GNSS Interference over Eastern Latvia, Revisited — 84% of Aircraft Flagging Degraded Navigation North-East of Riga over the Latest Week, 66% over Riga's Own Cell, Fifty-Three Days Without Change, September–October 2026

Why this brief exists — and why it is not the story the trigger told. Our detection layer raises a GNSS candidate when a cell near a large airport crosses 60% of aircraft flagged on the latest day, on a sample of at least 30 tracked aircraft, and the same cell was quiet a week earlier. The 57°N 25°E cell crossed the first bar on 21 September at 73.5%. It cleared the sample bar because 21 September happened to be its busiest day since 22 August — 34 aircraft where the September norm is around 22; the two days after it logged 14 apiece, then 18, 19, 20, 22, 27, 21, 27, 27 and 19 on 24 September–2 October. And it passed the “quiet a week ago” test on a technicality: the comparison day, 14 September, holds no data for any cell in the feed. Read against the actual series, that third condition fails outright — the cell has never posted a day below 60.9% since observation began on 10 August, and its September average is higher than the day that tripped the trigger. We already know this cell: it was the subject of our August brief, which closed on 19 August after eight days in a 69–85% band with no escalation. We are publishing again, because six further weeks of data are worth having on record and because the measured picture has a detail the August brief could sketch but not settle: the airport's own cell is now as heavily flagged as the trigger cell, on a far larger sample. But we label it as what it is. There is no escalation here to report, and a reader who arrives expecting news of a sudden jamming surge over Riga should leave with the opposite understanding.

Two hot flanks and a quiet corridor — the shape of the field around Riga. Written 4 October. Interference has a footprint, and around Riga it has a recognisable shape. Along the 57°N row on 2 October the flag rate ran 34.0% at 21°E over the Baltic off Courland, 22.4% at 22°E and 36.9% at 23°E over the Gulf of Riga and its northern shore, 45.0% at 24°E, 84.2% at 25°E over Vidzeme (16 of 19), then 100% at 26°E (18 of 18) toward Alūksne and the Russian border; the 26 September–2 October aggregates say the same with less noise — 36.6%, 27.4%, 36.7%, 48.6%, 84.0%, 100.0%. One row south, the 56°N row that holds the airport read 61.0% at 21°E, 57.4% at 22°E, 65.2% over Riga at 23°E (92 of 141), 69.4% at 24°E, 73.7% at 25°E and 93.3% at 26°E (15 aircraft) on 2 October — a few points below its own weekly figures across most of the row — and those weekly figures are 58–66% across western and central Latvia, 77.5% at 25°E and 94.9% at the eastern edge. South again, the 55°N row over northern Lithuania is saturated — 87.5%, 95.1%, 96.6% (119 aircraft over the Šiauliai area) and 87.0% across 21–24°E on 2 October, 82.0–93.2% over those cells for the week — and to the north the 58°N row over southern Estonia repeats the east–west slope, 41–55% at 21–23°E rising to 74.5% at 24°E, 89.6% at 25°E and 91.5% at 26°E. So the quiet corridor is the sea and the northern shore of the Gulf of Riga at 57°N; the hot flanks are the east toward Russia and the south toward Kaliningrad; and the airport sits on the hotter side of the line. In the August-to-September comparison almost nothing moved: the Vidzeme cell 84.0% to 83.4%, Riga's cell 73.9% to 71.9%, the cell east of the airport at 24°E 62.5% to 64.7%, the 57°N 24°E cell 53.3% to 54.5%, the Gulf cells at 57°N 22–23°E 40.2% to 35.1% and 40.0% to 37.5%. The one clear change is at the eastern edge, where 56°N 26°E rose from 81.8% to 92.8% — on 404 September aircraft, a firming of the saturated flank, not a spread toward the city.

FIR-average hides the flanks — why the cell-level number matters. Written 4 October. By the FIR-wide measure, the whole Latvia FIR (EVRR) sits at 8.4% over 27 September–3 October against a 10.6% 90-day average, reading 9.6% on 30 September, 7.5% on 1 October, 8.4% on 2 October and 7.9% on 3 October on roughly 1,780–1,850 tracked aircraft a day (FIR-level, medium confidence). This feed runs a day ahead of the per-cell one, which is why its week ends on 3 October. It has held a 7.3–9.7% band over the 14 days to 3 October — narrower than the 7.2–11.3% we quoted last update, because 18 September's 11.3% has rolled out of the window rather than because anything moved — and its monthly averages, 11.6% in July, 10.6% in August and 10.0% across September, describe a FIR that is, if anything, drifting very slightly quieter. The reason the FIR figure is an order of magnitude below the cells is that it pools the saturated eastern and southern flanks with the quiet sea corridor and, above all, with high-altitude transit traffic crossing the FIR on the Scandinavia–Central Europe and Baltic trunk routes, most of which reports healthy integrity. Zoom to Riga's own cell and the degraded share is nearly eight times the FIR figure. Both numbers are true; the cell-level one alone tells a dispatcher what the approaches to Riga look like today. The neighbouring FIRs on the same 27 September–3 October window: Tallinn (EETT) ~11.4% against a ~10.7% 90-day average (medium confidence), Vilnius (EYVL) ~20.8% against ~21.0% (medium), St Petersburg (ULLL) ~14.2% against ~12.0% (high), Helsinki (EFIN) ~9.0% against ~8.7% (high), and Kaliningrad (UMKK) ~27.3% against ~29.0% — on about 78 tracked aircraft a day and low confidence, so indicative rather than firm. One FIR-wide figure hides the flanks; a single cell alone overstates the FIR. We label which scale each number belongs to for exactly that reason.

Interference degrades navigation; it does not ground flights. Written 4 October. The figure counts aircraft broadcasting downgraded navigation integrity in their ADS-B position-quality fields — the consistent fingerprint of GNSS interference. It does not mean those aircraft are lost or off course. Airliners cross-check satellite positioning against inertial reference systems and ground-based aids — ILS, VOR/DME — and fall back to them when GNSS degrades, with published contingency procedures for GNSS loss. The real cost is precision and workload: satellite-based (RNP/RNAV) approaches can become unavailable, approach minima rise, and in poor visibility that can mean holding or a diversion to an alternate. That is exactly why Riga's departure board shows nothing: the traffic section above reads an ordinary early-autumn profile — a September mean of about 177 observed movements a day, and the window's two busiest days last, 200 on Thursday 1 October and 204 on Friday 2 October — while two in three aircraft over the airport's own cell flagged degraded GNSS across 26 September–2 October. For Riga's arrivals the relevant figure is that 66.3% weekly aggregate over the airport's own cell, not the thin trigger cell to the north-east; for traffic routing east toward the Russian border it is the 95–100% at 26°E and beyond. Weather matters for the practical outcome: Riga's 30-day record shows instrument conditions on about 14% of observations, so on most days a satellite-based approach becoming unavailable costs little; on the minority of low-visibility days it is when a diversion to Kaunas, Vilnius or Tallinn becomes plausible — and each of those sits under a flagged cell of its own (85%, 64% and 72% on 2 October).

Honest arithmetic on a thin cell, a large cell and a six-week series. Written 4 October. The Vidzeme cell sees between 14 and 38 tracked aircraft a day in our feed — thin enough that a single day can move by ten or fifteen points on the mix of traffic alone, which is why the 100% readings of 8 September (20 of 20) and 24 September (18 of 18), and the 60.9% of 19 September (14 of 23), are reported as sampling extremes and not as levels, and why we describe the state on weekly aggregates of 127–214 aircraft and monthly aggregates of 574 and 650. A cell that can reach 100% on eighteen aircraft can also fall twenty points the next day on the same physics — and did, to 78.9% on nineteen aircraft on 25 September, then 75.0% on twenty aircraft on 26 September before reading 90.9% on twenty-two on 27 September, 88.9% on twenty-seven on 28 September, 76.2% on twenty-one on 29 September, 88.9% on twenty-seven on 30 September, 81.5% on twenty-seven on 1 October and 84.2% on nineteen on 2 October; the 84.0% for 26 September–2 October is the number we stand behind. Riga's own cell is a different instrument: 117–147 aircraft a day, roughly 790–965 a week, so its weekly figures — 68.6%, 75.9%, 77.3%, 73.6%, 72.7%, 76.4% (six observed days) and 70.0% over the calendar weeks to 27 September, 66.3% over the latest seven days to 2 October and 65.2% over 24–30 September, still the softest seven-day window since observation began — carry real weight, and its 84.5% on 21 September is reported as a single day near the top of the range (the series high is 84.9%, on 22 August), not as a step — and the 58.3% of 24 September, 59.9% of 28 September and 57.5% (73 of 127) of 30 September are reported the same way, the three lowest September days and close to the 54.5% series low of 10 August; the entire 56°N row read low on each of those days, and each time the cell came back within a day or three — to 76.4% (97 of 127) by 27 September, 67.9% (93 of 137) on 29 September, 70.9% (95 of 134) on 1 October and 65.2% (92 of 141) on 2 October. We will not pretend the clustering is nothing: three of the four sub-60% days since 10 August fell inside one week, and they pull the latest seven-day figures below every calendar week before them — the five softest seven-day windows of the series now all end between 28 September and 2 October, at 65.2–67.0% against 68.6% for the next lowest, though consecutive seven-day windows share six of their seven days, so that is one soft fortnight counted five times rather than five independent facts. Two days of October have since read 70.9% and 65.2%, neither of them a fourth sub-60% day. We report all of that as measured and will call it a trend only if the calendar week of 28 September closes there — it stands at 64.3% (438 of 681) after five days, which would be the lowest calendar week of the series. A day with no data at all, like 14 September, is left out rather than filled in — and, as explained above, that same blank is part of what tripped the trigger. The series begins on 10 August, so it is 53 observed days deep and we make no claim about how long these cells have looked like this beyond that; the wider southern and eastern Baltic belt has carried elevated interference for well over a year (background: Baltic GPS-jamming escalation, how GPS jamming works and our GNSS explainer), and its other ends are covered in separate active briefs — south-eastern Finland, Gdańsk and east of Warsaw. The same discipline governs the traffic section, where 14 September is blank for the same reason, where 19 and 21 September are shown at their backfilled values rather than the partial readings first received, and where 3 October is withheld because the movement feed has not published it yet. Every figure here is an aggregate of crowdsourced receiver telemetry: a floor, not a total.

What would change the picture. Written 4 October; each of these tests is answered in the close-out block below. Three measured things are watched continuously by our automated multi-signal analysis. First, the quiet corridor: the 57°N 22–23°E cells over the Gulf of Riga (22.4% and 36.9% on 2 October, 27.4% and 36.7% across 26 September–2 October against 29.4% and 35.5% the week before — flat, with the 22°E cell marginally quieter) are the tell-tales — a step in those cells toward the 60–70% the 56°N row shows would mean the field had closed over the northern approaches and the sea routes to Stockholm and Helsinki rather than stopping at the shore. Second, the FIR-level EVRR ratio breaking decisively above the 7.3–9.7% band it has held over the 14 days to 3 October rather than oscillating inside it. Third, a shift from jamming (signal loss) toward spoofing (a coherent false position), or interference reaching approach paths at RIX in low-visibility conditions. A fourth signal — regulatory action, meaning NOTAMs suspending satellite-based approach procedures at Latvian airports — would be the clearest operational marker, and we flag its absence honestly: our NOTAM feed currently carries no items for the Latvia, Tallinn or Vilnius FIRs at all, so its silence here is a gap in our coverage, not evidence that nothing has been issued. No traffic-based airspace closure has registered for those three FIRs in our detection layer since the start of August. The picture can also change downward, and we would say so as plainly: the airport cell's 65.2% over 24–30 September is still its softest seven-day figure since observation began, 26 September–2 October follows at 66.3%, and three of its four sub-60% days since 10 August fell between 24 and 30 September — but each of those was a low day for the whole 56°N row, each reversed within days, and the two October days so far read 70.9% and 65.2%, inside the band. If the calendar week of 28 September, at 64.3% on its first five days, closes at that level, it would be the lowest calendar week of the series and a measured easing over Riga rather than sampling noise; that week closes on 4 October and we will say which way it went in the next update. This page updates from the same monitoring cycle as the RIX brief and the live regional GPS-interference map; if the Vidzeme or Riga cells drop below the bands they have held, or the Gulf corridor climbs into them, the page will say so.

What the data showed — close-out, 13 October. This brief opened on 23 September to put on record that an automated re-raise of the 57°N 25°E cell was a steady state, not a new zone, and to carry the series forward. It closes on 13 October with no recovery to report: the interference did not end, it held — and we close it the way its August predecessor was closed, as a documented steady state. The trigger cell. From 3 to 12 October the Vidzeme cell read 81.0%, 77.3%, 87.0%, 87.5%, 90.0%, 70.0%, 84.6%, 88.6%, 80.0% and 73.3% (22 of 30) on samples of 16 to 35 aircraft — 10 October's 31 of 35 being the second-largest sample of the series after 22 August. The calendar week of 28 September–4 October closed at 82.9% (136 of 164) and the week of 5–11 October at 84.4% (152 of 180); 1–12 October aggregates to 82.3% (246 of 299) against 83.4% across September and 84.0% across 10–31 August. Sixty-three observed days, every one of them between 60.9% and 100%; the two softest October days, 70.0% (14 of 20) on 8 October and 73.3% on 12 October, are the third- and sixth-lowest of the series, both inside the band and both on the thin sample this cell always carries. There is no inflection in it, in either direction. Riga's own cell, and the test this page set. The calendar week of 28 September–4 October closed at 66.6% (621 of 932) — the lowest calendar week of the series, below the 68.6% of 10–16 August — so by this page's own test it was a measured easing over Riga rather than sampling noise; the weekend pulled it up from the 64.3% it stood at after five days, and the week of 5–11 October came back to 69.2% (651 of 941), inside the range of every calendar week before it. The fourth sub-60% day did arrive: 56.7% (72 of 127) on 6 October, the second-lowest day of the series after 10 August's 54.5%, with the cell east of the airport at 24°E down to 43.5% the same day and the row back at 66–77% the next. Four days later the cell read 84.0% (105 of 125) on 10 October, its third-highest day — on the same day the Latvia FIR posted its one high reading and the Gulf of Riga cells their highest October readings — and 65.7% the day after. Across 1–12 October the airport cell aggregates to 69.5% (1,120 of 1,611), 68.0% over the final seven days; its monthly figures step down 73.9% (10–31 August), 71.9% (September), 69.5% (twelve days of October) — about two points a month. We report that as measured and do not call it a trend on twelve days; the softest rolling week of the series is still 24–30 September at 65.2%. The shape of the field. Over 6–12 October the 57°N row read 33.2%, 26.4%, 32.2% and 51.6% from 21°E to 24°E over the Baltic, the Gulf of Riga and its northern shore, then 82.4% at 25°E and 97.5% at 26°E; the 56°N row through the airport 63.7%, 67.2%, 68.0%, 59.7%, 77.2%, 87.7%; northern Lithuania on 55°N 85–97% across 21–24°E; southern Estonia on 58°N 53–55% at 21–23°E rising to 79.6%, 93.7% and 97.4%. Two hot flanks, a quiet corridor, the airport on the hotter side of the line — unchanged from 2 October, and from August. What did move, slightly and broadly, is the level west of 25°E: across 1–12 October every cell on the 56°N and 57°N rows from 21°E to 24°E reads below its September figure, by between one and nine points (Riga's cell 69.5% against 71.9%; the Gulf cells 26.0% and 33.3% against 35.1% and 37.5%), while the eastern edge did not move — 57°N 26°E 98.0% against 98.2%, 56°N 26°E 90.3% against 92.8% on 207 aircraft. Twelve days, a few points, reported and not called. The tests this page set, answered. The quiet corridor at 57°N 22–23°E stepping toward the 60–70% of the row south of it: it did not — 25.5% and 27.5% at 22°E for the calendar weeks of 28 September and 5 October, 35.8% and 33.3% at 23°E; its one high day, 44.1% and 41.2% on 10 October, reversed to 23.5% and 32.1% the next. The Latvia FIR (EVRR) breaking decisively above its 7.3–9.7% band: one day — 10.4% on 10 October, the highest since 18 September — followed by 7.3% and then 6.7% on 12 October, the lowest daily reading in this feed since at least the start of July; the seven days to 12 October average 8.1% against a 10.4% 90-day average, the first twelve days of October 8.1% after 11.6% in July, 10.6% in August and 10.0% in September, and the lowest seven-day windows since 1 September are the ones ending 9 October (8.10%) and 8 and 12 October (8.13%) — FIR-level, medium confidence, about 1,740 tracked aircraft a day. At FIR scale, then, Latvia drifted quieter, not louder. The neighbours on the same 6–12 October window: Tallinn (EETT) 11.2% against a 10.7% 90-day average, Vilnius (EYVL) 20.6% against 20.9%, St Petersburg (ULLL) 11.9% against 12.2% (high confidence), Helsinki (EFIN) 8.9% against 8.5% (high), Kaliningrad (UMKK) 27.3% against 29.0% on about 83 aircraft a day at low confidence on six of the seven days; the one-day spike of 8 October — Tallinn 15.4%, Helsinki 16.4% — did not reach Latvia, which read 7.3% that day. A shift from jamming toward spoofing, or interference on the approach paths in low visibility: the navigation-integrity measure on this page cannot separate the two, and nothing in our data identifies either — not fired, and not excluded by us. Regulatory action: our NOTAM feed still carries no items at all for the Latvia, Tallinn or Vilnius FIRs, so its silence remains a gap in our coverage and not evidence; no traffic-based airspace closure registered for those FIRs or for Kaliningrad in our detection layer between 1 August and close-out. Traffic. Riga (RIX) flew 170, 188, 185, 169, 166, 178, 199, 161, 183 and 182 observed movements on 3–12 October — about 182 a day across 1–12 October, its busiest stretch of the window, against 177 across September and 182 across 10–31 August; the window's three busiest days all fell in October (204, 200, 199), and its weekday profile is ordinary to the end. Tallinn (TLL) read 72.6% flagged over its own cell on 12 October with 124 observed movements, inside its 85–139 window band; Vilnius (VNO) 48.8% with 141; Kaunas (KUN) 85.5% with 28. No attribution of any traffic figure to the interference is made, and none is warranted by the pairings — the quietest October day at Riga fell on the airport cell's highest flag day. Why close now. The purpose of this brief was to show that the trigger had re-raised a steady state, and three weeks of further data say exactly that, as eight days did in August. A brief labelled active that has nothing to add is not an honest page, so this one ends. The Baltic interference field persists — it has for well over a year — and these cells remain tracked on the RIX brief and the live regional GPS-interference map; should the Vidzeme or Riga cells leave the bands they have held, or the Gulf corridor climb into them, that will be a new brief and not a reopening of this one. All cell-level figures are FlySafe aggregates of ADS-B Navigation Integrity Category reports from crowdsourced receiver networks — a floor, not a total; FIR-level ratios come from the same class of public data and are labelled as FIR-level wherever they appear; airport movement counts are observed ADS-B departures and arrivals, also a floor.