Search and rescue is one of the governmental functions Congress enumerated at 49 U.S.C. § 40125(a)(2). That line has convinced a lot of volunteer SAR teams they are public aircraft operators. Most are not. The enumerated function is necessary, not sufficient, and getting it wrong is the most common way a SAR drone program ends up flying without the authority it believes it has.
This guide covers what a SAR drone crew has to get right, in the order it bites: the authority you hold, the search theory that decides whether your flight hours produce a find, and the sensor and terrain realities behind both.
Search and rescue is a governmental function. That does not make your team a public aircraft operator.
Public aircraft status requires passing two tests on every flight. The entity and aircraft test comes from 49 U.S.C. § 40102(a)(41)(C): the aircraft must be "owned and operated by the government of a State... or a political subdivision." Subparagraph (D) extends this to aircraft "exclusively leased for at least 90 continuous days." The function test at § 40125(a)(2) lists search and rescue alongside firefighting, law enforcement, national defense, intelligence, aeronautical research, and resource management.
Both must be true. A 501(c)(3) SAR corporation, a volunteer fire company, or a mounted posse performs a governmental function on every callout and still fails the first test. AC 00-1.1B, Public Aircraft Operations — Manned and Unmanned is blunt: the status "is not automatic," must be determined on a flight-by-flight basis, and political subdivision status is "determined by the Attorney General of the State." That is a question for your state's AG, not one your board can vote on.
The trap. A member's personal drone flown on a callout is not "exclusively leased for at least 90 continuous days" by a government entity. Neither is a team-owned aircraft if the team is not itself a political subdivision. The FAA's Public Safety BVLOS Waiver Flow Chart, current 28 January 2026, says it plainly: "Many Public Safety Do NOT Qualify."
§ 40125(d) allows a narrow FAA waiver of the 90-day exclusive-lease requirement specifically for search and rescue under extraordinary circumstances. Read what it does: it relieves the lease clock, not the government-entity requirement. The Part 107 versus Part 91 comparison works that boundary, and the commercial-purpose trap that voids public aircraft status, in detail.
| Route | Who it fits | BVLOS? | Key constraint |
|---|---|---|---|
| Sheriff's office PAO umbrella | Teams designated under the county SAR authority | Yes, via the agency's 91.113 CoW/A | Title or 90-day lease sits with the agency |
| Part 107 PSO Shielded Operations | Volunteer SAR teams failing the PAO test | Yes — 200 ft AGL, Class G | PSO letter, Standard Remote ID, ADS-B In, lighting, parachute |
| Baseline Part 107 | Any certificated remote pilot | No | VLOS only; authorization in controlled airspace |
The two routes that work: a sheriff's umbrella or a PSO Shielded Operations waiver
Route A — fly under the sheriff. In most states the county sheriff is the legally responsible SAR authority; Oregon runs SAR under ORS Chapter 404 with 36 county SAR coordinators. If your team is formally designated a unit of that office and flies aircraft the office owns or exclusively leases for 90 continuous days, the flights are the sheriff's public aircraft operations under the sheriff's Part 91.113 CoW/A. Being on the callout roster is not enough. What makes it work on paper: title or lease in the agency's name, written designation of the unit, a Public Declaration Letter signed by counsel, and a per-flight determination that the mission is an enumerated governmental function.
Route B — Part 107 PSO Shielded Operations. This is the FAA's designed answer for organizations that fail the PAO test, and what most volunteer teams should pursue. It relieves 14 CFR 107.31 (visual line of sight), the altitude limit, and 107.145. From the FAA's PSO Shielded Operations checklist: at or below 200 feet AGL, or 100 feet above any obstruction, not to exceed 400 feet AGL; Class G only; under 55 pounds; 87 knots; Standard Remote ID; anti-collision lighting day and night; ADS-B In on 978 and 1090 MHz; and a parachute meeting ASTM F3322-18 or newer above 0.88 pounds. Eligibility rests on the § 926(e) Public Safety Organization definition from Public Law 118-63, not on PAO status.
Confirm where to file before you file. The FAA's Part 107 waivers landing page says applications moved to the Aviation Safety Hub (AVSHub), with a user guide dated July 2026. The PSO checklist still directs submission through FAADroneZone. Airspace authorizations did not move. Check the FAA waiver landing page the week you submit.
SGI, the 91.113 waiver, and Remote ID during an active search
For a live callout, Special Governmental Interest (SGI) is the mechanism. Call the System Operations Support Center at 202-267-8276 or email 9-ATOR-HQ-SOSC@faa.gov. Search and rescue is a listed qualifying operation, with firefighting, law enforcement, utility restoration, disaster damage assessment, and emergency media coverage.
SGI amends authority you already hold — it creates none. You must already hold a current Part 107 certificate or an existing COA. A team with neither cannot use SGI, and a callout is the wrong time to learn that. VLOS emergency approvals may be issued in minutes. BVLOS SGI typically requires a TFR and extended processing — do not build a night grid search around a phone call.
If your sponsoring agency holds a Part 91.113 CoW/A, understand what its two mitigation methods mean over terrain rather than over a city. The obstruction-shielded package, taken by roughly 87% of applicants, allows 200 ft AGL in Class G or 100 ft above an obstruction within a 100 ft radius up to 400 ft AGL, at 1 statute mile visibility clear of clouds. Shielding is defined relative to obstructions: generous downtown, nearly useless on a talus slope. In timber you shield off canopy; above treeline you work from a 200 ft AGL floor over rising ground. The DAA package allows 400 ft AGL but demands 3 SM visibility and 500 ft below / 2,000 ft horizontal cloud clearance, exactly the weather you do not have on a mountain search. And the waiver's airspace half is capped at the UASFM grid altitude or 200 ft AGL, whichever is lower; above that you need a separate ATO COA through CAPS, filed at least 60 business days ahead. The airspace and TFR guide covers that boundary.
Remote ID applies to SAR like everyone else, fully enforced since the FAA ended enforcement discretion on 16 March 2024. Part 89 provides a narrow path for government entities to obtain temporary authorization not to broadcast where operational security requires it, but per secondary reporting, routine categories including search and rescue, training, and firefighting generally do not qualify. Treat routine SAR as a broadcasting operation and check 14 CFR 89.510 and 89.520 and the FAA public safety toolkit. One constraint catches teams: a broadcast module is VLOS-only and cannot be used for BVLOS.
Search theory for drone crews: probability of area, detection, and coverage
A drone is a sensor assigned to a segment. It is not a search plan, and treating it as one is how programs burn 200 flight hours without a find.
Probability of area (POA) is the chance the subject is in a given segment, built from lost-person behavior data. Robert Koester's work is the reference; his published analysis of autism-spectrum subjects gives a median distance of 0.9 km from the initial planning point and a 95th percentile of 8.0 km, with urban subjects traveling nearly three times farther than wilderness subjects. Probability of detection (POD) is the chance you find the subject given they are in the segment you searched. Probability of success is the product. A perfect POD in a 3% POA segment is a wasted sortie.
Sweep width bridges the two. The National SAR Committee's Sweep Width Estimation for Ground Search and Rescue defines it as "a single number characterizing the average ability of a given sensor to detect a particular search object under a specific set of environmental conditions." Its field experiments produced effective sweep widths from 142 meters for a high-visibility adult in Virginia winter forest down to 17 meters for a low-visibility adult in dense Washington forest, with clue-sized objects at 8 to 20 meters. That 8:1 spread is the point: sweep width is a property of conditions, not of equipment.
For an aerial sensor, POD is resolution-limited before it is swath-limited. Climbing widens the footprint and covers more ground per minute, but every foot of altitude cuts pixels on target, and detection collapses below a pixel threshold no matter how much ground you covered. So fly the altitude your sensor needs and accept the coverage rate that produces. Report the segment, the coverage flown, and an honest POD back to plans — and if a hasty team already worked that segment, your POD does not add to theirs.
Thermal search: what it detects, and what it does not
Thermal is a detection aid with sharply bounded physics. Overselling it inside your own team is the fastest route to a bad after-action.
Thermal crossover is the first constraint. Around dawn and sunset the radiant temperature curves of different materials intersect — the point at which "there is no difference in the radiant temperature of materials." Contrast collapses, and those are the hours a callout most often launches.
Thermal does not see through foliage. It sees canopy. Peer-reviewed work on thermal drones for SAR found that vegetation obscuration "leads to a reduced thermal contrast... due to increased blending of tree branches and leaves with humans in each pixel." Woodland is the hardest case, and it is where a large share of wilderness subjects are.
The same Drones journal study gives the altitude arithmetic. To render a 0.5 × 0.5 m human as five pixels in thermal imagery, the maximum height is 397 m AGL for a subject 10 °C warmer than background, but only 119 m AGL at 3 °C warmer. Roughly ten pixels are needed for reliable identification. And the ceiling case: "If the variation in ground temperature is up to 3 °C, then the 11 °C human is never detectable."
| Condition | Effect on detection | Practical response |
|---|---|---|
| Dawn / dusk crossover | Contrast approaches zero | Fly thermal late night or pre-dawn; use EO in that window |
| Dense canopy | Signal blends with foliage in each pixel | Do not credit forested segments with a thermal POD |
| Wet or hypothermic subject | Reduced surface temperature | Fly lower and slower; accept less coverage |
| Subject in water to shoulder height | Strong signature against cooler water | Thermal is genuinely strong here |
| Sun-warmed rock, roads, livestock | Clutter exceeds subject contrast | Fly after the ground has radiated; verify hits with EO |
That third row deserves emphasis: wet clothing and hypothermic vasoconstriction both cool the surface the sensor sees, so the subject in most urgent need is systematically the hardest to find.
Mapping search versus live-video search
A single operator watching a live feed is one set of eyes, once, with no artifact afterward. That is right for a responsive, mobile subject when minutes matter.
For a defined segment with a static or unresponsive subject, a mapped grid reviewed by several people beats it. Fly nadir stills rather than video; published guidance for SAR imagery recommends about 60% overlap, lower than the 70–75% used for survey mapping, so each pixel appears in at least two images while conserving battery, with mid-day collection to minimize shadow. Reviewers, called "squinters" in the SAR literature, then work the tiles — a task researchers describe as "time consuming and subject to human error as squinters may be fatigued from the search." Several independent reviewers on the same imagery is therefore a POD multiplier, and you keep a record to re-examine when the subject description changes. The cost is latency you may not have.
Night, terrain, weather, and battery realities
Night no longer needs a waiver. Since the April 2021 rule, 14 CFR 107.29 requires that the RPIC completed an initial knowledge test or 107.65 recurrent training after 6 April 2021, and that the aircraft carry anti-collision lighting visible for at least 3 statute miles — the same in civil twilight, 30 minutes either side. The trap: night flight in controlled airspace still requires an airspace authorization. Under a 91.113 CoW/A, lighting operates day and night and "lights out" operations are prohibited.
- Density altitude. Hot, high, and humid degrades rotor performance exactly as the FAA's density altitude pamphlet describes for crewed aircraft. A published service ceiling assumes a sea-level standard day.
- Wind at ridge lines. Lee-side rotor builds far downwind of a ridge. Cross with altitude in hand, and never fly the last of a battery on the windward side.
- Cold batteries. Capacity falls and voltage sags in the cold, so endurance drops and low-voltage warnings arrive early. Pre-warm packs and treat published endurance as marketing in winter.
- C2 link in canyons. Terrain blocks control and video at short range. Plan a relay or elevated control position, and program a return-to-home altitude that clears terrain.
- Launch and recovery. Firm, level, clear of debris, outside the ICP footprint, with a briefed alternate.
Working under a helicopter: deconfliction and on-scene communications
If a crewed aircraft is working the search, the drone does not launch on the remote pilot's judgment. It launches when the incident's air operations function says so.
Whoever holds air operations — an Air Operations Branch Director, an ATGS on a fire, or the sheriff's aviation unit — owns the airspace over the incident. Before launch you need a named air boss, a dedicated tactical channel with one voice speaking for the UAS, an altitude block with a hard deck for the crewed asset and a hard ceiling for the drone, geographic separation if altitude separation will not work, and a "knock it off" phrase every crew member knows. Land and hold on that call immediately.
On wildland incidents this is formalized: NWCG and NIFC standards govern UAS on fires, dispatch evaluates a TFR whenever aircraft respond, and intrusions are reported through SAFECOM and to the nearest ARTCC. If your search area falls inside a fire TFR under 14 CFR 91.137, the 91.113 waiver does not get you in — there are no TFR operations under the waiver. That is an SGI request, coordinated with the incident.
Documentation, after-action, and evidence handling
Log the authority, not just the flight. For every sortie record what it was flown under (agency PAO, Part 107, the PSO waiver, or an SGI file number), the RPIC, aircraft and registration, segment, altitude AGL, sensor and settings, coverage flown, weather, times, and the POD reported to plans.
Some of it is mandatory. Under the 91.113 CoW/A you must retain a record of all flight times including monthly totals, available to the Administrator on request. NTSB reporting under 49 CFR Part 830 is not waived: report any event where the aircraft came within 500 feet of another aircraft vertically or horizontally, with closest-point-of-approach measurements, plus any loss-of-control event. Under Part 107, 107.9 requires reporting a serious injury, loss of consciousness, or property damage above $500 within 10 calendar days.
When a search becomes an investigation, the rules change under you. Establish chain of custody at capture: original media preserved unaltered, hashed, on write-once or access-controlled storage, with documented transfers. Eighteen states require a search warrant before drone deployment for surveillance or evidence-gathering, with exigency and rescue exceptions — and the exigency ends when the subject is located.
The after-action that improves a program compares the sweep width you planned against what the crew could actually resolve, and asks why a segment with a claimed high POD did not produce the subject. That record also answers the question if the team is ever asked why a segment was cleared.
Equipment and procurement for a small volunteer team
Buy the sensor first, the airframe second. For most wilderness teams an electro-optical camera producing mapping-quality stills plus a modest radiometric thermal payload outperforms a thermal-only aircraft, because EO mapping works in the daylight hours thermal is worst. Two ordinary aircraft beat one exotic one when the nearest replacement is four hours away, and batteries, not airframes, set your on-scene endurance.
Federal grant money changes the analysis. The American Security Drone Act, enacted in the FY2024 NDAA (P.L. 118-31), made the prohibition on operating covered UAS and on using federal funds to procure or operate them effective 22 December 2025. FAR 52.240-1 took effect 13 March 2026, and OMB memorandum M-26-02 imposes security requirements on non-federal grant recipients, including SAR teams receiving pass-through funds. The legally operative list is the FASC list published in SAM.gov, not the Blue UAS Cleared List, a DoD designation that does not bind state and local agencies. Several states have enacted their own restrictions with hard phase-out dates. The agency procurement guide works through the compliance path and the purchasing vehicles that produce a defensible paper trail.
Frequently asked questions
Can a volunteer search and rescue team fly as a public aircraft operator?
Usually not. Search and rescue is an enumerated governmental function under 49 U.S.C. § 40125(a)(2), but public aircraft status also requires that the aircraft be owned by a government entity or exclusively leased by one for at least 90 continuous days, and that the entity be a government or political subdivision. A volunteer nonprofit generally fails that test, and political subdivision status is determined by the state Attorney General.
Do we need a waiver to fly a SAR drone at night?
No. The April 2021 Operations Over People rule eliminated the night waiver. Under 14 CFR 107.29 the remote pilot must have completed an initial knowledge test or recurrent training after 6 April 2021, and the aircraft must carry anti-collision lighting visible for at least 3 statute miles. Night flight in controlled airspace still requires a separate airspace authorization.
How do we get authority to fly inside a TFR during a search?
Through a Special Governmental Interest request to the FAA System Operations Support Center at 202-267-8276 or 9-ATOR-HQ-SOSC@faa.gov. Search and rescue is a listed qualifying operation. The catch is that SGI amends authority you already hold, so you need a current Part 107 certificate or an existing COA. Visual-line-of-sight approvals may issue in minutes; BVLOS requests typically require a TFR.
Can thermal cameras see through trees?
No. Thermal sensors image the surface they can see, which under canopy is the canopy. Published research on thermal drones for search and rescue found that vegetation obscuration reduces thermal contrast, because branches and leaves blend with the human signature within each pixel. Woodland is the hardest detection scenario, so do not credit forested segments with a thermal probability of detection.
What altitude should we fly a thermal search?
Low enough that the subject occupies enough pixels to detect, which depends on thermal contrast rather than any fixed number. Research published in Drones found that rendering a 0.5 by 0.5 meter human as five pixels allows up to 397 meters above ground at a 10 degree Celsius contrast, but only 119 meters at 3 degrees. Ten pixels are needed for confident identification.
Does Remote ID have to be broadcasting during a search and rescue mission?
Assume yes. Part 89 applies to all drones requiring registration, including public safety aircraft, and has been fully enforced since 16 March 2024. A narrow path exists for government entities to obtain temporary authorization not to broadcast where operational security requires it, but routine categories such as search and rescue, training, and firefighting generally do not qualify.
Does a drone replace ground searchers?
No. A drone is a sensor assigned to a segment within an existing search plan, and its contribution is measured the way any resource is: probability of area multiplied by probability of detection. A high detection probability in a low-probability segment produces nothing. Aerial sensors are strongest on open terrain, water margins, and linear features, and weakest under canopy.
Where this leaves you
Settle the authority question before you buy anything. Get a written answer on whether your team is a political subdivision, and if it is not, stop forcing the PAO argument and file for the Part 107 PSO Shielded Operations waiver — it is the path the FAA built for you, and it gets you BVLOS to 200 feet in Class G. Confirm your remote pilots hold current certificates so SGI is available on a callout, and get your sponsoring agency's air operations function to write the deconfliction protocol before the night you need it.
Then work on the part that finds people: assign segments from a real probability-of-area model, fly the altitude your sensor supports rather than the one that covers ground fastest, and be honest in your POD reports. Flight records and per-flight authority documentation have to live somewhere retrievable, because the after-action and the eventual records request will both ask for them.