Three families of atomiser are fitted to spraying drones. Hydraulic flat-fan and hollow-cone tips are inherited from ground sprayers; their droplet size depends on orifice size and pressure, with smaller orifices and higher pressures giving finer sprays (Nuyttens et al., 2007). Air-induction versions of the flat fan draw air into the liquid and produce coarser, air-filled droplets intended to reduce drift. Rotary or centrifugal atomisers, now common on larger machines, fling liquid from a spinning disc or cage; droplet size is set mainly by rotational speed and is largely independent of flow rate, which gives a comparatively narrow spectrum of the kind sought in controlled droplet application (Matthews et al., 2014).

Droplet spectra are reported against reference nozzles so that results from different laboratories can be compared. The scheme of the American Society of Agricultural and Biological Engineers assigns a spray to a category from very fine to ultra coarse (ASABE, 2009), and an international standard now provides an equivalent classification (ISO, 2018). Table 1 summarises the atomiser options against the criteria that matter for variable-rate use. A weakness of the literature is that many drone studies give the nozzle model but no measured spectrum, and spectra measured in still air may not hold beneath a rotor.

A conventional sprayer changes its output by changing pressure, and with a hydraulic tip a change of pressure also changes droplet size and spray angle. For variable-rate work this coupling is a defect: the spray quality on which the label and the drift assessment rest shifts every time the rate does. Two remedies are available. Pulse-width modulation, in which a solenoid valve opens and closes the nozzle many times a second and the proportion of open time sets the flow, was shown long ago to vary output over a wide range with little change in droplet size (Giles and Comino, 1990), and it featured in early unmanned helicopter systems (Huang et al., 2009). The rotary atomiser offers the same independence by separating the pump, which sets flow, from the disc motor, which sets droplet size.

The control problem does not end at the nozzle. The aircraft must know where it is, how fast it is moving and what the prescription requires, and the liquid system must respond before the target has passed. Automatic systems that link flow to ground speed and position have been demonstrated (Xue et al., 2016). The logical extension, in which wind measured at the field corrects the flight line while the job is in progress, remains a research topic. Electrostatic charging has been proposed as a way to improve deposition at low volumes; the principle is well established for ground equipment (Law, 2001), but evidence for a consistent benefit on drones is limited.

Every rotorcraft supports its weight by driving air downwards, and the downwash carries droplets into the canopy. A study in rice that used a small single-rotor unmanned helicopter found that deposition within the crop and the control of planthoppers both depended on flight height and speed, which is consistent with a strong influence of the downwash (Qin et al., 2016). On a multirotor several smaller rotors interact with one another and with the airframe, and the wake is less orderly. It complicates matters most in forward flight. As speed rises the wake tilts backwards and rolls up into trailing vortices, and droplets entrained in them can be lifted and held aloft. Measurements on multirotor sprayers show that spray pattern and effective swath change with release height and ground speed (Martin et al., 2019). Because the downwash scales with the weight of the aircraft, the pattern also changes as the tank empties, which is an awkward property for a machine expected to deliver a prescribed dose.

Field comparisons agree on the direction of the main effects, if not their size. Coarser sprays and lower flight speeds reduce the drift potential of herbicide applications but at some cost in coverage (Hunter et al., 2020). Higher application volumes improve deposition and biological effect in wheat, yet they shorten the area covered per flight, and very low volumes with fine sprays give less reliable control (Wang et al., 2019). Crosswind is the dominant weather variable. Downwind deposits rise with wind speed in the trials collated for regulators, which is why any buffer distance has to be tied to a stated wind (OECD, 2021). Temperature and humidity act through evaporation, which shrinks droplets in flight and matters more for the finely atomised, low-volume sprays that small tanks impose (Matthews et al., 2014).

The reference method for field measurement of drift is an international standard that prescribes the layout of ground and airborne collectors, the tracer, and the weather records to be kept (ISO, 2005). It was written for boom and orchard sprayers, and its application to a small aircraft whose swath and height vary raises practical questions that authors have resolved in different ways. This is one reason why published drone drift curves are difficult to compare.

Regulatory assessment of aerial application has long relied on mechanistic models. AgDRIFT and its parent code AGDISP follow droplets released into the wake of a fixed-wing aircraft or helicopter and predict deposition downwind (Teske et al., 2002). Work to extend this approach to multirotors has coupled the droplet model to a rotorcraft wake code (Teske et al., 2018), but validation data remain sparse. The synthesis prepared for regulators concluded that the available trials were too few and too varied to support a general drift curve for unmanned aircraft (OECD, 2021).

In the European Union, unmanned aircraft operations are divided into open, specific and certified categories, and an operation that involves dropping material or carries a heavy payload generally falls into the specific category, which requires an authorisation based on a risk assessment (European Commission, 2019). In the United States, the rule for small unmanned aircraft covers machines below 55 pounds, and dispensing an economic poison additionally brings the operator under the long-standing rules for agricultural aircraft operations, with heavier spray drones needing exemptions (Federal Aviation Administration, 2016). India liberalised its framework with rules that classify drones by weight and zone the airspace (Ministry of Civil Aviation, 2021). The common thread is that payload mass, flight beyond visual line of sight and the operation of several aircraft by one pilot, each of which matters for the economics of spraying, are the points at which permissions become harder to obtain.

Pesticide law is often the tighter constraint. The European directive on sustainable use prohibits aerial spraying and allows it only by derogation where there is no viable alternative or a clear advantage for health and the environment (European Parliament and Council, 2009). Since a drone is an aircraft, the prohibition applies to it, however different its drift behaviour may be from a manned aeroplane. Elsewhere the difficulty lies in the label. A product is authorised for stated crops, doses, water volumes and application methods on the strength of residue, efficacy and exposure data generated with conventional equipment. Drone application at very low volumes and high concentrations may fall outside those conditions, and operator exposure during frequent mixing and loading of concentrated spray has scarcely been studied (OECD, 2021).

Variable-rate application raises a question that labels do not yet answer. A prescription that lowers the dose in part of a field is easily accommodated, since labels set maxima. A system that changes droplet size or volume on the move is harder to place, because drift mitigation measures such as buffer zones are tied to a declared spray quality. Atomisers that hold droplet size constant while varying flow therefore have a regulatory as well as an agronomic advantage.