Heat Pump Noise: Scoping Overview of Emission Characteristics and Systematic Review of Human Effects

Abstract

Objective:

The aim of this work is two-fold: to characterize air-source heat pump (ASHP) noise emissions and to systematically evaluate the scientific evidence regarding non-auditory health effects, thereby assessing the technology’s current and future public health impact.

Methods:

The primary component is a PRISMA-guided systematic review on noise impact. Utilizing a population–exposure–outcome framework, eight bibliographic databases were searched for English and German records. Of 501 identified records, 26 reports (covering 25 unique studies) met eligibility criteria. Risk of bias (RoB) was evaluated across four domains—sampling, measurement validity, confounding, and reporting—while overall evidence certainty was assessed using established frameworks. A secondary, non-systematic scoping overview was conducted to characterize ASHP noise emissions based on technical databases and literature.

Results:

ASHP sound emissions are highly variable, contingent upon specific components (fan and compressor), radiation angles, and operating conditions. The systematic review identified 10 laboratory studies, 14 observational surveys, and one prognosis study, predominantly published since 2011. While results indicate a non-negligible potential for annoyance, concentration impairment, and self-reported sleep disturbance, the overall certainty is low. This is attributed to small samples, methodological heterogeneity, and high RoB in critical domains. Preliminary data provide tentative indications that ASHP noise may exhibit higher annoyance potential than road traffic noise at equivalent sound pressure levels.

Conclusion:

Despite increasing research interest, the evidence base for long-term health impacts of heat pump noise remains limited and of low certainty. Future research should prioritize large-scale longitudinal studies and comparative investigations with other environmental noise sources to inform evidence-based noise control.

KEY MESSAGES

• (1)No “typical” heat pump noise: emissions vary across devices and conditions.
• (2)Heat pump noise shows annoyance potential even at relatively low sound pressure levels (SPLs).
• (3)Spectral and temporal characteristics influence heat pump annoyance.
• (4)Current evidence is largely based on short-term experiments and self-reported surveys, with very limited data on long-term health effects.
• (5)Comparative and long-term heat pump noise studies are urgently needed.

INTRODUCTION

Heat pumps (HPs) are systems that extract thermal energy from the outside surroundings (e.g., from the air [air-source heat pump, ASHP] or from the ground [ground-source heat pump, GSHP] to provide heat and/or hot water for residences. This technology is considered energy-efficient and central to achieving net-zero carbon-emission goals set by governments.[1] HP sales differ widely between countries,[2] but overall, the number of HPs is increasing, with 2.31 million units sold across 19 EU countries in 2024 and a total stock of 25.5 million HPs (roughly 12% of households).[3] Of these, ASHPs had a market share of around 80% across Europe in 2024 and were thus by far the best-selling HPs.[3]

The widespread adoption, in combination with the expected upward trend in the dissemination of HPs, “places new noise sources outdoors in a residential environment.”[4] This is particularly true for ASHPs: while residential GSHPs traditionally do not have outside noise-emitting components, ASHPs have their condenser unit (with its components: fan and compressor) often installed outdoors and thus can be relevant sources of noise, affecting residents. This applies to both ASHP constructions—the split system (fan outdoors and compressor indoors) and the monobloc unit (all components built into one outdoor system). The majority of the units sold have a heating capacity of 5–13 kW,[5] which is typical for use in single-family homes.[6] For densely populated urban areas and rural residential settings alike, this brings the following topics to the forefront: “What are the noise emission characteristics of HPs?”; “What is their noise impact on humans?”; and “What may be the consequences for public health?”

While ample measurements of HPs are conducted and results are published as sound power levels in various databases, detailed acoustic emission characteristics (frequency-resolved and/or directionally resolved data) seem much less accessible. Klingebiel et al.[7] and Langerova et al.[8] review the literature on this topic, but a systematic compilation of detailed acoustic emission characteristics is still lacking.

Similarly, there has been some research on the effects of HP noise on humans—also recently—which underscores the contemporary importance of the topic. However, no systematic literature review on HP noise specifically is available yet that consolidates the findings. Such a contribution would help evaluate the potential noise impact on the population, especially in the context of widespread adoption of this technology. In fact, knowledge on the noise effects of HPs is distinctly scarcer than knowledge of their acoustical characteristics. The recent systematic review on acoustical characteristics by Langerova et al.[8] did not report on noise outcomes. Other reviews that deal with the noise impact of various environmental sources in more general terms also do not sufficiently inform on HP noise impact: for example, the review on non-traffic noise related effects by Omlin et al.,[9] the review on environmental noise impact by van Kamp et al.,[10] or the review on effects of low-frequency (LF) noise exposure by Baliatsas et al.[11] all contain only one to two HP-source-specific papers (which are also included in the present review). In terms of effects related specifically to HP noise, the Department for Energy and Net Zero[12] conducted a systematic literature search but published only results relevant to the region of the UK, making it nongeneralizable. Consequently, a comprehensive and systematic review of the noise impact of HPs on human populations is still missing.

The objective of this study, therefore, is two-fold. The first objective is to give an ASHP-database overview and a literature review on noise emission characteristics of ASHPs (as the major HP type in residential areas), such as sound power level, frequency content, and directivity, to answer the question of what “typical” emission characteristics and/or potential variations are to be expected. Given the available reviews,[7,8] this overview is kept concise to complement available knowledge. The second objective is to provide a systematic review of HP noise effects in humans. Besides summarizing the current evidence base, we identify knowledge gaps to motivate possible future research on noise effects.

The article is structured as follows. First, we will give an overview of ASHP noise emission characteristics, summarizing insights from available ASHP databases and scientific literature and complementing these with our own measurements. The subsequent sections are dedicated to the literature review on the effects of HP noise on humans, describing the methods, results, and discussion, respectively, followed by overall conclusions.

ASHP NOISE EMISSION CHARACTERISTICS

This scoping overview focuses on ASHP noise emission characteristics. It does not cover measurement procedures; an overview of relevant standards such as EN 12102-1[13] is given by Langerova et al.[8]

ASHP Working Principle and Noise Generation

ASHPs transfer heat from a region of lower temperature to a region of higher temperature, for example, from the ambient air to the interior of a building, using electrical energy. The amount of transferred heat can be considerably greater than the required input work. Typical vapor-compression ASHPs operate using a vapor-compression cycle. They compress a gaseous refrigerant, thereby increasing its temperature and enabling heat transfer into the building interior. Subsequently, the refrigerant is expanded outside the building’s thermal envelope, which lowers its temperature below the ambient temperature. Thermal energy is then absorbed from the surrounding air through a heat exchanger, usually assisted by a fan to increase airflow through the exchanger and thereby improve the ASHP’s efficiency.[14] The compressor and fan are typically the dominant noise sources of an ASHP.[8]

Databases for Noise Emission Data

For HPs, extensive databases on the sound emissions at comparable operating points exist. However, manufacturers usually only publish total sound power levels, not detailed characteristics such as spectral data. The data are either published on manufacturers’ websites or in tools that are used for the prediction of sound exposure levels at receiver locations, as required by public authorities. Here, the European Committee for Standardization publishes the KEYMARK database.[15] In addition, there are national databases, such as those in the German-speaking European region, namely the Swiss Heat Pump Association,[16] the German Federal Heat Pump Association,[17] and the Austrian Heat Pump Association.[18] The data from the different sources are not directly comparable, as they provide sound power levels at different operating points according to EN 14825[19] (see above database),[16] non-standardized maximum sound power levels (databases [17] and [18], or the sound power levels according to EN 12102-1.[13,15] In addition, the data from these sources and the manufacturers’ websites are not available for bulk download, but must be accessed individually for each device. Here, the Heat Pump Test Centre Buchs, Switzerland,[20] is an exception. It publishes the A-weighted total sound power level (L_WA) according to EN 14825 on its website for various HPs.

Figure 1 shows an analysis of the latter database (version as of 28 November 2024), complemented with data by Stürenburg et al.[21] and measurements made by the authors. (Note that the latter two measurements were not performed at the same test points as those in [20]. The L_WA tends to increase with increasing heating capacity. However, the data scatter strongly, with differences in the L_WA of up to 25 dBA at the same heating capacity. This is in line with similar analyses,[7] and with measured sound exposure levels in residential areas as a function of the heating capacity.[22] Prediction of the L_WA based on heating capacity alone is thus not possible. Interestingly, for this dataset a higher (negative) correlation was found between L_WA and the coefficient of performance (COP). Using heating capacity and the COP in a regression model for L_WA resulted in a coefficient of determination R² of about 0.4. Nevertheless, at least favorable devices usually operate at relatively low emission levels, but increased levels may arise with less favorable devices.

Figure 1.

A-weighted sound power level (L_WA) of different ASHPs (monobloc and split design) as a function of heating capacity (A2/W35: air temperature 2°C, water temperature 35°C). Data (n = 38) from [20] complemented with [21] and the author’s measurements (both monoblocs).

Scientific Literature on Noise Emission Data

In addition to searching the databases mentioned above, the scientific literature was searched for more detailed sound emission data of AHPs (frequency-resolved and/or directionally resolved data). The literature was searched in close reference to the PRISMA Statement (Preferred Reporting Items for Systematic Reviews and Meta-Analyses),[23] but does not constitute a full systematic review.

The literature search was performed on 18 November 2025 using the databases Web of Science and Scopus. The following search strings were used: “HP noise,” “HP sound,” “hvac noise,” and “hvac sound.” The search yielded 4417 hits. Of these, 1116 duplicate records were removed, resulting in 3301 unique records. During title screening of the 3301 records, 2345 records whose titles did not contain the terms “heat pump,” “heat-pump,” “fan,” “compressor,” “valve,” or “hvac,” were excluded, resulting in 956 remaining records. During abstract screening of the 956 records, 812 records, which did not deal with ASHPs for residential buildings (e.g., GSHPs or HPs for industrial applications or vehicles), were removed, resulting in a total of 144 records. Of these, 51 full texts could be retrieved. In addition, a complementary search was performed on 18 November 2025 using the literature database Google Scholar with the search strings “Wärmepumpen Lärm” (HP noise) and “Wärmepumpen Geräusche” (HP sound) to cover German publications relevant for practical applications. This yielded 50 additional records (eligible full texts). Finally, nine records were included, which were known to the authors before the literature search and were not identified during the above screening. This resulted in a total of 110 eligible records, two of which were review articles on the same topic.[7,8]

The 110 records were screened for comprehensive sound emission data (frequency-resolved and/or directionally resolved sound recordings). The screening revealed that the range of topics covered by the available literature is very broad, covering (i) the optimization of HPs and their components (the topic of the majority of publications; e.g., [24], (ii) the characterisation of the (separate) main noise sources (but not the complete HP), (iii) limit values and regulations for HP noise,[25] (iv) psychoacoustic effects of HP noise (topic of the following noise effect review), (v) the control of individual components, complete devices or entire systems (HP plus consumer),[26] (vi) the market for HPs and how to influence it (e.g., [27], (vii) methods for measuring the acoustic properties of HPs or for fault diagnosis (e.g., [28], (viii) noise exposure in residential areas from HPs in outdoor areas (e.g., [22], and (ix) literature on acoustic emission characteristics of HPs.

Literature on the latter topic (ix), i.e., the focus of this review, is scarce. There are two overview articles. The systematic review by Langerova et al.[8] summarizes the acoustic properties of the fan and compressor and discusses relevant standards for noise assessment, but detailed emission characteristics are only exemplarily given. Klingebiel et al.[7] published a short, non-systematic review article focusing on strategies for acoustically optimizing HPs, including, among other aspects, the sound power level as a function of heating capacity (similar to Figure 1). Out of all relevant records on topic (ix), only nine studies document acoustic emission data beyond the total sound power level of commercially available, non-modified ASHPs. Table 1 gives an overview of the published data. Information on broadband and narrowband spectra and on radiation directivity is not systematically covered by all studies. Also, the data are neither fully comparable with each other nor with the ASHP databases discussed in the “Databases for noise emission data,” section as not all are measured according to standards, and not all devices and/or operating points are specified. A quantitative compilation of the data would thus be afflicted with large uncertainty, and a systematic compilation of detailed acoustic emission characteristics remains lacking. Nevertheless, a qualitative comparison of the data suggests that in many cases, the A-weighted emission spectrum is fairly uniform and has a weakly pronounced maximum between the 500 Hz and 1 kHz octaves, and that the compressor emits distinct tonal components (harmonics), which dominate the low-frequency emission spectrum. However, some compressors can also dominate the high frequencies in certain operating conditions.

Table 1.

Overview of studies included in the systematic review on acoustic emission data of commercially available, non-modified ASHPs.

Reference	Devices	Measurement	Broadband spectra (octave, 1/3-octave SPL)	Broadband spectra (octave, 1/3-octave LW)	Narrow-band spectra (SPL)	Directivity	Level-time histories	Sound recordings
Stürenburg et al.[21]	ASHP, model PUZ-WM112YAA by Mitsubishi Electric, Monobloc, 11.2 kW heating capacity	Laboratory, ISO 3745 [29] (semi-free field, 20 mics on hemisphere) and 12 mics 58 cm above ground for directivity	Determinable from sound recordings	No	Determinable from sound recordings	Determinable from sound recordings	Determinable from sound recordings	Yes
Krämer and Brandstätt[30]	ASHP by Bosch	Laboratory, ISO 3744 [31] (semi-free field, mic cube) and 73 mics on a hemisphere for directivity	Diagram for 5 mics	No	Diagram for 5 mics	Diagram for 100 Hz, 1.6 kHz, and total A-weighted SPL for ventilator operation	No	No
Schmidt and Müller[32]	ASHP prototype (close to commercial)	Laboratory (semi-free field, 20 mics on hemisphere)	No	No	No	Diagram for total A-weighted SPL	No	No
Reichl et al.[33]	ASHP with 8 kW heating capacity (“representative for a large part of the European market”)	Laboratory	No	Numerical values (test point A7/W35)	No	Numerical values (test point A7/W35)	Diagram	No
Troll and Gindre[34]	AHSP in compact design	Laboratory, ISO 3743-1 [35]	No	Diagram	No	No	Diagram	No
Poysat et al.[36]	3 ASHPs by Emerson: models ZXME-040, ZXLE-075E, OME-4MTL-07X	Laboratory (presumably)	No	Diagram (relative values)	No	Diagram (total A-weighted SPL, for ventilator or compressor)	No	No
Schulze et al.[37]	5 outdoor ASHPs in compact design (one without housing), 1 outdoor ASHP in split design, 1 indoor ASHP	Laboratory and field measurements, ISO 3746[38] and ISO 3747[39]	Diagram	Diagram	Diagram	Diagram	Diagram	No
BPC[40]	10 commercial ASHP	Field measurements, ISO 3746 [38]	No	No	No	Numerical values	No	No
Krämer and Rambausek[41]	2 ASHPs with 10 kW and 20 kW heating capacity	Laboratory (semi-free field), sound pressure level at 1 m distance and sound power measurement	Diagram	Diagram	No	No	No	No

Note: Abbreviations: SPL: Sound pressure level, LW: Sound power level. Reichl et al. ^[33] also cited in the review by Langerova et al. ^[8]

Acoustic Emission Measurement of an ASHP

To complement the above ASHP database and literature overview, an ASHP (alpha innotec, Switzerland) with a heating capacity of 5.8 kW, which is representative of a residential HP, was measured in the semi-anechoic chamber at the authors’ institution. Thirteen microphones were arranged around the ASHP on a virtual hemisphere with a radius of 2 m. Eight microphones (Microphones 1–8; see Availability of Data and Materials for equipment details) were positioned evenly around the ASHP at a height of 40 cm, and the remaining microphones (Microphones 9–13) on the upper part of the hemisphere. The setup has been verified with a reference source of known sound power, and a deviation of 0.2 dB between the measured and expected A-weighted sound power level has been determined. The fan of the ASHP was operated at rotational speeds between 75 and 500 revolutions per minute (rpm), and the compressor at frequencies of 45, 59, and 80 Hz, which cover typical operating conditions for this device. The measurements covered the source directivity as well as the spectral sound power level at different operating points, both for the compressor and fan separately and for the components together. The instrumentation did not allow for measurements according to a standard, but only for an estimation of the sound power levels.

Figure 2 shows the A-weighted sound power level of the fan and compressor as a function of their operating conditions. The sound power level increases with increasing fan rotational speed by up to 26 dBA. It also increases with compressor frequency, but much less pronounced, by only up to some 4 dBA. For fan rotational speeds above around 400 rpm, the fan dominates the total sound radiation of the ASHP. Figure 3 shows the A-weighted spectral sound power level of the fan and compressor at a representative operating condition. While the sound radiation of the fan is dominated by the frequency range from 250 to 2 kHz, the compressor mainly emits sound at low frequencies around 200 Hz. Furthermore, the fan mainly emits broadband sound, while the compressor emits more tonal sound.

Figure 2.

A-weighted sound power level (L_WA) of the fan and compressor as a function of the operating condition.

Figure 3.

(A) One-third-octave band spectra of the A-weighted sound power level of the fan at a speed of 420 rpm and the compressor at a frequency of 45 Hz. (B) A-weighted narrowband sound pressure levels. Recording: Microphone at a height of 40 cm, angle of 315° in the horizontal plane from the fan rotor axis.

Figure 4 shows the source directivity as the A-weighted sound emission level for different fan rotational speeds and compressor frequencies. The fan has a pronounced directivity. Its radiation is strongest along the fan axis and only slightly depends on rotational speed, with a difference between the emission levels along the rotor axis and perpendicular to it of some 6 dBA for rotational speeds of 300–500 rpm. The directivity of the compressor, in contrast, is much less pronounced. These findings correspond well to the literature, which reports very similar directivity differences of 6.2 dBA[33] and 4 dBA.[32]

Figure 4.

Source directivity patterns of the A-weighted sound emission level (in dB) of the fan (blue patterns) and compressor (red patterns) in the horizontal plane. The rotor axis extends along the 0/180° direction.

Summary

The current overview of emission characteristics of ASHPs revealed that, while there are various databases publishing total sound power levels, they are not necessarily comparable to each other and, within one database, rather allow for a comparison of different devices than for an assessment of the sound emission during operation. There is only a weak correlation with a large scatter between heating capacity and sound power level. This calls for noise limits for new devices to avoid the selection of unnecessarily loud devices (e.g., EU regulation 813/2013).[42] Overall, available data suggest that there is no such thing as a “typical” ASHP. The sound emission characteristics of a single HP depend on its components (fan, compressor), operating condition, and radiation direction, and further strongly vary between individual devices. All these strongly varying characteristics are expected to disparately affect human responses to HP noise.

HUMAN RESPONSE REVIEW: METHODS

This section details the methodological approach adopted for the systematic review on the noise effects of HPs on humans. The review adheres to the PRISMA guidelines.[23] Detailed search strategies for all databases, the full data extraction template, and the risk-of-bias (RoB) decision framework are provided in a public repository (see Availability of Data and Materials).

Research Question and Search Strategy

The research question to be answered in this review is the following: “Are individuals exposed to environmental noise from HPs at an increased risk of noise annoyance and/or for acquiring other non-auditory health effects?” To answer this question, we defined the eligibility of records following the PEO system (population, exposure, and outcome), for which the key criteria are listed in Table 2. The inclusion criteria are largely reflected by the keywords in the search string used for the database search. The search string was designed to identify a broad yet relevant body of literature by combining terms describing the noise source (HP), the domain of acoustics, and noise-related effects. To accommodate differences in database syntax and indexing, the exact search terms were adapted as necessary. A prototypical search string is shown in Table 3. The search was performed in two languages (English and German) to obtain a more comprehensive coverage of the subject matter. Thus, search terms were either translated manually or—where possible—translated automatically using translation and cognate detection features. Patents and meeting abstracts were excluded from the search records, as were documents that were not available in either English or German.

Table 2.

Key eligibility criteria for the selection of records for the systematic literature review on HP noise effects in humans.

Eligibility Criterion	Inclusion	Exclusion
Population	Human populations	Non-human populations (e.g., animals, calculation exercises, optimizations of HP or HP components, etc.)
Exposure	Residential HP noise	Not related to residential HPs (e.g., industrial facilities, cars, trains or other vehicles, water pumps, etc.)
Outcome	Noise effect (psychological, physiological, cognitive, behavioural, etc.)	No noise effect reported (e.g., only acoustic measurements); only qualitative reporting of noise effects (e.g., “people are annoyed by HPs”)

Table 3.

Prototypical search string used to identify records for the systematic literature review on noise effects of HPs on humans (adapted as required for each database).

Topic	Sub string
Source	“heat pump*” AND
Exposure	(“noise*” OR “sound*” OR “exposure*” OR “acoust*” OR “psychoacoust*” OR “loud*” OR “tone*” OR “tonal*” OR “sharpness” OR “roughness” OR “fluctuation strength”) AND
Outcome	(“annoy*” OR “disturb*” OR “bother*” OR “affective*” OR “preference*” OR “perception*” OR “psychol*” OR “psychi*” OR “health” OR “acceptance*” OR “noise problem” OR “community response” OR “complain*” OR “listening experiment” OR “human response” OR “acceptability”)

The literature search was conducted on 30–31 July and 5 August 2025, using the bibliographic databases Web of Science (Core Collection), Scopus, PubMed, PsycINFO, ProQuest Dissertations & Theses, GreenFILE, LIVIVO, and the TIB portal. Other information sources used to identify records were Google Scholar, OpenAlex, and Dimensions. Furthermore, the ICBEN proceedings (years 1973–2023) were searched manually using the keyword “heat pump”. Detailed search strategies for all databases are provided in a public repository (see Availability of Data and Materials for details and the access link).

The PRISMA flow diagram [Figure 5] outlines the number of retrieved records at each stage of the review process, as well as the number of and reasons for exclusions of records. A total of 501 records were initially identified as relevant for the present review, with 491 records from database searches and an additional 10 from organization or backward citation searches.

Figure 5.

PRISMA flow diagram for the systematic literature review on noise effects of HPs on humans in accordance with [43].

Data Extraction

The PRISMA flow diagram in [Figure 5] illustrates the record retrieval and screening process. All 491 citation entries retrieved during the database searches were imported into Rayyan (www.rayyan.ai) to remove duplicate entries (n = 160) and for subsequent screening. Duplicate detection was performed using Rayyan’s automated identification of potential duplicates, followed by manual verification of all flagged records. For each duplicate set, the most complete and final published version was retained. During the title and abstract screening stage, 252 of the 331 remaining entries were excluded. From the 79 entries sought for retrieval of full texts, 14 records could not be retrieved as the entries (i) did not constitute papers but meeting abstracts, (ii) were neither available in English nor German, or (iii) had no obtainable full texts. Full texts for the remaining 65 reports were retrieved and screened according to eligibility criteria (see Table 2), leaving 16 reports. Together with the 10 publications identified via organizations or citation searches, data were extracted from a total of 26 reports. Data extraction was performed using a customized, predefined template (for the link to the full template, see Availability of Data and Materials).

Quality Assessment

Given the heterogeneity of the included evidence (experimental, observational, predictive), no single established design-specific RoB tool could be applied uniformly across the full evidence base. The purpose of the RoB assessment in this review was to support a consistent cross-study appraisal within a single narrative synthesis spanning multiple evidence types, rather than to conduct separate design-specific quality appraisals. Applying different RoB tools to different study types was thus considered less suitable, as the resulting judgments would not have been directly comparable across the mixed-evidence pool. Therefore, RoB was assessed using a pragmatic, simplified four-domain framework (sampling, measurement validity, confounding, reporting), designed to capture major threats to validity across study designs. While the same four domains were applied across studies to preserve comparability, judgments were made with consideration of design-specific sources of bias (e.g., exposure controlled in experiments versus exposure assessed or misclassified in observational surveys; see Availability of Data and Materials for details of the assessment framework). Each domain was rated as low, moderate, or high RoB.

Overall RoB judgments were derived using a rule-based approach. Measurement validity and confounding were considered critical domains. Studies were classified as high RoB if at least one critical domain was judged high risk or if multiple important limitations were present across domains. Studies were classified as moderate RoB if no critical domain was high risk but some concerns were identified. Studies were classified as low RoB if the critical domains were rated as low risk and only minimal concerns were present in non-critical domains. Missing information in any of the domains was handled conservatively (i.e., resulting in higher RoB ratings). Detailed decision rules are provided in the repository (see Availability of Data and Materials).

For each noise-related outcome, quality of evidence was assessed based on key domains of the GRADE approach,[44] including assessment of RoB, consistency, precision, and directness. Assessment of publication bias was not feasible given the small number of studies contributing to each outcome and the heterogeneity of study designs, including observational studies, experimental studies, and gray literature sources.

General Overview

Included studies: The 26 included reports described 25 unique studies [Figure 5]. (Two reports included in this review described aspects of the same laboratory study.) The 25 studies covered 10 laboratory-based investigations, 14 observational survey studies, and 1 prognosis (prediction study). The surveys were mostly conducted online (five papers; 2017–25) or via mail (five papers; 1988–2021), with only two conducted by telephone interviews (2019–2023).

Time period: Although we did not set a lower bound on the publication year during the literature search, most of the retrieved papers were published within the last 15 years, in the 2010s and 2020s (namely, 23 of 26 reports between 2011 and 2025). This suggests a relatively recent research interest in the topic of noise effects of HPs on humans. Only three surveys were published earlier (between 1983 and 2001),[45,46,47] which raises the question as to how comparable the technology in the latter studies was in terms of acoustic emissions to the newer studies. Still, for the sake of comprehensiveness, these three studies were included in the current review.

Sources: ASHPs constituted the majority of the reported noise sources (namely, in 18 studies, both laboratory and field), mirroring their relevance in terms of market share and potential noise impact. While only one laboratory study investigated GSHP noise, GSHP users were included in a number of field surveys. Eight papers did not specify which type of HP was investigated.

HUMAN RESPONSE REVIEW: RESULTS

In the 26 retrieved reports, only a small number of papers reported on each health outcome (e.g., noise annoyance and sleep disturbance), as the overview in Table 4 illustrates. As this precluded quantitative analysis of the results (meta-analysis), the evidence is qualitatively synthesized and discussed in this section. A tabulation of the retrieved reports is presented in Table 5, summarizing key information on population, exposure, noise-related outcomes, and RoB assessments.

Table 4.

Overview of HP noise-related outcomes reported across the 25 unique studies examined in the systematic literature review on noise effects of HPs on humans.

Exposure	Noise-Related Effect/Outcome	Reporting Papers (n)*
	Noise complaints and concerns (HP audibility, satisfaction, and prognosis)	12
	HP adoption intention	1
Short-Term	Sound characterization/semantic differential	5
	Preference/unpleasantness Annoyance (absolute)	3 3
	Annoyance (relative)	2
	Cognitive performance	2
	Sleep disturbance; BP/HR; stress/mood; valence/arousal; other moderators	1
Long-Term	Annoyance (absolute)	2
	Difficulty concentrating	1
	Disturbed sleep/rest; medical and psychological symptoms	1

*Note that papers can report several outcomes. Thus, the sum of the reporting papers is higher than the number of records included in the human response literature review.

Table 5.

Characteristics of included studies in the systematic review, including study design, population, exposure, outcomes, measurement methods, and RoB.

Paper	Evidence Type	Study Design	Source Type	Region	Population	Exposure	Outcome(s)	Outcome Assessment	Risk of Bias
Acun et al.[48]	Experimental	Lab study	ASHP	UK	n = 50 (M: 35, F: 15), 19–57 y, mean 32.1 y	ST; LAeq,20s = 37–46 dB	Annoyance Valence Arousal	11-pt numerical scale	Moderate
Acun et al.[49] Note: Same study as [48]	See [48]	See [48]	See [48]	See [48]	Sub sample of [48] n = 35 [M: 24, F: 11] 19–58 y, mean 31.6 y	See [48]	See [48]	See [48]	Moderate
Benz et al.[52]	Experimental	Lab sleep study	ASHP	Germany	n = 40 (M: 19, F: 21), mean 37.5 y	ST; LAeq = 30 dB (“window tilted”)	Polysomnography Subjective sleep quality	Various measures	Low
		Lab daytime study				ST; LAeq,1h = 30 dB (“window tilted”)	Annoyance Cognitive performanceMood	5-pt verbal scale
							Sound characterization	7-pt semantic differential
							BP, HR	Physiological measures
Di et al.[60]	Experimental	Lab study	HP (+ added sounds)	China	n = 24 (M: 12, F: 12), 20–25 y	ST; LAeq,5s = 45–47 dB	Rel. annoyance	Paired comparison test	Moderate
Hellgren and Kaseß[62]	Experimental	Multiple lab studies (separately listed)	ASHP	Austria, Sweden	Austria: n = 20 (F: 10), mean 29.7 y Sweden: n = 20 (F: 10), mean 46 y	ST; LAeq,5s = 40 dB (equalized)	Rel. annoyance	Free magnitude estimation	Moderate
			ASHP	Not stated	n = 20 (F: 10), mean 28.6 y	ST; LAeq,5s = 37–44 dB	Rel. annoyance	Free magnitude estimation
			GSHP	Not stated	n = 14 (M: 10, F: 4), mean 40 y	ST; LAeq,3s = 42 dB (equalized)	Dissimilarity Preference	Scale unspecified Binary response
Feldmann and Carolus[63]	Experimental	Lab study	HP+fan-related	Germany	n = 20 (M: 10, F: 10), mean 23 y	Not stated	Sound characterization	7-pt semantic differential	Moderate
Feldmann et al.[51]	Experimental	Lab study + written interview	ASHP	Germany	n = 40 (M: 19, F: 21), median 23 y	ST; LAeq,5s = 55 dB*	Sound characterization	7-pt semantic differential	Moderate
Feldmann et al.[59]	Experimental	Lab study	HP+fan-related	Germany	n = 51 (M: 35, F: 16), 18–56 y, median 23 y	Not stated	Sound characterization	5-pt semantic differential	Moderate
Schneider and Feldmann[58]	Experimental	Lab study	HP	Germany	Group A: n = 20 (50% male), mean 28 y Group B: n = 20 (45% male), mean 25 y	ST; LAeq,10s = 55 dB	Unpleasantness	Paired comparison test	Moderate
Stürenburg et al.[61]	Experimental	Lab study	ASHP	Germany	n = 20 (M: 8, F: 12), 20–34 y, mean 25 y	ST; SPLStart = 60 dBA	Loudness, Preference	2AFC (white noise reference)	Moderate
Wang and Novak[50]	Experimental	Lab study	ASHP	USA	n = 30 (M: 15, F: 15), 19–61 y, mean 22.6 y	ST; HP: LAeq,55min = 44 dB	Objective task performance, Annoyance, Distraction, Sound characterizations	Performance score 7-pt numerical scale	Moderate
Caird et al.[66]	Observational	Postal survey	ASHP, GSHP	UK	n = 78 (M: 59, F: 19); 20.5% 25–44 y, 52.1% 45–64 y, 27.4% ≥ 65 y	Not stated	HP satisfaction	3-pt verbal scale	High
Department for Energy Security and Net Zero[12]	Observational	Online survey	ASHP	UK	n = 139	Not stated	HP audibility	Binary/verbal scale	High
		Telephone Interview			n = 2
Ghorbankhani[72]	Observational	Online survey	HP	Italy	n = 45 (M: 57.8%, F: 42.2%), 2 exclusions	Not stated	Noise concerns	Binary response	High
Gupta and Morey[74]	Observational	Survey (telephone, repeated)	ASHP	UK	n = 17 households	Not stated	Noise complaints	Binary response	High
Jagenäs[47]	Observational	Survey	ASHP, GSHP	Sweden	N = 279 (respondents n unknown)	Not stated	Noise complaints	Count	High
Lackova and Dinnie[67]	Observational	Postal survey (longitudinal)	ASHP	UK	n(Y1) = 120 n(Y2) = 91	Not stated	HP satisfaction	5-pt verbal scale	High
Lamacraft et al.[65]	Observational	Survey	ASHP	UK	n = 121	Not stated	Noise complaints	Count	High
Martens et al.[71]	Observational	Online survey	ASHP, GSHP	Belgium	n = 692 (M: 335, F: 357), mean 55.03 y	Not stated	HP adoption intention	5-pt verbal scale	Moderate
		Semi-structured interview (in-person)			n = 16 (M: 8, F: 8), mean 41 y			Open-ended questions
Persson and Rylander[45]	Observational	Postal survey	Indoor/Outdoor noise	Sweden	n = 279	LT; Indoors: LAeq,3min = 24–36 dB Outdoors: LAeq,24h = 44–49 dB	Annoyance, Sleep disturbance, Rest/Concentration disturbance	4-pt verbal scale	Moderate
Persson and Rylander[46]	Observational	Postal survey	HP+LF	Sweden	n = 233 (environmental health authorities)	Not stated	Noise complaints	Count	High
Sweetnam et al.[73]	Observational	Survey (in-person, longitudinal)	ASHP, GSHP	UK	n = 12	Not stated	Noise complaints	Binary response	High
Taylor et al.[69]	Observational	Postal/online survey	HP (ASHP, GSHP), GB	UK	n(HP) = 2395, n(GB) = 1025	Not stated	HP/GB satisfaction	4-pt verbal scale	High
UK Department for Energy and Climate Change[68]	Observational	Online survey	ASHP, GSHP	UK	n = 13,492	Not stated	HP satisfaction	7-pt verbal scale	High
Wang and Norbäck[57]	Observational	Postal survey	HP/Fan/Ventilation; other sources	Sweden	n = 5775 (M: 2483, F: 3219), 63.5% 18–65 y, 36.5% > 65 y	Not stated	Noise disturbance	5-pt verbal scale	Moderate
Eulitz et al.[70]	Predictive	Prognosis	ASHP+ other stationary sources	Germany	German residents	Modeled LF noise	Conflict potential	Threshold-based modeling	High

Note: ST: short-term; LT: long-term; ASHP: air-source heat pump; GSHP: ground-source heat pump; M = male; F = female; HP: heat pump; SR = self-reported; 2AFC = two-alternative forced choice task; LF = low-frequency; BP = blood pressure; HR = heart rate; GB: gas boiler; N = target population (i.e., all authorities receiving a questionnaire). * Approximation due to incomplete reporting in the publication.

Across the included literature, outcomes vary not only in type but also in their conceptual nature. While some studies assess established noise-related health outcomes (e.g., annoyance, sleep disturbance, and cognitive effects), others report measures such as noise complaints, concerns, satisfaction, or adoption intention. These latter constructs are not formal health endpoints but rather represent context-specific indicators of perceived impact, acceptance, or behavioral response. As such, they are interpreted as complementary indicators that inform the interpretation of perceived effects but do not constitute direct evidence of health outcomes.

Heterogeneity in study designs poses an additional challenge for the synthesis of evidence on human responses to HP noise. The included literature comprises controlled laboratory experiments, observational survey-based studies, and predictive studies, which differ substantially in their degree of exposure control, outcome assessment, susceptibility to bias, and external validity. Consequently, findings are not directly comparable across designs, even when the same outcomes are addressed. Table 6 therefore provides a structured overview, based on the reviewed literature, of typical methodological approaches (exposure and outcome measures) across evidence types and highlights their key strengths and limitations, including potential sources of bias and their contribution to the overall evidence base. These considerations should be taken into account when interpreting the synthesized results.

Table 6.

Framework for interpreting evidence types, typical outcomes, and methodological strengths and limitations in the reviewed HP noise literature.

Evidence Type	Typical Study Designs	Typical Exposure	Typical Outcomes	Key Strengths	Potential Sources of Bias	Contribution to Evidence Base
Experimental	Laboratory experiments or controlled field experiments	Usually controlled short-term exposure from synthetic sound sources (playback)	Subjective (e.g., annoyance ratings and perceptual judgments); Objective (e.g., cognitive performance and short-term physiological responses)	Strong internal validity; Allows comparisons under factor manipulations; Useful for isolating causal effects of acoustic features	Limited ecological validity due to short exposure durations, artificial listening context, small samples; inference to long-term residential exposure difficult	Helps identify potentially salient acoustic characteristics and perceptual mechanisms
Observational	Cross-sectional or longitudinal studies conducted in real-world settings	Limited control over exposure; often based on estimated, self-reported, or context-dependent exposure	Annoyance, sleep disturbance, self-reported health outcomes, complaints/concerns, attitudinal responses (adoption intention, satisfaction), etc.	Higher ecological validity; captures real-life residential context and multi-source exposure	Inconsistent outcome measures; exposure misclassification and confounding; recall, self-selection, and response bias; limited control over effect modifiers	Helps assess real-world relevance by capturing residential context, cumulative (long-term) exposure, and a broad range of self-reported responses
Predictive	Acoustic simulations, scenario-based comparisons, and prediction studies	Model estimations or scenario-based calculations	Usually acoustic indicators or predictions of (integral) noise effects	Useful for comparing installation scenarios, future developments, propagation conditions, or mitigation measures; can cover scenarios not yet studied empirically	Does not measure human responses; depends on modelling assumptions and input quality; not suited for causal inference on health effects	Helps understand exposure determinants and supports planning/risk assessment, should not be interpreted as direct health evidence

Noise Annoyance

In terms of short-term annoyance of HP sounds, four laboratory studies reported absolute mean ratings in some form (as [48,49] are two reports of the same experiment, they count as one “unique” study,[50,51,52]; other studies with relative—i.e., comparative—ratings are discussed further below). However, methodologically, the studies were very heterogeneous, using different rating scales (e.g., 5-pt verbal and 11-pt numerical), exposure durations (from several seconds to 1 h), and experimental contexts (e.g., presenting various HP and/or HVAC stimuli). Overall, A-weighted equivalent continuous sound pressure levels (L_Aeq) of the presented stimuli covered a range of 30–55 dBA, and corresponding mean annoyance ratings were moderately high, with 3.2–6.8 when linearly scaled to an 11-point numerical scale (0 = not annoying and 10 = extremely annoying).[53] [Figure 6] tentatively compiles these heterogeneously obtained results for short-term annoyance for illustrative purposes. Note, however, that due to the methodological differences, the compilation is afflicted with increased uncertainty and does not constitute a formal exposure–response synthesis. [Figure 6] shows generally relatively low SPLs for HPs, which is in line with the findings from the noise emission database [Figure 1]. Nevertheless, they appear to have a moderate to high annoyance potential. None of the studies retrieved in the noise effects literature search compared HPs with other environmental noise sources, making it difficult to assess the relative impact of HP noise. To provide a reference point for the relative annoyance potential of HPs, [Figure 6] also includes mean annoyance ratings for road traffic (RT) noise from a laboratory study.[54] The comparison of HP versus RT noise annoyance tentatively suggests that HPs may possess an increased annoyance potential compared with RT noise at equal SPL, depending on the respective scenario, which might be quantified with a psychoacoustic sound pressure level difference (ΔL; see, e.g.,[55]. At the same time, the figure also indicates that HP noise can yield lower annoyance ratings than RT noise at comparable SPL. However, this comparison should be interpreted with caution for two reasons: First, the RT reference in [Figure 6] is based on a single study, but annoyance responses to RT noise are known to vary depending on scenario characteristics such as the presence of quiet periods.[56] Second, the stimuli used in [51] (55 dB L_Aeq) are notably higher than levels typically expected for ASHPs in residential settings under representative conditions.

Figure 6.

Mean short-term annoyance for HP noise as a function of reported equivalent continuous sound pressure levels (L_Aeq) from four laboratory studies,[48,50,51,52] linearly rescaled to 0–10, and for RT noise from Schäffer et al.[54] for comparative purposes. Due to heterogeneity across studies and rescaling of outcome measures, this depiction is illustrative only and does not represent a formal exposure–response synthesis.

The quality of evidence for short-term noise annoyance was rated as very low. Although studies were generally at low to moderate RoB, indirectness was present due to the use of non-standardized and relative annoyance measures in several studies, limiting comparability and applicability to absolute annoyance outcomes. Accordingly, the available findings should be interpreted as indicating a possible annoyance signal rather than providing a reliable estimate of effect magnitude or exposure-response shape.

Long-term annoyance (not displayed in [Figure 6] is only assessed in two studies.[45,57] Noise exposure was assessed only in [45] and termed low-frequency (LF) noise by measuring in- and outdoor SPLs for a subsample of survey respondents. Both studies are methodologically disparate but suggest that a non-negligible proportion of residents are affected at relatively low SPLs. For example, Persson Waye and Rylander[45] report 14.7% annoyed in the “moderately” LF-exposed (27–33 dBA/49–50 dBC, inside), and 20% annoyed in the “highly” LF-exposed group (33–36 dBA/52–60 dBC, inside) group of residents. (Note that this study was conducted in 2001; see the “General Overview” section.)

The quality of evidence for long-term annoyance was rated as very low. This reflects serious concerns regarding imprecision (i.e., a small number of studies) as well as indirectness, as exposure was not consistently or directly assessed, with only one study providing approximate SPL measurements for a subsample.

Taken together, the reported annoyance potential at relatively low exposure suggests that the perception and evaluation of HP noise are captured less by its (A-weighted) SPL but to a large extent by its sound characteristics, for which the studies in the following section provide additional evidence.

Acoustic characteristics linked to noise annoyance: When looking at how HP sound is described, adjectives such as hissing (related to the fan), humming (related to the compressor), bellowing (suggestive of amplitude-modulated LF sound), or whistling (suggestive of tonal components) are repeatedly reported, which are negatively evaluated and found to be correlated with annoyance.[51,52,58] [Figure 3] illustrates the acoustic contributions of typical ASHP components: the LF and tonal content of the compressor and the broadband noise-like sound of the fan. However, there are also instances where HPs of a dark and low sound quality were rated as neutral[59] or not unfavourable[51] in semantic differential studies. Strong low-frequency content as well as tonal components with slow, periodic frequency modulation were shown to be unfavourable sound characteristics in laboratory studies.[60] Similarly, some studies investigated the effect of different HP acoustical characteristics on annoyance or preference ratings by changing the HP’s operating condition (i.e., compressor-fan configurations such as defrosting).[48,49,61,62] While annoyance was moderately associated with loudness or A-weighted SPLs in two studies,[48,62] low-loudness stimuli were among the least preferred in another study,[61] suggesting that loudness metrics alone may be insufficient predictors. This is also supported by Wang and Novak,[50] who found one HP stimulus with strong temporal fluctuation (repeatedly cycling on–off) but the second lowest loudness among the six stimuli to receive the second highest subjective loudness rating. Across studies, while the investigations into sound characteristics do not paint a very homogeneous picture, important psychoacoustic predictors appear to be sharpness, tonality, and roughness,[48,61,62] suggesting that spectral content as well as rapid amplitude modulations affect noise annoyance and preference.

Further influencing factors: Besides the above characteristics, HP placement and emission direction may affect noise annoyance[62] through modifying HP noise exposure. This is supported by [Figure 4], which illustrates the pronounced directivity of the fan. Furthermore, background noise levels were found to affect the perceptual prominence of HP sound, with lower background levels (L_Aeq = 31.5 dB versus 39.5 dB) being associated with lower masking and corresponding to higher affective arousal ratings.[48] Finally, extra-acoustic context may affect the evaluation of HP noise,[49,63] as is well known for other noise sources (e.g.,[64]. However, the two studies dealing with this topic showed substantial methodological shortcomings, as their experimental operationalizations were confounded (ownership, setting, and source type)[63] and/or appropriate control conditions were lacking,[49] limiting their ability to draw valid conclusions about contextual effects.

Acoustic characteristics and contextual modifiers were not formally graded using the GRADE approach, as they provide psychoacoustic and mechanistic insights rather than direct evidence on health-related outcomes.

Cognitive Performance, Sleep Disturbance, and Other Psychophysiological Effects

Further effects of HP noise on humans are very sparse and come from three isolated studies, namely two laboratory studies on short-term effects,[50,52] and one survey on various self-reported long-term effects.[45]

A negative association between HP noise exposure and self-reported cognitive outcomes (depending on the study, denoted as “cognitive performance,” “distraction,” or “disturbance of concentration”) was reported in all three studies. Here, correlation analysis showed “droning” and “not possible to fade out” characteristics to be most associated with increased difficulty to concentrate.[52] Objective cognitive task performance was only assessed in one paper,[50] where the authors reported no significant difference in the HP condition compared with other mechanical noise conditions, but pointed toward a possible negative contribution of LF spectral content and strong timescale fluctuations (HP cycling on/off) on executing cognitively demanding tasks. Overall, the certainty of evidence for cognitive outcomes was rated as very low, due to indirectness and imprecision arising from the limited number of studies, heterogeneous outcome measurements, and limited objective cognitive data.

The scarce evidence on sleep disturbance suggests a possible link between HP noise and increased self-reported sleep disturbance and morning sleepiness, even at relatively low indoor sound levels.[45,52] A sleep laboratory study found a higher number of physiological arousals related to HP noise under simulated “windows-tilted” scenarios at an L_Aeq,1h of 30.4 dBA, compared with “windows closed”.[52] In the field survey,[45] long-term exposure to combined HP/ventilation noise was found to be associated with self-reported disturbed rest. Moreover, the relationship between sleep disturbance and noise exposure was mediated by annoyance, with residents being annoyed by HP/ventilation noise inside their homes also reporting a higher frequency of difficulty falling asleep, besides other medical and psychosocial symptoms such as headaches, irritation, tiredness, and tenseness in the morning.[45] The certainty of evidence for sleep-related outcomes was rated as very low, limited by small study numbers and indirectness related to heterogeneous exposure assessment.

Regarding physiological reactions, no robust short-term effects of HP noise on measures such as heart rate or blood pressure were found under controlled laboratory conditions.[52] However, certainty of evidence was rated as very low due to imprecision (i.e., a single study).

Taken together, these findings point to a possible adverse effect on concentration and sleep-related outcomes, but the current evidence base is too limited to establish consistent effect sizes, thresholds, or causal pathways.

Noise Complaints and Noise Concerns

Several surveys seek to gauge the impact of HP noise by examining the number of noise complaints or by inquiring about noise-related concerns among residents or HP owners. However, neither complaints nor concerns are formally classified as noise (health) effects; rather, they provide contextual evidence on the perceived impact and acceptability of HP noise in real-world settings and can be interpreted as indicators of perceived annoyance or disturbance, and thus as proxies for noise effects more generally. Thus, noise complaints and concern-related findings were used qualitatively to complement the synthesis of noise-related outcomes but were not subjected to formal GRADE assessment.

Early evidence from the 1980s gathered in Sweden shows low numbers of complaints due to HP noise.[46,47] Despite accounting for a small share (∼9%) of LF noise complaints, authorities have already noted a disproportionately strong increase in complaints over time, especially in quiet, sparsely populated areas with limited background masking.[46] However, the transferability of these findings to modern HP technology is questionable.

Recent UK evidence suggests that HP noise is a concern, though not a widespread one. While complaints were raised, only a minority resulted in legal enforcement.[65] Here, identified problematic acoustical characteristics include tonal components, intermittency, roughness, and nighttime audibility,[65] which are in line with the literature discussed in the “Noise Annoyance” section. Several surveys conducted among HP owners in the UK indicated high overall satisfaction with noise levels (typically 75%–85%), but a non-negligible minority (15%–45%, depending on the study and context) reported intrusive or annoying noise, particularly for ASHPs.[12,66,67,68,69] Interestingly, noise complaints and dissatisfaction were found to be disproportionately higher in social housing and large-scale installations,[65,66] pointing to risks from cumulative noise impacts in residential areas as HP density increases and background sound levels gradually rise (described as “noise creep”). In fact, a 2020 prognosis for Germany predicts that the conflict potential due to LF noise in residential areas will increase strongly in the future with widespread ASHP adoption.[70]

Even if it is not the main outcome of interest, noise concerns surface as potential hindrances in studies assessing attitudes and adoption intentions toward HPs,[71,72] or even smart-grid technologies;[73,74] however, noise concerns are deemed as “minor” challenges in these sectors at the moment.

Overall, the above reports and reviews suggest that HP noise has caused some complaints and noise concerns, but the number still seems to be relatively low—at least as of 2025. However, as complaints are not a standardized health outcome, the reviewed literature does not allow a quantification of the HP noise impact, even more so since most surveys studied only single-source noise complaints (HPs) rather than comparisons with other environmental noise sources and did not assess the underlying noise exposure. This prevents the assessment of the current relative impact of HPs compared with other noise sources.

DISCUSSION

To the best of our knowledge, the current systematic literature review provides the first comprehensive overview of noise emission characteristics of HPs and related noise effects on humans. The following synthesis integrates the results of the reviewed studies and summarizes the main findings and highlights research gaps.

Synthesis

The database and literature search on noise emission characteristics revealed that there is no “typical” ASHP noise, as the acoustical characteristics depend on the device components (fan and compressor), operating condition, and radiation direction, and further strongly vary between individual devices. The literature review on noise effects identified that the number of studies is still quite limited. These studies reported signals for several potential psychological, cognitive, and physiological effects of HP noise; however, these findings were based on few methodologically heterogeneous studies, and the certainty of evidence was very low across outcomes. A rating of “very low certainty” means that the true effect may be substantially different from the observed findings, and conclusions should therefore be regarded as hypothesis-generating rather than confirmatory. Despite these limitations, the available literature provides the following insights.

Unsurprisingly, the studies demonstrate that both short-term and long-term noise annoyance from HPs increase with SPL;[45,48,49] in this respect, HPs do not differ from other environmental noise sources. Although HPs typically generate relatively low SPLs, the literature suggests a non-negligible potential for adverse effects, particularly psychological outcomes such as annoyance, perceived disturbed concentration, or self-reported sleep disturbance. Taken together, the available laboratory findings raise the hypothesis that some HP noise scenarios may be perceived as more annoying than road traffic noise at comparable SPLs [Figure 6]; however, this inference is indirect, based on limited and non-comparable evidence, and requires direct comparative validation. At present, empirical evidence remains too scarce to reliably estimate specific level corrections to account for an elevated annoyance potential of HP noise, particularly as direct comparative studies between HP noise and other environmental noise sources, such as RT, are still missing.

Similarly, laboratory findings further suggest that A-weighted SPLs are not sufficient to fully explain noise annoyance. Sound characteristics (spectral content, tonality, and roughness) also contribute to annoyance, as reflected by the negatively connoted semantic descriptions of HP sounds as “hissing,” “whistling,” or “bellowing.” Thus, the A-weighted SPL might not be sufficient as a sole indicator for threshold determination,[46,61] and correction terms might apply, as already accounted for in noise legislation (e.g., Swiss Noise Abatement Ordinance NAO).[75] Beyond acoustical characteristics, personal and situational factors such as ownership can affect the perceptual impact of HP noise and may help to explain negative effects and/or noise complaints even at comparatively low levels.

Looking ahead, a policy tension becomes apparent between the expansion of climate-friendly technologies such as ASHPs on the one hand and concerns regarding cumulative noise exposure (“noise creep”) and qualitative soundscape changes on the other. While acoustical modelling might not necessarily suggest a significant increase in background noise levels in ASHP-rich neighbourhoods,[76] even low-level sounds may become perceptually prominent, especially in quiet residential areas. A future shift in vehicle fleets—from internal combustion engines toward battery electric vehicles—may also alter the spectral composition of road traffic noise (less LF), potentially reducing low-frequency masking.[48] In such a scenario, stationary LF-dominant (and possibly tonally modulated) HP noise could become perceptually even more salient, increasing the annoyance potential (see [77].

Directions for Future Research on Noise Effects

With only a limited amount of literature available to date, there is a substantial need for further research on short- and long-term effects of HP noise exposure on humans. The literature review identified the following “gaps” in HP noise research that need to be closed in the future:

First, as no studies exist as of yet that compare the effects of HP noise on humans with other major (reference) environmental noise sources, such as road traffic, comparative baselines for policymakers are missing. For completeness, we want to cite here one conference paper identified during the literature search on noise emission characteristics, reporting laboratory results in which participants ranked HP sounds as less annoying than fountain, rain, children, and fan sounds.[78]

Second, annoyance is a central concept in noise effect research[79] and a major health effect.[80] The review has shown that HP-noise-related health effects, like sleep disturbance, can be mediated by annoyance. Considering annoyance as an additional outcome can thus make sense for noise impact researchers in general. In addition, studies on cognitive and (psycho)physiological effects are urgently needed as well.

Third, while some laboratory studies are available, studies on long-term effects of HP noise exposure, as well as the cumulative effects of HPs, are still largely lacking, with evidence for the former being over 25 years old and no evidence available for the latter.

Fourth, systematic studies on the influence of ASHP design and different operating states (e.g., defrosting), as well as their contribution to overall sound perception, are still lacking. In addition, the extent to which building type (detached houses versus multi-unit apartment buildings) and installation context (e.g., garden, façade, or balcony) impact residents’ responses (among others, through sound propagation) remains largely unexplored, limiting the generalizability of existing findings. Future research should therefore examine these contextual factors systematically.

Fifth, the currently available evidence is of limited comparability across studies. Several studies, even when investigating the same outcome (e.g., annoyance), used different rating scales (e.g., 4-point verbal and 7-point numerical), experimental paradigms (e.g., absolute versus relative ratings), exposure durations, and statistical approaches. It would therefore be beneficial for future studies to adhere to established standards (e.g.,[53] for the assessment of annoyance), and standardized acoustic and psychoacoustic metrics should be reported alongside health outcomes (e.g., L_Aeq, L_den, tonality, etc.) when establishing exposure-response relationships. Importantly, given that HPs often exhibit pronounced low-frequency components, which may strongly affect noise annoyance[81] but are insufficiently captured by A-weighted metrics, complementary weighting approaches (e.g., C-weighting or level differences such as L_Ceq–L_Aeq) should be included to better reflect potential annoyance and perceptual impacts. This point regarding A-weighted metrics was previously discussed in research on HP noise effects.[46,61]

Lastly, evidence from context-effect research remains scarce and inconclusive, necessitating well-controlled studies. For example, how far ownership biases, noise-source information, or possible audiovisual interactions in various situational contexts affect HP noise perception remains to be thoroughly explored.

Limitations

Literature: At both the study and outcome level, the systematic review revealed substantial heterogeneity among the included studies. Reporting was often incomplete, particularly regarding acoustic stimulus characteristics, measurement procedures, and statistical analyses. In addition, several factors that may influence the perception of HP noise remain largely unexplored, including installation context and cumulative exposure, which limit the generalizability of the available findings. As a result, the RoB for studies on HP noise effects is generally considered to be high, mainly due to insufficient methodological information. This lack of information may partly be explained by the fact that 23% of the contributions included in the noise effects review were conference papers (n = 6), where brevity of reporting is often encouraged.

Review: A limitation of the presented noise effect review is that the search was restricted to the specific keyword “heat pump” and did not include broader terms such as “HVAC.” As a result, some studies in which HPs were discussed under more general labels (e.g., HVAC technology or low-frequency noise sources) may have been missed. However, this choice was intentional to keep the review focused on HP technology. Although HVAC systems share technical and acoustical similarities, different technologies (e.g., refrigerators, industrial fans, vehicle air conditioning, HPs) differ in their typical sound characteristics, sound levels, and usage contexts. This limits comparability across systems and may reduce the relevance of such comparisons for the currently highly relevant issue of HP noise impact.

A further limitation concerns the pragmatic simplified cross-design RoB framework used in this review. While it enabled a consistent appraisal across heterogeneous evidence types, it is less granular than design-specific tools and may not fully capture biases unique to individual designs. In studies with sparse methodological reporting (particularly conference papers and government or technical reports), conservative judgments (pointing toward increased RoB) may therefore reflect incomplete methodological information rather than definite flaws in study conduct. The overview on noise emission characteristics was conducted as a scoping overview of acoustic properties of HPs, prioritizing breadth and technical characterization over exhaustive coverage. A formal RoB assessment was therefore not implemented.

CONCLUSION

For this study, two reviews were performed to establish an overview of (i) noise emission characteristics and (ii) effects on humans associated with HP noise. The overview of emission characteristics indicates that HP emissions vary strongly by components, operating conditions, radiation direction, and device. These differences likely also lead to disparate effects on humans depending on the situation. The noise effect review indicates elevated annoyance potential of HPs even at relatively low sound pressure levels, hinting toward a higher annoyance potential at comparable noise exposure than that of other environmental noise sources such as road traffic, but such hypotheses require direct comparative validation. Given the typically LF, at times fluctuating and/or tonal, noise emissions of HPs, A-weighted levels alone may underestimate their perceptual impact. However, the overall certainty of evidence is very low due to the limited number of short-term and only very few long-term studies, demonstrating the need for future research.

Availability of Data and Materials

No primary data were generated for this study. Detailed methodological materials supporting the systematic review process (including full search strategies, data extraction templates, and risk-of-bias assessment procedures) are available in a public repository at https://doi.org/10.5281/zenodo.20828501.

Author Contributions

Claudia Kawai: Conceptualization, Methodology, Formal analysis, Investigation, Data Curation, and Writing—Original draft. Urs Pachale: Conceptualization, Methodology, Formal analysis, Investigation, Data Curation, and Writing—Original draft. Reto Pieren: Visualization, Investigation, Formal analysis, and Writing—Reviewing and Editing. Beat Schäffer: Investigation, Supervision, Project administration, Funding acquisition, and Writing—Reviewing and Editing.

Ethics Approval and Consent to Participate

Not applicable. This study synthesizes previously published data and does not involve new human participants or animals; therefore, ethical approval and informed consent were not required.

Conflicts of Interest

The authors have no conflicts of interest.

Funding Statement

This work was supported by the Swiss Federal Office for the Environment (FOEN, Assignment No. 5211.02518.100.01, Contracts No. 22.0001.PJ/0CE07F745 and 22.0001.PJ/B8CED6295).