Science1 distinct publisher3 min readPublished
Calibrating penguin guano color against guano chemistry gave one team a krill-share model they could run backwards through an archive that starts in the 1980s. It is cheap, it covers a whole continent, and it is correlational about krill, the number everyone wants pinned down.
The Scientist · Science desk
Compiled by The ScientistSomething wrong?How this is made
The calibration is where this method stands or falls, and it is the part worth copying. Guano samples went into a spectrometer in a makeshift shipboard lab, which measured each one as a satellite sensor would see it, and the same samples went through environmental chemistry to establish which diet had produced that color; statistical modeling then mapped color onto the krill fraction of the diet [6]. Taking two measurements from one sample is the methodological point: it turns guano color from a proxy for a proxy into a direct link to diet. The pink comes from krill's undigested exoskeleton, which puts it there [5], and the chemical assay of the same sample confirms it [6]. The authors say that, to their knowledge, nobody has previously monitored what animals eat directly from space [7].
About 4,000 images carry the analysis, drawn from an archive that begins in the 1980s [4][3]. Call that four decades [21], and the average works out to roughly 100 images a year for the entire range of the species [20]. Spread across colonies scattered around a continent, individual colony series are thin, and the reported variation within a season as well as between years [9] comes out of that same limited pool. The account here is a first-person essay by the researcher who led the group, and it does not report validation error for the color model or how many colonies went into the population comparison [19].
Krill share is a ratio, which is the honest limit of the readout. A colony whose guano turns pinker could be finding more krill or finding fewer fish, and the ratio moves the same way either way. The ice relationship is measured across colonies and years [8], not manipulated, and the step from less ice to fewer fish is the author's interpretation of that association [11].
The population result is the most useful finding and the easiest to over-read: colonies where penguins eat less fish have declined over the past few decades across the whole range, while colonies eating more fish are more often stable or increasing [13]. Chick mass supplies a mechanism that fits, since fish-fed chicks are heavier and heavier chicks are more likely to reach adulthood [12]. It remains a comparison between colonies that differ in more than diet, and the essay does not report what else was controlled for [19].
What transfers to other programs is the design: an optical archive that already exists, plus a pigment that survives digestion and can be pinned to diet by chemistry [6]. That is most valuable exactly where field work is hardest, which is the case the authors make for Antarctica [18].
Sea ice around the continent has declined sharply since the 2010s, with several years setting historical minima [10], and further declines are projected [14]. Meanwhile the same colonies face krill trawlers taking large volumes for fish feed and supplements [17], baleen whale populations that rebounded after commercial whaling ended [15], and gentoo penguins moving south as the climate warms [16]. The next decline can now be read against a forty-year record of prey use rather than a handful of field seasons [21].
Ranked by verification strength, evidence, and original report placement.
Adelie penguins are among the most abundant predators in the Southern Ocean; they feed in the ocean but nest on rocky outcrops around the Antarctic continent.
Adelie colonies often hold hundreds of thousands of nesting birds and generate so much guano that the patches can be readily identified from space using NASA's Landsat satellites.
A multidisciplinary group of scholars, led by the essay's author, used thousands of satellite images dating to the 1980s to examine penguin guano patches and infer what the birds were eating.
The team used information from around 4,000 satellite images to examine how penguin diets varied over time and space.
Krill have a pink exoskeleton that penguins do not fully digest, which contributes a rosy shade to their guano.
To build the color-to-diet link, the team collected guano at several locations, used a spectrometer in a makeshift shipboard lab to determine each sample's color as a satellite would see it, analyzed the guano's contents with environmental chemistry tools to determine the diet that produced that color, and used statistical modeling to link color to the krill share of the diet.
Distinct publishers with included, body-backed reporting in this cluster.
phys.org
1 article · September 3, 2026
Follow any of these and your For You feed starts watching them — no settings page required.
science
The krill fishery hits its limit on time and in the wrong places1 distinct publisher
science
H5N1 is in Australian resident birds, and overseas the first wave does most of the killing1 distinct publisher
science
Tides vary by nearly a metre inside one bay, and gauge-based flood maps smooth it out1 distinct publisher
science
Landsat's guano record backdates 18 emperor penguin colonies by 17 years on average1 distinct publisher
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
One voice, no numbers to check
The whole story rests on a first-person essay by the scientist who ran the study, carried verbatim by Phys.org from The Conversation. The calibration chain it describes is coherent and the mechanism — undigested pink krill shell tinting guano — is physically sensible. But nothing checkable is offered: no fit statistic for the colour model, no count of colonies or images that survived screening, no controls behind the claim that fish-poor colonies shrank. A reader who wanted to know how tight the inference is has nowhere to look.
Nothing yet shows anyone else using it
Our coverage contains no sign of anyone beyond the original team touching this: no released dataset or code, no second group replicating the colour calibration, no fisheries or conservation body citing the diet series. Absence this early is not failure, it simply leaves nothing to measure.
The framing runs ahead of the measurement
The findings are described with reasonable restraint; the sentences around them are not. 'First effort to directly monitor what animals are eating from space' is a priority claim hedged with 'to our knowledge' and left entirely untested, and 'less sea ice means fewer fish' arrives as settled causation in a study that measured guano colour, not fish stocks. Strip those two lines and what remains — regional and seasonal diet variation, plus a correlation between low-fish diets and shrinking colonies — is modest and interesting.
The method's inventor is also its narrator
The Conversation's format is researchers explaining their own work in their own voice, and Phys.org passes it through untouched. That buys the good detail — the makeshift shipboard lab that probably still smells — and costs the caveats paragraph, which is exactly the part a principal investigator has least reason to write. No funder is disclosed, no peer review is pointed to, and the krill industry named in the closing section is given no chance to reply.
Confident on direction, blind on magnitude
Two things hold up well: the physical basis for reading krill share off colour, and the reported pattern that ice-rich colonies eat more fish. What we cannot judge from here is how much error the colour model carries, how much of the population trend the diet variable actually explains, or whether the projected ice declines behind the outlook come from any specific model. That gap between a plausible mechanism and unquantified results is what holds this near the middle.