Monitoring fire danger in near real-time using field-based agrometeorological measurement systems.

Abstract

Africa has been termed the ‘Fire Continent’ due to its high annual fire frequency. Wildfires are considered one of the most common disasters in South Africa resulting in a high number of human fatalities and financial loss on an annual basis. It is believed that increased population growth, as well as more concentrated settlement planning is likely to result in increased fire disasters and increased human fatalities as a direct result of wildfires. The high number of human fatalities and high financial loss associated with wildfires served as the main motivation for the research throughout all studies. While wildfires may provide beneficial environmental service, increased wildfire activity can result in a number of adverse effects on the environment, for example the removal of vegetation, fascilitating/aggravating floods and soil erosion but do bring with them positive effects such as nutrient recycling and removal of alien species. In order to better understand the spatial and temporal variations and characteristics of wildfires in South Africa an 11-year dataset of MODIS-derived Active Fire Hotspots was analysed using an open source geographic information system. The study included the mapping of national fire frequency over the 11-year period. Results indicate that the north-eastern regions of South Africa experience the greatest fire frequency, in particular the mountainous regions of KwaZulu-Natal, Mpumalanga and the Western Cape. Increasing trends in provincial fire frequency was observed in eight out of the nine provinces with Mpumalanga being the only province where a decrease in annual fire frequency over the study period was observed. Temporally, fires have been observed in all months for all provinces although distinct fire seasons were observed, largely driven by rainfall seasons. The South-Western regions of South Africa (winter rainfall patterns) experienced higher fire frequencies during the summer months with the rest of the country (summer rainfall) experiencing higher fire frequencies during the winter months. Regions which experience bi-modal rainfall seasons did not display distinct fire seasons. The study included an investigation into the likely effects of climate change on South African fire frequency. Three of the 11 years were identified as being climatologically anomalous. Fire frequencies in 2005 and 2010 (two of the warmest years in recent history) were significantly greater than normal years. Observed fire frequencies in 2008 were also significantly greater. The increased fire frequency was attributed to a severe La Niña event which may have resulted in increased vegetation growth prior to the dry season. A current issue with the mitigation of wildfires is the lack of proper real-time monitoring and measurement systems which can aid decision makers in the timing of controlled

fires. The development of a system for improved monitoring of meteorological conditions conducive to fire was investigated. The traditionally used nomogram and lookup table used by Lowveld fire danger index (LFDI) system was replaced by mathematical functions which were then programmed into an automatic weather station datalogger. Near real-time results of the calculated LFDI were displayed in a web-based teaching, learning and research system found at: http://agromet.ukzn.ac.za:5355/?command=RTMC&screen=Fire%20danger%20index. Warm, dry mountain winds known as Berg winds have a direct link to fire weather, enhancing the danger of uncontrollable fires. Berg winds are associated with periods of increased air temperature, decreased relative humidity and increased wind speeds. This increases the potential for the development and spread of wildfires. The effects of Berg winds on the microclimate and fire danger were quantified. For this purpose, historic hourly meteorological data, local and international, were used together with a fuzzy logic system for determining Berg wind conditions in near real-time. This included the use of modelled diurnal sinusoidal functions for solar irradiance, air temperature, relative humidity, wind speed and direction, for various locations. Application of the system demonstrated that out of four sites in the KwaZulu-Natal Midlands, South Africa, Baynesfield was the most vulnerable to Berg wind occurrences, followed by Ukulinga (near Pietermaritzburg) and lastly Cedara. Boulder in Wyoming, USA, experienced a larger frequency of Berg wind events compared to Sidney in Montana, USA, despite being located at a higher altitude. The McArthur Fire Danger Meters (FDM) have been used to monitor and measure fire danger in Australia since the 1950s when they were first developed following severe wildfires. The McArthur FDM has traditionally made use of nomograms to calculate grassland and forest fire danger. In the 1980s mathematical descriptions of the McArthur FDM were developed. These then allowed for the sub-daily and near real-time measurement of grassland and forest fire danger. While commercial forestry contributes only a small percentage of South Africa’s GDP, forest fires have been identified as a common occurrence in commercial forestry plantations and natural forests, which can at times spread to neighbouring grasslands or non-forest vegetation. Forest fire disasters in South Africa are well documented by the mainstream media and it is believed that proper monitoring of forest fire danger, along with grassland fire danger, may mitigate potential wildfire disasters and limit human fatalities and financial loss. The McArthur FDM also provides opportunity to model fire behaviour – a component of wildfire mitigation lacking in the Lowveld fire danger index. The study utilized the equations developed to investigate the use of the McArthur FDM in South Africa by applying the

equations to historical hourly meteorological data recorded at four locations in the KwaZulu-Natal Midlands, South Africa. Modifications to the McArthur grassland fire danger index (GFDI) were needed and resulted in the development of a method to calculate vegetation curing as a function of micrometeorological variables. Datasets were analysed to investigate seasonal curing at the four locations with results indicating high monthly curing averages in all months, with the greatest range of curing averages experienced in the winter months. Frequency analysis of both the GFDI and the forest fire danger index (FFDI) indicate that lower fire danger ratings are more common than high danger ratings as one would expect. The GFDI and FFDI displayed acceptable sensitivity to changes in the microclimate as observed through conducting sensitivity analysis and by plotting diurnal variations of GFDI and FFDI during Berg wind events. The fire behaviour modelling component of the McArthur FDM does not yield realistic results on steep topography but does provide a baseline on fire characteristics on gentler slopes