AppendicesText A.1

Heat flux was averaged over the whole study period for the sheltered panels, and Analysis of Variance (ANOVA) was used to compare heat flux between substrate depths (0 for controls, 7.5 and 15 cm green roofs), with the two sensor locations in each panel serving as replicates. These analyses were repeated for the following subsets of the data, by 15-minute sampling interval: when substrate surface temperatures in the 7.5 cm depth panels were > 0 ° C, <= 0 ° C, and during sunny conditions (solar radiation > 8 KJ/m2; vs. <= 8 8 KJ/m2 (cloudy or nighttime)). Average snow depths over the snow study period were analysed in the same way, but included sheltered, raised and exposed roofs. As an index of the duration of snow cover, the number of days each panel had snow depth > 3 cm. was also compared among the roof types. Mixed linear modeling with repeated measures was used to predict heat flux from snow depth, air temperature, and substrate depth, with sensor as the within-subject random factor, in the sheltered panels, using data from the days on which snow depth was measured, and air temperature for the 12:00-1:30 pm period when snow depth was measured. All subsets selection was used based on the AIC (Akaike's information criterion) to pick the best fitting model (nlme package in R). Since only the sheltered panels had heat flux transducers, one of the two sensor locations in the 7.5 cm substrate depth panel was used to determine whether substrate temperature (measured from the thermocouples at the base of the substrate layer) is a linear predictor of heat flux and can be used as an indicator of winter performance. Air temperature, snow depth and solar radiation from the intervals when snow depth was sampled were examined as potential predictors of heat flux, and the resulting best model (using all subsets selection with AIC) was validated using data from the other sensor location in the same green roof panel (TSA package in R). Substrate temperatures were compared between the roof types using ANOVA models, with substrate depth and roof type (sheltered, raised, exposed) as fixed factors. As the exposed roof only had substrate temperature data for a short period, substrate temperature from the entire sampling period was also compared just between the sheltered and raised systems. The relationship between average substrate temperature and average snow depth was compared using linear modeling. In the modular system, snow depth and duration, and substrate temperatures were compared among species, and among growth form groups using two-way ANOVA with species or group and row as fixed factors. Substrate temperature data in the modules was compared separately for the entire study period, the average across three days where there was no snow cover, and averaged across 10 days where snow depths were > 3 cm; for the latter subset, the relationship between average substrate temperature and average snow depth is assessed using a linear regression model. All analyses were conducted with the R package, v. 2.15 (R Core Team 2013).

Table A.1. Species planted in four green roof systems. * Fragaria does not have woody tissue but included as a creeping shrub as it is most similar to the species in this group.

“Sheltered” and “raised” roofs

Plant Group Species Proportion of plugs planted

Succulent Rhodiola rosea 25%Tall forb Solidago bicolor 15%Grass (bunch-forming) Danthonia spicata 13%Creeping shrub Sibbaldiopsis tridentata 13%Grass (bunch-forming) Deschampsia flexuosa 10%Succulent Sedum acre 10%Tall forb Plantago maritima 3.6%Grass (sod-forming) Poa compressa 2.3%Creeping shrub Empetrum nigrum 1.7%Succulent Sedum X rubrotinctum 0.5%Succulent Sedum spurium 0.5%

“Exposed” roofs

Plant Group Species # of plots

Tall forb Solidago bicolor 4Grass (bunch-forming) Deschampsia flexuosa 4Mixture S. bicolor and D. flexuosa 4Creeping shrub Sibbaldiopsis tridentata 4Grass (bunch-forming) Danthonia spicata 4Mixture D. spicata and S. tridentata 4

Green roof modules

Plant Group Species Figure abbreviation

Grass (sod-forming) Carex argyranthra Cara

Grass (sod-forming) Carex nigra CarnGrass (bunch-forming) Danthonia spicata DansGrass (bunch-forming) Deschampsia flexuosa DesfGrass (sod-forming) Festuca rubra FesrGrass (bunch-forming) Luzula multiflora LuzmTall forb Symphyotrichum novae-

belgiiAstn

Tall forb Solidago bicolor SolbTall forb Solidago puberula SolpCreeping shrub Fragaria virginiana* FravCreeping shrub Arctosatphylos uva-ursi ArcuCreeping shrub Empetrum nigrum EmpnCreeping shrub Sibbaldiopsis tridentata SibtCreeping shrub Vaccinium macrocarpon Vacm

Table A.2. Climate summary for study period (Winter 2010-2011) in Halifax, Nova

Scotia, Canada compared with climate norms for 1971-2000. Source: Environment

Canada, 2013. http://www.climate.weatheroffice.gc.ca. Solar radiation normals (1983 -

2005) were obtained from the NASA Langley Research Center Atmospheric Science

Data Center Surface meteorological and Solar Energy (SSE) web portal supported by the

NASA LaRC POWER Project. https://eosweb.larc.nasa.gov

Year 2010-2011

Mean Temp (°C)

Average Max Temp (°C)

Average Min Temp (°C) Rain (mm) Snow (cm)

Daily Solar Radiation- Horizontal kWh/m2/d

Nov 4.1 7.5 0.7 226.3 5.4 0.82Dec 0.8 3.5 -2.0 191.0 30.9 0.52Jan -4.6 -1.3 -7.9 112.4 61.4 0.82Feb -5.4 -0.8 -10.0 166.2 87.2 1.76Mar -0.9 4.2 -6.1 67.8 4.9 3.65

Norms 1971-2000*

Mean Temp (°C)

Max Temp (°C)

Min Temp (°C) Rain (mm) Snow (cm)

Daily Solar Radiation- Horizontal kWh/m2/d

Nov 4.5 7.9 1.2 146.8 6.9 1.58Dec -1.3 2.6 -5.1 131.7 28.5 1.26Jan -4.4 -0.2 -8.6 112.3 38.4 1.55Feb -4.1 -0.1 -8.1 76.2 37.7 2.44Mar -0.3 3.5 -4.2 106.0 28.4 3.51

Table A.3. ANOVA for heat flux comparisons. Substrate depths: conventional (0 cm); 7.5 cm; 15 cm.

DF SS F value P

All light and soil conditions Substrate depth

2 12.637 6.297 0.043

Residual 5 5.017

Non-frozen soil; all light conditions Substrate depth

2 12.868 6.764 0.038

Residual 5 4.756

Non-frozen soil; sun Substrate depth

2 383.6 44.81 0.001

Residual 5 21.4

Non-frozen soil; no sun Substrate depth

2 125.0 103.7 0.000

Residual 5 3.02

Frozen soil Substrate depth

2 12.010 4.307 0.082

Residual 5 6.971

Frozen soil; sun Substrate depth

2 19.113 7.829 0.029

Residual 5 6.103

Frozen soil; no sun Substrate depth

2 31.90 10.99 0.015

Residual 5 7.25

Table A.4. ANOVA for snow depth and # Days > 3cm snow in built-in green roof systems. The 'position' factor refers to the sheltered, raised and exposed roof systems.

Df SS F-value P

Snow Depth Treatment 2 8.44 3.616 0.038Position 2 6.91 2.963 0.066Treatment X Position

2 4.29 1.841 0.175

Residuals 33 38.48

# Days > 3cm Snow Treatment 2 75.42 6.621 0.004Position 2 130.58 11.463 0.000Treatment X Position

2 9.82 0.862 0.432

Residuals 33 187.96

Table A.5. ANOVA for snow depth and # Days > 3cm snow in modular green roof system. 'Position' refers to location on roof along solar exposure gradient.

Df SS F-value

P

Snow Depth

Species Species 14 94.95 3.116 0.000Position 4 212.01 24.350 0.000Residuals 122 265.55

Life form group Group 4 78.56 9.394 0.000Position 4 221.93 26.539 0.000Group X Position 16 46.30 2.89 0.16Residuals 125 261.33

# Days > 3cm SnowSpecies Species 14 297.6 5.78 0.000

Position 4 769.4 52.30 0.000Residuals 122 448.7

Life form group Group 4 155.2 9.153 0.000Position 4 800.8 47.219 0.000Residuals 132 559.6

Table A.6. Mixed-model linear regression of mean snow depth, mean air temperature, and substrate depth on mean heat flux in built-in green roof systems (heat flux = snow depth X air temperature X substrate depth; random effect: roof plot).

Table A.7. Multiple regression of substrate temperature, snow depth, air temperature and solar radiation on heat flux for 7.5 cm depth treatment in built-in green roof system, using time series data. Model presented is best model using AIC all-subsets selection criterion (heat flux = substrate temperature - snow depth + air temperature + solar radiation). R2

adj=0.78; F4,20=21.73; P = 0.000.

Standardized coefficient±St.Err

t-value P

intercept -3.053±0.369 -8.276 0.000

Substrate temperature

3.239±0.617 5.254 3.84e-5*

Snow depth -0.605±0.209 -2.900 0.009*Air temperature 0.449±0.250 1.762 0.093Solar radiation 0.573±0.236 2.432 0.024*

Snow depth -0.737±0.452 186 -1.630 0.105Air temperature 2.240±0.481 186 4.655 0.000Substrate depth -1.331±0.620 186 -2.147 0.075Snow depth X Air temp. -2.368±0.483 186 -4.905 0.000Snow depth X Substrate depth

0.386±0.450 186 0.788 0.432

Air temp. X Substrate depth -2.539±0.517 186 -4.908 0.000Air temp. X Substrate depth X Snow depth

2.949±0.547 186 5.396 0.000

Standardized coefficient±St.Err

DF t-value P

Table A.8. ANOVAs for maximum and minimum substrate temperature in built-in green roof systems. Top: entire winter season (Nov. 5, 2010-March 31, 2011) with sheltered and raised systems making up the levels of the 'position' factor; Bottom: end of winter only (Feb. 21-March 31 2011) with sheltered, raised and exposed levels of 'position'.

All winter, sheltered and raisedDF SS F-value P

Maximum Substrate depth 2 1243.2 189.34 0.000Position 1 24.1 7.36 0.017Substrate depth X Position 2 18.6 9.3 0.092Residual 14 46.0

Minimum Substrate depth 2 214.4 9.765 0.002Position 1 535.7 48.786 0.000Substrate depth X Position 2 22.7 1.031Residual 14 153.7

End of winter only, sheltered, raised and exposedDF SS F-value P

Maximum Substrate depth 2 1812.0 113.50 0.000Position 2 86.2 5.402 0.010Substrate depth X Position 3 185.3 7.739 0.000Residual 28 223.5

Minimum Substrate depth 2 788.7 174.80 0.000Position 2 818.1 181.31 0.000Substrate depth X Position 3 134.9 19.93 0.000Residual 28 63.2

Table A.9. ANOVAs for maximum and minimum substrate temperature in modular green roof system (separate analyses for species and life form groups). Bottom two are subsets for snow sampling days where no snow cover was present, and for the 10 days where snow depths were greater than 3cm.

Df SS F-value P

Whole Winter Maximum

Species 14 1399.3 8.708 0.000Position 3 350.8 10.187 0.000Residuals 86 987.1

Group 4 1121.4 21.116 0.000Position 3 341.1 8.564 0.000Residuals 96 1274.6

Whole Winter Minimum

Species 14 245.7 14.53 0.000Position 3 81.2 22.40 0.000Residuals 86 103.9

Group 4 198.5 30.64 0.000Position 3 76.8 15.81 0.000Residuals 96 155.5

No Snow Maximum Species 14 373.1 7.811 0.000Position 3 71.9 7.021 0.000Residuals 78 266.1

Group 4 323.8 22.65 0.000Position 3 72.7 6.778 0.000Residuals 88 314.6

No Snow Minimum Species 14 49.8 1.027 0.436Position 3 252.3 24.280 0.000Residuals 78 270.2

Group 4 26.19 1.868 0.123Position 3 237.61 22.598 0.000Residuals 88 308.44

Snow depth > 3 cm maximum

Species 14 5.257 1.183 0.305Position 3 14.168 14.884 0.000Residuals 78 24.750

Group 4 2.321 1.783 0.139Position 3 13.208 13.525 0.000Residuals 88 28.646

Snow depth > 3 cm minimum

Species 14 11.07 4.476 0.000Position 3 21.27 40.139 0.000Residuals 78 13.78

Group 4 8.85 12.31 0.000Position 3 21.45 39.80 0.000Residuals 88 15.81

Figure A.1 Built-in green roof system ("sheltered"); cross section not to scale.

Level 1

Level 2Level 3

Level 4

Level 5

Figure A.2 Built-in green roof system ("raised"); cross section not to scale.

Figure A.3 Built-in green roof system ("exposed")

Figure A.4 Time series of temperature, heat flux, and snow cover for sheltered roof system (Levels as in Figure A.1).

a) Air temperature and precipitation for Halifax, Stanfield Airport (approximately 35 km from field site) during study period.

b) Control roof

c) 7.5 cm substrate depth green roof

d) 15 cm substrate depth green roof

Figure A.5Solar radiation and heat flux over two six-day intervals for green and conventional roof panels. Top figure depicts a week (February 2-7 2011) where there was snow cover on the roof; bottom figure shows a week (March 2-7 2011) having no snow cover.

Figure A.6 Mean net heat flux for green roof (7.5 cm and 15 cm substrate depth panels combined) and conventional roof panels versus snow depth. Data for 21 days between January 1-March 31 with snow depths and heat flux values taken between 12:00-1:00 pm.

Figure A.7. Average snow depth (left) and duration (right) grouped by growth form: Bram= Bunch-forming graminoids; Cshrub= creeping shrubs; Control= substrate-only controls; Sgram= Sod-forming graminoids; Tforb= tall forbs. Means sharing letters are not significantly different at α = 0.05.

a

c

acb

aaa

ab

abb

Figure A.8. Mean temperature at different roof levels in the sheltered system (Nov. 5-March 31). Sensor locations as indicated in Figure A.1.

Figure A.9. Maximum and minimum substrate temperatures in modular green roof systems planted with different vegetation. Top: Averages over three days during snow sampling study where there was no snow cover; Bottom: averages over 10 days when snow depths were > 3cm minimum. i: maximum by species; ii: maximum by growth form group; iii: minimum by species; iv: minimum by group. For growth form group comparisons, means sharing letters are not significantly different at α = 0.05. "ns." means that the ANOVA was not statistically signficant.

c

b

a

a

a

ns.ns.

aa

a

b

a

ns.ns.

Figure A.10. Mean substrate temperature vs. mean snow depth in different roof settings. a) Built-in

green roof systems (sampled Feb. 21-March 7. trendline: (substrate temperature = 0.81(snow depth) -

2.3; df=26; R2=0.54; P<0.0001). b) Modular system planted with different species, grouped by

growth-form (averaged across 10 days where snow depths were > 3 cm): Bgram= bunch-forming

graminoids; Cshrub= creeping shrubs; Control= substrate-only controls; Sgram= Sod-forming

graminoids; Tforb= tall forbs. trendline: substrate temperature = 0.058(snow depth) - 0.759; R2=0.40;

R<0.0001).