*2.1. Rainfall Simulation Trial*

Rainfall runoff trials were undertaken using a nutritionally deficient (low total and available N) coarse-textured sand (Table 1). This material was selected as the growing medium to reflect the potential N runoff loss from a sandy-textured soil, to minimise any potential extraneous N sources for plant uptake (e.g., mineralisation of resident organic N), and to minimise suspended colloid concentrations that could preferentially transport fertiliser N in runoff via ion adsorption. This coarse sandy material therefore maximised the potential N losses in runoff as a means of evaluating the benefits of organic versus inorganic sources of N in maximising plant uptake.


**Table 1.** Selected properties of the sand. EC = electrical conductivity, ECEC = effective cation exchange capacity.

The sand (total sand ≈ 93%) was slightly acidic (pH 6.15), non-saline, and contained extremely low organic carbon (organic C = 0.22%) and nitrogen (total N = 0.022%) concentrations. The effective cation exchange capacity (ECEC = 1.12 cmol+/kg) was very low, with calcium being the dominant exchangeable cation, along with lesser concentrations of magnesium and potassium.

Nitrogen was added as either a poultry-based organic fertiliser (CropUpTM) or as inorganic ammonium sulphate ((NH4)2SO4). CropUpTM is a mixture of composted manure, molasses, humates and natural minerals (14.5% zeolite), and is slightly soluble in water (Sustainable Organic Solutions Pty Ltd., Brisbane, QLD, Australia—Safety Data Sheet). CropUpTM contains 3.07% N (as total N; LECO CN Dumas analyser, St Joseph, MI, USA), 23.25% C (as total C; LECO CN Dumas analyser), a C:N ratio of 7.56, 4544.49 mg/kg of available NH4 <sup>+</sup> (2M KCl-extractable; [24]; Method 7C2), 40.67 mg/kg of available NO3 − (2M KCl-extractable; [24]; Method 7C2), a pH of 8.41 (1:5 CropUpTM:18.2 MΩ deionised water) and an EC of 9.18 dS/m (1:5 CropUpTM:18.2 MΩ deionised water).

Ammonium sulphate, rather than urea [20], was used as the inorganic N source to minimise the potential for NH3 volatilisation. Inorganic N was applied at rates of 0 (Control), 50 (ASLow), 75 (ASMedium) and 100 (ASHigh) kg N/ha as (NH4)2SO4 (Table 2), which reflects the application rates used by previous workers [20]. The poultry manure-based organic material CropUpTM was applied to achieve N rates of 25 (CULow), 37.5 (CUMedium) and 50 (CUHigh) kg/ha, and was supplemented with (NH4)2SO4 (25, 37.5 and 50 kg N/ha, respectively) to match the amount of N added in inorganic N treatments, and to ensure that N availability was not limited during the plant establishment and growth phases.


**Table 2.** Treatments used in the rainfall simulation experiment. "AS" indicates inorganic N source (NH4)2SO4 and "CU" indicates combined (NH4)2SO4 + CropUpTM.

The sand was packed into stainless steel trays (1045 × 457 × 40 mm; *n* = 3) to achieve an approximate bulk density (ρb) of 1110 kg/m3. The sand was initially packed to a height of 30 mm, and the various treatments (Table 2) were uniformly surface applied. The treatments were covered with an additional 7 mm of sand, and lightly compacted to produce a relatively uniform surface. Ryegrass (*Lolium multiflorum*) seed was spread across the sand surface at a rate, equivalent on a surface area basis, of 200 kg/ha. The grass seed was covered with an additional 3 mm of sand, and the soil tray was slowly moistened with water. Water was applied to achieve an approximate gravimetric water content (θg) corresponding to 60% of field capacity (θfc). The soil trays were maintained at this moisture content for a period of 42 days prior to undertaking rainfall simulation trials.

To minimise the likelihood of nutrient deficiencies limiting the ryegrass growth, each tray received a basal nutrient application equivalent, on a surface area basis, to 20 kg P/ha, 100 kg K/ha, 28 kg Mg/ha, 70 kg S/ha, 0.43 kg Cu/ha, 0.84 kg Zn/ha, 7.7 kg Mn/ha, 0.97 kg B/ha, 0.33 kg Mo/ha and 30 kg Ca/ha.

Nutrient runoff from the treated soil trays was generated using a rainfall simulator built in accordance with published specifications [25], using a similar procedure described by [26]. The rainfall simulator was positioned centrally over two flumes, and the simulation was conducted at a nozzle pressure of 28 kPa (the design pressure required to deliver a rainfall intensity of 70 mm/h) over a runoff period of 20 min.

Prior to commencing each simulation run, the soil surface was photographed as a measure of grass cover, as this parameter can influence surface runoff rates, and hence the nutrient loading of the runoff water. The photographs were taken orthogonally to the soil surface under uniform light conditions. The software was written in Python using the OpenCV library (Python Software Foundation, 2019) to enable a uniform set of pixel colours (defined by hue, saturation, and value, the standard HSV digital colour space) to be selected as either soil or plant matter across all the images after gamma-balancing each image. Plant coverage was calculated from the ratio of background to plant pixel counts and validated automatically via the total percentage of area covered by soil and the percentage of area covered by plant pixels. The set of pixel characteristics selected was modified to minimise discrepancies across these three methods iteratively, and then applied uniformly across all images.

After being placed in the flume, each treatment tray was manually wet up to saturation prior to commencing rainfall. A sample of rainfall water was collected for analysis as outlined below for the runoff samples. A composite sample of the water exiting the flume of each treatment was automatically collected (WS750 water sampler, Global Water Instrumentation Inc., College Station, TX, USA) at the commencement of flow and subsequently every 5 min (100 mL aliquot composited for each 5 min collection event). The height of the water at the flume discharge point was measured at each 5 min sampling period, and the volume flow rate (V) was calculated using [27]:

$$\mathbf{V} = 341 \times \mathbf{H}^{2.31} \tag{1}$$

where V = volume flow rate (L/s) and H = head of water (m).

At the end of each simulation period, eight cores (internal diameter = 110 mm) were removed from the treatment tray. The cumulative area of the eight cores was, on a surface area basis, approximately 8% of the treatment area. The composited sand core materials (soil + grass) were thoroughly mixed and then stored at <4 ◦C prior to analysis.

The runoff water, soil and plant materials were analysed as follows. An aliquot of each sample of runoff water was initially filtered (<0.45 μm), and samples of unfiltered and filtered water were analysed for total N (APHA 5310B). The filtered samples were also analysed colorimetrically for NH4 <sup>+</sup> [28] and NO3 − + NO2 − using a modified Griess method [29] with a microtiter plate reader (BioTek EPOCH<sup>2</sup> Microplate Reader) at a wavelength of 625 and 540 nm, respectively. Runoff nutrient and particulate concentrations were converted to mass loss to account for small variations in flow rate between the two flumes using Equation (1).

Sub-samples of soil (*n* = 3) were analysed for mineral N (2M KCl extractable; [24]; Method 7C2). Plant material (above- and below-ground material) was separated from the soil by washing with Milli-Q deionised water. The retained plant material was ovendried at 60 ◦C, weighed to estimate the dry matter (DM for whole plant biomass), and ground prior to analysis for total N and total C by high-temperature combustion (LECO CN Analyser), and for aluminium, boron, calcium, copper, iron, potassium, magnesium, manganese, sodium, phosphorus, sulphur and zinc by nitric acid digestion and ICPOES.
