# Is It a Bird?

> Set a radar's bar low and birds flood the screen; set it high and drones slip through. Layered sensing is the way out.

- Published: 25 September 2026
- Author: Ghulam Mujtaba (https://drosonic.com/about.html)
- Publisher: DroSonic OSINT (https://osint.drosonic.com/), the research arm of DroSonic (https://drosonic.com/)
- Series: Radar vs Small Drones, part 5
- Confidence: moderate
- Web page: https://osint.drosonic.com/assessments/radar-false-alarms-birds-drones.html

## Key questions

**Why do birds cause false alarms on drone radar?**
Birds return about the same radar echo as small drones (around 0.01 m²) and fly at the same heights. A threshold low enough to catch drones also catches birds; Sandia National Laboratories describes bird false alarms as not yet adequately dealt with.

**What is layered drone detection?**
Combining sensors that fail in different ways: radar for range and tracking, RF detection for control links, cameras for confirmation, and acoustic sensors for drones that are small, low, hovering or radio-silent.

## Key numbers

- **1 flock** can hold hundreds of drone-sized radar echoes (each bird about 0.01 m²)
- **> 90%** drone-vs-bird accuracy from track behaviour (machine learning on radar tracks (SPIE 2023))
- **≈ 20%** of targets entering Ukraine tracked by acoustics alone (per United24)

## Bottom line

**A radar never "sees a drone". It sees echoes, and has to decide which ones are worth reporting.** Every echo is compared with a threshold. Set it low and the radar catches faint, distant drones, but also every bird, rain cell and patch of ground clutter. Set it high and the screen is clean, but bird-sized drones fall below the line. Adaptive thresholds, track filtering and machine-learning classifiers move that trade-off but cannot remove it. Against targets as small and low as consumer drones, among thousands of birds of the same echo size, there is no single setting that works. The practical answer is layering: radar combined with RF, optical and acoustic sensors that each catch what the others miss.

## Five problems, one conclusion

| Part | The problem for radar | What another sensor adds |
|---|---|---|
| 1 — The fourth-root problem | Range collapses as targets shrink | Dense, cheap sensors close to the threat |
| 2 — Radar cross section | Plastic, bird-sized echoes | Sensing that doesn't rely on reflection: sound, RF |
| 3 — Under the horizon | Low drones hide behind Earth, terrain and clutter | Distributed sensors that don't need line of sight from afar |
| 4 — Doppler and micro-Doppler | Hovering drones vanish into clutter filters | Sensors that hear rotors whether the drone moves or not |
| 5 — Is it a bird? | Thresholds trade missed drones for false alarms | A second, independent signature to confirm or reject |
Legend: Layering does not make radar obsolete: it makes each radar detection worth more

## Where do you draw the line?

A radar receiver always hears something: its own electrical noise, plus echoes from the ground, weather and wildlife. Detection means deciding that a particular echo is strong enough to be a real target. That threshold sets the trade-off. Set it too low and "more real targets will be detected, but at the expense of increased numbers of false alarms." Set it high and you get the opposite. ([Wikipedia](https://en.wikipedia.org/wiki/Constant_false_alarm_rate))

Because the background changes from place to place, over hills, in rain or near a town, a fixed threshold produces too many alarms in some areas and misses targets in others. Most modern radars therefore use a **constant false alarm rate (CFAR)** detector. It estimates the local background from the cells around each one under test and sets the threshold just above it. Guard cells next to the test cell are left out, so a target's own energy does not raise its own bar. ([Wikipedia](https://en.wikipedia.org/wiki/Constant_false_alarm_rate))

CFAR keeps the alarm rate steady, but it does so by raising the bar wherever the background is busy. A small drone flying through clutter, exactly where Part 3 says it will be, faces the highest bar of all.

![Schematic of radar echo strength along one beam: a noisy background with a hump of clutter, a drone and a bird producing similar spikes, a weak distant drone, a dashed fixed threshold, and an adaptive threshold that rises over the clutter, leaving the drone inside it undetected](https://osint.drosonic.com/assets/img/assessments/radar-false-alarms-birds-drones/threshold-trade-off.png)

Figure: **Where to draw the line — *fixed versus adaptive thresholds*** Schematic, not measured data. A fixed line fires on clutter and treats the bird and drone alike. An adaptive line quietens the clutter but raises the bar over it, hiding a drone inside. Credit: Diagram: DroSonic OSINT · original · threshold behaviour from Wikipedia "Constant false alarm rate"

## The false alarm that flies

Clutter can be mapped and filtered because it mostly stays put. Birds don't. They move, they fly at drone heights, and each one returns about as much energy as a small drone (Part 2). A single flock can put hundreds of drone-sized echoes in front of a radar at once.

That is a known, under-solved problem. A Sandia National Laboratories publication notes that radar can detect drones at more than 1 km, but that nuisance and false alarms from natural sources, birds in particular, have not been adequately dealt with. ([Sandia National Laboratories](https://www.sandia.gov/research/publications/details/polarized-radar-for-detection-and-automatic-non-visual-assessment-of-unmann-2019-10-01/)) Research presented at SPIE in 2023 describes counter-drone radars being swamped by bird tracks. It reports that machine learning on track behaviour (how a target moves over time) can recognise target drones with better than 90% sensitivity and specificity. ([SPIE](https://www.spiedigitallibrary.org/conference-proceedings-of-spie/12535/125350X/UAV-discrimination-from-birds-using-radar-track-information/10.1117/12.2666281.full))

Better than 90% is impressive in a trial, but a sensor that misjudges one target in ten will still miss drones and raise false alarms when hundreds of birds pass by. Classification helps, and it works best with a second, independent piece of evidence.

> Every bird is a test the radar has to pass. *Over a flock, thousands of tests a minute.*

## Layering sensors that fail differently

Each sensor family fails in a different way, which is exactly why they work together:

| Sensor | Finds drones by | Blind to |
|---|---|---|
| Radar — reflection | Its echo, at kilometres, in any light | Low or hovering drones in clutter; telling drones from birds |
| RF detection — emissions | The control and video links | Fibre-optic and autonomous drones that do not transmit |
| Electro-optical / infrared — sight | A picture an operator can confirm | Bad weather, darkness without thermal, narrow field of view |
| Acoustic — sound | Propellers, motors and engines | Long range, loud environments, strong wind |
Legend: Our assessment · each row's blind spots are covered by at least one other row

The U.S. military's own counter-drone vehicles are built this way. The M-LIDS system pairs a Ku-band radar with a 30 mm gun and Coyote interceptor drones on one vehicle. ([Wikimedia Commons](https://commons.wikimedia.org/wiki/File:6-56_ADA_validates_counter-UAS_training_during_UAV_demonstration_(1).jpg)) The U.S. Defense Innovation Unit's 2025 low-cost counter-drone sensing challenge named acoustic, passive RF, radar, optical, infrared and multi-sensor technologies together, with a target of cutting ownership costs by 50–80%. ([The Defense Post](https://thedefensepost.com/2025/07/23/us-low-cost-counter-drone/))

Ukraine's air defence shows the payoff at national scale. Its passive acoustic networks are fused with radar data. According to United24, about **20%** of targets entering Ukrainian airspace are tracked by acoustic data alone, which is coverage radar was not providing. ([United24 Media](https://united24media.com/war-in-ukraine/sky-fortress-ukraines-acoustic-detection-system-that-tracks-drones-cheap-and-fast-9451))

![A U.S. Army M-LIDS counter-drone vehicle with its radar and sensor mast during a live-fire exercise in Romania, with soldiers standing beside it](https://osint.drosonic.com/assets/img/assessments/radar-false-alarms-birds-drones/m-lids-romania.jpg)

Figure: **M-LIDS in Romania, 2024 — *radar, optics and weapons on one vehicle*** The U.S. Army's counter-drone vehicle during a live-fire exercise at Capu Midia. Credit: Photo: Lt. Anca Medrea · Public domain · [Wikimedia Commons](https://commons.wikimedia.org/wiki/File:M-LIDS_Live_fire_exercise_of_the_U.S._Army%E2%80%99s_newest_counter-drone_system_in_Capu_Midia,_Romania,_April_10,_2024.jpg)

![A tan U.S. Air Force M-LIDS vehicle on a raised dirt firing range under a pale sky](https://osint.drosonic.com/assets/img/assessments/radar-false-alarms-birds-drones/m-lids-range.jpg)

Figure: **Counter-drone training, 2025 — *several sensors, one decision*** A U.S. Air Force M-LIDS on a live-fire range during counter-small-drone training in the CENTCOM area. Credit: Photo: U.S. Air Force, Master Sgt. Luke Olson · Public domain · [Wikimedia Commons](https://commons.wikimedia.org/wiki/File:M-LIDS_system_at_firing_range_for_counter-UAV_exercise.jpg)

## What five weeks of radar physics add up to

Across the five parts, the same conclusion keeps returning:

1. **Small targets shrink radar range** with the fourth root of echo size (Part 1).
2. **Small drones are bird-sized echoes**, and plastic or foam makes them smaller still (Part 2).
3. **Flying low hides them** behind the Earth, terrain and ground clutter (Part 3).
4. **Hovering can defeat Doppler filtering**, leaving only a faint rotor signature (Part 4).
5. **Thresholds cannot separate them from birds** without another source of evidence (Part 5).

None of this makes radar obsolete. It remains the best sensor for range, speed and all-weather tracking, and nothing else gives the same air picture. But against small, cheap, low drones, radar is one layer of a stack, not the whole defence.

### Confidence

#### How sure we are, and what would change it

**High confidence** in the threshold trade-off and CFAR behaviour, which are standard radar theory. **High confidence** that birds are a major false-alarm source for counter-drone radar, from national-laboratory and conference research.

**Moderate confidence** in the layered-sensing comparison, which is our assessment. Real performance depends on specific products, sites and integration. The United24 acoustic figure is reported, not independently verified.

## The layer that listens

Five weeks of radar physics end in one place. The targets that matter most now, small, cheap, low drones, are the ones radar handles worst, and a growing share of them emit no radio signal for RF detectors to find. A defence built only on radar is short of range, short of height and buried in birds. A defence that adds independent sensors gets a second opinion on every echo.

Acoustic sensing is the natural partner. It is cheap enough to deploy densely, passive so it reveals nothing, unaffected by a target's echo size or material, and sensitive to exactly the rotors and engines that radar strains to classify. A bird's wings and a drone's propellers make very different sounds. At [DroSonic](https://drosonic.com/) we build [passive acoustic drone detection](https://drosonic.com/#technology) to be that layer. If this series was useful, the reports [Listening to the Sky](https://osint.drosonic.com/assessments/listening-to-the-sky.html) and [When the Detector Stays Silent](https://osint.drosonic.com/assessments/fibre-optic-drones-acoustic-detection.html) show acoustic detection at work in Ukraine today.