Condicions d’estudi

The key findings are summarised for each ignition source. Each ignition source was

investigated through experiments that were developed in the laboratory. Tussock and exotic grass samples were tested in their natural vertical orientation, and no significant difference in ignition behaviour was found. Wind speed and sample MC were varied, while ambient temperature and RH were kept relatively constant. The ignition thresholds are presented below and in Table 4.18. Comparison with previous work indicated that both fuel orientation and ignition source location have an effect on ignition behaviour. Field experiments were conducted to validate findings from laboratory results.

6.1.1 Hot Metal

Off-road vehicle exhaust systems can reach up to 585ºC, as indicated by previous studies (Chapter Two). According to temperature measurements in this study, an unloaded, fully maintained Nissan Navara 2006, manual diesel reached 398ºC at the manifold while driving.

This temperature corresponds to the lowest temperature required for a 50% probability of ignition from hot metal, as predicted by the logistic regression model. However, it is unlikely that grass will contact the manifold of a utility vehicle, as it is further away from the ground than other parts of the exhaust system. Data from driving on off-road tracks and gravel roads suggested that most locations on the Nissan‘s exhaust system remained below 300ºC while driving, and decreased to less than 230ºC when in idle. These observations, and results from the hot metal experiments imply that utility trucks similar to the Nissan have a very low to no probability of igniting dead grass at any MC level. However, further work is required to test vehicle exhaust system temperatures under a greater range of conditions, such as fully laden vehicles that may reach higher exhaust system temperatures due to a higher engine workload.

Conversely, ATVs have high potential to cause ignitions in grassland fuels. The field experiment recorded exhaust system temperatures between 427 and 512ºC at the manifold, after only 17 minutes of driving on gravel roads. Successful ignitions were observed for all samples tested. The temperatures were within the ignition thresholds reported for the vertical hot plate orientation with wind speeds of 1 or 2 m/s. Hot exhaust systems of trail bikes and industrial equipment should therefore be considered to pose high ignition risk.

The ignitibility of grass samples was tested by using a copper hot plate in two orientations:

vertical and horizontal. At a MC of 1%, the ignition thresholds for a probability of ignition of 50% were 398 and 421ºC for a vertical hot plate orientation and wind speeds of 2 and 1 m/s respectively. For a horizontal hot plate orientation with wind speeds of 2 and 1 m/s, the thresholds (for a MC of 1%) were 429 and 452ºC respectively. For both hot plate orientations, there was very low to no probability of ignition when wind was not present. The results and model (which correctly predicted 77% of the observations) indicated that ignitions can occur at MC levels up to at least 100%, for a five minute exposure time.

Comparison with other studies suggested that grass orientation affects ignition success. Grass in its natural standing orientation appeared to require higher contact temperatures before ignition could occur, which was attributed to a lower flow of oxygen through the sample. Hot plate orientation also affected ignition probability, where samples required progressively higher metal temperatures for ignition from hot metal at the following locations: on top of the

hot plate < adjacent to the hot plate < underneath the hot plate. This is likely due to the buoyancy of the convective heat plume in relation to hot plate orientation.

6.1.2 Hot Carbon Emissions

Laboratory and field experiments were designed to investigate ignition capabilities of hot carbon emissions from vehicle exhausts for grass fuels at various MC levels. In the

laboratory, hot carbon particles of 1.0 mm diameter ignited grass samples at up to 116% MC;

however, this was only true for a hot air flow of 3.7 m/s at 200ºC. In the field, no ignitions were successful, with hot exhaust gas at an average of 105ºC. Air flow was not measured in the field. It was difficult to compare results between laboratory and field trials, as

experimental methods varied slightly. In this study, the vehicle (Nissan Navara 2006) used for the field experiment was properly maintained, and was not emitting sparks, nor exhibiting high exhaust gas temperatures. Poorly maintained vehicles may reach higher exhaust gas temperatures. Further research is required for this ignition source, as outlined in section 6.3.

Two models were statistically significant, but they were weak and had large error associated with predictions. The preferred model was not as statistically significant as the secondary model, but can be applied to a wider range of environmental conditions. The secondary model can only be used when conditions are between 18.5 and 20.1ºC ambient temperature, and 31 and 54% RH. The ignition thresholds for a probability of ignition of 50% were 65% MC for the preferred model and 62% MC for the secondary model, indicating that there was little difference between models. The preferred model correctly predicted 69% of the experimental observations, and the secondary model correctly predicted 78%.

Findings from two previous studies were in accordance with results, where hot carbon particles between 1 and 3 mm were found to ignite grass fuels. Tussock grass blades have a smoother texture than exotic grasses, which influenced the ability of hot carbon particles to land and remain on tussock grass samples. The higher percentage of ignitions observed for exotic samples was attributed to this difference in sample characteristics.

6.1.3 Metal Sparks

Surprisingly, metal sparks ignited samples at higher MC levels than expected (up to 69% vs.

15%). Results affirmed citations that metal sparks are a significant ignition source from grinding operations. No previous studies have explored the ignition behaviour of metal sparks with grassland fuels; therefore, findings have furthered scientific knowledge tremendously, providing considerable insight for fire management.

The probability of ignition success model was highly significant, and predicted a 50% ignition probability of 37% MC, regardless of wind speed. The model correctly predicted 90% of the experimental observations. During the experimental trials, some sparks did not land on the samples. This was attributed to variability in the stream of sparks, as grinding caused sparks to fly in many directions. Unfortunately this could not be quantified; but, the model was highly significant regardless of this observation. Furthermore, this variability was representative of field conditions. All samples tested in the field ignited, with the model correctly predicting ignition success.

6.1.4 Organic Embers

Further research is required before conclusions can be made regarding ignition thresholds of organic embers, which were meant to simulate smouldering organic matter that had fallen off a vehicle. Laboratory simulation was difficult, and the organic disks did not contain enough fuel to sustain smouldering; therefore, once removed from the heat source and placed on the samples, they cooled and none of the samples ignited. No other experiments similar to this have been previously conducted, and further investigation should consider field tests with several different vehicles and situations, as outlined in section 6.3.

6.1.5 Open Flame

The lighter-sized flame ignited samples at MC levels up to 54% for light wind (1 m/s) conditions, and at MC levels up to 32% without wind. None of the previous studies had investigated the ignition behaviour of dead grass under the same conditions. One study reported a threshold (38% MC) for live and dead grass with various wind speeds up to 11.1 m/s. Without wind, other studies reported higher thresholds in comparison with this study, but this was attributed to horizontally oriented grass samples (which seem to ignite more readily than vertically oriented samples), and longer flame application which allowed samples to dry sufficiently for successful ignition. This study agrees with reports that flame location influences ignitibility of fuel, where ignition occurs more readily when the flame is located within or adjacent to the fuel.

The probability of ignition model was highly significant, and was the strongest of all models in this study. However, predictions for a wind speed of 2 m/s were not statistically significant, so this wind speed was not included in the model. The non-significance was attributed to the wind blowing the flame out during trials, causing inconsistency in the results. The ignition thresholds for a probability of ignition of 50%, for conditions with a wind speed of 1 m/s and without wind, were 28 and 55% MC respectively. The model correctly predicted 97% of the

experimental observations. All samples tested in the field ignited, with the model correctly predicting ignition success.