Edge Localised Mode (ELM)-induced heat loads, in particular divertor heat loads present one of the main challenges in adopting nuclear fusion as an industrial-scale energy source. At present, the type-I ELMy plasma regime is chosen as the most suitable operational scenario because of its energy confinement and steady-state capabilities, however, with the understanding that the enormous quasi-periodic heat loads deposited by the ELMs will need to be controlled in order to avoid intolerable damage to heat- bearing elements. At present, a detailed theoretical understanding, and in particular, a sufficiently reliable way of predicting ELM behaviour based on basic discharge parameters is missing. The fusion community has been expending a great deal of effort in order to improve this situation by gathering as much experimental data as possible from a range of different-sized tokamaks, in order to develop empirical scalings and refine theoretical models, with the ultimate aim of mitigating their destructive power below thresholds where they won’t affect continuous scientific exploitation of the reactors. This experimental thesis joins into this effort via the installation and commissioning of a fast infrared camera on TCV viewing the outer divertor. Near the end of the thesis, an effort of similar magnitude has been carried out with a camera on temporary loan from the MAST group, in order to image the inner divertor region as well. The cameras possess a sub-array recording capability, enabling acquisition frequencies up to 25 kHz. Both cameras were calibrated with the combination of a low- and high-temperature blackbody source with large and small radiation surfaces, respectively. By estimating the surface emissivity as 0.85 and employing a gray-body model, target surface temperatures could be inferred. The resulting images were processed to produce 1D poloidal temperature profiles serving as an input to the THEODOR (THermal Energy Onto DivertOR) 2D finite-element code, with the goal of calculating the heat fluxes impinging on plasma facing components. The VIR (Vertical InfraRed) outer divertor system images a flat, horizontal surface covered by relatively large tiles from a perpendicular viewing angle. Deposited layers, microscopic surface changes and very small dust particles cause the appearance of microscopic hot-spots in the field-of-view (FOV), leading to a substantial deviation from the assumed model of a single-temperature (the tile bulk graphite) gray body radiator. Near the middle of the thesis, nearly all tiles inside TCV have been removed for cleaning via sandblasting. With 3 of the tiles having been left untouched, this presented a good opportunity to contrast the behaviour of the cleaned and uncleaned (~12 years of operation) surfaces. It has been found that already after 10 days of operation, the temperature signal significantly exceeds what could be expected from a pure graphite surface. This, though increasing error bars of the final heat flux figures, is also beneficial in the sense that it serves as a signal amplifier – e.g. most observed ELM filaments near the main divertors carry very little energy, and were it not for this effect, they would not be visible at all. In contrast to the outer divertor, the inner divertor geometry is more complicated, as the tiles’ shape causes a gradual toroidal change in the magnetic field line attack angle from ~20 to 0 degrees, excluding toroidal averaging. Sequences recorded by this system also require a correction for mechanical vibrations, which, though also present for the vertical system, are much more pronounced, affecting the outcome of heat flux calculations to a great degree. Best results have been achieved with a method based on cross-correlations. Apart from eliminating the effects of the horizontal shaking, as the global vertical movement of the strike point leg on the target needed to be retained, the vertical shaking was either left uncorrected or parametrized in a way to remove the oscillations about the obtained vertical displacement trendline. In the process of calculating the heat flux from the temperature spatio-temporal evolution, the surfaces’ deviation from a one-temperature blackbody radiator manifests as an overshooting during transients, in both the positive and negative directions. This leads to the appearance of unphysical results (heat fluxes conducted outwards from the tiles, whilst plasma is still depositing heat on it). By introducing a simple boundary condition in THEODOR, a crude correction for this effect could be applied. The resulting tile deposited energies were cross-checked with rudimentary heat capacity calculations based on tile-installed thermocouples, yielding ~25% agreement. Two discharges, one in FWD-, the other in REV-B configuration, were selected for detailed analysis of ELM divertor heat loads. Both were heated by 3rd harmonic ECRH heating, the former exhibiting large- and the latter type-III ELMs. A dataset of various parameters such as maximum heat flux, profile width, total divertor deposited powers, ELM deposited energies and in-out power balance during and in between the ELMs, as well as the fraction of ELM energy recovered at the divertors, the rise-to-peak times and integral energy-to-peak was built up and evaluated in detail. The TCV outer divertor power loads dominate those of the inner under all circumstances – a finding that contradicts empirically established ELM characteristics from other tokamaks. The observed power imbalance is larger during the ELMs, and smaller in between. The ratio of ELM deposited energies at the inner- and outer divertors is 1:3, always in favour of the outer divertor. ELM profile broadening of a factor in the range ~1-2 has been found to occur consistently in both field directions, at both divertors. The total fraction of ELM energy recovered at the divertors (EIR/WELM) has been found to decrease with growing ELM size. The ELM pedestal energy fraction versus collisionality, the rise-to-peak time versus parallel convection time and integral rise-to- peak energy versus collisionality were compared to values from other tokamaks. With the exception of the ELM pedestal energy fraction of the “giant” ELMs in FWD-B discharge #38014, all values are in agreement with what would be extrapolated from other devices. In line with the target of predicting ELM heat loads, comparisons with 1D kinetic PIC simulations and the so-called free-streaming particle (FSP) approach (an analytical expression for the power density based on a relatively simple model) were performed. With the exception of the early rise phase, PIC simulations of type-III ELMs reproduce the experimental time evolution well. They give insight into finer details of ELM heat transport: a very fast electron heat pulse (carrying about 5% of total ELM energy), is followed by the main pulse on a slower time-scale. About 3⁄4 of the total energy is carried by the ions, of which 0.3-0.4 is deposited in the rise phase. The FSP expression, based on a collisionless approximation, provides a reasonable fit to the temporal evolution of divertor-deposited power for TCV ELMs of various sizes. In most cases, both the rise- and decay phases can be described by the single time parameter within the expression. This parameter is at least twice as large as the parallel transport time based on pedestal parameters, implying a delaying of particles near the X-point region. In contrast to other tokamaks, filamentary ELM heat deposition patterns are only occasionally recorded at the outer divertor. The observed spatial structure is consistent with a plasma release at discrete toroidal locations in the outer midplane vicinity, but as very few are visible on the IR (when compared to the Langmuir probes), the determination of quasi-mode numbers was deemed unfeasible. A crude estimation of the energy content of individual filament spirals yielded 1% of the total ELM energy in each, consistent with existing non-divertor ELM heat load predictions. Finally, a new divertor configuration, the snowflake, was recently demonstrated in TCV experimentally, and, subsequently, the creation of H-modes in it. This configuration presents many advantages, most prominently the possibility of distributing the divertor heat load to 4 (as opposed to 2 for standard single-null divertors) regions. Infrared measurements at one of these additional legs have confirmed substantial heat loading and a lack of ELM profile broadening.