Building and Environment 243 (2023) 110731

Contents lists available at ScienceDirect

Building and Environment

journal homepage: www.elsevier.com/locate/buildenv

Climate adaptive urban measures in Mediterranean areas: Thermal

effectiveness of an advanced multilayer green roof installed in
Palermo (Italy)
Dario Pumo *, Francesco Alongi , Marcella Cannarozzo , Leonardo V. Noto
Dipartimento di Ingegneria - Università degli Studi di Palermo, Viale delle Scienze, Ed. 8, 90128, Palermo, Italy

A R T I C L E I N F O A B S T R A C T

Keywords: Several nature based and climate adaptive solutions have been proposed to improve cities resilience to the effects
Green roof of global warming and restore natural processes in strongly anthropized areas. Green roofs are among the most
Climate change efficient nature based solutions to address recurrent urban challenges, such as pluvial floods and urban heat
Urbanization
islands. Various benefits offered by green roofs are rather known, such as their capacity to enhance buildings
Green infrastructure
Nature based solution
thermal insulation; green roofs also favor urban biodiversity, improving buildings aesthetic value and human
Urban heat island well being. Multilayer green roofs (MGRs) are green roofs with an additional layer that increases their water
Building thermal insulation storage capacity. Deep analyses on MGRs are still lacking due to their recent development, and the few works in
literature are prevalently focused on their stormwater retention primary function. This work explores the thermal
function of an experimental MGR prototype installed in Palermo (Italy), comparing its response to local climate
with that of an unaltered portion of the rooftop through the analysis of surface temperature time series collected
over a two years monitoring period. Performances are evaluated thought various daily thermal indices, also
analyzing the role of the water stored into the system. Results contribute to raise awareness about the benefits
arising from the use of MGRs in semi-arid Mediterranean urban areas, confirming, as main thermal advantage,
their cooling effect, with mean daily surface temperature reduced by 8.4% outdoor and 5.8% indoor; perfor­
mances increases with water storage and are particularly evident during the hot and dry summers that typically
characterize such regions.

1. Introduction (LW) radiation. The reduced sky view factors could partially limit the
LW radiation released outside the urban areas. Urban canyons might
Climate change and urbanization are profoundly modifying the at­ also impede cooling processes during the nighttime. The increasing
mospheric energy balance [1] and many of the processes involved in the temperature due to climate change and urbanization, as well as the
hydrological cycle [2,3], with important fallouts on several environ­ anthropogenic heat produced by human activities in high density ur­
mental and socio-economic aspects (e.g. Refs. [4,5]). The population banized areas, are in fact the main causes of the well-known Urban Heat
and extent of cities continue to grow [6] and this implies an increasing Island (UHI) effect, generating elevated temperatures at urban nuclei
demand for water, energy, construction materials, and food [7] and new compared to the outlying rural areas [8]. UHI is becoming an important
challenges for the scientific community and policy makers. matter today, since it could further exacerbate the economic impacts of
Climate change is causing heating and frequent heat waves, that are heat waves and represents an environmental stressor causing serious
particular evident in urban areas. Urban surfaces typically have low risks to human health and related social costs [9].
albedo and high solar absorptance, and urban geometry could influence Moreover, high air temperatures can modify the distribution of
shade conditions and create urban canyons where the solar short-wave pollutants in the air [10,11] and increase intensity and frequency of
(SW) radiation, after multiple reflections between urban surfaces, can rainfall extremes [12,13]. Ground impacts of rainfall extremes are
be trapped, absorbed by building walls and re-radiated as long-wave further exacerbated by urban expansion at the expense of green areas,

* Corresponding author.
E-mail addresses: [email protected] (D. Pumo), [email protected] (F. Alongi), [email protected] (M. Cannarozzo), leonardo.noto@
unipa.it (L.V. Noto).

https://doi.org/10.1016/j.buildenv.2023.110731
Received 22 June 2023; Received in revised form 11 August 2023; Accepted 12 August 2023
Available online 13 August 2023
0360-1323/© 2023 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-
nc-nd/4.0/).
D. Pumo et al. Building and Environment 243 (2023) 110731

causing frequent severe stress to the urban drainage systems. of MGRs are still rather rare, especially in warm-temperate climates.
Several researcher carried out specific studies (e.g. Refs. [14,15]), The analysis presented here is based on an experimental system
aimed to demonstrate the advantages that could arise from the imple­ installed in 2019 in Palermo (Italy), fully described in Sect. 2. The sys­
mentation in urban areas of the so-called Nature-Based Solutions (NBSs). tem could be assimilated to an extensive GR, considering its lightness
NBSs are “green” infrastructures inspired and supported by nature that and the limited thickness of the soil layer, and it has been monitored by a
include a series of sustainable and low-cost solutions aimed to restore complex network of sensors, also described in Sect. 2, since December
pre-development conditions. NBSs, given their multipurpose nature 2020, also monitoring an equal size portion of the rooftop bordering the
[16], are ideal solutions to achieve many of the Development Goals system, used as benchmark “grey” roof for comparison.
(SDGs) proposed in the Sustainable Agenda 2030 [17]. The present research integrates a previous work [42], where the
Green Roofs (GRs) are climate adaptive NBS particularly appreciated hydrological effectiveness of the experimental site was deeply evalu­
by the society, since they allow to exploit otherwise unutilized spaces ated, demonstrating from the analysis of the first year of monitoring, the
and offer a series of aesthetical as well as environmental, social and high performance of the system in terms of retention capacity at rainfall
economic advantages [18,19]. Incorporating vegetation, growth me­ event scale. The new analysis presented in this paper focuses on the
dium and other landscape components on the rooftop of buildings pro­ thermal response of the system, exploring possible advantages of MGRs
vides several direct and indirect benefits, such as: i) stormwater in terms of thermal insulation capacity and roof thermal inertia.
attenuation [20] and heat stress mitigation [21], with possible potable Different daily performance indices, defined in Sect. 2, are used and
water and energy consumption reduction; ii) water quality enhance­ they are based on the comparison among air and roof surface temper­
ment, since green roofs may buffer acidic rain and retain pollutants [22]; ature data measured at both outdoor and indoor sensors placed at the
iii) attenuation of noise levels in urban spaces arising from road, rail and MGR and the benchmark grey roof. Such indices are analyzed at the
air traffic; iv) mitigation of air pollution, by directly consuming gaseous level of entire two-year monitoring period and at the seasonal level, and
pollutants through vegetation or indirectly, by modifying microclimates the results are described in Sect. 3. A discussion on the main outcomes of
[23]; v) restoring of biodiversity lost due to urban development, offering the present study and their possible implications is reported in Sect. 4,
a safe recovery place for birds and insects; vi) extending of rooftop life, while some concluding remarks conclude the paper.
since GRs protect roof membrane from extreme heat, wind and Some aspects extremely innovative and specific for multilayer green
ultra-violet radiation [24]; vii) providing social benefits including cre­ roofs are treated, such as an analysis on the influence of the water
ation of educational and employment opportunities, recreational green storage on the thermal response of the system. Thermal benefits for
areas and potential spaces for biological food production [25], with a MGRs are rarely quantitatively supported by field measures; sharing this
series of positive implications for public health [26–28]. type of information is fundamental for the development of reliable
Green roofs have been traditionally used as thermal insulation models able to reproduce the thermal response of MGRs to different
measures for buildings. Actually, several experimental studies demon­ climate forcings, which are key tools for various possible specific ap­
strated how GRs are able to contribute to roof insulation of buildings, plications, such as extending the analysis from the plot scale to larger
absorbing 60% of the direct solar radiation and reflecting about 20–30% scales, or evaluating the potentialities under future scenarios impacted
[29], reducing daily temperature fluctuations, and consequently by climate change.
allowing energy to be saved from the air conditioning systems (e.g.,
Refs. [30–32]. In Mediterranean areas, thermal effectiveness of exten­ 2. Materials and methods
sive, semi-intensive and intensive GRs was investigated in Portugal by
Ref. [33]; showing that GRs enable up to 20% of energy saving for 2.1. Experimental site: the Polder Roof system
extensive solutions and up to 70% for intensive ones.
Several studies have been recently focused on the potentialities of Multilayer-Green Roofs (MGRs) are characterized by the presence of
GRs under different perspectives and at different scales, ranging from a high-capacity storage layer, often referred to as blue layer (BL), located
plot-scale, homes, buildings, and district levels (e.g., Refs. [34–36]). The in the bottom part of the system. Rainwater is infiltrated into the upper
direct monitoring and the identification of appropriate performances green layer (GL), where it feeds the growing medium soil moisture
indicators, as well as the development of accurate methods and models useful for the vegetation physiological activities; when field capacity is
to assess the response of such systems to past, present and future climate exceeded, percolation water is conveyed toward the lower BL where it is
forcings, are essential aspects that might drive us toward the identifi­ retained up to the fulfillment of the storage retention capacity; water
cation of new GRs design criteria oriented to the concepts of “design exceeding the BL retention capacity is finally released out of the system
with nature”, “resilience paradigm” and “building sustainability”. as outflow discharge.
This work contributes to the understanding and quantitative Water volume retained by the system is extremely important, since it
assessment of some potential thermal benefits related to the use in a can reduce the pressure on urban drainage systems during intense
Mediterranean climatic area of a new typology of GRs, that is the so- storms, and represents, at the same time, a useful water resource for
called Multilayer Green Roof (MGR). The Mediterranean basin, which possible non-potable reuses (e.g., flushing toilets and clothes washing,
can be considered as a hot spot for climate change [37,38], urgently car washing and garden irrigation) and passive irrigation of the same
needs a big effort to promote and implement climate adaptive solutions system vegetation. Water storage and system outflow can be controlled
such as GRs. Compared to traditional GRs, MGRs have a further “blue” dynamically through a retention control weir, which in advanced sys­
layer for water storage, which allows for water retention and possible tems such as that considered for this paper, can be regulated remotely
reuse, with a storage capacity comparable to that typical of Rainfall and automatically according to predefined targets [42].
Harvesting Systems [18], such as blue roofs. In MGRs, the blue layer is The hydrological and thermal efficiency of MGRs strongly depends
connected to the upper “green” layer via capillary cones that allow for on the properties of the system, such as the GL and BL depth, the type of
passive irrigation, supporting physiologic activities of the plants and substrate and vegetation, the retention control weir management [42].
reducing water stress during the driest periods. Due to their recent also demonstrated how the retention rate of MGRs relies on the system
diffusion, potential advantages of MGRs are much less explored initial conditions, in terms of GL soil moisture content and BL water
compared to traditional GRs [39]. Most of the scientific literature was storage immediately before the rainfall occurrence, and the character­
focused on the main advantage that typically MGRs can provide, namely istics of the incoming rainfall (i.e., total amount, intensity, duration).
the ability to slow down and reduce stormwater volume in urban areas, The experimental site under analysis (Fig. 1a), called Polder Roof
due to their augmented water storage capacity with respect to tradi­ was developed by the Dutch company Metro-Polder and installed in
tional GRs [40,41]. Observational analyses to assess possible co-benefits June 2019 within the project Polder Roof Field Lab, supported by the

2
D. Pumo et al. Building and Environment 243 (2023) 110731

Fig. 1. (a) Experimental site after about 3 years from installation (Aug 2022); (b) sedum species in the shallower region (Apr 2022); (c) shrub and aromatic species in
the deeper region (Apr 2022).

European Institute of Innovation and Technology (EIT) Climate-KIC experimental site and it is equipped with three different sensors for
(Knowledge and Innovation Community) program. Other three similar rainfall monitoring and a meteorological weather station, both trans­
prototypes have been also installed within the same project in the Italian mitting data to a local server. Rainfall is monitored, with 1 min data
cities of Cagliari, Perugia and Viterbo, and are described in Ref. [43]. acquisition frequency, by a weighing rain gauge (OTT Pluvio2-400), a
The installation covers a rectangular portion of 32.1 m2 of the laser-optic disdrometer (OTT- Parsivel2) and a tipping bucket rain gauge
rooftop of the Engineering Department of the University of Palermo (LSI LASTEM - DQA130.1). The weather station is equipped with a
(Italy). An equal size area adjacent to the MGR, hereafter referred to as thermo-hygrometer (Lambrecht 8096) and an air pressure sensor
“grey roof” (GrR), is used as reference unaltered roof surface for com­ (Lambrecht 8128). More detailed information about such sensors can be
parison. Two sub-regions can be distinguished: a central sub-region of found in Ref. [44]. All data from the weather station are published on­
18,0 m2 with soil depth of 20 cm and a border sub-region of 14,1 m2 with line at the HYCLIC (HYdrology & Climate change ImpaCts Lab) website
soil depth of 10 cm. Soil constituent materials meet the requirements of (http://idrologia.unipa.it/) and freely available on request.
the Italian standard UNI 11235 (2015). In particular, the substrate for A system of two external plastic rain barrels with a capacity of 1000 l
both the sub-regions is made of a fertile and extremely light soil con­ receives separately discharge outflows from the MGR and the GrR; each
sisting of a mixture of volcanic lapillus (90%) and pumice (10%). rain barrel is equipped with a couple of pressure sensors (i.e., vanEssen
Vegetation is composed by a selection of typical autochthonous species micro-diver and baro-diver barometric pressure loggers), which allows
for Sicily, rather drought-tolerant once established and selected in order for measuring, with 5 min acquisition frequency, water levels and,
to favor their adaptability with respect to local climate. The shallower consequentially, volumes released from the two different portions of the
sub-region (Fig. 1b) has been realized by various sedum species, and roof.
typical Mediterranean meadow species while the deeper sub-region Roof surface temperatures are measured with 30 min acquisition
(Fig. 1c) is covered by different perennial and aromatic species. frequency by a set of four thermometers (MX2203 by HOBO); more
The GL is supported by the BL and separated from this by a capillary specifically, two couples of sensors are installed at vertical corre­
geotextile drainage membrane. The BL is a 10 cm height water retention sponding positions in the upper (outdoor) and bottom (indoor) surface
box consisting of a modular system produced by the Permavoid Com­ of both the MGR and the GrR. Outdoor sensor for the MGR was installed
pany (http://www.permavoid.co.uk/) that combines plastic support at depth of about 5 cm below the soil surface, in the central point of the
units, enabling stormwater storage and conveyance, with capillary col­ system, while the other outdoor sensor was placed on the central point of
umns, enabling natural capillary irrigation. The system is completed by the external GrR roof surface. Indoor sensors for both the MGR and the
a protective geotextile filter fabric and a waterproof membrane. GrR are installed on the ceil of two separated, conterminous, not air-
conditioned and scarcely utilized rooms, where the presence of a sec­
ondary modular ceiling system, placed 33 cm below the roof, minimizes
2.2. Monitored variables the effects of potential different thermal and ventilation conditions in
the two rooms. The two rooms have equal surface (34.1 m2) and height
The MGR is equipped with its own integrated monitoring system, (4.1 m), and both have an equal size (3.0 × 1.6 m), permanently locked,
named Sm^rtMILL, which includes a set of sensors for rainfall, temper­ window facing north-west.
ature, and water height into the BL; Sm^rtMILL collects data with 10 min
frequency, and transmits them to a Metro-Polder dashboard accessible
on the internet, called AQORA. A dynamic control of the water storage 2.3. Local climate in Palermo
into the BL and outflow discharge release is operated by a semi-
automatic outflow control system, called Smart Flow Control (SFC), by Palermo is the fifth most populated city in Italy. It lies on Sicily’s
which the control weir can be remotely regulated at any time from the northwestern coast at the head of the Bay of Palermo, facing east. Inland
dashboard or programmed, setting opening degrees and timing. During the city is enclosed by a large plain and the surrounding mountains that
the entire period of monitoring, the weir was set constantly at the rise up to an elevation of about 1.000 m a.s.l north of the city. Local
maximum closure corresponding to a height of 7 cm. climate is classified as Csa - Hot-summer subtropical Mediterranean climate
Besides the Sm^rtMILL system, the MGR and the GrR are monitored according to the Köppen-Geiger climate classification, with long, hot
by an external weather monitoring station, a system of two rain barrels and dry summers and rainy winters. A characterization of rainfall and
equipped with water depth sensors and a set of thermometers. The temperature regime of the city based on the analysis of the last two
weather station is located in close proximity (i.e. 65 m) to the decades (2002–2022) of data collected at the gauge station ID 276-

3
D. Pumo et al. Building and Environment 243 (2023) 110731

Uditore of the SIAS (Servizio Informativo Agrometeorologico Siciliano) The system antecedent conditions are evaluated through indices
regional agency, about 3 km far from the experimental site, is reported derived from literature [42,47–49] and slightly modified for the pur­
in Fig. 2, from which it can be observed a mean annual air temperature poses of our analysis, with the objective of assess the amount of water
of 18.9 ◦ C, with mean daily values around 26.2 ◦ C during the summer present within the different layers of the system at the day of evaluation.
and 12.4 ◦ C during the winter. The hottest month is August (mean daily The adopted indices (Table 1) are: i) the Antecedent Dry Weather Period
temperature 27.2 ◦ C), while the coldest one is January (mean daily (ADWP), given by the number of days to the last rainy day before the day
temperature 11.9 ◦ C). The daily temperature excursion is rather relevant of evaluation; ii) the Degree of Water Storage (DWS), given by the
(mean equal to 7.4 ◦ C and maximum of 20.3 ◦ C), with low temperature percent ratio between the mean daily water level into the BL at the day
during the night due the heat exchange and the thermal inversion, and of evaluation and the thickness of the BL (i.e. 10 cm); iii) the 5-days
with temperature during the diurnal hours that can reach very high Antecedent Precipitation Index (API5), here defined as the cumulative
value, especially during the summer, under non-windy conditions precipitation over the last 5 days, including the day of evaluation; iv) the
(maximum record 44.2 ◦ C in June 2007). 5-days Antecedent Temperature Index (ATI5), given by the mean daily
The mean annual precipitation is 740 mm, with on average 554 mm temperature over the last 5 days, including the day of evaluation; and v)
(75%) during autumn and winter. The mean monthly precipitation the 5-days Antecedent Runoff Index (ARI5), equal to the cumulative
spans from about 13 mm/month (in June, July and August) to about 97 runoff produced by the MGR in the last 5 days, including the day of
mm (for the months from October to January). Rainfall retention mea­ evaluation. Due to a sensor malfunction, data of BL water levels are
sures are particularly useful in Palermo, which is periodically affected available only from Jan. 15th, 2021 to Jun. 29th, 2021, and thus the
by urban pluvial floods with significant economic damages [45]. DWS has been calculated only over a short initial time window of the
monitoring period.
The analysis of thermal benefits has been carried out on 30-min
2.4. Dataset and performance indices
temperature time series recorded by the four surface thermometers
(outdoor/indoor sensors at the MGR and the GrR) and air temperature
In this study, a two-years reference period, from Dec. 22nd, 2020 to
time series measured by the weather station, computing the thermal
Dec. 31st, 2022, is analyzed. Rainfall series was obtained using data
indices defined in Table 1. Original temperature time series present very
from the weighing pluviometer, that among the various sensors is the
few missing data: 97 for the outdoor GrR sensor, 9 for each of the other
one offering the highest accuracy. Some missing data (1.7% in total)
three surface temperature sensors, and 14 for the air temperature sensor,
have been filled with data from the disdrometer and the tipping-bucket
on a total of 35,520 data per series. Daily temperature data, arising from
gauge. Besides the daily rainfall depth (P in mm), the weather at each
time aggregation of 30-min data with more than 13% of missing data per
day is also characterized by the mean daily air temperature (Tair in ◦ C)
day (i.e. more than 3 h per day) have been excluded from the analysis,
measured by the weather station.
leading to a total of only three daily data removed from the outdoor GrR
The two monitored years, especially 2022, were characterized by less
sensor (i.e. a period from 22 to 24 Feb. 2020 during which the sensor
rain (on average 684 mm/year) and higher temperature (on average
was under maintenance).
19.6 ◦ C) compared to the long period averages for Palermo (Fig. 2).
Surface Temperature Reduction (STR), External Temperature Ratio
Fig. 3 shows the frequency histograms of P (left) and Tair (right) over the
(ETR) and Temperature Excursion Reduction (TER) reported in Table 1
monitoring period, distinguishing in this last the histogram relative to
are typical indices of literature [31,50]; the first compares the average
only rainy days (blue bars) and that relative to no-rainy days (red bars).
daily surface temperature at the MGR with that measured at the GrR,
It can be noticed how most of the significant (i.e., P > 2 mm) rainy days
with regard to both indoor (STRin) and outdoor (STRout) sensors. The
(i.e. 106 out of a total of 121 rainy days) were characterized by light/
External Temperature Ratios (ETRmax and ETRmin) are given by ratio
moderate daily rainfall depth, i.e. lower than 20 mm according to the
between the maximum (or minimum) daily surface temperature from
[46] classification. Moreover, rainy days are characterized by low
the outdoor sensor and the maximum (or minimum) daily air temper­
temperature, since they are mainly concentrated during the coldest
ature, and they are computed separately for both the MGR and the GrR.
winter/spring seasons.

Fig. 2. Climate in Palermo (data from the SIAS gauge

station ID 276-Uditore: 2002–2022): mean monthly
rainfall (mm/month, blue solid line), mean daily
temperature (◦ C, red solid line) and its monthly
variation range (shadowed area). Dashed lines refer
to data of rainfall (blue) and temperature (red)
collected at the experimental site during 2021 (cir­
cles) and 2022 (squares). (For interpretation of the
references to color in this figure legend, the reader is
referred to the Web version of this article.)

4
D. Pumo et al. Building and Environment 243 (2023) 110731

Fig. 3. Monitoring period: from Dec. 22nd, 2020, to Dec. 31st, 2022. Frequency histograms of daily rainfall depth (left plot) and mean daily air temperature (Tair,
right plot). Blue color refers to rainy days, while red refers to no-rainy days. (For interpretation of the references to color in this figure legend, the reader is referred to
the Web version of this article.)

TER index refers to the daily surface temperature excursion (Texc), given temperature (21.8 ◦ C), approaching the mean air temperature value
by the difference between the maximum and the minimum roof surface (19.5 ◦ C). Nevertheless, it can be noticed how the maximum daily sur­
temperature measured in a day; more specifically, TER compares MGR face temperature at the MGR indoor sensor exceeds 32 ◦ C in only 33 days
and GrR roof temperature excursions, and it is computed considering out 740, with maximum value equals to 32.8 ◦ C, while at the GrR indoor
both the indoor (TERin) and the outdoor (TERout) data. sensor this value was overcome in 123 days, with a maximum of 35.1 ◦ C.
Mean seasonal values and mean values over the entire period of the
3. Results minimum, average and maximum daily temperature recorded at the
four surface temperature sensors and at the weather station are sum­
3.1. Temperature comparison at the MGR and the GrR marized in Table 2, where it can be noticed how, as it was expected, on
average the minimum daily temperature at the MGR outdoor sensor is
Green roofs are known to be capable of reducing roof surface tem­ always relevantly higher than that at the GrR outdoor sensor, especially
perature and daily temperature excursion compared to grey roofs [21]. during the spring (+6.6 ◦ C) and, at the same time, the maximum tem­
At equal substrate depth, MGRs in a semiarid environment are expected perature at the MGR outdoor surface is always relevantly lower than
to further improve building thermal insulation, especially during hot those at the GrR outdoor sensor, especially during the hottest season of
summer conditions, due to the presence of a further insulation layer the year, i.e. the summer (-15.2 ◦ C). The indoor MGR surface is always
filled by air and water with a dynamic ratio. The mean daily temperature colder than the indoor GrR surface, especially during the spring, with
and the daily temperature excursion over the entire reference period reductions in terms of maximum temperature higher than for the mean
measured on the roof surface by the MGR and GrR sensors are reported and the minimum daily temperature, leading to an overall reduction by
in Fig. 4, where they are also compared with the mean daily air tem­ the 30% of the daily temperature excursion for the MGR compared to the
perature measured by the weather station (red line). Upper graph refers GrR.
to the outdoor sensors, while bottom graph to the indoor sensors. As it Fig. 5 shows a comparison between the daily surface temperature
clearly appears from the upper graph, the outdoor GrR average surface profiles for the “mean day” of the four seasons. Each profile is derived
temperature follows approximately the average air temperature during averaging all the data recorded during a given season by a given sensor,
the colder months of the year, while it is on average sensibly higher than according to the original 30-min time-resolution of acquisition. In each
air temperature during the hotter summer months; moreover it is graph, the corresponding mean daily air temperature profile is also
characterized by a pronounced daily excursion, highlighted by grey reported.
shadow area, especially during the summer where it reached almost From the analysis of the left plots of Fig. 5 referring to the outdoor
32 ◦ C. The overall effect of the MGR was to reduce the outdoor surface sensors, it can be observed how, for all the seasons, the GrR surface
temperature (on average − 1.8 C◦ over the entire monitoring period temperature follows the average solar irradiance daily profile, with roof
compared to the GrR), especially during the summer season (on average surface colder than the air temperature during the night, and relevantly
− 2.8 ◦ C) where in some days of August the outdoor average surface hotter during the daylight hours. The MGR shows an inverse behavior,
temperature of the GrR was also over 8 ◦ C higher than in the MGR. The with surface temperature higher than air temperature during the night,
decay of the daily thermal excursions in the vegetated roof, highlighted especially during the hottest spring/summer seasons, and lower during
by green shadow area, compared to the unaltered roof is even more the hottest daylight hours, especially during the coldest autumn/winter
evident; the mean daily excursion recorded at the outdoor MGR sensor seasons, reaching, for all the seasons, a minimum around noon and a
was equal to 2.9 ◦ C (vs. 21.4 ◦ C for the GrR), while the maximum values, maximum approximately 2 h after the sunset.
occurring at the end of the spring when the differences between diurnal This behavior could be explained by the fact that the MGR outdoor
and nocturnal temperatures are usually more relevant, were around sensor is installed 5 cm below the GL surface and soil-vegetation systems
8.8 ◦ C for the MGR (vs. 31.7 ◦ C for the GrR). This effect can be inter­ are typically characterized by higher albedo and thermal inertia
preted as clear evidence of the increased thermal inertia due the MGR compared to conventional rooftop paving systems [51], which implies
presence. an attenuation of the solar irradiation thermal loads and temperatures
The analysis of indoor temperature in the bottom graph of Fig. 4 daily variability, respectively. This, together with the augmented ther­
shows how the mean daily MGR surface temperature (20.5 ◦ C) over the mal insulation due to the presence of an extra layer with respect to the
entire reference period is lower than the mean daily GrR surface GrR, also explains the differences between the mean daily profiles of the

5
D. Pumo et al. Building and Environment 243 (2023) 110731

Table 1 thermal comfort for internal spaces, this could appear as a possible
Monitored variables and indices (temp. = temperature; surf. = surface; BL= Blue disadvantage of the MGR since, during the winter, roof surfaces are
Layer). frequently below the internal comfort temperature, and, therefore, a
Symbol Unit Definition lower surface temperature for the MGR could further exacerbate po­
System Antecedent Dry ADWP day Time to the last rainy
tential thermal discomfort. However, this is partially compensated by
Antecedent Weather Period day before the day of the lower temperatures occurred during nighttime at the GrR because of
Conditions evaluation the intense heat exchange with the sky; this aspect explains why indoor
Initial Degree of IDWS % Percent ratio between mean daily temperatures at the MGR are not so dramatically distant
Water Storage the mean daily BL water
from the reference roof during the winter. Moreover, the gain in terms of
and the BL thickness (10
cm) thermal insulation for the MGR, demonstrated by the significant
Antecedent API5 mm Cumulative reduction of the daily temperature excursion also during the winter,
Precipitation precipitation over the limits negative heat fluxes from indoor spaces to the outdoor environ­
Index last 5 days, including the ment; this favors a slower thermal dispersion with a more stable internal
day of evaluation
Antecedent ATI5 ◦
C Mean daily temperature
air temperature, which implies a potential lower effort and energy
Temperature over the last 5 days, consumption for heating internal spaces.
Index including the day of
evaluation 3.2. Evaluation of the thermal performances
Antecedent ARI5 mm Cumulative outflow
Runoff Index from the MGR over the
last 5 days, including the The quantitative evaluation of the thermal benefits of the MGR is
day of evaluation here performed through the computation of the indices listed in Table 1;
Rainfall Daily Rainfall P mm Total amount of daily the results, in terms of daily indices over the entire period and each
Depth rainfall
season, are displayed as boxplots in Fig. 6. STR and TER indices (upper
Temperature Air Temperature Tair ◦
C Air temp. from the
weather station
plots) provides an immediate comparison of the external/internal sur­
Outdoor MGR Tout, ◦
C Outdoor temp. on the faces temperature between vegetated and unaltered roofs; red boxplots
Temperature MGR external surface of the refer to indices computed using data from the outdoor sensors, while
MGR black boxplots are computed using data from those indoor. The STR
Outdoor GrR Tout, ◦
C Outdoor temp. on the
values (top-left plots) for all the analyzed periods are prevalently lower
Temperature GrR external surface of the
GrR than unit, denoting average daily surface temperature for the MGR
Indoor MGR Tin,MGR ◦
C Indoor temp. on the lower than for the GrR for both outdoor and indoor roof surfaces. The
Temperature internal surface of the outdoor indices behavior (red boxplot) derives from the fact that heat
MGR exchange between the sky and the GrR surfaces is more intense than for
Indoor GrR Tin,GrR ◦
C Indoor temp. on the
Temperature internal surface of the
the MGR, since the absorptance of the concrete surfaces, measuring the
GrR effectiveness in absorbing radiant energy, is more elevated than for
Daily Texc ◦
C Difference between max vegetated surfaces. This can generate a significant diurnal heat gain for
Temperature and min daily temp. the GrR outdoor surface, as it can be also noticed from the comparison
Excursion
among the outdoor temperature daily profiles in Fig. 5. The outdoor STR
Thermal Surface STR - = (Indoor/Outdoor)
Indices Temperature MGR/(Indoor/Outdoor) median value over the entire period is equal to 0.91, while it ranges from
Reduction GrR mean daily surf. 0.85 in winter to 0.94 in autumn, with the lowest variability of the index
temp. during the summer (Inter-Quartile Range, IQR = 0.07) and the highest
External ETR - = (MGR/GrR) Outdoor (IQR = 0.18) during the winter. The indoor STR boxplots (black boxplots
Temperature surf. (Min/Max) daily
Ratio temp./Air (Min/Max)
in top-left panel of Fig. 6) confirm the well-known cooling effect of
daily temp. vegetated roofs. In this case, indices are less variable than outdoor STR
Temperature TER - = (Indoor/Outdoor) values for all the seasons; the indoor STR median value over the entire
Excursion MGR/(Indoor/Outdoor) period is equal to 0.94, and its seasonal value ranges from 0.91 in spring
Reduction GrR daily temp.
to 0.98 in autumn, with corresponding values of IQR about from one-
excursion
quarter to one-half of the values referring to the outdoor STR, and the
lowest and the highest variability again during the summer (IQR = 0.02)
indoor roof surface temperature for the MGR and the GrR, shown in the and the winter (IQR = 0.10), respectively.
right plots of Fig. 5. For all the seasons, the two profiles for the MGR and ETR indices, shown in bottom plots of Fig. 6, provide a comparison
the GrR appear almost specular, with the former laying lower than the between outdoor roof surface temperature and air temperature,
latter, except for 3 h in autumn, especially during the spring and summer considering both ETRmin and ETRmax daily values with green and blue
seasons, when differences in the mean temperature of some night hours boxplots for both indices referring to the MGR and the GrR, respectively.
reached values over 2 ◦ C. Differently from the GrR profiles, in the MGR, From the boxplots analysis, it is possible to notice that the ETRmax was
the shape of the indoor mean daily surface temperature profiles follows constantly above the unit for the GrR and prevalently below one for
that of the corresponding outdoor profiles, even if with much less MGR, while an opposite behavior can be observed with respect to the
marked hourly temperature fluctuations; the maximum temperature ETRmin. These results can be addressed to the fact that, on the one hand,
excursion from the indoor mean daily profile of the MGR is about 0.8 ◦ C, the maximum values of the outdoor daily surface temperature for the
occurring in spring and summer, while it reaches 4.6 ◦ C during the conventional grey roof were constantly higher than air temperature
summer for the outdoor profile. maxima due to the marked influence of solar irradiance, and, on the
As it emerges from Fig. 5 and Table 2 and in winter months, MGR other hand, the vegetated roof, especially during the colder seasons,
showed outdoor/indoor maximum surface temperatures markedly could be also subjected to a cooling effect due to negative heat fluxes
lower than the GrR. Indeed, in winter, the elevated absorption coeffi­ from indoor to outdoor spaces, which, as it can be observed from Fig. 5,
cient of the reference surface can generate significant heat gain through could even nullify the diurnal heat gain due to solar irradiance in the
the roof cover, which may eventually lead to mean daily surface tem­ MGR outdoor surfaces. Conversely, ETRmin prevalently below the unit
perature for the GrR higher than for the MGR. With specific regard to the for the GrR and above the unit for the MGR can be explained by the
higher thermal insulation of the MGR compared to the GrR, which

6
D. Pumo et al. Building and Environment 243 (2023) 110731

Fig. 4. Average daily roof surface temperature (Tavg) measured at the two pairs of outdoor (upper plot) and indoor (bottom plot) sensors of the MGR and the GrR.
Daily temperature excursions (Texc) for the MGR (green shadow area) and the GrR (grey shadow area) are also reported, as well as the average daily air temperature
(red dashed line). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

temperature at the MGR compared to that at the GrR, which implies

Table 2
more stable temperatures during the day. Actually, at the indoor sensors
Air temperature (Tair) and outdoor/indoor roof surface temperature for the MGR
(black boxplots) we found a median TER equal to 0.66 over the entire
and the GrR. Mean seasonal values and mean over the entire monitoring period
(FULL) for minimum (Tmin, in italics), average (Tavg, in bold), maximum (Tmin) monitoring period, with median values for the hot and dry seasons equal
daily temperature, and daily temperature excursion (Texc, in bold italics) are to 0.77 and 0.73 for the spring and summer, respectively and higher
reported in ◦ C. than those relative to the wetter and colder seasons, which are equal to
0.53 and 0.55 for the autumn and winter, respectively, and character­
AIR OUTDOOR INDOOR
ized by a similar variability over the various seasons; this emphasizes the
Tair Tout,MGR Tout,GrR Tin,MGR Tin,GrR
augmented capability in presence of the system to prevent heat loss from
FULL PERIOD Tmin 15.9 18.8 13.8 20.2 21.3 indoor spaces through the roof cover, especially during the colder
Tavg 19.5 20.2 22.1 20.5 21.8 months. TER values for the outdoor sensors (red boxplots) were much
Tmax 23.1 21.8 35.2 21.0 22.4
Texc 7.15 2.94 21.41 0.72 1.04
lower and less variable than indoor TER values. Again, the lowest TER
WINTER Tmin 9.6 10.4 6.6 12.5 13.3 median values were found during the colder seasons with values equal to
Tavg 12.7 11.2 12.9 12.7 13.6 0.07 and 0.08 for the autumn and winter, respectively, while during the
Tmax 16.2 11.9 24.6 13.0 14.1 hotter seasons the medians were equal to 0.17 for both spring and
Texc 6.59 1.49 17.95 0.54 0.88
summer. The seasonal IQR values of outdoor TER indices were, on
SPRING Tmin 15.0 20.0 13.4 20.2 22.0
Tavg 19.0 22.0 23.7 20.5 22.5 average, one order of measure lower than those for the indoor TER. It is
Tmax 22.7 24.2 37.9 21.2 23.3 worth emphasizing that the daily outdoor TER never exceeded 0.54,
Texc 7.74 4.17 24.56 1.02 1.31 which implies that the daily temperature excursion at the outdoor MGR
SUMMER Tmin 24.0 28.0 22.4 29.6 31.1 surface was always more than halved compared to that relative to the
Tavg 28.0 30.3 33.1 29.9 31.7
outdoor surface of the unaltered grey roof. TER indices analysis high­
Tmax 31.8 32.7 47.9 30.4 32.3
Texc 7.77 4.65 25.52 0.86 1.20 lights how the main thermal advantage of the use green roofs in Medi­
AUTUMN Tmin 15.4 17.0 12.9 19.0 19.2 terranean semi-arid region is its capacity to attenuate the daily
Tavg 18.4 17.8 18.8 19.2 19.5 temperature fluctuations at the outdoor roof surfaces, whose combined
Tmax 21.9 18.5 30.5 19.5 20.0
effect with the augmented thermal insulation capacity and thermal
Texc 6.50 1.48 17.64 0.47 0.80
inertia, eventually leads to a significant reduction also in the daily
temperature excursion at the indoor surface of roof.
results less sensitive to the external air temperature; this implies a
smoothing effect of the daily surface temperature excursion compared to 3.3. Influence of stored water on the thermal response of the system
the air temperature excursion that is more relevant for the MGR than for
the GrR, with a minimum temperature at the vegetated surface signifi­ Compared to traditional GRs, the water storage capacity of a MGR is
cantly higher (on average + 5 ◦ C) than at the grey surface. higher and residence times of the water stored into the system, which
Among the different indices, TER, shown in the top-right panel of depends on the operational rules of the control weir, are usually longer.
Fig. 6, is probably the most indicative of the gain in terms of thermal In order to assess the role of water in determining the thermal response
insulation capacity and thermal inertia provided using MGRs; values of the MGR to the external climate forcings, we analyzed the impact of
lower than one denote a reduction in the daily excursion of surface the indices of system antecedent conditions defined in Table 1 on the

7
D. Pumo et al. Building and Environment 243 (2023) 110731

Fig. 5. Mean daily temperature profiles for the “mean day” of the four seasons. Each plot refers to the season indicated in the title and shows daily curves relative to
mean values of the 30-min temperature records at outdoor (left plots) and indoor (right plots) sensors of the MGR (green) and the GrR (blue), and the corresponding
mean daily air temperature profile (red dashed curves). (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of
this article.)

Fig. 6. Boxplots of the daily thermal indices defined in Table 1. Boxplots are computed over the entire period (All) and the four seasons. STR and TER are displayed
in the upper plots, where boxplots referring to outdoor sensors are reported in red, while black boxplots refer to indoor sensors data. In the bottom plots, ETR derived
using maximum (left) and minimum (right) daily temperature are reported separately; green and blue boxplots refer to indices computed using outdoor sensors data
at the MGR and the GrR, respectively. Outliers are not displayed to help visualization. (For interpretation of the references to color in this figure legend, the reader is
referred to the Web version of this article.)

STR and TER indices. Long ADWP values, usually greater than three-four moisture conditions in the GL, regardless to the BL degree of fullness.
weeks, depending on the season, are always associated to a null ARI5, Figs. 7 and 8 show scatter matrices relating paired thermal and
and indicative of almost absence of water in both GL and BL; on the antecedent condition indices. In particular, scatterplots with black
contrary, not null ARI5 values occur only when the BL has been filled up points relates thermal indices with each other, scatter plots with red
to its maximum retention capacity at least once in the previous 4 days or points relates thermal indices with antecedent condition indices, scatter
at the day of evaluation, and this, in turn, implies the overcoming of the plots with green points relates antecedent condition indices with each
field capacity in the GL triggering the percolation processes that feed the other, while bar plots provide frequency histograms for each one of the
BL. Null API5 associated to high ATI5 could be representative of low soil indices.

8
D. Pumo et al. Building and Environment 243 (2023) 110731

Fig. 7. Scatter matrix considering two thermal indices (STRin and TERin) and two antecedent condition indices (ADWP and ARI5). Figure shows frequency histograms
for all indices, and correlation plots, with the associated trend line, for all the possible pairs of indices, except for ARI5 vs. ADWP, due to its scarce significance.

Fig. 8. Scatter matrix considering two thermal indices (STRout and TERout) and two antecedent condition normalized indices, API5*and ATI5*; the former is the API5
normalized with respect to the maximum, while the latter is the ATI5 normalized with respect to mean daily temperature over the reference period. Figure shows
frequency histograms for all indices, and correlation plots, with the associated trend line, for all the possible pairs of indices.

Fig. 7 refers to the indoor STR and TER indices and the ADWP and MGR. Presence of water into the system, demonstrated by ADWP in the
ARI5 indices. It can be noticed how, TERin values show a clear decreasing order of few days or not null ARI5, often leads to very variable, and
trend with STRin, assuming value above unit prevalently when STRin is sometimes over unit, STRin. The 92% of STRin higher than one occurred
below one; this implies that a daily temperature excursion at the MGR during the wetter autumn-winter months, when water is likely present
indoor surface higher than at the GrR surface often occurs in days when, into the system and STRin time series is more variable day by day. In very
however, the indoor roof surface of the GrR is hotter than that of the wet periods (i.e., ARI5>65 mm), STRin tends to assume values around

9
D. Pumo et al. Building and Environment 243 (2023) 110731

the unit; this could indicate that, during the wettest periods, the possible that, differently from the indoor thermal indices shown in Fig. 7, the two
occurrence of a significant amount of ponding rainwater onto the GrR outdoor thermal indices of Fig. 8 are concordant, with TERout increasing
for a prolonged time, generates similar conditions in the MGR and the with STRout and this last always above one in the few cases with TERout
GrR, minimizing the thermal differences between the outdoor roof above 0.3.
surfaces and leading to similar indoor roof surface temperatures in the A direct measure of the water present into the system is given by the
two roofs. On the contrary, for ADWP over 1 month that exclusively Degree of Water Storage, DWS (Table 1), whose influence on the STR and
occurred during the spring-summer period, and thus under a condition TER indices is analyzed in Fig. 9. In particular, this index was classified
of likely absence of water in the MGR, both the thermal indices are less using the following three ranges: “Empty”, when DWS = 0%; “Inter­
variable and lower than unit; it is possible then to conclude that dry mediate”, when DWS is between 0% and 50%; and “High”, when DWS is
antecedent conditions leads to more stable thermal indices, less influ­ higher than 50%, and, thus, the mean daily water level (over 5 cm) is
enced by the external air temperature daily variability and always close to the maximum weir closure position (7 cm). For this analysis, the
indicating a clear gain of the MGR compared to GrR in terms of thermal available limited sample of “direct records” of DWS was extended with
insulation and inertia. At the same time, the lowest values for STRin in “reconstructed data”, assuming an “Empty” BL when ADWP is longer
the series have been reached only in presence of water (e.g. STRin<0.85 than 30 days and “High” condition of the BL at the days with not null
only for ADWP< 10 days) and TERin tends to decreases as ARI5 outflow discharge from the MGR.
increases. From the analysis of the outdoor and indoor indices in Fig. 9, it
Fig. 8 is similar to Fig. 7 and refers to the outdoor STR and TER emerges how despite STR indices are scarcely affected from the presence
indices and the API5 and ATI5 indices which have been normalized for of water into the BL in terms of mean and median values, the dispersion
sake of visualization, the former (API5*) with respect to the maximum of the values around the mean increase with increasing water storage,
API5 in the series and the latter (ATI5*) with respect to mean daily leading to some occurrences of STRout and STRin above the unit that
temperature over the reference period. imply temperature at the MGR surfaces higher than at GrR surfaces; this
High values of API5 are indicative of recent precipitation, causing is not necessarily a disadvantage in terms of internal comfort, consid­
wet soil moisture conditions in the GL and possible water also into the ering that “High” water storage conditions prevalently occurs during the
BL, while on the contrary, dry soil moisture conditions of the GL could wetter and colder wintry months, when hotter roof surfaces could
occur when API5 is null and ATI5 is high, since both conditions favor soil reduce heating energy demand. Such a behavior could be explained by a
dryness processes. From the figure we can notice how STRout slightly gain in terms of thermal inertia due to water storage in the BL that slows
increases with API5*; thus, the possible presence of water in the GL of down heat dispersion from internal to external environments, which
the MGR, which prevalently occurs during the winter, could improve its also justifies what can be considered as the most evident impact of the
thermal performances, leading to roof surfaces of the MGR hotter than presence of stored water on the thermal response of the MGR, that is the
those of the GrR during the coldest season of the year. On the contrary, marked reduction of the daily thermal excursion shown in the right plots
higher ATI5 values, typically occurring during the hottest summer of Fig. 9. From the figure, it can be observed how passing from “Empty”
months, when API5 values are frequently null, are characterized by to “High” conditions, the distribution of both TERin and TERout indices
lower variability of the STRout, which is stably lower than one, con­ changes markedly, shifting downwards and assuming a positive skew.
firming the outcomes of Fig. 7. TERout index, which always denotes daily
temperature excursion at the outdoor MGR significantly lower than at 4. Discussion
the outdoor GrR, appears increasing with ATI5* and scarcely sensitive to
API5*, probably because other factors here not considered, such as the The analyses carried out in the present work have allowed to quan­
solar irradiance, the air humidity and the wind conditions, as well as tify the thermal performances of a relatively small MGR under a typical
their daily variability, may play a prominent role compared to the semi-arid Mediterranean climate, focusing on the modifications induced
presence of water into the GL in determining the daily temperature on the indoor/outdoor roof surface temperature.
excursion at the outdoor roof surfaces. Finally, it worth emphasizing All the analyzed indices are concordant in highlighting an overall

Fig. 9. Frequency histogram (left plot) of the three classes of DWS: Empty (DWS = 0%), Intermediate (0% ≤ DWS ≤50%), High (DWS>50%). Distribution plots
(middle and right plots) of the corresponding STRin and TERin (top), and STRout and TERout (bottom), with indication of mean (black lines) and median (red). (For
interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

10
D. Pumo et al. Building and Environment 243 (2023) 110731

cooling effect (Table 2) for the MGR outdoor surface, especially during the first attempt to assess quantitatively the influence of water storage
the warm periods, and a heat storage effect demonstrated by the on the thermal response of a GR, an aspect that assumes particular
increment of the minimum daily temperature, especially in the colder relevance for the cases of MGR, which have the potential for storing for
periods, that might favoring an easier reaching and keeping of comfort long times considerable amounts of rainwater. All the analyses that we
temperatures in the indoor environments [52]; this aspect, in turn, im­ have performed are concordant in highlighting that the MGR thermal
plies a potential consistent reduction of the yearly energy demand, and efficiency, especially in terms of daily temperature excursion reduction,
the consequent costs, for air-conditioning. Moreover, as it was demon­ increases with water storage, with median TERout for the system with
strated by the daily profiles shown in Fig. 5, the MGR layers mitigate the “High” water storage condition of the BL reduced by the 54% compared
influence of weather on the outdoor rooftop, with effective protection of to the system with an “Empty” BL (bottom-right plot of Fig. 9); this is
roof membrane against both rapid cooling during the cold winters and also coherent with [50]; where the most irrigated green roof among the
the influence of intense solar radiation during the hot summers, thus three compared was that showing the most important excursion reduc­
prolonging its lifetime [24]. tion and the lowest TER values. Nevertheless, the response of the MGR in
The mean daily surface temperature of the conventional grey roof presence of significant water storage resulted much more variable than
was found to be considerably higher than air temperature (on average + under dry conditions. Water is characterized by an order of measure
2.6 ◦ C), especially during the summer, and even more relevant differ­ higher thermal conductivity compared to air, with heat transfers
ences resulted for the maximum daily temperature (up to +16 ◦ C in the occurring at a higher rate that could determinate, in presence of water, a
summer); this might influence considerably the microclimate conditions higher sensitivity of the thermal indices to the external air temperature
above the roof. Our analysis has demonstrated how the presence of a and solar irradiation, and thus to their daily and seasonal fluctuations;
vegetated roof may significantly reduce the external mean daily roof this could explain the higher variability of the performance indices in
surface temperature; in our case we observed a reduction over the entire presence of water, when both the lowest and the highest values for both
year of almost 2 ◦ C with a reduction of − 3 ◦ C during the summer. STRin and TERin indices have be obtained.
Vegetation present in the MGR act as an insulation layer by reducing Finally, it is worth emphasizing that during the monitoring period we
convective heat flow, while soil may act as a thermal resistance and a have not applied any operational rules by the SFC system, never opening
heat sink; all this is reflected in a considerable lowering of the daily the control weir. Actually, our results could offer a new perspective for
temperature peaks, which in our experimental site reached a mean an optimal management of the control weir in a MGR, not only aimed to
reduction of about − 15 ◦ C during the summer. Moreover, the dominant minimize/prevent/delay system outflow during the intense storms in
way for vegetated surfaces to dissipate the absorbed heat is by evapo­ wet seasons and/or maximize water detention during the driest periods
transpiration [53], which contributes to the reduction of the surface for passive irrigation and/or later water reuses [42,58]. The real time
temperature of the materials and the increase of the external air hu­ monitoring of climate and water storage into the system could in fact
midity, producing an overall cooling effect for the air surrounding the allow for the implementation of automatic procedures to optimize the
building [54], which confirms the potential of green roofs to reduce UHI use of the SFC system also in terms of best thermal efficiency, to reduce
effect [55]. In multilayer GRs, this effect is even more pronounced than energy consumptions for air conditioning during both the coldest and
in conventional GRs due to the increased water storage that can supply hottest periods, varying conveniently the amount of water stored into
evapotranspiration processes. the storage layer as a function of the external weather conditions.
The reduction of the outdoor roof surface temperature is also re­
flected in a lowering of the indoor surface temperature; we observed, on 5. Conclusion
average, a difference of − 1.3 ◦ C between the mean daily temperature
recorded at the indoor MGR and GrR sensors. Moreover, the analyzed Climate change and increasing urbanization are posing growing
MGR produced a significant reduction of the mean daily temperature challenges in urban areas for which new paradigms of climate adaptive
excursion compared to the benchmark grey roof, for both outdoor and design, oriented to the criteria of resilience, sustainability, hydraulic and
indoor sensors equal to 86% and 31%, respectively, with maximum hydrological invariance are urgently needed. New measures defined
percent reduction during the summer equal to 92% and 39% at the within the concept of NBSs, represent multi-purpose techniques capable
outdoor and indoor sensors, respectively; this last reduction, in partic­ to simultaneously address various future threats in a resource-efficient
ular, provides clear insight of a higher thermal insulation capacity and and adaptable manner, also allowing for urban regeneration and
thermal inertia, which can potentially contribute in improving the en­ renewal of areas unused or subjected to the combined effect of endog­
ergy efficiency class of existing buildings. enous and exogenous criticalities.
For a conventional GR installation in Chicago (USA) [56], showed This study has presented a new and advanced green infrastructure,
that the roof surface temperature on the hottest day was 22 ◦ C lower that is the MGR that, combining the advantages of traditional green
than a reference grey rooftop. In semi-arid Mediterranean regions, such roofs with those of rainfall harvesting systems, represents a measure
as Sicily, the thermal advantages of vegetated roofs during the hotter particularly suitable in semi-arid Mediterranean regions, where the
months are more evident than during the colder season, as it is also climate could generate serious issues for the implementation of tradi­
confirmed by other studies in climatologically similar zones [33]. [57]; tional green roofs and the marked intermittent nature of rainfall regime
reported, for an installation of GR in France, an outdoor surface tem­ requires an efficient management of urban stormwater and water re­
perature reduction up to 30 ◦ C during the summer and, for the same sources. Stored water in MGRs can supply capillary water to the system
season, an indoor air temperature reduction of 2 ◦ C; in the same study, vegetation itself, preventing water stress and reducing maintenance
authors quantified a 6% reduction of the annual building energy de­ costs, and it might be also reutilized as grey water, reducing building
mand [50]. analyzed three extensive green roofs in Calabria, southern water demand and potentially mitigating water scarcity problems. In
Italy, using the same performance indices explored in our study. They this work we also highlighted as vegetated roofs can be thought as
found for outdoor sensors values of STRavg between 0.70 and 1.1, medium of temperature moderation and energy saving in the building
consistent with our results. In Ref. [50], the reference roof experienced sector; the presence of the storage layer in MGRs that may act as a
daily temperature excursions higher than the green roof, with TER further insulation layer, has the potential to increase roof thermal
values for the winter ranging from 0.44 to 0.61 and for the summer inertia, improving the mitigation effect with respect to heat waves and
ranging from 0.31 to 0.52; here, our system has experienced even lower UHI that frequently characterize the largest Mediterranean urban areas.
TER values, around 0.09, demonstrating a higher potential of the MGRs The design of such a kind of solutions is often oriented primarily to
compared to traditional extensive GRs. an efficient management of stormwater in urban areas, trying to reduce
The present study represents, to the best of the authors knowledge, the pressure on existing urban drainage system during intense storms,

11
D. Pumo et al. Building and Environment 243 (2023) 110731

and the possibility of grey water recycling. A previous analysis [42] on Declaration of competing interest
the hydrological performances of the experimental site of Palermo here
tested, in fact, highlighted how MGRs outperform traditional GRs in The authors declare that they have no known competing financial
terms of capacity to retain/detain rainwater, estimating that a large interests or personal relationships that could have appeared to influence
fraction of retained water, equal to about the 70%, could be potentially the work reported in this paper.
reused. In this work, it was explored an aspect much less investigated in
literature up to now, that is the thermal benefits that MGRs can provide Data availability
also in consideration of the degree of fullness of the water storage layer.
Due to their recent commercialization, thermal efficiency of MGRs in Data will be made available on request.
semi-arid environments had never been tested before; thus, a compre­
hensive evaluation of the full range of benefits that MGRs could pro­ Acknowledgments
vides, supported by observational studies, is essential to support the
current policy directions aimed at promoting climate adaptive urban The experimental prototype was realized within an agreement with
design paradigms. Delft University of Technology (Netherlands) as part of the EIT Climate-
Our analysis has shown how the MGR of Palermo was able to pro­ KIC program (funding received from the European Union Climatic KIC
duce a reduction of the mean daily temperature equal to 8.4% and 5.8% programme with grant number EIT SGA 2018 supporting project ‘Polder
for outdoor and indoor spaces, respectively, with a significant decre­ Roof Fieldlabs - 180522’). The authors also thank anonymous reviewers
ment of the daily maxima (up to 38.2% for outdoor spaces) especially and editor-in-chief for their suggestions on the quality improvement of
during the diurnal hours of the hottest months. All the daily thermal the present paper.
indices showed performances significantly higher than those referring to
traditional extensive green roofs with similar thickness and under References
similar climate conditions [50], also highlighting how the amount of
water stored into the storage layer may have a not negligible influence [1] C. Milesi, G. Churkina, Measuring and monitoring urban impacts on climate change
from space, Rem. Sens. 12 (21) (2020) 3494, https://doi.org/10.3390/
on the thermal response of MGRs. With this regard, a general outcome rs12213494.
was that, despite the thermal performances of the MGR in general in­ [2] D. Pumo, E. Arnone, A. Francipane, D. Caracciolo, L.V. Noto, Potential implications
crease with water storage, especially for what concerns the reduction of of climate change and urbanization on watershed hydrology, J. Hydrol. 554 (2017)
80–99, https://doi.org/10.1016/j.jhydrol.2017.09.002.
the outdoor daily surface temperature excursion, the presence of water [3] E. Arnone, D. Pumo, A. Francipane, G. La Loggia, L.V. Noto, The role of urban
into the system may also induce a high instability of thermal indices, growth, climate change and their interplay in altering runoff extremes, Hydrol.
whose distribution under wet conditions shows a dispersion larger than Process. (2018), https://doi.org/10.1002/hyp.13141.
[4] N.W. Arnell, S.N. Gosling, The impacts of climate change on river flood risk at the
that relative to days with empty storage layer and/or dry green layer. global scale, Climatic Change 134 (3) (2016) 387–401.
This effect probably occurs since the thermal response of the system in [5] I.B. Karlsson, T.O. Sonnenborg, J.C. Refsgaard, D. Trolle, C.D. Børgesen, J.
presence of water becomes more sensitive to other factors here not E. Olesen, E. Jeppesen, K.H. Jensen, Combined effects of climate models,
hydrological model structures and land use scenarios on hydrological impacts of
explored, such as solar irradiance, wind, and air relative humidity. A
climate change, J. Hydrol. 535 (2016) 301–317, https://doi.org/10.1016/j.
more complete understanding of the water influence on thermal jhydrol.2016.01.069.
response of the MGR probably would require accounting for some cli­ [6] UN, United Nations Final Report on World Urbanization Prospects 2018, 2018.
matic variables here ignored and their mutual influence, and to adopt [7] M. Swilling, M. Hajer, T. Baynes, J. Bergesen, F. Labbé, J. Musango,
A. Ramaswami, B. Robinson, S. Salat, S. Suh, et al., The Weight of Cities: Resource
indices defined on finer time-scales. Requirements of Future Urbanization; A Report by the International Resource
The reduction of the roof outdoor/indoor surfaces temperature daily Panel, United Nations Environment Programme, Nairobi, Kenya, 2018.
excursion due to presence of water into the blue layer provides a clear [8] M. Santamouris, Recent progress on urban overheating and heat island research.
Integrated assessment of the energy, environmental, vulnerability and health
indication of an augmented thermal inertia due to the installation of a impact. Synergies with the global climate change, Energy Build. 207 (2020),
MGR; this effect, coupled with others typical of vegetated roofs, such as 109482, https://doi.org/10.1016/j.enbuild.2019.109482.
shading through foliage and branches, the increase in air humidity and [9] F. Estrada, W.W.J. Botzen, R.S.J. Tol, A global economic assessment of city policies
to reduce climate change impacts, Nat. Clim. Change 7 (6) (2017) 403–406,
cooling by evapotranspiration, may be also interpreted as an extremely https://doi.org/10.1038/NCLIMATE3301.
positive outcome in terms of UHI attenuation. Other research directions [10] C. Sarrat, A. Lemonsu, V. Masson, D. Guedalia, Impact of urban heat island on
to further stress the importance of encouraging future analyses on this regional atmospheric pollution, Atmos. Environ. 40 (2006) 1743–1758.
[11] C. Cao, X. Lee, S. Liu, N. Schultz, W. Xiao, M. Zhang, L. Zhao, Urban heat islands in
type of technology include the exploration of other possible co-benefits China enhanced by haze pollution, Nat. Commun. 7 (2016), 12509.
not analyzed in the present work, such as the reduction of air and water [12] J. Liu, D. Niyogi, Meta-analysis of urbanization impact on rainfall modification,
pollution, the attenuation of noise levels, and the increase of buildings’ Sci. Rep. 9 (2019) 7301.
[13] D. Pumo, L.V. Noto, Exploring the linkage between dew point temperature and
aesthetic value and, consequentially, of their commercial value.
precipitation extremes: a multi-time-scale analysis on a semi-arid Mediterranean
region, Atmos. Res. 105508 (2021), https://doi.org/10.1016/j.
Funding atmosres.2021.105508. ISSN 0169-8095.
[14] L. Ruangpan, Z. Vojinovic, S. Di Sabatino, L.S. Leo, V. Capobianco, A.M.P. Oen, M.
E. McClain, E. Lopez-Gunn, Nature-based solutions for hydro-meteorological risk
The present work was funded by the European Union – NextGener­ reduction: a state-of-the-art review of the research area, Nat. Hazards Earth Syst.
ationEU – with grant MUR D.M. 737/2021 – research project “Multi­ Sci. 20 (2020) 243–270, https://doi.org/10.5194/nhess-20-243-2020.
layer Green Roofs: multipurpose nature based solutions towards [15] N. Seddon, A. Smith, P. Smith, I. Key, A. Chausson, C. Girardin, J. House,
S. Srivastava, B. Turner, Getting the message right on nature-based solutions to
sustainable and resilient cities”. climate change, Global Change Biol. 27 (2021) 1518–1546, https://doi.org/
10.1111/gcb.15513.
CRediT authorship contribution statement [16] C.M. Raymond, N. Frantzeskaki, N. Kabisch, P. Berry, M. Breil, M.R. Nita,
D. Geneletti, C.A. Calfapietra, A framework for assessing and implementing the co-
benefits of nature-based solutions in urban areas, Environ. Sci. Pol. 77 (2017)
Dario Pumo: Writing – review & editing, Writing – original draft, 15–24, https://doi.org/10.1016/j.envsci.2017.07.008.
Visualization, Validation, Supervision, Software, Resources, Project [17] UN General Assembly, Transforming Our World : the 2030 Agenda for Sustainable
Development, 2015, 21 October 2015, A/RES/70/1 available at: https://www.re
administration, Methodology, Investigation, Funding acquisition, fworld.org/docid/57b6e3e44.html. (Accessed 3 May 2023).
Formal analysis, Data curation, Conceptualization. Francesco Alongi: [18] E. Cristiano, R. Deidda, F. Viola, The role of green roofs in urban Water-Energy-
Writing – review & editing, Data curation. Marcella Cannarozzo: Food-Ecosystem nexus: a review, Sci. Total Environ. 756 (2021), 143876.
[19] G. Mihalakakou, M. Souliotis, M. Papadaki, P. Menounou, P. Dimopoulos,
Writing – review & editing. Leonardo V. Noto: Writing – review & D. Kolokotsa, J.A. Paravantis, S. Papaefthimiou, Green roofs as a nature-based
editing, Supervision, Resources, Investigation, Conceptualization.

12
D. Pumo et al. Building and Environment 243 (2023) 110731

solution for improving urban sustainability: progress and perspectives, Renew. [41] D.G. Cirkel, B.R. Voortman, T. van Veen, R.P. Bartholomeus, Evaporation from
Sustain. Energy Rev. 180 (2023), 113306, https://doi.org/10.1016/j. (Blue-)Green Roofs: assessing the benefits of a storage and capillary irrigation
rser.2023.113306 art. system based on measurements and modeling, Water 10 (9) (2018) 1–21, https://
[20] G. Ercolani, E.A. Chiaradia, C. Gandolfi, F. Castelli, D. Masseroni, Evaluating doi.org/10.3390/w10091253.
performances of green roofs for stormwater runoff mitigation in a high flood risk [42] D. Pumo, A. Francipane, F. Alongi, L.V. Noto, The potential of multilayer green
urban catchment, J. Hydrol. 566 (2018) 830–845, https://doi.org/10.1016/j. roofs for stormwater management in urban area under semi-arid Mediterranean
jhydrol.2018.09.050. climate conditions, J. Environ. Manag. 326 (2023), 116643, https://doi.org/
[21] K.L. Getter, D.B. Rowe, J.A. Andresen, S.I. Wichman, Seasonal heat flux properties 10.1016/j.jenvman.2022.116643.
of an extensive green roof in a Midwestern U.S. climate, Energy Build. 43 (2011) [43] E. Cristiano, A. Annis, C. Apollonio, D. Pumo, S. Urru, F. Viola, R. Deidda,
3548–3557, https://doi.org/10.1016/j.enbuild.2011.09.018. R. Pelorosso, A. Petroselli, F. Tauro, S. Grimaldi, A. Francipane, F. Alongi, L.
[22] D. Suszanowicz, A. Kolasa Więcek, The impact of green roofs on the parameters of V. Noto, O. Hoes, F. Klapwijk, B. Schmitt, F. Nardi, Multilayer blue-green roofs as
the environment in urban areas—review, Atmosphere 10 (2019) 792, https://doi. nature-based solutions for water and thermal insulation management, Nord.
org/10.3390/atmos10120792. Hydrol 53 (9) (2022) 1129–1149, https://doi.org/10.2166/nh.2022.201, 2022.
[23] J. Yang, Q. Yu, P. Gong, Quantifying air pollution removal by green roofs in [44] F. Lo Conti, A. Francipane, D. Pumo, L.V. Noto, Exploring single polarization X-
Chicago, Atmos. Environ. 42 (31) (2008) 7266–7273, https://doi.org/10.1016/j. band weather radar potentials for local meteorological and hydrological
atmosenv.2008.07.003. applications, J. Hydrol. 531 (Part 2) (2015) 508–522, https://doi.org/10.1016/j.
[24] M.G. Rasul, L.K.R. Arutla, Environmental impact assessment of green roofs using jhydrol.2015.10.071.
life cycle assessment, Energy Rep. 6 (1) (2020) 503–508, https://doi.org/10.1016/ [45] A. Francipane, D. Pumo, M. Sinagra, G. La Loggia, L. Noto, A paradigm of extreme
j.egyr.2019.09.015. rainfall pluvial floods in complex urban areas: the flood event of 15 July 2020 in
[25] K. Vijayaraghavan, Green roofs: a critical review on the role of components, Palermo (Italy), Nat. Hazards Earth Syst. Sci. 21 (2021) 2563–2580, https://doi.
benefits, limitations and trends, Renew. Sustain. Energy Rev. 57 (2016) 740–752. org/10.5194/nhess-21-2563-2021.
[26] M. Hanzl, Urban Forms and Green Infrastructure – the Implications for Public [46] P. Alpert, T. Ben-Gai, A. Baharad, Y. Benjamini, D. Yekutieli, M. Colacino,
Health during the COVID-19 Pandemic, Cities & Health, 2020, pp. 1–5. L. Diodato, C. Ramis, V. Homar, R. Romero, S. Michaelides, A. Manes, The
[27] A. Jenkins, Biotic Systems as a Critical Urban Infrastructure during Crisis: Learning paradoxical increase of Mediterranean extreme daily rainfall in spite of decrease in
from the COVID-19 Pandemic, Cities & Health, 2020, pp. 1–3. total values, Geophys. Res. Lett. 29 (2002) 1–31, https://doi.org/10.1029/
[28] Jing Xie, Shixian Luo, Katsunori Furuya, Dajiang Sun, Urban parks as green buffers 2001GL013554.
during the COVID-19 pandemic, Sustainability 12 (2020) 6751. [47] I. Gnecco, C. Berretta, L.G. Lanza, P. La Barbera, Storm water pollution in the urban
[29] U. Berardi, A. GhaffarianHoseini, A. GhaffarianHoseini, State-of-the-art analysis of environment of Genoa, Italy. Atmos. Res. 77 (1–4) (2005) 60–73, https://doi.org/
the environmental benefits of green roofs, Appl. Energy 115 (2014) 411–428. 10.1016/j.atmosres.2004.10.017.
[30] A. Niachou, K. Papakonstantinou, M. Santamouris, A. Tsangrassoulis, [48] H. Kasmin, V.R. Stovin, E.A. Hathway, Towards a generic rainfall-runoff model for
G. Mihalakakou, Analysis of the green roof thermal properties and investigation of green roofs, Water Sci. Technol. 62 (4) (2010) 898–905, https://doi.org/10.2166/
its energy performance, Energy Build. 33 (2001) 719–729. wst.2010.352.
[31] A. Teemusk, Ü. Mander, Greenroof potential to reduce temperature fluctuations of [49] M. Razzaghmanesh, S. Beecham, The hydrological behaviour of extensive and
a roof membrane: a case study from Estonia, Build. Environ. 44 (2009) 643–650. intensive green roofs in a dry climate, Sci. Total Environ. 499 (2014) 284, https://
[32] M.A. Polo-Labarrios, S. Quezada-García, H. Sánchez-Mora, M.A. Escobedo- doi.org/10.1016/j.scitotenv.2014.08.046, 269.
Izquierdo, G. Espinosa-Paredes, Comparison of thermal performance between [50] P. Bevilacqua, D. Mazzeo, R. Bruno, N. Arcuri, Surface temperature analysis of an
green roofs and conventional roofs, Case Stud. Therm. Eng. 21 (2020), 100697. extensive green roof for the mitigation of urban heat island in southern
[33] C.M. Silva, M.G. Gomes, M. Silva, Green roofs energy performance in mediterranean climate, Energy Build. 150 (2017) 318–327, https://doi.org/
Mediterranean climate, Energy Build. 116 (2016) 318–325. 10.1016/j.enbuild.2017.05.081. ISSN 0378-7788.
[34] T.B. Carson, D.E. Marasco, P.J. Culligan, W.R. McGillis, Hydrological performance [51] G. Mutani, V. Todeschi, The effects of green roofs on outdoor thermal comfort,
of extensive green roofs in New York City: observations and multi-year modeling of urban heat island mitigation and energy savings, Atmosphere 11 (2) (2020) 123,
three full-scale systems, Environ. Res. Lett. 8 (2) (2013), 024036, https://doi.org/ https://doi.org/10.3390/atmos11020123.
10.1088/1748-9326/8/2/024036. [52] L. Pastore, R. Corrao, P.K. Heiselberg, The effects of vegetation on indoor thermal
[35] P.A. Versini, D. Ramier, E. Berthier, B. de Gouvello, Assessment of the hydrological comfort: the application of a multi-scale simulation methodology on a residential
impacts of green roof: from building scale to basin scale, J. Hydrol. 524 (2015) neighborhood renovation case study, Energy Build. 146 (2017) 1–11, https://doi.
562–575, https://doi.org/10.1016/j.jhydrol.2015.03.020. org/10.1016/j.enbuild.2017.04.022.
[36] G. Krebs, K. Kuoppamäki, T. Kokkonen, H. Koivusalo, Simulation of green roof test [53] R.M. Lazzarin, F. Castellotti, F. Busato, Experimental measurements and numerical
bed runoff, Hydrol. Process. 30 (2016) 250–262. modeling on a green roof, Energy Build. 37 (2005) 1260–1267.
[37] L.V. Noto, G. Cipolla, A. Francipane, D. Pumo, Climate change in the [54] B. Barozzi, A. Bellazzi, M.C. Pollastro, The energy impact in buildings of vegetative
Mediterranean basin (Part I): induced alterations on climate forcings and solutions for extensive green roofs in temperate climates, Buildings 6 (2016) 33,
hydrological processes, Water Resour. Manag. 37 (2023) 2287–2305, https://doi. https://doi.org/10.3390/buildings6030033.
org/10.1007/s11269-022-03400-0, 2023. [55] A. Solcerova, F. van de Ven, M. Wang, M. Rijsdijk, N. van de Giesen, Do green roofs
[38] L.V. Noto, G. Cipolla, D. Pumo, A. Francipane, Climate change in the cool the air? Build. Environ. 111 (2017) 249–255, https://doi.org/10.1016/j.
Mediterranean Basin (Part II): a review of challenges and uncertainties in climate buildenv.2016.10.021.
change modeling and impact analyses, Water Resour. Manag. 37 (6–7) (2023) [56] J. Sonne, Evaluating green roof energy performance, ASHRAE J. 48 (2006) 59.
2307–2323, https://doi.org/10.1007/s11269-023-03444-w, 2023. [57] I. Jaffal, S.-E. Ouldboukhitine, R. Belarbi, A comprehensive study of the impact of
[39] E. Andenæs, T. Kvande, T.M. Muthanna, J. Lohne, Performance of blue-green roofs green roofs on building energy performance, Renew. Energy 43 (2012) 157–164.
in cold climates: a scoping review, Buildings 8 (2018) 55, https://doi.org/ [58] T. Busker, H. de Moel, T. Haer, M. Schmeits, B. van den Hurk, K. Myers, et al., Blue-
10.3390/buildings8040055. green roofs with forecast-based operation to reduce the impact of weather
[40] M. Shafique, R. Kim, D. Lee, The potential of green-blue roof to manage storm extremes, J. Environ. Manag. 301 (2022), 113750.
water in urban areas, Nat. Environ. Pollut. Technol. 15 (2) (2016) 715–718.

13