Simulated rainfall

In Experiments 1 and 2, drip-system simulators were used. The application of water in this manner did not truly represent rainfall, but provided a quick and convenient method of applying water to the plants in order to quantify the volumes captured by different species. Based on the design described by Nagase and Dunnett (2012), rainfall simulators of this type were constructed for the experiments by drilling 1 mm-diameter holes in the bottom of a plastic box and inserting syringe needles (21G: 0.8 x 40 mm; Terumo UK Ltd., Surrey, UK) into the holes. These were sealed in place with Blu tack (Bostick Ltd.) to prevent water leaking out around the needles, and produced water droplets with consistent size and frequency. Reverse osmosis (RO) water was used for rainfall simulation with this type of simulator in Experiments 1 and 2 to prevent precipitates and air bubbles from mains tap water blocking the needles (Clarke and Walsh, 2007).

2.6.4.1 Experiment 1 simulator tests

Six small rainfall simulators were constructed for Experiment 1, using 1.6 L plastic boxes (internal dimensions 15.8 x 11.5 x 9.7 cm; Really Useful Products Ltd., Normanton, West Yorkshire, UK) each with 16 needles (Figure 2.2 A), which were 48 mm apart in rows 28 mm apart. Simulated

Experiment 3

Treatments: 6 replicates of Heuchera, Salvia, Stachys, Sedum and control (bare substrate). Measurements: SMC; canopy height and diameter; leaf area; fresh root volume; root and

48 rainfall covered a rectangular area of approximately 180 cm2_{, which was just slightly smaller than}

the area of the 2 L containers thus ensuring that the total volume of water applied fell onto the plant/container. Simulators were supported approximately 60 cm above the ground, and one plant/container was positioned below each simulator for rainfall application (Figure 2.2 B).

Initial tests demonstrated that at least 200 mL of water was required to generate runoff from planted containers; 500 mL of rain was therefore applied to each container to ensure that there was enough runoff for accurate measurement. This was achieved by simulating rainfall for 10 minutes, with an initial water depth of 8 cm, which equated to a target rainfall intensity of 165 mm hour-1_{. This very high rainfall intensity was chosen to reflect climate change projections}

of more intense precipitation events in the UK in the future (IPCC, 2013).

Figure 2.2: One of the six rainfall simulators used in Experiment 1 (A), and the position of the containers and trays below each simulator in the experimental runs (B).

The mean volume of simulated rainfall actually applied from each simulator in all of the experimental runs (as described in Section 2.6) was statistically similar (P = 0.820; LSD (5%) = 63.22; data not shown). There was little variation between simulators within each run, so all treatments received almost exactly the same volume of rainfall in each of the runs, allowing fair comparison of rainfall capture and runoff.

2.6.4.2 Experiment 2 simulator tests

For Experiment 2, three larger drip-system rainfall simulators were constructed using 32 L plastic boxes (internal dimensions 35 x 55 x 17 cm; Asda Stores Ltd., Leeds, UK). Each box had 72 needles arranged in 6 rows 45 mm apart (Figure 2.3 A), so that simulated rain fell in an area of 1925 cm2_,

49 above the ground, and one tray was placed below each simulator for rainfall application (Figure 2.3 B). Trays were rolled in and out of position at the beginning and end of the rainfall application on a board, so that they were approximately 8 cm above the ground; the actual height of the simulators above the trays was therefore 62 cm.

Ideally, the rainfall intensity with this simulator would have been the same as in Experiment 1 to enable comparison of results at individual plant and canopy scales. However, with this experimental setup and needle configuration, which was partially dictated by the design of the mesh bench on which the simulators were supported, the maximum rainfall intensity that could be achieved was 115 mm hour-1_{. The duration of rainfall was kept as 10 minutes, which equated}

to a target rainfall application of 3700 mL per tray. Testing of the simulators indicated that trays of bare substrate required application of at least 750 mL of water before runoff production began, and so 3700 mL was considered great enough to generate runoff even from planted trays.

Figure 2.3: One of the drip-system rainfall simulators used in Experiment 2 (A), and the position of the tray below the rainfall simulator during testing (B).

Actual volumes of water applied from each simulator during each experimental run (described in Section 2.6) varied, both between simulators and also between runs (data not shown). Average rainfall in all runs applied with simulator 1 was significantly lower than with simulator 3 (P = 0.033; LSD = 651.9), and the average volume from all simulators applied in the third run (‘unsaturated’) was significantly less than applied in the first run (‘saturated’) (P = 0.03; LSD =

50 648.7). The reduction in rainfall volume over the course of the experiment (i.e. between runs) could have been a result of the syringe needles becoming blocked with either air bubbles or debris in the water over time, despite the use of RO water, slowing the rate of dripping (Clarke and Walsh, 2007). As a result of this, only comparisons between treatments within each run were made. Additionally, the rainfall simulator used to test each treatment was varied in all experimental runs, so that rainfall from each simulator was applied to two replicates of each treatment in each run, and any differences in application volumes between simulators should therefore average out.

2.6.4.3 Experiment 3: Sprinkler rainfall simulator

In order to bring the characteristics of the simulated rainfall closer to those of natural rainfall, a sprinkler system based on the design described by Iserloh et al. (2012) and designed ‘in house’ by an irrigation specialist at RHS Garden, Wisley, was used to simulate rainfall in Experiment 3. The system consisted of a Lechler 460 608 nozzle attached by a 2 m length of hosing (Tricoflex) to a flow control, which was a series of pressure gauges and filters that ensured that the water flow and the characteristics of the droplets produced were constant. This was connected to the mains water supply by hosepipe, and ‘rainfall’ could be turned on and off directly on the simulator. The optimum flow pressure to achieve consistent rainfall in terms of droplet size and distribution was found to be 0.15 bars (15 kPa), and so this pressure setting was used for all rainfall simulations. The nozzle, hosing and simulator were fastened to an L-shaped timber support 2.4 m high and 1 m across (Figure 2.4), which was then secured to a metal pole in the glasshouse to keep the simulator stable and ensure that the rainfall always fell in the same area.

Trays were placed on a trolley and rolled into position under the rainfall (Figure 2.4 B), so that the actual height of the nozzle above the trays was 1.6 m; this is in line with the heights of other rainfall simulators cited in the literature, typically used in soil erosion and runoff studies, which vary between 0.7 and 3 m above the ground (e.g. Humphry et al., 2002; Fister et al., 2012). The height of the rainfall simulator determines the kinetic energy and terminal velocity of the water droplets produced; since the fall height is much lower than for actual rainfall, the velocity and energy of large drops in particular will be lower than in natural rainfall (Iserloh et al., 2012).

51

Figure 2.4: The sprinkler rainfall simulator used in Experiment 3, showing the pressure regulators and filters connected to the nozzle (A) and the experimental setup of the simulator secured to an L-shaped timber support with a tray positioned below the nozzle on a trolley for testing (B).

The spatial distribution of rainfall was found to vary and so tests were initially carried out to identify which position within the rainfall area could be used to ensure consistent rainfall intensity. Fifty-three numbered, empty buckets (24.2 cm diameter) were weighed and then positioned in concentric circles on the floor under the rainfall (Figure 2.5 A). Rainfall was simulated for 10 minutes with the flow pressure set to 15 kPa, and then all buckets were weighed again. The volume of water applied in each position was determined as the weight gain of each bucket, allowing spatial rainfall intensity to be mapped (Figure 2.5 B). Rainfall intensity was lowest directly below the nozzle and higher around the perimeter, similar to the spatial rainfall distribution described by Iserloh et al. (2012) and Fister et al. (2012) with the same nozzle type. The area of buckets 1-2-6-7 (Figure 2.5 A) was identified as having the most consistent volume and intensity of rainfall in every rainfall simulation in further testing of the simulator, and this was therefore chosen as the position for the trays during the experiments. Unfortunately, no other areas were identified as having similar and consistent rainfall intensity, and so it was decided to

52 test one tray at a time during the experiments, always in location 1-2-6-7 to ensure that all treatments were exposed to the same rainfall intensity. At trolley height (i.e. 1.6 m below the nozzle), the average rainfall intensity in the chosen location was consistently 28 mm hour-1_.

To further characterise the simulated rainfall, average raindrop size was measured using the flour pellet method described by Clarke and Walsh (2007). Flour was sieved into a tray to a depth of approximately 2 cm and lightly compressed, and the tray was then placed under the rainfall simulator at the selected height and position, and exposed to the rainfall for 5 seconds. After

< 10 mm hr-1 10 – 20 mm hr-1 20 – 30 mm hr-1 > 30 mm hr-1 Timber support Nozzle Rainfall intensity:

B

A

Figure 2.5: Setup of buckets (A) for the spatial rainfall distribution tests of the sprinkler simulator used in Experiment 3, and the ‘mapped’ rainfall intensities at ground level (B).

53 drying in the oven at 70°C for 24 hours, the raindrops formed pellets in the flour which were photographed. The diameters of all raindrops in three representative 4 x 4 cm areas were then measured using Image J software (National Institutes of Health, USA). Raindrop sizes ranged from 0.21 to 2.76 mm with the majority of droplets (70%) smaller than 1 mm diameter, similar to the simulated raindrops produced in other studies (e.g. Fister et al., 2012; Iserloh et al., 2012).

The time taken for runoff to be generated from trays with bare substrate was tested with the chosen settings, and found to vary between 2 and 8 minutes, depending on initial SMC. To ensure that adequate runoff was always generated from all planted treatments and all SMC conditions, it was therefore decided to simulate rainfall for 20 minutes for each tray. Since rainfall could only be applied to one tray at a time, this required rainfall duration meant that only 12 trays could be tested in a day. Each experimental run was therefore conducted over three consecutive days, testing 2 replicates from each treatment each day so that results were not affected by any differences in environmental conditions between days.

The actual volume of water applied on each day of each run of the experiment (as described in Section 2.6) was measured three times per day (beginning, middle and end of the experiment) by placing an empty tray of the same dimensions in position under simulated rainfall for 20 minutes (data not shown). The average volume of rainfall captured in the tray was similar in the ‘saturated’ and ‘unsaturated’ runs (P = 0.154; LSD = 148.1), although there were significant differences between experimental days in both runs (P = 0.017, LSD = 84.5 for the ‘saturated’ run and P = <0.001, LSD = 121.7 for the ‘unsaturated’ run). However, as two replicates per treatment were tested each day and there was very little variation in the volume of rainfall captured within each experimental day (data not shown), any differences in rainfall volume should average out for all treatments.