12 Designing the soundscape
Eleven chapters have been spent on a harm. This one ends on a possibility. For most of its regulatory history, noise policy has been subtractive, a matter of limits, barriers and the removal of unwanted sound. The soundscape turn of Chapter 1 opens a different verb: not only to reduce sound but to compose the acoustic environment, to ask what a place should sound like and build toward it. The repertoire is already in use — masking a traffic drone with water and birdsong, protecting quiet areas as public goods, augmenting a street’s soundscape with AI-selected natural sound, and, at the level of principle, treating a right to quiet as worth naming. This is the affirmative counterpart to the diagnosis, and it keeps faith with the governing ambition, which is to sketch the components of a scientifically current philosophy of noise without issuing regulatory certainties.
12.1 From abatement to design
The subtractive regime can ask only one question of a place: how much sound is too much. That is the question the whole legal architecture was found to be asking in Chapter 11, and asking of the wrong quantity. The soundscape framework asks a second one. Where the acoustic environment is defined not as a level but as what is perceived and experienced in context (International Organization for Standardization, 2014), the object of concern stops being the decibel and becomes the character of a place as heard. Character admits of improvement in a direction that mere subtraction cannot reach. Less sound and a better soundscape are not the same goal. A street scrubbed to silence can read as sterile, even uneasy; a courtyard with a fountain and birds can be experienced as calm at a level no quieter than the street it adjoins. The measure that matters is not how much is removed but what remains, and how it is heard.
Once perception is the measure, the descriptors of a soundscape become quantities that can be predicted rather than merely recorded, and what can be predicted can be aimed at (Aletta et al., 2016). That is the whole of the design move: the passage from measuring what is present, through modelling what would be experienced, to composing toward a chosen result. It does not retire abatement. The drone must still be capped, the barrier still built, the source still turned down; a soundscape cannot be composed over a din that drowns it. What design adds is a second axis the level-based regime lacks — the very axis found absent from the law in Chapter 11, quality set beside loudness rather than loudness alone.
12.2 Masking and the natural soundscape
Masking has a narrow acoustic sense and a broader perceptual one, and the difference matters for what it can be asked to do. In the narrow sense — energetic masking — one sound raises the threshold at which another becomes audible, literally covering it. In the broader sense — informational masking — a sound need not cover another to change how it is heard; it can draw attention away from it, or recast it as one element in a scene rather than the intruder in a quiet one. Natural sounds work by both, but their distinctive contribution is neither: it is affective. Water and birdsong carry positive valence, and introducing them shifts the whole scene toward the pleasant pole of the circumplex. The physiological trace is measurable — recovery from a stressor is faster during nature sound than during environmental noise (Alvarsson et al., 2010). And the aggregate is by now well attested: across a body of studies, exposure to natural sound lowers stress and annoyance and improves health and affect (Buxton et al., 2021).
The effect is real but neither uniform nor unconditional. In controlled listening, adding a fountain reduces the perceived loudness of road traffic only when that traffic is steady; against fluctuating traffic it does far less, and birdsong raises pleasantness and eventfulness by more than water does in either case (De Coensel et al., 2011). Masking is therefore source-, level- and type-dependent — a design instrument with a response curve, not a switch. Which natural sound, at which level, against which drone, are questions with different answers.
Two limits keep the technique honest, and both matter more the more capable augmentation becomes. The first is arithmetic: a masker is itself sound, so adding it raises the overall level, and beyond a point that added loudness costs more pleasantness than the wanted sound buys (De Coensel et al., 2011). There is an optimum, and past it more nature is simply more noise. The second is evidential: in the aggregate, the weakest benefit is found where natural sound is mixed with the noise rather than substituted for it (Buxton et al., 2021). Masking complements abatement; it does not replace it. A data-centre hum overlaid with birdsong is still a data-centre hum, now at a higher overall level, and the temptation to treat the overlay as a solution is the licence problem raised in Chapter 11, arriving now in constructive dress.
12.3 AI-assisted augmentation
Field trials of automatic masker-selection systems report gains in perceived quality and restorativeness, an early demonstration of design rather than mere abatement (Lam et al., 2024). The system is an algorithm choosing natural sounds to augment a traffic-exposed residential soundscape. What makes the automation possible is the predictive step already described, made operational: models that estimate perceived pleasantness from the balance of sound sources present (Mitchell et al., 2021) allow a system to judge, and continuously adjust, a masker in situ without a single person being surveyed. The soundscape stops being something assessed after the fact and becomes something steered in real time — the fullest expression, so far, of the move from measuring sound to composing it.
The critical reservations follow the technique rather than opposing it. Automation inherits every limit of masking and adds its own, the first of which is whose experience it optimises. The pleasantness a predictive model maximises is an average, fitted to aggregate responses (Mitchell et al., 2021), and an average can be raised while a minority is left worse off. The sound-sensitive tail is the clear case — the migraine sufferer of the threshold note in Chapter 11, or anyone for whom the remedy for unwanted sound is less sound and not more. An instrument that improves the mean by adding sound cannot help those the addition harms, and it has no way of knowing it has not helped them. Optimisation quietly encodes a choice about whose ears count.
The second reservation is the licence problem raised to system scale. The same tool that composes a genuinely better street can be deployed to make a non-compliant source tolerable enough to leave in place — to buy the operator time rather than to protect the resident. And a reflexivity runs underneath, real but easily overstated: the compute that trains and runs such systems is itself among the machine-hum sources of Chapter 7, so that artificial intelligence may end up enlisted to soften a soundscape its own infrastructure helps to harden. None of this voids the method. It locates it — a real addition to the designer’s repertoire, and no substitute whatever for turning the source down.
12.4 Quiet areas and a right to quiet
The positive turn is not wholly without a foothold in law. The Environmental Noise Directive already carries, beside its mapping machinery, an obligation to identify and preserve quiet areas of good acoustic quality (European Parliament and Council of the European Union, 2002) — a rare clause in the whole apparatus concerned with sound worth keeping rather than sound to be capped. It is the design idea present in the statute in embryo. But the clause is largely unimplemented and, where implemented, unenforced (European Environment Agency, 2025) — a good instinct lodged in the law as aspiration, without the duty or the metric that would make it bite. Preserving a quiet area presupposes a way of saying what makes it worth preserving, and that is a claim about quality, not level — the quantity the governing instrument cannot read.
Strasbourg has gone further than the directive’s soft language. In Moreno Gómez v. Spain, the European Court of Human Rights held that persistent night-time noise in a district the city had itself declared acoustically saturated, tolerated by authorities that then failed to act on their own designation, violated the applicant’s right to respect for her home and private life under Article 8 of the Convention (European Court of Human Rights, 2004). The reasoning matters as much as the result: quiet in the home is treated not as an amenity to be traded away but as a protectable interest, whose sustained denial is a rights violation the state must answer for. It is a Spanish case, and its logic reaches directly the zonas acústicamente saturadas that Granada and other cities designate — the local instrument through which an abstract right acquires a map.
Yet the case marks the limit as clearly as the possibility. A right built on the home and on gross, measurable excess still leans on the metrics found blind in Chapter 11 — a tonal hum sitting within the A-weighted limit fits no easy Article 8 claim, however unliveable the bedroom, because the instrument that would evidence the harm cannot register it. And a right framed only as freedom from noise secures just one axis of the good. Sonic justice, on the fuller account of Chapter 9, concerns access to wanted sound as much as protection from the unwanted (Trudeau et al., 2023); a right to quiet answers the second and is silent on the first. The distributional argument returns here with force, for quiet is a scarce good unevenly held, and a right that only subtracts noise does little for those who had no good soundscape to begin with. To name a right to quiet is to take a first, real step and to see, in the same movement, how partial a step it is.
12.5 An augmentation demonstrator
To keep augmentation from becoming a conjuring trick, two features have to stay visible in the arithmetic, and the demonstrator below is built around them. A traffic drone is set at the bedroom window and a natural masker chosen and set beside it (the drone as an immission level, the sound as received, not the source’s emission). No audio is played: what the tool returns is an estimated soundscape pleasantness on the ISO scale, computed from the balance of sound sources after the predictive model of Mitchell et al. (2021) and the masker-level behaviour measured by De Coensel et al. (2011). Past an optimum, additional masker only raises the overall level and the estimate falls, so that more nature becomes more noise. And the pleasantness the best masker can reach is bounded from above by the drone itself, so that a source loud enough can be made less unpleasant but never pleasant. The licence problem, put out of reach of rhetoric, is left standing plainly in the numbers.
Source · pleasantness estimated from the balance of perceived sound sources after the predictive model of Mitchell et al. (2021), expressed on the ISO 12913-3 pleasantness scale (International Organization for Standardization, 2019); the masker-level trade-off and the type ordering (birdsong above water) after De Coensel et al. (2011); augmentation-as-design after the field results of Lam et al. (2024). Levels are illustrative immission values at the façade; model coefficients are indicative, not a reproduction of the source regressions. Accessed 1 August 2026.
12.6 The sound of a decarbonised, dense, liveable city
Three movements have carried the argument this far. A measurable object, whose harm is real and graded and whose measure both reveals and conceals it; a set of sources in rapid change, the old giant of traffic now joined by the steady hum of the energy and digital transitions; and a politics of distribution, governance and design, in which the harm falls unevenly and the instrument that governs it cannot see much of what it does. The three do not resolve into a single recommendation, and were never meant to. They converge instead on a claim about the near future, and it is a claim about choice.
The decarbonised city is not, by default, the quiet city. Combustion recedes as the fleet turns electric, and with it the loudest register of the twentieth-century street; but the transition arrives humming, in heat pumps on party walls and data centres at the edge of town, sources steady and tonal and low where traffic was intermittent and broad. Granada is not a spectator to this. A seventy-megawatt AI compute campus proposed on its metropolitan edge (DataCenterDynamics, 2026) is the machine-hum future landing locally, in a city whose own strategic map already records where sound falls and on whom. What that campus, and its like across Europe, will add to the acoustic environment is not yet written; it is being decided now, in permits and setbacks and the choice of which quantity to hold it to.
And that choice, in the end, is the one durable finding turned, at last, from diagnosis toward possibility. What is measured decides what can be governed; it decides, equally, what can be designed. Govern and design by the A-weighted level alone, and only loudness can be managed — the hum stays invisible, the quiet area unprotected, the augmentation ungoverned. Adopt instead a measure of what a place is like to be within, and a different practice becomes available, one that can protect a quiet worth having, catch a harm the meter discounts, and tell composition from cover. The argument about instruments arrives, at the last, not as a complaint but as a condition for doing better — hearing first what is there.
None of this is a regulation to be passed, and the discipline observed throughout forbids issuing it as one. The sound of the liveable city is a design decision being taken already, for the most part by default, in the aggregate of choices no one made deliberately. Taking it on purpose asks something modest and hard: a way of listening precise enough to register what the current measure leaves out, and the will to let what is heard count. The harm has been diagnosed, and this last turn points past it — not to a prescription, but to the one prerequisite every better soundscape shares. Hear it first.