The Polar Researcher

The processing trailer is twenty-three meters from the main building and at this time of year the walk between them is the only outdoor exercise Elin gets. She makes it four times a day. Three minutes each way at minus thirty-six, in the wind, with the goggles fogging if you breathe up into them. The fog is the worst part. The cold is the cold.

She is thirty-eight, Norwegian, a radioglaciologist at NPI in Tromsø, on her fourth austral summer at Vostok under the bilateral arrangement the institute still calls the bilateral arrangement in internal emails. The arrangement gives her bench time on the bedrock-radar dataset in exchange for a co-authored paper a year on the basal hydrology, senior-author slot reserved for whichever Russian PI is on the publication. Currently this is Mikhail Vasilyevich Yegorov, of the Arctic and Antarctic Research Institute in St. Petersburg, who has co-authored eleven of her papers and read approximately two.

The bedrock radar this season is the AWI EMR-4 — an ice-penetrating system mounted on a Pisten Bully traverse that ran the eastern margin grid in late November, before she arrived. The grid is the same grid they have run for nine consecutive seasons. Line spacing 500 m. Fold 24. Center frequency 60 megahertz. The geometry is fixed precisely so that this exact comparison — year-over-year delta on a slow-moving subglacial feature — is the comparison the dataset is designed to support.

She has been doing this comparison for nine seasons. The processing chain is hers. She wrote half of it in MATLAB and the other half in Python and she has been quietly migrating the MATLAB half for two years without telling Yegorov because he prefers MATLAB. The Python half is faster and produces the same answer.

This morning she runs the 2025 acquisition through the chain at the standard mask. She runs it again with the seasonal navigation correction. She runs the migration. She loads the 2024 result into the second viewer. She tiles them.

The Bell footprint, the part of it the grid covers, is the same. Then it is not.

She sits back. She looks at the two panels for the time it takes to be sure that what she is seeing is not, as she initially assumed, a misalignment.

The englacial reflectors line up. The basal interface lines up. The internal layering across the firn and the deep ice lines up to within a wavelength along the whole profile. The navigation is clean — RTK-corrected against the station benchmark, sub-decimeter residual across the traverse. Two loops cross within five meters of last year’s loops. The match at the crossings is good.

The basal feature on the east margin — the orthogonal radar-bright reflector at the bedrock interface, the one that the literature has been carefully calling a probable basal accretion artifact for twenty years and that her supervisor in Tromsø privately calls the Bell thing — has changed.

Not its position. Its geometry.

Last year the reflector presented as a planar return, dipping shallowly to the east, with diffraction hyperbolae at both terminations that suggested clean edge geometry. This year the eastern termination is still where it was. The western termination has migrated. Not the location of the edge — the character of it. Last year’s western edge was a clean, single hyperbola. This year there are three hyperbolae, stacked vertically, at roughly seventy-meter spacing. The middle one is the brightest.

She zooms in on the stack. She runs the deconvolution. The hyperbolae sharpen.

She reaches for the metadata pane to flag the segment as suspected instrument noise.

She stops.

She runs the alignment again with a tighter mask. She narrows the spatial window to ±200 m of the affected segment and re-migrates. The hyperbolae are still there. She checks the source amplitude across the rest of the line. It is consistent. The instrument was not having an episode. It was working normally before the segment and normally after.

She pulls the raw shot gathers. There are three hyperbolae in the raw data. The processing is not adding them.

She checks the navigation logs against the 2024 traverse. The two lines are 4.7 meters apart at this segment. Within the lateral resolution of the system, this is the same line. The bedrock should look the same. The bedrock does not look the same.

She makes the spreadsheet she always makes when she is being cautious. Columns for hypothesis, prediction, supporting evidence, falsifying evidence, status. She types six hypotheses. Instrument drift. Processing artifact. Navigation error. Real englacial change (basal melt, water intrusion, sediment redistribution). Real bedrock change. Real basal-interface change of unknown character. By the time she reaches the bottom of the spreadsheet the first three are crossed out. The fourth is plausible but the geometry is wrong. The fifth is implausible on this timescale. The sixth is the one that fits.

She does not type the sixth is the one that fits. She types insufficient evidence to distinguish 4 from 6; recommend additional acquisition pass.

It is 0814. She has not eaten.

---

In the dining hall she gets the kasha and the cup of tea from the urn and sits near the window. The light through the rime is bluish and even. Pavel Sergeyevich, the Russian glaciologist who runs the basal hydrology side of the project, is at the other end of the table working through a plate of sausage. They have been collegial across four seasons. They often default to Russian because it makes him more relaxed and her more careful and the conversations end up being shorter, which they both prefer.

He looks up. He says, Elin. Ty videla EMR? — Have you looked at the radar?

She says, da, smotrela. Yes, I have looked.

He waits. He is not asking idly. Pavel processes a different slice of the same dataset and the same survey lines. If there were anything obviously wrong with the acquisition he would know.

She says, in Russian, the navigation is good. The fold is correct. I’m finishing the standard comparison. I’ll send you the basal layer when it’s done.

He nods. Khorosho.

He goes back to the sausage. She drinks the tea. The thing she does not say, that she had a brief impulse to say, is that the eastern margin segment looks different from last year. She does not say it because she is not yet sure. She is not yet sure because — she puts this into words for herself only after she has left the dining hall — she does not yet know what sure would consist of. She knows what sure looks like for instrument noise. She knows what sure looks like for a basal melt event. She does not know what sure looks like for the Bell thing has changed shape.

The walk back, the goggles fog. She wipes them with the glove, which makes it worse. She walks the last fifty meters by the feel of the path.

---

In the trailer she opens her own archive.

Her MSc thesis is on the laptop. She wrote it in 2014, at UiT, on subglacial radar signatures in East Antarctic catchments. There is a footnote on page 71 about the east-margin basal reflector. She had noted its orthogonal character. She had suggested it might be a basal accretion artifact of the kind described in Bell et al. 2002, and that further acquisitions would be needed to characterize it.

She had, in the two years after submitting, come to believe that the feature was indeed an artifact. Her advisor had told her so. The argument had been: if it were structural, the geometry would have evolved between the Russian acquisitions of the 1990s and the 2000s, and it has not. The argument had relied on the assumption that the geometry was stable.

She looks at the footnote now. She had been right at twenty-four to be cautious. She had been wrong at twenty-six to be reassured.

She opens the literature folder. There is the Bell ‘95 paper, the Bell 2002 paper on the magnetic anomaly, the Studinger 2003 paper on the bedrock topography, the Filina 2008 gravity inversion, the Popov 2012 basal hydrology. There is also, in a folder she does not label, a PDF of a 2017 internal report from the AARI that had circulated, in the discreet way these things circulate, through three or four polar-radar working groups before settling onto her hard drive. The report has a redacted appendix. The header reads, in Russian, Anomalous returns at the eastern margin, 1998–2014. She has read the header maybe thirty times. She has never been able to do anything with it.

She has also, somewhere on the drive, a scanned page-and-a-half summary from the 2014 anomalous-inclusions report — the cryolab incident that circulated in samizdat for a couple of years before being formally retracted by AARI in 2017 on the grounds of contamination. The contamination explanation was plausible. The retraction was clean. The memo’s circulation had stopped.

She is, at this moment, aware that she is the kind of person who has these documents on her hard drive. The awareness is itself information about her position. She closes the literature folder.

---

She works through the morning. Standard comparison product. Eastern-margin sub-product. Difference map. The difference map is unambiguous. A discrete region, roughly 1.4 km by 0.8 km, where the bedrock-interface return character has changed in a way the rest of the survey has not. The change is localized. The change is geometric. The change is at the scale of the radar’s resolution, which is itself information — whatever did this is at least as coherent as the instrument can see.

She writes the report. She uses the word anomalous, which is the word the protocol uses. She uses the phrase not consistent with documented basal-hydrology mechanisms at this site. She does not use the phrase not consistent with documented geophysical mechanisms. She had written it. She deleted it.

The report has a metadata block at the top with a field called Recommended classification. The pulldown options are: Routine. Anomalous — instrument. Anomalous — environmental. Anomalous — requires re-acquisition. Hold for PI review.

She leaves the field empty. She will come back to it.

---

The data transfer window opens at 1900 station time and closes at 2230. Bandwidth is metered and the system administrator — Konstantin Ivanovich, a quiet man in his fifties who has been at Vostok for nine of the last fifteen years — reviews the outbound packet list. Officially, to manage bandwidth across the seven research programs on station. Unofficially, in a way everyone knows about and no one names. There was an Australian biogeochemist two seasons ago whose outbound packets were delayed for review for eleven days running. She had switched to carrying data out on a personal SSD when she rotated. The SSD did not make it through customs handling at Progress. She received it back six weeks later with the housing intact and the drive wiped. She had not made a fuss. She had been on a postdoctoral fellowship.

Elin is not on a postdoctoral fellowship. She is a permanent staff researcher at NPI with two children, ages nine and six, and a husband whose architectural practice in Tromsø has been quietly losing money for three years. Her salary is the household’s stable line. Her promotion case to senior researcher goes to committee in 2027. The Vostok work is the load-bearing item on the case.

She knows all of this without having to think about it.

---

At 1530 the station director, Dmitri Anatolyevich Volkov, comes by the trailer.

This is not unusual. Volkov circulates daily. He is sixty-one, courteous, fluent in English to the point of using the subjunctive correctly. He served on the Russian side of the IPY oversight committee in 2007. He has, in conversations across four seasons, never said a thing that was not appropriate.

He knocks. He comes in. He stamps the snow off his boots. Elin, how is the season going?

Well, thank you. The acquisition came in clean.

He nods. He looks at the two screens — the difference map on the left, the report draft on the right. He does not appear to be reading them. He is, she has noticed over the years, the kind of man who can stand near a screen and not appear to read it while having read it.

And the comparison with last year?

The basal hydrology in the upstream catchment is consistent. I’m finishing the eastern margin section now.

The eastern margin. He pauses. You know, Elin, I have been at this station a long time. The eastern margin is a difficult section. The radar has many artifacts there. The basal returns near the magnetic anomaly are particularly noisy. We have, over the years, learned to be cautious with what we publish from that region.

Yes. Bell and Studinger flagged the same thing in 2003.

They did. He smiles. The smile is not unkind. I always appreciated those papers. Cautious work. Cautious authors.

He looks at the screen for another second. I will not keep you. The kasha tonight is the good one — with the dried apricots. Don’t be late.

He goes out. He closes the door behind him. She hears his boots on the path for some seconds and then she does not.

Volkov did not threaten her. Volkov did not instruct her. He made an observation about the difficulty of the eastern margin and an aesthetic appreciation of cautious authors. The observation was correct. The appreciation was sincere. Nothing he said could be repeated to anyone outside the station as evidence of anything.

She understands, sitting in the chair, that there is a system in place around what she has seen, and that the system is not made of any one person, and that it predates her and will outlast her, and that no one at the station is going to tell her what it is, because the system does not need to be told. It operates by everyone correctly inferring what it is from the shape it leaves in the air.

She understands, further, that she is now inside it. She has been inside it since 1430, when she ran the migration. She did not enter it by choice. She entered it by looking.

---

She gets up. She makes a cup of tea from the kettle in the trailer. She drinks half of it. She sits back down.

She writes the metadata.

In the Recommended classification field she selects Anomalous — requires re-acquisition. This is the second-mildest of the three anomalous options. It will trigger a recommendation for an additional traverse next season. It will not trigger a Hold for PI review, which would route the report to Yegorov before the data archive sees it, which would mean Yegorov reads it, which would mean Yegorov decides what happens next. The re-acquisition recommendation goes directly to the AWI traverse planning committee and is processed as a logistical matter.

In the Comments field she writes, in English: Eastern margin segment shows altered scatterer geometry relative to 2024 acquisition. Cause not determined. Possible candidates include localized basal hydrology change; instrument-related processing artifact not yet excluded; environmental factors. Additional acquisition recommended at standard fold with overlap of ±200 m to constrain.

She does not write the geometry is structural. She does not write the change is at the spatial scale of the source. She does not write the western edge of the Bell footprint has acquired three vertically-stacked diffraction hyperbolae at seventy-meter spacing.

She attaches the difference map. She does not attach the deconvolved shot gathers.

She copies the deconvolved shot gathers, the raw radargrams for the affected segment, and a snapshot of her hypothesis spreadsheet to the encrypted volume on the personal SSD in the bottom drawer of her desk. The SSD is the one she is not officially supposed to have. It is a 2-terabyte drive in a metal housing that she bought in Tromsø three years ago and has carried out and back through Progress six times. It has never been inspected. She does not know whether this is because it has never been noticed or because it has been noticed and judged to contain nothing actionable. She has thought about this question and does not have an answer for it.

She queues the report for the 1900 transfer.

At 1847 she walks to the dining hall. The kasha tonight is the one with the dried apricots. Volkov is at the head of the table. He nods to her as she sits down. She nods back.

After dinner she walks back to the trailer. The wind has dropped. The sky is the color of nothing — not light, not dark, just the white-blue absence the high plateau gives at this latitude in January. She stops on the path for a second because she has the brief sense that the station is the wrong shape — that the dome of the main building, viewed from this angle, is not the geometry she has walked under four hundred times before. The sense passes. She finds her footing. She walks on.

At 2147 the system administrator’s review log shows her packet cleared at 2031. The packet went out clean. The report is in the AWI queue. The re-acquisition request will be filed on Tuesday.

She closes the laptop. She brushes her teeth. She gets into the bunk.

In the metadata field where she had written Anomalous — requires re-acquisition, the pulldown now reads, in the archive’s confirmation email, Anomalous — requires re-acquisition.