Strategic Integration of a Vegetative Component on a Metal Roof Base: An Evaluation of Its Impacts on Thermal and Acoustic Performance in the Tropics

Abstract

This paper attempts to ascertain the thermal and acoustic impacts of introducing a vegetative roof layer on insulated and uninsulated metal roofs for tropical climates, through field measurements in Skudai, Johor, Malaysia, that were conducted for both dry and wet days. Four small-scale roof modules were tested, namely an uninsulated metal roof (uiMDR), an insulated metal roof (iMDR), and two identical corresponding modules with an additional vegetative component (uiGR and iGR, respectively). Outdoor ambient temperature (T_amb) was the most influential correlated variable affecting the roof outer surface temperature (R_TOS) in 50% of the assessed scenarios. On the selected dry day, the inter-quartile ranges (IQR) of iGR, iMDR, uiGR, and uiMDR were 6.21 °C, 8.32 °C, 6.69 °C, and 1.66 °C, respectively; the IQRs were 1.6 °C, 4.11 °C, 2.59 °C, and 1.78 °C, respectively, on the selected wet day. Based on design U-value calculations, iGR was better than iMDR and uiMDR for both dry and wet days. The U-value of uiGR was also better than iMDR under dry-day conditions. The Wilcoxon signed-rank test also indicated a statistically significant difference in the roof inner surface temperature (R_TIS) measurements (p-value = 0.00) during Malaysian daylight hours, between 8.00 a.m. and 6.00 p.m., regardless of the weather. In terms of sound level reduction under dry-day conditions, the Kruskal–Wallis and Wilcoxon signed-rank tests showed statistically significant differences in sound level reductions, with iGR and uiGR performing better than iMDR and uiMDR (p-values = 0.00). The sound level reductions for iGR, iMDR, and uiGR were 51%, 32%, and 31%, respectively, while uiMDR experienced sound level amplifications by 6%, possibly due to the acoustic resonance effect. This proof of concept may encourage a broader application of extensive GRs in Malaysia using metal roofs, beyond the conventional RC base construction method.

1. Introduction

In recent decades, Malaysia has seen an increased interest in the thermal comfort performance of buildings to mitigate rising global temperatures and the urban heat island effect in cities. The green roof (GR), also known as a living roof, eco-roof, roof garden, or vegetated roof, is one green building strategy that addresses the above-mentioned concerns. It also provides various benefits, such as improved thermal and acoustic performance [1,2], increased biodiversity [2], aesthetics [1], and better retention of surface runoff [1,3]. Depending on their composition, GRs are capable of reducing indoor temperatures [3], heat gain on roof surfaces [3,4], and sound transmission through the roof [1].

An analysis of the assessment methods employed by past studies for the thermal performance of roofs, described in Table 1, suggests that field measurement is a viable assessment method for various roof types and climates. Studies on acoustic performance, however, tend to be conducted in laboratory-based environments [5,6,7], such as a reverberation chamber; field measurements are uncommon, although they have been used to assess noise exposure through windows [8] and roofs [9]. While laboratory tests may indicate the acoustic performance of the roof under ideal, controlled conditions, field measurements may verify its performance under more realistic, non-isolated conditions.

Furthermore, most thermal performance studies report their findings based on averages, such that the results are weather-agnostic. However, Malaysia is a hot and humid tropical country along the equator that experiences between 150 and 200 days of wet days a year [10], with an annual mean monthly rainfall amount between 1500 mm and 4000 mm [11]. GRs have been ascertained to perform differently when the substrate layer is water-logged, which occurs when the roof is exposed to rain [12,13]. As Malaysia experiences between 41% and 55% days with rainfall per annum, this study addresses a research gap in tropical GR studies by separately analysing the thermal performance of roofs on both wet (rainy) and dry (non-rainy) days.

Table 1. Past research methods and elements of assessment, along with main performance indicators and climate of interest.

Discussion Aspect	Reference	Assessed Element(s)	Climate	Method(s) of Assessment	Main Performance Indicators
Thermal	[14]	Green roof system; RC roof system	Tropical	Simulations	Roof thermal transfer Value (RTTV) and overall thermal transfer value (OTTV)
	[4]	Green roof plant species	Tropical	Field measurement	Surface and ambient temperature (°C)
	[15]	Green roof system	Tropical	Field measurement, interview, and case study	Roof surface temperature, analysis of interviewees’ responses
	[16]	Green roof system with mineral wool substrate	Sub-tropical	Field measurement	Roof surface temperature, reduction in conductive heat flow, and delayed heat transfer
	[17]	Intensive and extensive green roof systems	Sub-tropical	Field measurement	Roof surface temperature, roof heat flux, and cooling/heating load
	[18]	Green roof system	Irrelevant	Simulations	Energy performance of simulated green roofs using leaf area index, plant height, and soil thickness
	[19]	Extensive green roof system	Sub-tropical	Simulations and field measurement	Maximum and minimum indoor temperature
	[20]	Extensive green roof system	Tropical	Field measurement	Roof surface temperature, ambient temperature, and temperature reduction
Acoustic	[8]	Windows	Temperate	Field measurement	Sound level differences between outdoors and indoors
	[9]	Green roof system	Temperate	Field measurement	Sound transmission loss of roof system

Table 1. Past research methods and elements of assessment, along with main performance indicators and climate of interest.

Discussion Aspect	Reference	Assessed Element(s)	Climate	Method(s) of Assessment	Main Performance Indicators
Thermal	[14]	Green roof system; RC roof system	Tropical	Simulations	Roof thermal transfer Value (RTTV) and overall thermal transfer value (OTTV)
	[4]	Green roof plant species	Tropical	Field measurement	Surface and ambient temperature (°C)
	[15]	Green roof system	Tropical	Field measurement, interview, and case study	Roof surface temperature, analysis of interviewees’ responses
	[16]	Green roof system with mineral wool substrate	Sub-tropical	Field measurement	Roof surface temperature, reduction in conductive heat flow, and delayed heat transfer
	[17]	Intensive and extensive green roof systems	Sub-tropical	Field measurement	Roof surface temperature, roof heat flux, and cooling/heating load
	[18]	Green roof system	Irrelevant	Simulations	Energy performance of simulated green roofs using leaf area index, plant height, and soil thickness
	[19]	Extensive green roof system	Sub-tropical	Simulations and field measurement	Maximum and minimum indoor temperature
	[20]	Extensive green roof system	Tropical	Field measurement	Roof surface temperature, ambient temperature, and temperature reduction
Acoustic	[8]	Windows	Temperate	Field measurement	Sound level differences between outdoors and indoors
	[9]	Green roof system	Temperate	Field measurement	Sound transmission loss of roof system

Generally, GRs can be classified as either extensive or intensive. Extensive GRs are comparatively light-weight, weighing 80–150 kg/m², and comprise a 50–150 mm layer of soil, little or no irrigation systems, and limited groundcovers or shrub species, allowing them to be installed on sloped roofs. Intensive GRs may weigh between 300 and 1000 kg/m² and comprise a 200–2000 mm layer of soil, irrigation systems, and a wider range of shrubs and trees. For cost-effectiveness or ease of maintenance, extensive GRs are advantageous [2].

The adoption of GRs has become common in many developing and developed countries [21]. Among ASEAN countries, Singapore and Malaysia have made significant developments towards the implementation of GRs [22]. The degree of adoption and enforcement, where statutorily obligated, varies across countries. In Malaysia, the federal Town and Country Planning Department published a general guideline for roof gardens in 2012 but delegated its implementation to local authorities, who only indirectly encouraged their utilisation [23]. Generally, local authorities allow developments to apportion the provided open spaces at grade and on podiums or building rooftops to comply with the required “open spaces” (“kawasan lapang”), comprising softscape, hardscape, and a small percentage of utilities.

Several factors may have contributed to its lower adoption rate in Malaysian buildings, including preference for low-maintenance roofs, high wind speeds on high-rise rooftops, and, for residential buildings, the construction sequence of the roof level to facilitate early payment claims for stage completion under the Third Schedule of the Schedules G and H standard Sales and Purchase Agreement formats for new housing developments [24]. Lack of proper technical guidelines, government incentives, and high installation costs have also been found to be deterrents to GR adoption [22,25].

Most GRs in Malaysia are constructed on a reinforced concrete (RC) base [14]. However, roofs that are constructed purely for protection from the weather without the need for accessibility may be constructed using alternate materials, such as metal sheets, polycarbonate, acrylic, slates, or tiles. Among these materials, the use of metal roofs (MDRs) has become more widespread for their ability to achieve low gradients and long spans, features that facilitate modern aesthetics while being relatively light-weight.

In view of this trend, this study investigates the potential thermal and acoustic improvements that may be achieved by the integration of vegetation into a typical metal roof system in order to ascertain its feasibility as an alternative GR base material. The analyses will be conducted on both wet and dry days to account for Malaysia’s weather. This proof of concept may encourage a broader application of extensive GRs in Malaysia beyond the conventional RC base construction method.

2. Literature Review

Studies of both the thermal and acoustic performance of green roofs are available in the existing literature across various climates. Generally, green roofs were found to perform differently under dry or wet conditions due to the changed properties of their components [26,27,28]. The U-value [28] and heat flux [29] of the assembly are negatively affected by moisture content. Water fills the air gaps in porous components such as the substrate and increases the thermal conductivity of such components; a similar effect occurs in frozen green roofs [30]. GRs in warm climates also experience evapotranspiration, which improves the cooling effect [4,31].

The following studies compared green roofs against a type of conventional roof, which typically comprises a reinforced concrete flat roof, a metal roof, or a small-scale test module with similar compositions as the afore-mentioned systems.

Gagliano et al. [32] found that green roofs greatly reduced the peak outer surface temperature (from 56.3 °C to 28.6 °C) and daily temperature fluctuations (from 33.2 °C to 5.6 °C) compared to standard roofs in a Mediterranean climate. They also noticed a slower transmission rate of heat from the outer surface to the inner surface of green roofs. For Mediterranean and temperate climates, as well as mountainous and polar environments, insulation may play a more significant role than substrate and vegetation due to the fluctuating and often cold weather, which affects the properties and performance of the green roof [33].

In Singapore, a tropical country located next to Malaysia, a similar comparison by Wong et al. [4] ascertained a maximum surface temperature difference of 18 °C and a 60% reduction in heat gain. Another Singapore-based study by Qin et al. [34] found that green roofs had a lower surface temperature (7.3 °C on average) and lowered the ambient air temperature by 0.5 °C on average during daytime hours. In Indonesia, another neighbouring tropical country, Yuliani et al. [14] used software simulations to compare the thermal performances of concrete- and corrugated zinc-based green roofs; their findings suggested that metal-based green roofs are a viable alternative to reinforced concrete for temperature reduction through the roof system.

For local Malaysian findings, Rahman et al. [15] found that the roof surface temperature of extensive green roofs was significantly lower than conventional flat roofs, at 22.9 °C and 46.7 °C, respectively. Mirnezhad et al. [26] compared a bare roof and two green roofs of different substrate depths (12 cm and 28 cm); they discovered that a green roof with grass and a 12 cm-thick substrate resulted in the lowest mean ambient air temperature. During periods of drought, the peak soil temperature was recorded at 35.2 °C, which was significantly lower than the surface temperature of the bare roof at 60.7 °C.

Thermal performance may be measured using several different parameters, such as reduction in surface temperature [4,13,15,16,17], thermal conductivity (k-value), thermal resistance (R-value) [35], thermal transmittance (U-value) [13], and heat flux [17]. Temperature and heat flux are among the direct evaluation indices available to measure thermal performance. Thermophysical indices include R-value, U-value, and OTTV assessments. Finally, energy-related indices, such as cooling loads, heating loads, or energy consumption, may also be used.

Studies on green roofs have been conducted based on steady state heat transfer [31,36], although green roofs experience transient state heat transfer due to variations in the weather throughout the day [31,37]. The equation for transient state heat transfer, also known as the heat diffusion equation, describes the relationship between heat transfer, heat storage, and time. It takes into consideration the thermal diffusivity of the material (which studies its thermal conductivity, or k-value, density, and specific heat), temperature, and time. The equation for heat flux under steady-state heat transfer, on the other hand, may be calculated using the U-value, surface area, and temperature differential on the outer and inner surfaces of the assembly, as described in Equation (1):

$$q=UA(R_{TOS}-R_{TIS})$$

(1)

where:

q = heat flux (W/m²);

A = area (m²);

R_TOS = roof outer surface temperature (°C);

R_TIS = roof inner surface temperature (°C).

The U-value represents the effectiveness of an assembly in conducting heat from one surface to another. It is inversely proportional to the sum of the thermal resistances (R-values) of an assembly or system [38], including the thermal resistances of the internal and external faces. The relationship is described through the following equations:

$$R=\frac{d}{\lambda }$$

(2)

$$U=\frac{1}{\sum R}$$

(3)

By calculating the R-value to arrive at the U-value, a “design” or calculated U-value may be derived. The thickness (d) of each material or layer in the assembly shall be divided by its individual thermal conductivity value, or k-value (λ), then summed up to obtain the R-value. Equation (2) may then be used to obtain the U-value of the assembly.

Furthermore, in Malaysia, By-Law 38A of the Uniform Building By-Laws 1984 (Amendment 2021) (UBBL) [39] and Malaysian Standard MS1525: Energy Efficiency and Use of Renewable Energy for Non-Residential Buildings—Code of Practice (Third Revision) (MS1525:2019) [38] contain requirements for maximum roof U-values. Different maximum limits are imposed on light-weight and heavy-weight roofs. However, the definitions of “light-weight” and “heavy-weight” differ between these documents, as tabulated in

Table 2. Comparison of light-weight and heavy-weight roof definitions under the UBBL and MS1525:2019.

Roof Type	Document
	UBBL [39]	MS1525:2019 [38]
Light-weight roof
Definition	Below 50 kg/m2	Non-concrete roof construction
Maximum U-Value	0.4 W/m2K	0.4 W/m2K
Heavy-weight roof
Definition	Above 50 kg/m2	Concrete roof construction
Maximum U-value	0.6 W/m2K	0.6 W/m2K

.

The weight of an extensive green roof was likely to exceed 100 kg/m². As such, the discrepancy in the definition of roof types between UBBL and MS1525:2019 complicates matters: the proposed green roof on a metal deck base qualifies as a light-weight roof under MS1525:2019, but it is considered a heavy-weight roof under UBBL. A streamlining exercise between these two documents may be proposed, as UBBL By-Law 38A explicitly refers to MS1525:2019. However, as a lower U-value is generally considered to be more beneficial, this paper assumes 0.4 W/m²K to be the maximum U-value permissible for MDRs and GRs.

With regards to acoustics, the acceptable noise levels for human hearing ranged from 55 dB to 60 dB [40]. However, daytime exposure to noise levels averaging 55 dBA was found to annoy a quarter of the population studied by Kaltenbach et al. [41], while exposure to 50 dBA noise levels was associated with higher learning difficulties in schoolchildren [41].

Connelly and Hodgson [9] found that a light-weight metal roof increased transmission loss by up to 10 dB, 20 dB, and >20 dB in the low-, mid-, and high-frequency ranges, respectively. When it rained, Idris et al. found that uninsulated MRs not only had negligible sound insulation properties, but also amplified sound up to 10 dB higher [42].

Unlike MRs, the sound insulation properties of GRs may be more challenging to study as GRs are multi-layered systems; the introduction of more layers would allow sound transmission loss, increase damping, and potentially reduce the coincidence effect [43]. Sound transmission loss increased with substrate depth in a non-linear way, becoming more effective at high frequencies [9].

However, past studies have found that the moisture content of the substrate did not significantly affect sound transmission loss [9,44]. Galbrun and Scerri [44] found that varying the soil distribution, water content, and compaction level resulted in less than 1 dB variations in sound reduction levels. Therefore, this paper conducted the air-borne acoustic performance test by measuring reductions in sound levels under dry-day conditions only.

Generally, green roofs achieve better thermal and acoustic performance than standard bare roofs. For seasonal countries, green roofs exhibited different behaviours during summer and winter. As Malaysia is a tropical country, the performance of green roofs would be more closely aligned to summer findings, although localised contextual studies are desirable for more accurate assessments.

3. Methodology

The thermal and acoustic performance of the test modules were investigated via field measurements using small-scale test modules, a method that has been used in past studies [5,6,7,20,45]. As illustrated in Figure 1, the study was conducted within the Nichias Eco Experiment House compound at Universiti Teknologi Malaysia, Skudai, Johor, Malaysia (1°33′51.9″ N 103°37′43.7″ E). The site was located at least 45 m away from any inhabited structures. Climate-wise, Malaysia is a hot and humid tropical country with daily and annual variations in temperature of 5 °C–10 °C and less than 2 °C, respectively [46].

A total of four tests were performed: a thermal performance test and an air-borne acoustic performance test, conducted on both wet and dry days. The first round of tests was conducted on insulated GR and MDR roof modules (iGR and iMDR, respectively). The thermal performance test for iGR and iMDR was conducted between 9 November 2022 and 16 November 2022, while the acoustic performance test was conducted on 9 November 2022. The insulation layers were then removed before the second round of tests to assess uninsulated GR and MDR roof modules (uiGR and uiMDR, respectively). The thermal performance test for uiGR and uiMDR was conducted between 10 December 2022 and 17 December 2022, while the acoustic performance test was conducted on 14 June 2023.

3.1. Test Module and Sensor Set-Up

The small-scale test modules developed for this study comprise a square hollow section (SHS) frame on castor wheels, with internal clear dimensions of 1.0 m (W) × 1.0 m (D) × 1.0 m (H). The floor of the enclosure was raised 0.3 m above ground level. The frame was then clad with metal sheeting for walls, followed by 50 mm thick mineral wool (density 40 kg/m³) and then 12 mm thick cement boards as the internal finish. The wall insulation was included to limit heat flux through the walls and ensure that the heat flux measured through the roof would not be compromised. One side was fitted with a hinge and handle to allow access to the internal space in the enclosure. The roof modules were then installed on roof trusses at a 5° slope. All modules were positioned outdoors and exposed to the weather. The set-ups are illustrated in Figure 2, Figure 3 and Figure 4, with field measurement settings as per Table 2 and

Table 3. Field measurement settings for the thermal performance test.

Thermal Performance Field Measurement Settings
HOBO U30 outdoor weather station	27 m from roof modules
HOBO UX120-006M (4-channel) surface temperature data loggers and sensors	Sensors are located at midpoint (0.5 m from all edges) of internal surface of roof module and 0.025 m above the external surface of roof module
HOBO U12 air temperature and humidity data logger	Midpoint of test frame interior (0.5 m above the floor and equidistant from all side walls)
Data logging intervals	5 min
Data logging duration	Minimum 7 days
Extracted normal day data from total logged duration	1 typical day comprising 24 continuous hours of no rainfall within a calendar day
Extracted rainy day data from total logged duration	1 typical day with some rainfall within 24 continuous hours in a calendar day
Specimen test sequence	Concurrent. Phase 1: iGR with iMDR Phase 2: uiGR with uiMDR
Roof condition	Dry (dry day) Water-logged (wet day)
Devices’ manufacturer’s name and address	Onset HOBO, United States

.

Throughout the duration of both test phases, a HOBO U30 Outdoor Weather Station located approximately 27 m away from the test specimens logged the outdoor weather data. The weather station, installed 1.8 m above ground level, was used to collect outdoor ambient temperature (T_amb), wind speed (WS), solar intensity (SI), and relative humidity (RH) data.

3.2. Thermal Performance Test

The thermal performance assessment was carried out through a combination of literature review, field measurements, the Kendall rank correlation coefficient [47], and the Wilcoxon signed-rank test [48]. Details of the field measurement assessment are included in Table 3.

Field measurements were carried out over seven continuous days for both insulated and uninsulated roof module tests. From that data, a dataset for one full day with no rain (dry day) and a dataset for one full day with some rainfall (wet day) were extracted. As Malaysia’s annual variation in temperature is less than 2 °C [46], the findings from the sample wet and dry day data can represent typical-day performance of the roof modules.

This paper examined the reductions in surface temperature, comparing the performances of the metal roofs (iMDR and uiMDR) with and without the green roof assembly (iGR and uiGR). Dry day evaluations were carried out on the data of one typical day comprising 24 continuous hours of no rainfall within a calendar day. However, as there were no 24-hour periods of rainfall throughout the duration of the study, wet day evaluations were carried out from the moment of identified rainfall within a calendar day to distinguish peri- and post-pluvial findings from dry day findings.

The Kendall rank correlation coefficient [47] was used to non-parametrically assess the impacts of four weather parameters—namely wind speed, solar intensity, relative humidity, and outdoor ambient temperature—on the roof outer surface temperature of the roof modules. For the full 24 h of the selected wet and dry days, the assessment was conducted for two distinct temporal periods: daylight hours, from 8.00 a.m. to 6.00 p.m., and non-daylight hours, from 12.00 a.m. to 8.00 a.m. and from 6.00 p.m. to 11.55 p.m. This allows observations of the roof modules during a period of greatest heat gain and the periods preceding and succeeding it in a given day.

The Wilcoxon signed-rank test [48] was used to compare the inner-surface temperatures of the roof modules, in pairs (GRs and MDRs) over a period of 10 h, between 8.00 a.m. and 6.00 p.m., coinciding with Malaysia’s typical daylight hours and also the period of the highest heat gain. A total of 189 points per dataset were assessed. The Wilcoxon signed-rank test evaluates whether the mean values of two dependent groups differ significantly from each other.

Finally, a literature review exercise was conducted to determine the k-value of each material used in the roof module assemblies. These were extracted from standards and past research. Statutory requirements or recommendations for roof U-values were ascertained as well. Based on this information, the R-value and, subsequently, the U-value for each roof module were calculated.

3.3. Air-Borne Acoustic Performance Test

The air-borne acoustic performance assessment was carried out through a combination of a literature review, field measurements, and descriptive and inferential statistics. Findings from past research conducted on roofs were summarised to determine parameters or benchmarks for the field measurement test.

Details of the assessment are included in

Table 4. Field measurement settings for the air-borne acoustic performance test.

Air-Borne Acoustic Field Measurement Settings
Outdoor ambient sound measurement: Cirrus Optimus+ Green Sound Level Meter CR:1720	1.1 m above ground level and equidistant from both roof modules
Internal ambient sound measurement: CEM DT-8851 Sound Level Meter	Midpoint of test frame (0.5 m above the floor and equidistant from all side walls)
Air-borne sound source	One (1) BOSE 732522-2110 Sound Bar with the following settings: - suspended 0.4m at the midpoint above roof module; - oriented with speaker side facing down towards the roof module, volume set at 100%.
Description of air-borne sound	Audio clip of rain noises averaging at 65 dBA
Data logging intervals	1 s
Data logging duration	10 min
Extracted data duration	5 min
Specimen test sequence	Non-concurrent. The test for any one roof module is concluded before the next roof module is tested.
Roof condition	Dry (dry day) Water-logged (wet day)
Sound level meter manufacturer’s name and address	Cirrus Research UK, United Kingdom
Sound bar manufacturer’s name and address	Bose Corporation, United States

. Field measurements of dBA readings were conducted by playing rain sounds directly above the roof module being tested for a period of 10 min at one-second intervals, out of which five continuous minutes of data were extracted for further analysis. This removes any recordings of interfering noises that were not part of the test, such as during the set-up and conclusion of the sound tests. The average source sound level was set at approximately 65 dBA to ensure sufficient outdoor noise could be detected under the roof.

Statistical analyses were carried out with Python version 3.11, using the pandas version 2.2.1, matplotlib version 3.8.3, and SciPy version 1.11.4 libraries. The Shapiro–Wilks [49] and Levene’s [50] tests were used to test the normality and variance of the sound levels detected under the roof modules, respectively. In the event that the data was non-normal with significant variance, the non-parametric Kruskal–Wallis test [51] and Wilcoxon signed-rank test [48] would be used for further inferential statistical analyses. A comparative boxplot for each scenario was then plotted and analysed for visual verification of the non-parametric test results.

4. Results and Discussion

4.1. Thermal Performance Test

Field measurement for thermal performance was carried out from 9 November 2022 to 16 November 2022 for insulated roof modules (iGR and iMDR) and from 10 December 2022 to 17 December 2022 for uninsulated roof modules (uiGR and uiMDR). An overview of the temperature data recorded over those time periods is illustrated in Figure 5.

The Kendall rank correlation coefficient study was conducted over the full 24-hour period of all selected days. For the insulated roof modules test (iGR and iMDR), 15 November 2022 was selected for a dry-day scenario analysis for a 24-hour period; 13 November 2022 was selected for a wet-day scenario analysis. Based on data from the weather station, there were two approximate periods of rainfall on 13 November 2022, namely between 1.30 p.m. and 3.30 p.m., then 11.00 p.m. and 12.00 a.m. However, for the temperature assessments of the roof modules, the wet day scenario analysis was conducted on data collected between 1.30 p.m. and 12.00 a.m. to remove pre-pluvial conditions.

For the uninsulated roof modules test (uiGR and uiMDR), 16 December 2022 was selected for a dry-day scenario analysis for a 24-hour period. 14 December 2022 was selected for a wet-day scenario analysis. Based on data from the weather station, there were two approximate periods of rainfall on 14 December 2022, namely between 5.00 a.m. and 3.30 p.m., then 8.30 p.m. and 9.00 p.m. However, for the temperature assessments of the roof modules, the wet day scenario analysis was conducted on data collected between 5.00 a.m. and 12.00 a.m. to remove pre-pluvial conditions.

With reference to

Table 5. The Kendall rank correlation coefficient results between the roof outer surface temperatures of iGR, iMDR, uiGR, and uiMDR and the measured weather parameters.

Time of Day	Roof Type and Dry/Wet Day	Solar Intensity, SI	Relative Humidity, RH	Wind Speed, WS	Outdoor Ambient Temperature, Tamb
Daylight Hours (8.00 a.m.–6.00 p.m.)	Dry day
	iGR	0.72	−0.097	0.27	0.30
	iMDR	0.62	−0.320	0.36	0.51
	uiGR	0.35	−0.76	0.35	0.79
	uiMDR	0.40	−0.58	0.35	0.61
	Wet day
	iGR	0.35	−0.54	0.36	0.45
	iMDR	0.40	−0.51	0.30	0.42
	uiGR	0.56	−0.76	0.08	0.88
	uiMDR	0.60	−0.68	0.13	0.75
Non-Daylight Hours	Dry day
	iGR	0.13	−0.16	−0.29	0.42
	iMDR	0.10	0.24	0.01	−0.01
	uiGR	0.04	−0.59	0.11	0.83
	uiMDR	0.16	−0.61	0.29	0.67
	Wet day
	iGR	0.11	N/A	−0.055	0.01
	iMDR	0.18	N/A	−0.042	0.06
	uiGR	0.28	−0.18	−0.041	0.69
	uiMDR	0.49	−0.092	0.360	0.24

, T_amb is most strongly correlated with changes in the roof outer surface temperature (R_TOS) of iGR, iMDR, uiGR, and uiMDR, in 8 out of 16 assessed scenarios, with p-values of < 0.01 (daylight hours, dry day: uiGR, uiMDR; daylight hours, wet day: uiGR, uiMDR; non-daylight hours, dry day: iGR, uiGR, uiMDR; non-daylight hours, wet day: uiGR). Generally, it may be inferred that the higher thermal mass and lower thermal conductivity of iGR made it less susceptible to changes in T_amb throughout the day, such that it exhibited lower correlations than iMDR with T_amb during daylight hours, and higher correlations during non-daylight hours as the R_TOS of iMDR dropped more rapidly than T_amb. However, the combination of rain and lower T_amb at night inversed the performance of iGR relative to iMDR. This may be due to the changed properties of the roof module when water-logged, as suggested by findings from past research [26,27,30].

The highest correlation between T_amb and R_TOS occurred in uiGR, with a maximum correlation on wet days during daylight hours (tau = 0.88 and p = < 0.01), as it was both slower to gain and lose heat than uiMDR. It may also be worth noting that the maximum R_TOS of uiGR during rainy daylight hours was 28 °C, compared to 31.4 °C of uiMDR (refer to

Table 6. Temperature recordings for iGR, iMDR, uiGR, and uiMDR throughout the analysed periods.

Weather	Roof Module	Roof Surface	N	Min (°C)	Max (°C)	Mean (°C)	Std. Dev.
Dry Day	iGR	Outer	289	21.5	37.8	27.5	4.8
		Inner	289	23.6	33.6	28.2	3.2
	iMDR	Outer	289	22.3	58.2	30.7	9.9
		Inner	289	23.3	37.2	29.1	4.6
	uiGR	Outer	289	22.2	44.8	29.5	7.3
		Inner	289	24.3	35.3	29.0	3.8
	uiMDR	Outer	289	22.1	59.4	32.5	11.8
		Inner	289	23.2	57.5	32.9	10.8
Wet Day	iGR	Outer	127	23.2	35.5	27.3	4.2
		Inner	127	26.7	32.2	29.4	1.7
	iMDR	Outer	127	23.9	35.6	28.5	4.2
		Inner	127	26.6	34.5	29.9	2.3
	uiGR	Outer	229	23.1	28.0	25.0	1.5
		Inner	229	24.3	26.5	25.2	0.7
	uiMDR	Outer	229	23.1	31.4	26.0	2.3
		Inner	229	24.4	28.4	25.8	1.2

Note: N = number of observations, taken at five-minute intervals. Min = lowest temperature recorded during the analysed period (°C). Max = highest temperature recorded during the analysed period (°C). Mean = average temperature recorded during the analysed period (°C). Std. Dev = standard deviation of the recorded temperature.

). The stronger significance of the correlation for uiGR compared to uiMDR persisted for both dry and wet conditions throughout the day, becoming most discrete when it rained during non-daylight hours (tau = 0.69 and 0.24, respectively, and p = < 0.01). Besides T_amb, SI and RH were also able to explain changes in R_TOS, except when RH was constantly at 100% throughout the assessment period.

Upon studying the impacts of weather parameters on R_TOS, the temperature differences between the outer and inner surfaces of the roof modules were then analysed to study the degree of temperature change through the roof assemblies. A tabulation of the temperatures recorded on the outer and inner surfaces of each roof module, including the number of data points per dataset, minimum and maximum recorded temperatures, average temperature, and standard deviations according to dry day and wet day weather categories, is recorded in Table 6.

Temperature variance (R_TOS − R_TIS) was calculated and plotted as a boxplot (refer to Figure 6). The range of temperature variance was greater on dry days than wet days across all roof modules. For both dry and wet days, uiMDR exhibited a smaller temperature variance and inter-quartile range (IQR) than iMDR, suggesting that a high amount of heat was transmitted from its outer surface to its inner surface.

On a dry day, uiGR exhibited a slightly greater range of overall temperature variance than iGR, at 13.6 °C and 10.37 °C, respectively. The IQR of iGR’s temperature variance was 6.21 °C, compared to 8.32 °C for iMDR. The presence of insulation in iGR created a positive temperature variance between its R_TOS and R_TIS throughout approximately 35% of the day, between 7.35 a.m. and 3.30 p.m. The R_TIS of iMDR was lower than its R_TOS for almost 50% of the day, but many outliers (unusually high temperature variances) occurred during the period of highest heat gain, between 7.30 a.m. and 3.40 p.m. The maximum daily temperature for the R_TOS of iMDR was higher than that of iGR, at 58.2 °C and 37.8 °C, respectively. This concurred with findings on tropical green roofs, which found the R_TOS of green roofs to be lower than the benchmarked roof types [15,34].

The uiGR boxplot had a longer tail, with a median variance value of −2.13 °C but an upper quartile variance value of 4.32 °C and an IQR of 6.69 °C, indicating that the data was positively skewed while the temperature changes for iGR were more evenly distributed. The IQR of uiMDR was much smaller than that of iMDR at 1.66 °C. The absence of insulation in uiMDR greatly decreased its thermal insulation properties over iMDR. In fact, the inner surface of uiMDR tended to be warmer than its outer surface; when considered in conjunction with the overall higher R_TOS of uiMDR, which recorded a maximum R_TOS of 59.4 °C compared to 44.8 °C on uiGR, this was an undesirable outcome.

On a wet day, from the onset of rain, the R_TIS of iGR was generally lower than its R_TOS, with a median and maximum temperature variance of −1.17 °C and 1.21 °C, respectively. Heat loss on its R_TOS was expedited by direct exposure to rain and lower outdoor temperatures, but the higher thermal mass of iGR meant that its inner roof surface was slower at releasing heat towards the outer surface. In comparison, iMDR revealed a higher temperature variance (IQR of 4.11 °C against iGR’s 1.6 °C) despite recording almost similar maximum R_TOS as iGR (35.6 °C against iGR’s R_TOS of 35.5 °C). Without mineral wool and aluminium foil, uiGR had less thermal mass than iGR and recorded a larger IQR of 2.59 °C. Interestingly, uiMDR had a smaller IQR (IQR = 1.78 °C) but a larger temperature variance than uiGR, suggesting that water-logged soil on an otherwise uninsulated metal roof may confer only limited improvements to thermal insulation. However, R_TOS should be considered as well, as the R_TOS of uiGR was generally lower than uiMDR.

Wilcoxon signed-rank tests for both insulated and uninsulated roof modules showed that the addition of a green roof component resulted in a statistically significant difference in the R_TIS measurements (p-value = 0.00) during daylight hours, between 8.00 a.m. and 6.00 p.m., regardless of the weather. Generally, the R_TIS of iGR and uiGR were more stable than their iMDR and uiMDR counterparts within that time frame.

Annex D in MS2680: Energy Efficiency and Use of Renewable Energy for Residential Buildings—Code of Practice (MS2680:2017) includes a list of k-values for building materials commonly used in Malaysia, from which reference k-values for dry soil, wet soil, metal, and mineral wool were obtained [52]. The k-values for building materials that were not part of Annex D were further obtained from past research. The materials, corresponding k-values, thicknesses, and resultant R-values are tabulated in

Table 7. Thermal conductivity values and calculated R-values for each material in the roof module assemblies.

No	Material	K-/R-Value References	K-Value, λ (W/mK)	Thickness, d (m)	R-Value (d/λ)
1	Pearlgrass	[53]	0.25	0.02	0.080
2	Dry Soil Potting Mix (Loam Soil)	[54]	0.10	0.05	0.500
3	Wet Soil Potting Mix (Loam Soil)	[13]	2.10	0.05	0.105
4	Dry Non-Woven Geotextile	[55]	0.10	0.002	0.02
5	Wet Non-Woven Geotextile	[55]	0.62	0.002	0.003
6	Dry Hygroscopic Rock Mineral Fibre Felt	[56]	0.037	0.02	0.541
7	Wet Hygroscopic Rock Mineral Fibre Felt	[56]	0.80	0.02	0.025
8	Dry Moisture Mat	[57]	0.0352	0.012	0.341
9	Wet Moisture Mat	[57]	0.0379	0.012	0.317
10	Subsoil Drainage Mat	[54]	0.02	0.002	0.100
11	Root Shield Membrane	[58]	0.04	0.001	0.025
12	0.48 mm BMT Metal Deck as Roof Structure Bottom Plate	[52]	52.0	0.00048	0.0000092
13	0.55 mm BMT Metal Deck as Roof Panel	[52]	52.0	0.00055	0.0000106
14	Mineral wool	[52]	0.035	0.05	1.429
15	Aluminium Foil	[59]	237.0	0.000016	0.0000001
16	Rsr	[60]	-	-	0.044
17	Rsi	[60]	-	-	0.160

and

Table 8. Design R-values and U-values of the roof modules, calculated using the k-values derived from the literature review.

Roof Module	Materials Used (Based on Table 7)	Total R-Value (m2K/W)	U-Value, 1/R (W/m2K)
iGR (Dry)	Items 1–2, 4, 6, 8, 10–17	3.239	0.309
iGR (Wet)	Items 1, 3, 5, 7, 9–17	2.206	0.453
uiGR (Dry)	Items 1–2, 4, 6, 8, 10–13, 16–17	1.810	0.552
uiGR (Wet)	Items 1, 3, 5, 7, 9–13, 16–17	0.778	1.286
iMDR	Items 12–17	1.633	0.613
uiMDR	Items 12–13, 16–17	0.204	4.901

.

All GR roof modules had lower design U-values than MDRs, except uiGR under wet conditions. Due to the higher k-values of wet soil, geotextile, hygroscopic rock mineral fibre felt, and moisture mat, the U-values of iGR and uiGR when wet were higher than when they were dry. These findings concur with past studies [13,61,62].

The control roof modules, iMDR and uiMDR, had higher U-values than the GR modules (except uiGR under wet conditions), at 0.613 W/m²K and 4.901 W/m²K, respectively. Although the U-value of uiMDR exceeded the maximum U-value required under UBBL and MS1525:2019, the addition of a green roof component (with reference to the uiGR roof module) greatly improved its design U-value, even in the absence of mineral wool insulation and aluminium foil. In its dry state, the U-value of uiGR was 4.349 W/m²K lower than uiMDR and 0.061 W/m²K lower than iMDR. These findings suggest that a green roof component may be a viable complement to conventional roof insulation materials in terms of thermal performance.

4.2. Air-Borne Acoustic Performance Test

As variations in soil distribution, water content, and compaction level resulted in insignificant variations in sound reduction levels [44], this paper conducted the air-borne acoustic performance test by measuring reductions in sound levels under dry day conditions only.

Field measurement for air-borne acoustic performance was carried out on 9 November 2022 for insulated roof modules (iGR and iMDR) and on 14 June 2023 for uninsulated roof modules (uiGR and uiMDR). A total of 301 data points per dataset were collected for each assessment.

Although the field measurements were conducted outdoors, the site was approximately 45 m away from any inhabited structures. As such, the man-made sound produced by the speakers, averaging 65 dBA, was the dominant sound source during the experiments. The findings of the air-borne acoustic performance test are as per

Table 9. Results of the air-borne acoustic performance test.

Test Specimen	μSLo (dBA)	μSLint (dBA)	μSLred (dBA)	Deviation (%)
iGR	65	32	−33	−51
uiGR	65	44	−21	−32
iMDR	65	45	−20	−31
uiMDR	65	69	+4	+6

Note: μSL_o = average outdoor sound level. μSL_int = average sound level measured within the test module. μSL_red = average sound level reduction between outside and within the test module. Deviation = μSL_red/μSL_o × 100%.

.

iGR recorded the greatest degree of sound reduction, with an average reduction of 51%, followed by uiGR and iMDR. uiGR achieved nearly similar sound reduction levels to iMDR, at 32% and 31%, respectively, suggesting that a green roof component may be a viable alternative insulative layer for the conventional combination of mineral wool and aluminium foil. The sound level under uiMDR was consistently louder than the outdoor sound. This phenomenon might be partially attributed to the acoustic resonance effect within the metallic small-scale test setup. Although the air-borne acoustic performance test would not test any additional effect of the physical impact of raindrops, the findings of uiMDR did concur with those of Idris et al. [42], as the sound levels were higher than the outdoor sound.

The normality of the sound levels measured under the roof modules was assessed using the Shapiro–Wilk test. The p-values of iGR, iMDR, uiGR, and uiMDR are less than 0.05, suggesting that the data distribution is non-normal (refer to

Table 10. Shapiro–Wilks normality test results for all roof modules.

Roof Module	Shapiro–Wilk Test Results
	Statistic	Nos. Observations	Significance (p-Value)
iGR	0.789	301	0.000
iMDR	0.985	301	0.003
uiGR	0.746	301	0.000
uiMDR	0.971	301	0.000

). Levene’s test for equality of variance based on both mean and median returned p-values of 0.000, indicating a significant difference between the variances in sound levels of the roof modules. The analysis therefore proceeded using the non-parametric Kruskal–Wallis test by ranks to compare all roof modules, and the Wilcoxon signed-rank test to conduct pair-wise comparisons for insulated and uninsulated roof module pairs (iGR against iMDR and uiGR against uiMDR).

The Kruskal–Wallis tests resulted in p-values of 0.00, rejecting the null hypothesis of similar medians across all groups. The use of different roof assemblies did lead to statistically significant differences in the sound levels recorded under the roof modules. Wilcoxon signed-rank tests for both insulated and uninsulated roof modules showed that the addition of a green roof component elicited a statistically significant improvement in sound level reductions compared to the control modules, iMDR or uiMDR (Z-value = −15.039, p-value = 0.00).

The findings were visually verified by plotting a comparative boxplot of the aforementioned variables for iGR, iMDR, uiGR, and uiMDR (refer to Figure 7). For uiGR, the addition of a green roof component on an uninsulated metal roof resulted in significant improvements over uiMDR; it also mitigated the sound amplification effect detected for uiMDR. While iGR conferred a less drastic improvement over iMDR when compared to uiGR (only 19%), it did result in the narrowest range of recorded sound levels among the four roof modules, which may contribute to a more controlled internal acoustic environment.

5. Conclusions

The introduction of a green roof component on a metal roof, insulated or otherwise, improved the thermal and acoustic performance of MRs. uiGR (uninsulated green roof on a metal base) achieved nearly similar sound reduction levels to iMDR (insulated metal roof), suggesting that a green roof component may be a viable alternative insulation besides mineral wool and aluminium foil for acoustic purposes. These findings are consistent with the outcomes of similar research on green roofs in the tropics [4,9,14,15,26,34,42].

Thermal and acoustic performance varied under dry-day and wet-day scenarios, suggesting that it may be prudent to compare and assess roofs separately for both weather conditions. This strategy departs from the conventional approach of reporting averaged observations that generalise the findings into weather-agnostic data. GRs, in particular, would benefit from such an assessment method as water-logged substrates develop different properties than dry substrates, which may affect the outcome of the investigation. However, field measurements of wet-day scenarios may introduce limitations on the reproducibility of results as natural weather conditions are less predictable, necessitating strategic simultaneous testing.

Furthermore, for the calculation of the design U-value for thermal performance, our literature review identified a need to streamline the definition of “light-weight” and “heavy-weight” roofs in UBBL and MS1525:2019 in order to determine the appropriate reference U-values for iGR and uiGR.

Further studies would be required to determine the thermal and acoustic effects of manipulating other variables in the composition of the roof assembly, such as the thickness of insulation, type of radiant barrier, depth of substrate, or the introduction of air gaps, and their impacts on the comparative effectiveness of metal-based GRs. The impact of the GR layer on the urban heat island effect, structural load, framing size, and the overall cost of the roof system also merits investigation.