Nitrous oxide (N₂O) is a potent greenhouse gas (IPCC 2013) and is now the most significant contributor to the depletion of stratospheric ozone (Ravishankara et al. 2009). Its atmospheric concentration has increased by 20 % since the mid-19th century, most particularly since the production of synthetic nitrogen (N) fertilisers began, upon which modern agriculture which is highly dependent. Consequently the limiting of agricultural N₂O emissions is of great importance. Biochar – charcoal made by the pyrolysis of biomass – has been shown capable of reducing N₂O emissions when incorporated into soil (Cayuela et al. 2014; Lehmann and Joseph 2015). This Ph.D. study was undertaken to examine the mechanisms related to the production and movement of N₂O in soil and whether, and by what means, biochar addition to soil might mitigate N₂O emissions.

In all experiments the biochar (BC) used was made, at a highest treatment temperature of 550 oC, from the woody residue of oil mallee trees (E. Polybractea) after steam extraction of eucalypt oil. In each experiment the BC was mixed with soil(s) at rates of 0 %, 1 % and 5 % (w/w). One experiment used x-ray computed tomography, at a resolution of 70 μm, to examine the effect on soil structure resulting from the addition of BC to samples of an Arenosol, a Ferralsol and a Vertisol. For each soil bulk density (BD) decreased with increasing biochar content. Significant increases were found in porosity, pore connectivity and mean pore radius with addition of 5 % BC to the Vertisol and Ferralsol. The 1 % BC amendments produced no significant changes in those soils. Over a 15-month incubation at high moisture content the Ferralsol containing 5 % BC showed significant temporal changes in porosity, pore connectivity, mean pore radius and fractal dimension (FD) – the increase in FD suggesting greater pore space homogeneity. These changes suggested biochar-soil interactions, possibly related to reactive clay minerals and/or microbial activity. Amendment of the Arenosol with both 1 % and 5 % BC resulted in significant increases in pore connectivity and mean pore radius. A significant temporal increase in pore connectivity resulted from the 5 % amendment. Addition of BC significantly decreased the rate of water loss through evaporation and drainage of all incubated soils, reduced volumetric water content at field capacity in the two clay-rich soils and increased the available water content (– 0.01 to –1.5 MPa) of the Arenosol by 20 % (Chapter 2 and Quin et al. 2014).

In a laboratory experiment the Ferralsol, containing the BC as before, was repacked into PVC columns of 37 mm internal diameter (ID), sealed and incubated at 3 water regimes (12 %, 39 % and 54 % water-filled pore space (WFPS)) following gamma irradiation to render the contents abiotic. After N₂O was injected at the base of the soil column, in the 0 % BC control 100 % of injected N₂O was released into the headspace, declining to 67% in the 5% amendment. In a 100 % BC column at 6 % WFPS, only 16 % of the expected N₂O was released. X-ray photoelectron spectroscopy identified changes in BC surface functional groups from the 5 % amendment that suggested reactions between N₂O and the carbon matrix upon exposure to N₂O. Scanning transmission electron microscopy showed formation of an organomineral layer coating an external surface of a BC particle from that group. With increasing rates of BC application, higher pH adjusted redox potentials were observed at the lower water contents. Evidence suggested that the BC from soil had taken part in redox reactions, reducing N₂O to dinitrogen (N₂), in addition to adsorption of N₂O in (at least) the 100 % BC columns (Chapter 3 and Quin et al. 2015).

A field trial was established in north-eastern New South Wales, with the same Ferralsol repacked into PVC columns of 240 mm ID and 585 mm in height (Chapters 4 and 5). The columns were installed vertically in the ground, save for the top, soil-free 50 mm. The upper 100 mm of soil contained BC at dosage rates as before. Each column had a removable airtight cap for headspace gas sampling and silicone tubing installed at three depths for the sampling of soil gas content by diffusion. The columns were also fitted with ceramic cup lysimeters at two depths for the sampling of soil water, and thin tubes for the injection of liquid fertiliser at a depth of either 75 mm or 200 mm. With five replicates of each design (BC dosage rate and injection depth: 3×2×5 = 30) and five controls the central column of the 7×5 grid was fitted with soil moisture and temperature sensors. Three months after installation the trial commenced (on Day 0) when columns were injected with 62.8 % ¹⁵N potassium nitrate (1.68 g ¹⁵N-KNO3) at one of the two depths. Nine days prior to injection soil water contained mostly undetectable quantities of ammonium (all < 0.05 mg L^-1 ) and those of native nitrate (NO₃ ^-) ranged from 2.2 to 120 mg L^-1 , so it was assumed that denitrification would effectively be the sole pathway of N₂O production through reduction of NO₃ ^-. Following persistent rainfall in-soil concentrations of N₂O rose by approximately 2 orders of magnitude as soil WFPS increased to > 80 % on Day 10. This coincided with periods of high hydraulic conductivity, equivalent to drainage of 13.0 L m^-2 h ^-1 . Drained at that rate the downward carriage of (calculated) dissolved excess ¹⁵N-N₂O (¹⁵N₂O) in 75 mm and 200 mm injected columns containing 0 % BC would be respectively 189 and 30 times the surface fluxes on that day. Such drainage of dissolved N₂O suggests that offsite transport of N₂O by leaching from some soils may be greatly underestimated, and could possibly explain some of the discrepancy between ‘top down’ estimates of emissions of N₂O of ~ 4 % of applied N (Smith et al. 2012) and the Intergovernmental Panel on Climate Change’s (IPCC) default ‘bottom up’ estimate of ~ 1.3 %.

Throughout the trial no emitted excess ¹⁵N₂ was detected, although some was detected within the soil, indicating that some full denitrification had occurred. For each depth of NO₃ ^- injection there were no significant differences in relation to biochar content of surface fluxes of N₂O on any day of measurement, nor of their cumulative emissions during the 89 day period of the trial. Total emissions of ¹⁵N₂O for all biochar contents (0, 1 and 5 %, n = 15) of 75 mm and 200 mm injected columns for Days 1–23 (after which emissions were minimal) were respectively 9.74 mg ¹⁵N-N₂O m^-2 and a significantly lower (p = 0.0002) 1.60 mg ¹⁵N-N₂O m^-2 – corresponding to 0.46 % and 0.075 % respectively of total N denitrified from ¹⁴⁺¹⁵NO₃ ^- injected and below the IPCC default of 1 % for direct emissions. The effect of deeper fertiliser placement on indirect emissions remains unclear as, while there was considerable leaching of ¹⁵NO₃ ^-from all columns, it was greater from those injected at 200 mm and its fate undetermined.

In summary, the possible and hitherto unrecognised drainage of significant quantities of dissolved N₂O from some soils would seem to warrant further study. Overall, amendment with the eucalypt BC clearly affected soil structure. While the BC lowered N₂O emissions from Ferralsol in the laboratory, through both adsorption and redox reactions, it was plainly ineffective in lowering emissions in the field. It is apparent that there may not be a single explanation for this outcome – the activity of field soil biota, the acidity of the Ferralsol, the weathering of the BC and greater variability of conditions in the field being factors possibly contributing to the difference. This emphasises the need for further field trials of biochars to determine their effectiveness in mitigating N₂O emissions, their effect on N cycling in soil, and the longevity of any effects prior to their widespread use.

Machine vision is an essential function of autonomous robotics, especially those which use visual mechanisms to navigate the complexities of the outside world. Agricultural environments such as pastures presents diverse and complex visual scenes containing flora, fauna and farm machinery. The ability to detect key objects within this environment greatly assist autonomous robotic navigation and operations. Operation of agricultural robotics such as quad-copters requires real-time (milliseconds) analysis of visual data to ensure performance. Current machine vision systems lack performance in processing time or detection accuracy within such environments. To address current machine vision limitations, this thesis presents a customised class of extreme learning machine algorithms intended for use within remote laptop or fog computing settings.

Colour was observed to often be a key visual cue for object detection in pasture scenes. The colour-feature extreme learning machine (CF-ELM) was introduced for image classification and was demonstrated to out-perform existing extreme learning machine (ELM) algorithms which did not use colour information for object detection. The CF-ELM utilised the small memory structure and fast training times of the ELM to develop a real-time classification algorithm with the added benefit of colour information. This allowed the CF-ELM to classify objects within pastoral scenarios in 0.06 to 0.18 seconds and between 82% to 96% accuracy. These scenarios included, weed detection, cattle detection and farm vehicle detection.

The multiple expert colour-feature extreme learning machine (MEC-ELM) was then introduced to both enhance detection and further reduce processing time. The MEC-ELM used multiple instances of the CF-ELM and a summed area table to produce real-time classification of objects within video frames. Object detection was performed on both quad-copter and surveillance camera video to demonstrate the wide utility of the MEC-ELM algorithm. Detection scenarios included stock monitoring, weed scouting and vehicle tracking with the MEC-ELM producing 78% to 95% precision and recall with processing times between 0.5 and 2.0 seconds per frame. Performance of the MEC-ELM was compared and contrasted to other suitable machine vision algorithms. The results in this research indicate that the MEC-ELM is a highly competitive algorithm suitable for real-time object detection in video, particularly for agricultural robotics applications.