It asserts the existence of a “sudden and stark rise,” presumes that classification as a marine heat wave is a meaningful exercise, and then moves directly to impacts. What it does not do is pause to ask whether the quantities being connected—“the MHW” and “marine ecosystems”—are defined with sufficient precision to support causal inference. That omission is not incidental; it permeates the paper.

The first weak link is the marine heat wave itself. Despite the strong connotations of the term, the paper does not identify discrete heat-wave events in the ocean in any physically intuitive sense. Instead, it constructs a metric based on threshold exceedances. Temperatures above the 99th percentile of a historical reference period are counted across 88 different time series and summed by year. The resulting aggregate is labeled “MHW frequency.”

The authors describe this as follows:

“We define an MHW as any temperature exceeding the 99th percentile of the 1870–1969 reference period and count the number of MHWs each year.”

What is being counted here is not events, but exceedances. A single basin-wide warm anomaly can inflate this number dramatically if many correlated locations cross their thresholds at once. The metric is therefore sensitive not just to temperature, but to spatial coverage, data density, and correlation structure. Calling this “frequency” suggests a count of independent occurrences. That is not what the method produces.

This matters because frequency implies repetition, recurrence, and increased likelihood of impact. But what the paper is really tracking is the spatial coherence of warmth relative to a chosen percentile. A broad warm year will look like many heat waves, even if it is physically one phenomenon. The distinction is blurred, then forgotten, as the narrative advances.

The construction becomes even more problematic when the temperature data themselves are examined. The paper combines gridded sea-surface temperature fields with point-based, subsurface observations at approximately 100 meters depth. These are not interchangeable measures of “heat.” They respond to different processes, operate on different time scales, and interact differently with biological systems. Yet they are treated as equivalent contributors to a single index.

The ecological side of the ledger is even less well defined. The paper repeatedly refers to “ecosystem impacts,” “abrupt changes,” and “widespread responses,” creating the impression of a well-established baseline from which deviations can be measured. In reality, the baseline is an assemblage of disparate observations drawn from incompatible sources.

The paper itself hints at this breadth:

“Evidence for ecosystem change was compiled from a wide range of peer-reviewed publications and stock assessment reports spanning multiple trophic levels.”

That breadth is presented as strength. But breadth without standardization does not produce a baseline; it produces a mosaic. Fish stock assessments, plankton indices, benthic surveys, and fisheries-dependent data are all governed by different assumptions, sampling strategies, and human influences. Each has its own implicit reference period, often shifting over time. The paper does not reconcile these differences. It aligns them temporally and treats concurrence as coherence.

Many of the changes described are qualitative judgments elevated through repetition. Terms like “abrupt,” “sudden,” and “marked” recur throughout the cited literature, yet formal tests against null models of natural variability are rare. A change that appears striking within a short observational window is implicitly treated as exceptional, even though the historical envelope of variability is poorly constrained.

The problem is compounded by the retrospective nature of many ecological baselines. Species “expansions” are often inferred from first detection rather than systematic absence. Abundance shifts are inferred from improved surveys compared against earlier, less reliable data. Once an event like 2003 is identified as climatically noteworthy, ecological observations are naturally reinterpreted in its shadow.

The paper leans into this alignment explicitly:

“The 2003 MHW coincided with abrupt ecosystem changes across multiple trophic levels.”

Coincidence is doing a great deal of work here. Temporal overlap is treated as explanatory, even though the mechanisms linking a basin-scale temperature anomaly to diverse ecological responses are not rigorously established. Fishing pressure, regulatory changes, survey redesigns, and market dynamics—all of which profoundly shape marine data—are acknowledged only peripherally.

Nowhere is this more evident than in the treatment of fisheries data. Stock redistributions and abundance changes are presented as ecological responses, even though fisheries systems are deeply socio-ecological. Changes in quotas, targeting strategies, and detection technology can produce “abrupt” shifts in reported distributions without any underlying biological reorganization. The baseline, such as it is, quietly assumes that fisheries outputs reflect ecological truth rather than human behavior layered onto ecosystems.

The paper goes further, invoking regime-shift language that carries strong causal implications:

“The observed changes are consistent with the characteristics of regime shifts previously identified in marine ecosystems.”

Yet no formal regime-shift detection is performed. There is no change-point analysis, no state-space modeling, no statistical demonstration that 2003 represents a structural break rather than a conspicuous point in a noisy, evolving system. Consistency with a regime-shift narrative is asserted, not demonstrated.

Perhaps the most revealing moment comes when the authors acknowledge a tension they do not resolve. Later periods show similarly high values of the heat-wave metric, yet the ecological responses appear less dramatic or less well documented. The paper notes this, then moves on. But this observation undermines the central claim. If the same “forcing” produces different outcomes, either the forcing is not well characterized, or it is not the dominant driver.

At no point does the paper seriously consider the possibility that the baseline itself is unstable. North Atlantic ecosystems were not in equilibrium prior to 2003. They were already responding to decades of fishing pressure, nutrient changes, and circulation variability. Treating the late 20th century as a stable reference state is a narrative convenience, not an established fact.

In the end, the paper offers a compelling story, but not a disciplined causal argument. A loosely defined marine heat wave is paired with a loosely defined ecological baseline, and the two are joined by temporal coincidence and confident language. The synthesis feels objective because it is large and comprehensive, not because its foundations are secure.

What is presented as evidence of causation is better understood as plausibility structured into a narrative. That distinction matters. When weakly constructed quantities are allowed to support strong claims, uncertainty is not reduced—it is concealed.

The danger is not that this paper alone overreaches. The danger is that this style of reasoning becomes standard: correlation framed as causation, baselines inferred rather than established, and skepticism replaced by accumulation. That is not how complex systems are understood. It is how stories harden into doctrine.