See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/237247828

Seismic Response of Green Roofs

Conference Paper · April 2009

DOI: 10.13140/2.1.2525.8564

CITATIONS READS

3 1,551

5 authors, including:

Piotr Omenzetter George Charles Clifton

University of Aberdeen University of Auckland
179 PUBLICATIONS   1,937 CITATIONS    291 PUBLICATIONS   2,967 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Cyclic, plastic shear and the effects of strain ageing on eccentrically braced frame active links View project

Structural Fire Engineering View project

All content following this page was uploaded by George Charles Clifton on 27 May 2014.

The user has requested enhancement of the downloaded file.

Seismic Response of Green Roofs

M.O. Carmody, M. Jasarevic, P. Omenzetter, G.C. Clifton &

E. A. Fassman
Department of Civil and Environmental Engineering, University of
Auckland, Auckland, New Zealand

2009 NZSEE
Conference

ABSTRACT: Green roofs consist of vegetation with a light-weight substrate planted

over a drainage layer and waterproof membrane. The green roof retains rainwater in the
plants and substrate and releases the water through evapotranspiration and some surface
drainage. This research explored the green roof’s ability to resist seismic forces and the
potential for the green roof to be used as an energy dissipater to reduce the response of
the building to lateral dynamic loading, such as wind loading. Shake table tests showed
the substrate layer was capable of resisting most seismic events and with the addition of
plants it became resistant to all but the most severe earthquake events. Varying the
moisture conditions showed that up until the substrate started responding to excitations as
a liquid, the additional water was beneficial to the survivability of the substrate layer. The
mixing and sloshing of free water within the drainage layer will provide additional
damping to a structure. Laboratory-scale dynamic tests proved the concept of using free
water within the drainage layer as a source of additional damping and enabled
quantification of the damping effects. Numerical simulations demonstrated that
significant additional damping could be achieved in full-scale buildings susceptible to
excessive wind-induced vibrations.

1 INTRODUCTION
With increasing imperviousness of urban areas, stormwater runoff is becoming a prominent issue. An
emerging solution in stormwater retention is the green roof. Green roofs are partially or completely
covered in vegetation with a light-weight substrate planted over a drainage layer and waterproof
membrane. For smaller storms, the green roof retains precipitation in the plants and substrate and
releases the water through evapotranspiration. During larger events, some runoff may be discharged
via the drainage layer. In either case, stormwater runoff volume is eliminated or reduced, and peak
flow and runoff timing is mitigated, thus reducing the load on stormwater networks. Green roofs offer
many other benefits, including insulation and CO2 absorption (Getter and Rowe, 2006) among others.

The two main types of green roofs are extensive and intensive. Typically, extensive green roofs have a
substrate depth of no more than 150mm (Figure 1 and 2), and have lower maintenance requirements.
Intensive green roofs have a much greater substrate depth and are more structurally demanding
(Fassman and Simcock, 2007). Extensive green roofs are better suited for retrofit of existing buildings
due to their usage of light weight materials, but are also effective for new buildings.

Paper Number 31
<150 mm

Figure 1: Components of an extensive green roof Figure 2: Typical green roof drainage mat
(http://www.alumasc-exteriors.co.uk/)

Green roofs offer many benefits, but a lack of knowledge in how they behave under severe earthquake
excitation, and the influence they have on the structures supporting them may restrict their wider
application throughout New Zealand and in many other countries. This research project focussed on
the response of extensive green roofs to levels of acceleration associated with severe earthquakes and
their ability to reduce building accelerations under in-service lateral loading through increased
damping.

Substrate composition used for a green roof may vary, but typically, it is characterised by low density
with high field capacity and permeability to keep the load on the roof low while capturing optimal
amounts of water. Fassman and Simcock (2007) have developed a substrate blend comprised of
locally (New Zealand) sourced materials, including 50% pumice, 30% zeolite and 20% organic
material, which is currently subject to field trial for stormwater retention in Auckland. The plants used
have to be hardy and capable of surviving long dry periods (Getter and Rowe, 2006). A variety of
plants are currently being trialled on Auckland green roofs, including non-native sedum varieties and
several native species.

Tuned Liquid dampers (TLDs) are used on structures to dampen the response during excitations.
TLDs are essentially pools of water that utilise the sloshing of water, during excitations, to dampen
movement (Reed et al., 1998). TLDs share many of the same properties as a green roof’s drainage
layer raising the possibility of the green roof acting like a TLD with the drainage layer having multiple
barriers to smooth flow, creating turbulent flow to dissipate energy.

The purpose of this research has been to establish how the presence of a green roof on a structure
affects its structural response during lateral excitation and to better understand the performance of the
green roof itself under these excitations. To this end, the forces and frequencies that cause failure in
the substrate layer, the influence of water content and plants on substrate performance and the level of
damping provided by the green roof to the structure have been investigated.

2 METHODOLOGY
2.1 Definitions
Testing was performed at various conditions of substrate moisture content. Engineered substrates
common to green roof applications have different moisture retention properties compared to typical
soils; hence, it is important to define parameters in this context. A “good” green roof substrate
combined with proper drainage layer will have very high permeability in order to minimize structural
loading and prevent detrimental impacts to plant health. It is unlikely (and unintended) that an
extensive green roof should ever be fully saturated. Saturation is specifically defined as the condition
where all pore space is occupied by water. A very wet substrate is described by its field capacity. Field
capacity is a density term that gives a measure of the ability to hold (store) water against gravity. In
terms of stormwater management, field capacity is used to determine the maximum probable substrate
moisture content and weight during and immediately after a significant storm event (significant
meaning the storm depth exceeds the storage capacity of the substrate). In the current research, the

2
system used for the substrate strength tests (refer to Section 2.2) was not designed to release water in
the drainage layer. In some tests, water rose into the substrate, exceeding saturation and allowing the
system to respond to excitations as a liquid.

2.2 Substrate strength test

The substrate strength test experiment is designed to test how the substrate holds together and to what
extent the substrate can resist the forces acting upon it during dynamic excitation, while maintaining
its functionality. The substrate investigated is a mix (described above) that is used on the University of
Auckland’s green roof (Fassman and Simcock, 2007 and Fassman et al., 2008). Movement in the
substrate greater than 15mm was deemed to be detrimental to the integrity of green roof.

A small-scale green roof model was constructed for the substrate strength test (Figure 3). The dimen-
sions of the green roof model base are 1200mm long by 800mm wide plywood with holes drilled to al-
low a rigid connection with the shake table. The portion holding the substrate and other green roof
layers was 1200mm long, 600mm wide and 150mm high to allow movement unaffected by end ef-
fects. Acrylic, a transparent material, was used for the sides so the movement in the layers could be
observed.

First, the worst case scenario was tested, where only 50mm of substrate is placed on top of the drain-
age layer with no plant material. To track the movement in the substrate layer, skewers were set up-
right and to the bottom of the substrate layer at even intervals. String lines were placed 15mm either
side of the skewers to judge the displacement in the layer. The skewers were cut such that they only
reached the level of the string; this negates the effects of a top heavy skewer amplifying the effects felt
in the substrate. The moisture contents were measured for all the substrate tests to ensure adequate re-
cording of the testing done on the substrate and to determine the effect of the moisture in the substrate.

The green roof model was placed on the shake table and tested at a range of sinusoidal accelerations
starting at 0.05g, which were then incrementally increased until detrimental forces were found at
frequencies of 0.5, 0.75, 1, 1.25, 1.5, 1.75 and 2Hz. The accelerations and frequencies were chosen to
represent the range of excitations experienced in New Zealand for serviceability and ultimate limit
state earthquakes. After testing the substrate was re-laid to ensure that each test was unaffected by any
previous damage to the substrate. Once the detrimental frequencies and forces were determined for the
plant free green roof at the initial moisture content, two litres of water was uniformly distributed over
the entire layer and the testing continued until the substrate system began to respond as a liquid. The
liquid state is characterized by little resistance provided to flow as the system is excited. This
experiment showed how the moisture conditions impact on the ability to resist horizontal acceleration.

The subsequent testing used native Selliera plants; it was expected that the root systems of the plants
would hold the substrate together and result in much greater resistance to the seismic forces (Phillips,
2005). The plants were given six weeks to get established in the substrate layer and then tested at each
of the detrimental frequencies and forces found for the system without plants. Where the green roof in-
tegrity was maintained, the force was subsequently increased for each frequency. The ability to deter-
mine the detrimental force for a system with plants was limited by the testing equipment.

2.3 Damping enhancement test

A damping enhancement experiment was designed to quantify the effectiveness of the green roof in
mitigating a structure’s response to excitation through generating increased damping. Initial moisture
conditions were measured to determine the equivalent mass of the green roof. Subsequent tests
stepped through a range of moisture conditions until field capacity was reached, as per Table 1. The
substrate used in this test was the same mix used in the first test. The sloshing phenomenon of the
water within the drainage mat (Figure 1) was expected to be the mechanism of any possible additional
damping. The drainage mat was approximately 8mm thick and had drainage pockets into which the
water could flow (Figure 1 and 2).

3
To find the damping ratio, a simple structural model was used, consisting of flexible aluminium col-
umns, steel floors and a wooden base. A small green roof model of dimensions 650mm square was
chosen to fit our model structure. The portion holding the substrate and other green roof layers was a
450mm square, 65mm high. The base was made of plywood and the sides of timber with holes drilled
to allow a strong connection with the bookshelf model. The structure was attached to the floor and
calibrated accelerometers were placed on the green roof model (Figure 4). A MATLAB code was used
to capture the accelerometer output at a rate of 200Hz. The accelerometers were attached to the struc-
ture on opposite corners to allow for calculations to remove any torsional effects.

An initial equivalent rigid mass, in the form of immovable weights, clamped symmetrically to the
model, was tested to establish the baseline properties. Adopting the snap back test methodology, the
structure was pulled back 30-60mm (to ensure the model remains in the elastic range), suddenly re-
leased by cutting the pulling chord and the free vibration acceleration response recorded. Each test was
repeated three times to provide enough values for an accurate average. Testing was then repeated for
the system with the substrate at its initial moisture condition and then with additional water until there
was free water in the drainage layer. This process was followed for the bookshelf structure with one
storey and for three stories to get a range of natural frequencies. To capture the desired frequency for
the three storey model required a band pass, or low pass, filter to be designed using MATLAB to
eliminate all other frequencies of response. The first translational mode of the three storey bookshelf
structure was at 1.3Hz; therefore the filter was designed to encompass frequencies close to this value.

Figure 3: Shake table model set up. Figure 4: Snap back test model set up

3 RESULTS

3.1 Substrate strength testing

Each set of tests conducted on the shake table had a frequency and ground accelerations recorded. For
each frequency the ground acceleration which was deemed to be detrimental, i.e. resulting in substrate
displacement of more than 15mm, was plotted as shown in Figure 5 and fitted with a solid line. For the
system with plants, the limitations of the shake table did not allow to subject the system to ground
accelerations which would result in failure, therefore the dotted line shows only the maximum
accelerations at each frequency applied to the system.

The equation of best fit for the substrate system without plants, y = 0.1172e0.8757x, and for the planted
system, y = 0.101e1.262x, where x is the frequency in Hz and y ground acceleration in g’s, both show an
excellent statistical fit as the R2 value exceeds 94% for both equations.

4
 Detrimental Ground Accelerations for each Frequency

1.4 1.262x
y = 0.101e
R2 = 0.9902
1.2
Ground Accelerations (g's)

0.8
0.8757x
y = 0.1172e
R2 = 0.9426
0.6

0.4
No Plants

0.2
Planted- lower limit of
detrimental forces
0
0 0.5 1 1.5 2 2.5
Frequency (Hz)

Figure 5: Detrimental ground accelerations for each frequency at nominal moisture conditions.

3.2 Damping ratio test results

Table 1 shows the damping ratios, for varying moisture conditions, for the single storey structure with
a first mode frequency of approximately 4.6Hz. Table 2 shows the damping ratios for the three-storey
bookshelf structure with a first mode frequency of approximately 1.3Hz. For each water content
conditions three or five tests were conducted. The damping ratio  and frequency f for each test were
determined using Equations (1) and (2) (Chopra, 2007). Two separate peaks of voltage measured by
accelerometers, ui and ui+n, are selected which are n peaks apart and the time at each peak, ti and ti+n, is
recorded.

f = n / (ti+n – ti ) (1)
1 u
  ln i (2)
2n u i  n

4 DISCUSSION

4.1 Substrate strength test

The results from the first round of testing on the green roof substrate showed an exponential
relationship between the frequency and ground acceleration at which the substrate failed (Figure 5).
The results showed that the substrate system without the plants was able to maintain its integrity
during the serviceability level excitations but failed readily when exposed to ultimate limit state
excitations.

For the green roof with established plants, the dotted line in Figure 5 effectively shows the limit of the
shake table but can also be considered as a (conservative) lower strength limit of the substrate that is
planted. The results show an improvement of over 70% at the higher frequencies tested. Significant
movement was seen in the substrate layer for the 2Hz tests and failure appeared imminent slightly
above the 1.30g maximum acceleration that could be applied. This means the results give a
conservative envelope at which the planted substrate layer would fail.

5
Table 1- Damping ratio and frequency of the single storey structure with the green roof model with varying
moisture conditions.

Avg. Freq.
Moisture Conditions Sample No. (Hz) Damping % Average Damping %

Translational Mode Translational mode

Mass 1 4.7 1.954

(Lump mass without 2 4.7 1.789

any water or substrate) 3 4.7 1.888 1.9

Initial 1 4.7 1.816

Moisture Content (MC) 2 4.7 1.830

(MC= 25.1%) 3 4.7 1.934 1.9

Plus 0.5L 1 4.6 1.935

2 4.6 1.838

(MC= 31.3%) 3 4.6 1.840 1.9

Plus 1.0L 1 4.6 1.982

2 4.6 1.950

(MC= 34.7%) 3 4.6 1.909 1.9

Plus 1.5L 1 4.5 2.176

(Field Capacity) 2 4.6 2.145

3 4.6 2.116 2.1

Table 2- Damping ratio and frequency of the three storey structure with the green roof model with varying
moisture conditions.
Moisture Conditions Sample No. Avg. Freq. Damping % Average Damping %

(Hz) Translational Mode Translational mode

Mass 1 1.4 0.982

(Lump mass without 2 1.4 0.794

any water or substrate) 3 1.4 0.905

4 1.4 0.773

0.8
5 1.4 0.747

Plus 1.5L 1 1.3 1.455

(Field Capacity) 2 1.3 1.279

3 1.3 1.556

4 1.3 1.296

1.4
5 1.3 1.406

6
Due to the plants being regularly watered to help them become established, the water content of the
substrate was higher, 54%, compared to when the first test on the substrate was conducted at 14%. As
a result, some of the improvement could be attributed to this increase in water content. Once the
critical frequency and amplitude of excitation was determined for the system at nominal moisture
conditions, subsequent tests were conducted at increasing moisture content. The testing showed this
additional water content would only have had a small beneficial effect. The system continued to fail at
the determined excitations but was able to resist failure for a longer period of time.

These tests demonstrated the resilience of the green roof system to horizontal accelerations. When the
green roof is first planted, it is susceptible to large seismic events but after two months of ideal
growing conditions, no matter the season (i.e. the moisture conditions within the substrate layer
provided the drainage layer is functioning and the substrate does not become saturated) the green roof
will survive most ultimate limit state events. This gives confidence in green roof applications in
seismically active locations like Wellington in New Zealand, or countries like Japan.

4.2 Damping enhancement test

The results section shows that the damping ratios of the laboratory structures without substrate or
water for the translational responses are 1.9% for the single storey model (Table 1) and 0.8% for the
three-storey model (Table 2). These results indicate that the damping ratios are significantly smaller
for the three storey system.

During testing water was added in 0.5L increments to the substrate until the model reached field
capacity and there was the maximum amount of free water in the drainage layer. For the single storey
model, little if any enhancement of damping occurred until this moisture state was reached. Under
these conditions, the translational damping ratio increased to 2.1%. As a result of the first tests, only
the state without substrate or water and the field capacity state on the three-storey model were tested.
When the model was at field capacity, the translational damping ratio increased to 1.4% from 0.8%.

In order to provide an assessment of the likely damping enhancement provided by green roof to full-
scale buildings the following numerical simulations were conducted. The single storey, three storey
laboratory structures or a full scale building with a green roof can be modelled as multi degree of
freedom lumped mass oscillators as shown in Figure 6, where n is the number of storeys and m1, m2,
…, mn are masses of each storey. m represents the mass of the green roof, and damping properties of
the green roof are represented by a viscous damper with damping coefficient c. Note there is no
potential energy change associated with the displacements of m and it is assumed it moves exactly as
the roof of the building. Using the results of snap back tests on the single storey and three storey
laboratory structures and performing modal analyses of their lumped mass models with assumed linear
1st mode shape yielded the values of the additional damping coefficient c of 2.9Ns/m at 4.6Hz and
2.7Ns/m at 1.3Hz, respectively.

Simulations were conducted for a 15 storey full-scale building of natural period of 1.5s. The building
was assumed to have 200mm concrete slab floors. Two cases of the building damping ratio without
water in the drainage layer of the green roof of 0.5% and 1% were considered, which represent typical
damping ratios in buildings susceptible to wind-induced vibrations (Kijewski-Correa and Pirnia,
2007). The green roof was assumed to cover the whole roof area. For the full-scale simulations the
damping coefficient c from the small scale experiments had to be extrapolated. It was decided that
for the full scale green roof its damping coefficient was proportional to the volume of water in the
drainage mat. This gave the value of 2765Ns/m per 1m3 of water averaged form the two previously
reported small scale results. The simulation results are reported in Figure 7 which shows the relative
increase in damping ratio due increased depth of water in the drainage mat compared to the damping
ratio without the green roof. Practically linear increase can be observed in the both cases of initial
damping ratio and 50mm or 100mm of water more than double the damping ratio for the 0.5% and 1%
case of initial damping, respectively. This simulation demonstrates that the supplementary damping
expected in real structures could be high for even moderate and practically feasible depths of water.

7
 c
m 600
mn 1.0% initial damping
0.5% initial damping
500

Increse in damping ratio [%]

400

300
m2

200

m1 100

0
0 50 100 150 200
Water depth [mm]

Figure 6: Lumped mass oscillator model. Figure 7: Increase in damping ratio for full-scale structure.

The results reported show that an increase in damping can be provided, especially when there is free
water within the drainage system. The drainage mat used was very thin, but a typical specification for
extensive green roofs. Drainage mats designed for heavier, intensive green roofs use a mat that is
>20mm thick and therefore would be able to carry greater volume of free water which could lead to
even more favourable results if applied to an extensive green roof. Keeping water in the drainage
system could be simply achieved by elevating the outlet of the drainage layer to approximately 2/3 its
height. Such a condition would provide ample opportunities for turbulent flow in advance of an
incoming weather system forecast to generate in-service winds sufficient to cause noticeable
accelerations. This, added with the potential for water tanks used to store water for watering the green
roof during dry periods, means there is further damping potential in green roofs.

These tests have demonstrated that free water present in a green roof drainage layer can add damping
to the structure in a similar way as a TLD does. Green roofs also provide other additional benefits such
as stormwater control, prolonged life of the roof, and cooling effects in summer and insulation over
winter. The green roof should become a more popular feature of today’s multi-storey buildings.
However, because the substrate loses its strength when it begins to act as a liquid, designers need to
ensure that the drainage layer is only filled so that free drainage from the substrate is maintained. This
will also ensure the health of all the plants. This configuration will provide the ideal balance between
the integrity of the green roof, the maximum potential damping provided to the structure and the mass
of the system. Designers will have to structurally design for the additional weight provided by the
permanent water in the system and depending on the structure a compromise between the level of
damping and the mass of the system may need to be reached. Careful control would need to be
adopted to control the volume of water within the drainage layer to both provide optimum damping
while maintaining life of the plants.

5 CONCLUSIONS
To test the ability of the green roof substrate layer to resist shaking in an earthquake event, tests were
undertaken on a pumice and zeolite substrate mix lightly compacted to a depth of 50mm. This
involved the worst case scenario for substrate stability, i.e. substrate without any plants. The tests
showed the strength in the substrate was sufficient to resist earthquake events close to the
serviceability design event. The substrate was also tested after Selliera plants had become established
on the layer. The addition of the Selliera plants held the substrate together and gave an increase in the
force resisted of over 70% at high frequencies. The addition of the plants means that a planted
substrate layer would be able to withstand without damage most ultimate limit state earthquakes.

8
 The results showed that the water in the substrate provided only a slight benefit to its ability to resist
the motion caused by shaking. When the substrate layer began to act as a liquid, the substrate stability
was compromised at a lower force than was previously resisted. A green roof added to a structure will
increase the damping ratio and could be used instead of, or to supplement a small TLD. A compromise
between the amount of water (weight and plant health) in the system and the level of damping will
need to be made.

To fully realise the damping potential of the green roof, water will need to be retained in the drainage
layer. This will require the raising of the outlet to the drainage layer so that water will stay in the
system. Care will need to be taken in maximising the damping potential of the roof. A fully saturated
system will dampen the movement but will also compromise the substrate layers ability to resist
motion caused by an earthquake. Leaving too much water in the system will also negatively impact on
the plants. If the plant roots are sitting in water they will rot and the plants will die. Most importantly,
if the system is constantly saturated (or even very wet), significant limitations on stormwater retention
are imposed. Finding the balance between the amount of free water in the system, the integrity of the
substrate layer, and stormwater retention will provide designers with a stormwater solution that will
dampen movement and survive earthquakes, thus making green roofs ideally suited for applications in
seismically active areas.

6 ACKNOWLEDGEMENTS
We would like to thank Dr. Robyn Simcock from Landcare Research for all her help with the green
roof materials and for the growing of the plants used in the experiments, and the Auckland Regional
Council for their financial assistance with this project. Thank you too to Quincy Ma, Sujith Padiyara
and Faisal Shabbir for their assistance with the shake table operation and Tony Daligan for help in the
Test Hall.

7 REFERENCES
Chopra A.K. 2007. Dynamics of Structures: Theory and Applications to Earthquake Engineering, 3rd edition,
Upper Saddle River, NJ, Prentice Hall.
Fassman E.A. and Simcock R. 2007. Progress Report. Extensive Green Roof Design for New Retrofit
Application in the Auckland Region.
Fassman E.A., Simcock R., Voyde, E.A. and Oldendorf, P. 2008. Progress Report. Extensive Green Roof Design
for New Retrofit Application in the Auckland Region.
Getter K.L. and Rowe D.B. 2006. The Role of Extensive Green Roofs in Sustainable Development. HortScience,
41, 1276-1285.
Kijewski-Correa T. and Pirnia J.D. 2007. Dynamic Behaviour of Tall Buildings under Wind: Insights form Full-
scale Monitoring, Structural Design of Tall and Special Buildings, 16, 471-486
Phillips C. 2005. Erosion and Sediment Control Using New Zealand Native Plants- What Do We Know? Erosion
Control Seminar.
Reed D.A., Yeh H., Ju H., Yu J. and Gardarsson S.M. 1998. Performance of Tuned Liquid Dampers under Large
Amplitude Excitation, Journal of Wind Engineering and Industrial Aerodynamics, 74-76, 923-930.

View publication stats