Air Defence Log
Russian drone and missile strikes on Ukraine, logged wave by wave from 28 September 2022 to the present from Ukrainian Air Force daily reports. More than 145,000 aerial weapons have been launched — overwhelmingly Shahed/Geran-class loitering munitions. Roughly 87% have been neutralised: shot down, or electronically suppressed so that they failed to reach their target.
Read this before quoting a rate
Ukrainian Air Force reports carry two separate loss columns: destroyed (kinetic kills) and not_reach_goal (drones downed or diverted by electronic warfare and never reaching target). A weapon counted in either column did not strike. The defensible measure is therefore:
neutralisation rate = (destroyed + not_reach_goal) ÷ launched
Dividing destroyed by launched alone understates Ukrainian performance by up to 33 points in some months, and produces artefacts that look like findings. Between July 2024 and July 2025 the Air Force reported EW suppression as a separate line; before and after that window those kills are folded into destroyed. A kinetic-only rate therefore shows a collapse and recovery across that boundary which did not happen. The chart below plots both bases so the reporting change is visible rather than hidden. Full detail on the methodology page.
Live data — campaign totals
—Weapons launched
—Neutralised
—Neutralisation rate
—Reached target
—Waves logged
—Largest single night
—
NEUTRALISATION RATE OVER TIME — WHY THE DENOMINATOR MATTERS
The shaded band is the window in which EW suppression was reported on its own line. The dashed series is the kinetic-only rate — the number you get from destroyed ÷ launched. Its dip and recovery are a reporting change, not a change in Ukrainian performance.
NEUTRALISATION RATE BY WEAPON CLASS
Both series on the corrected basis. Loitering munitions are neutralised at a far higher rate than ballistic and cruise missiles, and the gap is the central fact of the campaign: Russia substituted cheap mass for scarce precision. Months in which a class saw fewer than 15 launches are left blank rather than plotted, since a handful of weapons produces a rate that looks like signal and is not.
Does mass saturate the defence?
Attack nights grouped by size. The third column is the corrected rate; the second is what a kinetic-only calculation returns. The fourth shows how much of each band falls inside the EW-reporting window — which is what drives the apparent dip in the middle bands.
No saturation threshold is visible in the corrected data. Neutralisation rises with attack size, not falls. Within each reporting period taken separately the relationship is monotonic or flat, and the largest raids are neutralised at the highest rate of all. An earlier version of this page reported a "saturation threshold" at 61–150 weapons per night; that band is roughly half-composed of nights inside the EW-reporting window — double any other band — and the finding was an artefact of the denominator. It has been withdrawn.
Key findings — original analysis
Mass does not overwhelm the defence
The largest raids are neutralised at the highest rate: 92.5% for nights of 600+ weapons, against 83.6% for nights of 20 or fewer. The relationship holds inside both reporting periods separately, so it is not an artefact of when large raids happened. The common assumption that swarm attacks saturate Ukrainian air defence is not supported. What large raids do achieve is absolute leakage — 7.5% of 1,449 weapons is still over a hundred impacts in a night.
Shahed-as-decoy hypothesis not supported
Mixed salvos combining loitering munitions with missiles (n=628 nights) are neutralised at 86.6%, against 88.5% for drone-only nights (n=450). The 1.9-point difference is within error margins. Mixed nights are larger on average (139 weapons against 125), so what little difference exists tracks volume rather than any decoy function. Missile-only nights, by contrast, are the hardest of all to defend at 74.4% (n=83), holding at that level both inside and outside the EW-reporting window. This conclusion survived recalculation on the corrected basis unchanged.
Weapon mix: loitering munitions are now almost the entire threat
Shahed-class launches grew from 35% of the total in late 2022 to 93% in 2025. The 2026 labelled share reads lower only because a growing number of drone rows are logged as "Unknown UAV"; counting all drone rows, loitering munitions are roughly 96% of launches. Monthly volume peaked at 10,098 in May 2026, around 20× the 2023 baseline. Russia has substituted cheap mass for expensive precision as its primary strategic instrument.
Withdrawn — weather effect on interception rates
This page previously reported that night-time weather has no statistically significant effect on interception rates (r=0.051, p=0.21, n=603 Shahed attack days, Open-Meteo ERA5). That null result is not currently established. It was computed on the kinetic-only basis, and a third of Shahed attack nights fall inside the EW-reporting window, across whose boundary the kinetic-only rate steps by 22.3 points for reasons that have nothing to do with weather. Adding a large weather-independent variance component to the dependent variable (per-night dispersion of 17.9 points against 12.6 on the corrected basis) attenuates any correlation toward zero. A near-zero r is what this test would return whether or not a weather effect exists. The analysis will be re-run on the corrected basis; until then no claim is made in either direction.
Methodology & data source
Underlying collection by Petro Ivaniuk (Kaggle: piterfm/massive-missile-attacks-on-ukraine), compiled from Ukrainian Air Force official daily reports; independently verified by CSIS as matching official figures. Please credit that dataset directly in any downstream use, not only this tracker. Coverage from 28 September 2022 to present, refreshed daily via an automated GitHub Actions export. Rate methodology, confidence tiers and known artefacts are documented on the methodology page. Original analysis by Andro Mathewson, PhD Candidate in War Studies, King's College London. Weather data: Open-Meteo Historical Archive API.
Licensed CC BY-NC 4.0. Bulk extracts of the underlying wave-level data are available on request for non-commercial research — get in touch. Commercial use, including the training or enrichment of counter-UAS systems, requires separate written permission.