Applied Geography xxx (2009) 1–9

Contents lists available at ScienceDirect

Applied Geography
journal homepage:

Characterization of the July 2007 Swaziland fire disaster using satellite remote sensing and GIS
Wisdom M. Dlamini*
Swaziland National Trust Commission, P.O. Box 100, Lobamba, Swaziland

a b s t r a c t
Keywords: Disaster Fire GIS MODIS MSG-SEVIRI Swaziland

Data from the NASA’S MODIS (Aqua and Terra) and EUMETSAT’S MSG-SEVIRI satellite sensors is analysed to characterise the geographic and temporal (including diurnal) evolution of the July 2007 fire disaster in the Kingdom of Swaziland using a geographic information system (GIS). Significant fire activity was observed during a three-day period beginning on the 27th July 2007. A total of 1358 and 4365 active fire hotpots were detected by MODIS and MSG-SEVIRI, respectively, mainly concentrated in the Highveld (70.91% for MODIS, 89.89% for MSG) and Middleveld (11.27% for MODIS, 5.23% for MSG) with MSG/MODIS active fire count ratio ranging from a high of 3.69 in the Highveld to a low of 0.06 in the Lubombo Plateau. The results indicate complex differences in spatial fire distribution, behaviour and risk within the country and the effect of sensor differences. A pronounced fire diurnal cycle with a broad afternoon peak centred on 14:00 local time is observed, in general agreement with observations from the region. Despite their limitations, the study demonstrates the importance and usefulness of remotely sensed data and GIS technology for fire disaster and risk assessment for a developing country, where fire monitoring resources are scarce. Ó 2008 Elsevier Ltd. All rights reserved.

Introduction Scientific evidence exists to suggest that savannah ecosystems evolved and adapted with fire as an agent of ecological change (Bond & Keeley, 2005). However, human activities have altered many natural landscapes and placed human beings in direct contact with fire or fire sources. Fires are known to cause loss of human life and personal property, and economic upsets, whilst heat, smoke and aerosol particles are known to cause disturbances in regional and global atmospheric composition and chemistry, and ultimately on climate (Andreae & Merlet, 2001; van der Werf et al., 2004). Research and ˜ ˜ results from several studies also suggest that fire activity is linked to the El Nino cycle (Randerson et al., 2005; Riano, Moreno ´ ´ Ruiz, Baron Martınez, & Ustin, 2007; van der Werf et al., 2006). Since such conditions are likely to become more frequent in the future, the magnitude and frequency of fires may increase (Lavorel, Flannigan, Lambin, & Scholes, 2007). Fires, therefore, have become an area of interest for scientists dealing with climate change and this makes fire detection also useful for climatologists. The three elements necessary for combustion, namely heat, oxygen and fuel (Trollope, De Ronde, & Geldenhuys, 2004) form what is commonly known as the fire triangle. Other conditions such as topography, fuels and weather, collectively called the fire environment, can perpetuate extreme wildfire behaviour (Perry, 1998). Drought conditions, such as those that have

* Tel.: þ268 6024716; fax: þ268 4161875. E-mail address: 0143-6228/$ – see front matter Ó 2008 Elsevier Ltd. All rights reserved. doi:10.1016/j.apgeog.2008.10.007

Please cite this article in press as: Wisdom M. Dlamini, Characterization of the July 2007 Swaziland fire disaster using satellite remote sensing and GIS, Applied Geography (2009), doi:10.1016/j.apgeog.2008.10.007

cloud cover. through its Spinning Enhanced Visible and Infrared Imager (SEVIRI) instrument. and emission estimation. developing countries are still lagging behind in taking advantage of such technologies (Flasse.apgeog. at a spatial resolution of 3 km at nadir for the short wave infrared (SWIR) and thermal infrared (TIR) bands (Jahjah. 2005). San-Miguel-Ayanz et al. 2002). biome. which may overcome the logistical and financial constraints of ground-based. or when no other significant fires are nearby. or when the scene is free of clouds. & Feltz. sun position. The emergence of satellite remote sensing provides opportunities for continuous. the performance of these sensors is somewhat satisfactory and such conditions occur when a fire is observed at (or near) nadir on a fairly homogeneous surface. However. 2006). Zhou. High potential for a unique contribution was recognised. 2006). limits the detection of small fires and validation of the minimum detectable fire size for MSG-SEVIRI in southern Africa is still under investigation (Frost & Vosloo. among others and countries with significant wildfire activity. Remote sensing instruments on polar-orbiting and geostationary satellites allow fire observations at a broad range of spatial and temporal scales.ARTICLE IN PRESS 2 W. land surface temperature. & Justice. Polar-orbiting satellite.e. make a small fire potentially disastrous. However. The Swaziland fire disaster During the month of July 2007. periods of critical fire weather and fire behaviour can develop. geostationary satellites have better temporal resolution. Although the capabilities of current geostationary satellites are limited. all satellite-based fire detection systems have their advantages and shortcomings. Castronuovo. Kelha. & Wang. MODIS and MSG-SEVIRI. & Navarro. Raimadoya. given the specific Sun-target sensor geometric conditions. Dlamini / Applied Geography xxx (2009) 1–9 been prevalent in Swaziland and the region over the past few years. 2003). 2006). and are critical for fire detection and monitoring in remote locations and developing countries. requiring constant and consistent monitoring. These satellite-based fire and thermal anomaly detection systems are indispensable for both research and operational use. especially developed countries. (2005) provide an overview of capabilities and limitations of remotely sensed data applications for fire emergency management. MODIS Terra scans the Southern African region between 10:00–11:30 am and at night around 22:00pm whereas MODIS Aqua scans in the afternoons between 14:00–15:30pm and also in the early morning at 03:00 am (Giglio. coupled with windy conditions. Applied Geography (2009). including a few others such as the DMSP-OLS (San-Miguel-Ayanz et al. The high temporal resolution makes it is very useful for detecting continuously changing phenomena like active fires detection as also seen by Schmetz et al. but worse spatial resolution and have difficulties when scanning at high viewing angles. near 100% probability of detection) is approximately 50 m2 in size. fire risk assessment. 2007). large-scale monitoring. Hot. they can provide valuable local. Ceccato. have developed ground and air-based monitoring networks (San-Miguel-Ayanz. however. on one hand. 1997). to take images of the Earth every 15 min. The continuous monitoring of fire involves observation of fires and fire-causing processes. Hillger. and Cadau (2006) report a minimum detectable fire size of 1000 m2 in the Mediterranean. As the synoptic weather pattern changes within the climatic conditions. (2002). Fire-related processes were clearly identified in the list of applications that can be derived from MSG-SEVIRI data. Csiszar. the smallest flaming fire that can be routinely detected (i. The accuracy of the fire detection by MODIS has been observed to reach up to 90% and even higher (Wang. Downey. land cover dynamics analysis. and windy conditions are generally ideal for the rapid growth and spread of wildfires and steeper slopes tend to further increase the rate of fire spread. the fires Please cite this article in press as: Wisdom M. The MODIS sensor also includes bands specifically selected for fire and cloud detection and allows the retrieval of sub-pixel fire area and temperature. Characterization of the July 2007 Swaziland fire disaster using satellite remote sensing and GIS. and good spectral resolution of the instrument. dry. which possesses enhanced spectral and temporal capabilities with respect to the then existing generation of Meteosat spacecraft. is of the order of 100 m2 (Frost & Vosloo.M. Validation in the USA on similar satellites have shown detectable sizes in the region of 500 m2. operational systems have been using data from the NOAA Advanced Very High Resolution Radiometer (AVHRR) and GOES (Geostationary Operational Environmental Satellite). regional and global fire products in near real time. For a long time. 2001). due to technological and financial difficulties. Satellite instruments that can simultaneously utilise 3. Ravail. Under ideal conditions. (Giglio. 2005). trigger fire and.9 mm and 11 mm channels can be used for fire detection due to the 3. doi:10. depending on factors such as scan angles. Flynn. Laneve. Schmetz.10. Propelled by high wind speeds and a prolonged dry season. Laneve. cannot see short-duration fires that take place between the satellite overpasses. Dlamini.. In these circumstances. The main purpose of this study is to describe the geographical occurrence and temporal (daily and diurnal) variation of the July 2007 fires as recorded by the two satellite-based fire detection systems. including the present.9 mm channel’s strong thermal sensitivity even if only a small portion of the pixel is covered by fire (Matson & Dozier.2008. Validation results for MODIS indicates that a minimum detectable fire size.1016/j. Its geostationary orbit allows MSG. & Marzoli. the high revisit frequency. & Ollero. amount of smoke and wind direction (Prins. a series of devastating fires blazed throughout Swaziland resulting in widespread destruction of property and even the loss of lives. Significant progress was made with the launch in 1999 and 2002 of the Moderate Resolution Imaging Spectroradiometer (MODIS) instrument on board the morning descending Terra and afternoon ascending Aqua polar-orbiting earth observation satellites thus offering the opportunity to observe fire activity both day and night.007 . heavy smoke or sun glint. The coarse resolution. 1981). These conditions often persist over Swaziland ˜ during the months of July/August more especially in western half of the country and in association with the El Nino Southern Oscillation (ENSO) warm phase (Dlamini. In 2002. flaming at between 800 and 1000 K and typically detectable at 50% probability. 2007). and air-/satellite-borne observations. on the other hand. EUMETSAT launched Meteosat-8 or Meteosat Second Generation (MSG). One of the common shortcomings is the lower limits for fire sizes that space-borne instruments can detect and the incapability of detecting fires through clouds or thick smoke.

or unknown. It is also endowed with tremendous natural diversity and complex topography with elevation that decreases from an average of 1400 m above sea level on the west to below 100 m on the eastern part of the country giving rise to four major eco-climatic (commonly referred to as agro-ecological) regions. Applied Geography (2009). A full description of the algorithm is given by Giglio et al. Middleveld. Descloitres.10. Goudie and Price-Williams (1983) and Remmelzwaal (1993) further subdivide the Middleveld into two zones . Variations in temperature also follow the altitudinal gradients. Smoke pollution was evident as the sky was darkened by plumes from the same at a spatial resolution of 1 km (Justice et al. namely the Highveld.e.M.. and Kaufman’s (2003) enhanced contextual fire detection algorithm which uses the 4 mm and 11 mm bands and classifies every pixel as missing data. ArcGISÔ was used for its versatility and its ability to simultaneously handle both spatial and temporal data. MODIS and MSG-SEVIRI. the driest months of the year occur in the winter. The country is sandwiched between South Africa to the north. 1983). Methods Study area The Kingdom of Swaziland.007 . similarly. Mean annual rainfall also varies extensively from above 2000 mm per annum in the Highveld to below 500 mm in the Lowveld. fire. However. Dlamini / Applied Geography xxx (2009) 1–9 3 resulted in damage to thousands of hectares of forest and other vegetation. Dlamini. The MSG-SEVIRI system currently uses a contextual algorithm-based on a threshold technique and statistical analysis of each potential fire pixel developed by Flasse and Ceccato (1996) for the AVHRR. http:// maps. which also houses the Disaster Management Agency. The focus of this analysis is on the July 2007 fires during which a majority of the fires were large-sized fires. is 17 364 km2 in extent straddling latitudes 25 400 and 27 200 South and longitudes 30 400 and 32100 East (Fig. i. from around April to September (Goudie & Price-Williams. The MODIS dataset is based on the version 4 contextual fire detection algorithm from the MODIS Rapid Response System (Web Fire Mapper.ARTICLE IN PRESS W. the Lowveld is partitioned into the Western and Eastern Lowveld (Fig. west and south. The MODIS active fire data is based on Giglio. When the ENSO warm phase is established. located in southern Africa. Whilst these may not be ignored. From the MOD14A1 (MODIS active fire) data. puts the number of homesteads destroyed at 169 and the total number of affected people at 934 with two casualties (MORDYA. water.umd. MODIS fire pixels with a confidence level of 0% were also considered as individual fire counts. 2005). 1). 2007). Therefore. especially the higher altitude areas. The confidence levels reported in MODIS active fires during the period under investigation were explored and it was found that even pixels in areas where the fires were intense (plantations) were given a zero confidence. 2002). can often be the catalyst for extreme weather conditions which may amplify fire behaviour such as was observed during the disaster. The active fire hotspots from both the MODIS and MSG-SEVIRI datasets were obtained in the form of ASCII files which were imported into the geographic information system (GIS) software ArcGISÔ 9. and fire confidence. and Mozambique to the east. the MSG-SEVIRI data was found inappropriate for separate geographical analysis of fire distribution. cloud.1016/j. Various media reports suggest acts of arson and negligence. Satellite data analysis Two active fire datasets. The country’s topography.apgeog.geog. To answer some of the important ecological questions it was necessary to relate fire activity information to the country’s eco-climatic zones so as to uncover knowledge on which ecosystems or landscapes were most affected. 1981. Characterization of the July 2007 Swaziland fire disaster using satellite remote sensing and GIS. 2007). Climatic variations within the country are largely controlled by topography and within a year there are four seasons with December being mid-summer and June mid-winter. Justice. non-fire. most likely due to the obscuration by the thick smoke cover over almost the whole country. and direct economic losses estimated in the tens of millions of US dollars. Typically in Swaziland. A report by the Ministry of Regional Affairs and Youth Affairs.Middleveld grassveld or Upper Middleveld on the west and Middleveld bushveld or Lower Middleveld on the east. An improved approach using a Kalman filter model of the normal diurnal cycle is being developed (Van Den Bergh & Frost. The ratio was calculated to provide information on the behaviour of fires where a large MSG/MODIS ratio is used as a proxy of a high degree of large short-duration and intense fires that were detected by the geostationary MSG-SEVIRI Please cite this article in press as: Wisdom M. doi:10. due to the low spatial resolution (w4 km) and the spatial scale of the analysis. The fire confidence is calculated by a system of equations within the algorithm and is expressed as a percentage (Giglio. 1).2008. observed brightness temperature. Goudie & Price-Williams. MODIS active fire hotspots were then counted for each eco-climatic zone as an indicator of the spatial or geographic distribution of fire activity. the Highveld being temperate and seldom hot while the semi-arid Lowveld can record temperatures of up to 40  C during summer. The 15-min temporal resolution MSG-SEVIRI active fire data was obtained from the Council for Scientific and Industrial Research (CSIR) – Satellite Application Centre (SAC)/Eskom’s Advanced Fire Information System (AFIS). as it has been in the past 2–3 years (Dlamini. Lowveld and the Lubombo Plateau (Gibbons. from 1 to 31 July 2007 were collected and collated. This culminated in the declaration of these fires by the country’s Prime Minister as a national disaster on the 1st August 2007. pixel size. an anomalous dry and warm pattern can be significantly amplified. big enough to be detected even by the low spatial resolution MSG-SEVIRI sensor. (2003). 2006) for spatial and temporal analysis.2 (ESRI. it is critical to understand the underlying eco-climatic factors which may have accelerated and exacerbated the impact of these fires. 2007). Basic analysis of the active fires detected was undertaken through the calculation of a basic ratio of MSG-SEVIRI to MODIS active fires for each ecoclimatic zone. 1983). the smoke was found to have been classified as ‘‘nonfire clear land’’. The MODIS active fire product contains information about the detected fire pixels including location.

indicates the presence of smaller fires that are undetected by MSG. has become a very important temporal metric in fire analysis. The diurnal fire cycle. The smoke plumes from these fires were spread throughout almost the whole country with most of the fires Please cite this article in press as: Wisdom M. The high temporal resolution MSG-SEVIRI dataset was used to study the diurnal cycle of fire activity for each eco-climatic zone in Swaziland. A ratio of less than unity.1016/j.10.ARTICLE IN PRESS 4 W. or the systematic variation in fire activity with respect to time of the day. Characterization of the July 2007 Swaziland fire disaster using satellite remote sensing and GIS. eco-climatic zones and elevation of the Kingdom of Swaziland. that could have been missed by the polar-orbiting MODIS. Active fires detected by both the MODIS and MSG-SEVIRI during the month of July were counted for each day of the month to ascertain the temporal evolution of the fire activity and to confirm the days of intense fire activity and relating these to actual observations on the ground.M. These were binned into 1-hour intervals to provide more meaningful interpretation. on the other hand. Results and discussion Geographic distribution A total of 1358 hotspots were detected by MODIS over the period under investigation whilst MSG-SEVIRI detected a total of 3926 hotspots. Location.007 .2008. Applied Geography (2009). 1. Dlamini. Dlamini / Applied Geography xxx (2009) 1–9 Fig. A ratio of unity is supposed to indicate the occurrence of fires that are most likely simultaneously or equally detected by both sensors.apgeog. doi:10.

007 . (2004) observe that the increased risk of accidental fires is also due to landscape fragmentation and land cover change which exposes forests and woodlands to fires. Applied Geography (2009). 2. The Highveld and Upper Middleveld are the areas where hundreds of hectares of plantation forests. and windy environment needed for extreme fire behaviour and accelerated fire spread rates. dry. gusty conditions were prevalent in the country creating perfect conditions for the fires. therefore. MODIS active fire hotspots during July 2007 in Swaziland.10. Dlamini / Applied Geography xxx (2009) 1–9 5 detected in the Highveld followed by the Middleveld regions. Although the exact cause or ignition point of these fires has not yet been ascertained. where the Piggs Peak Timber plantations company reportedly lost tens of millions of US dollars worth of property and an estimated 80% of the total plantation forest area in addition to significant destruction of other plantations in the south-western part of the country. confirming Dlamini’s (2005) observations that this zone is most prone to fires.apgeog. The most evident cluster of fires is in the north-western part of the country. The steep and rugged topography of the country. respectively. more especially the Highveld and Middleveld (Remmelzwvaal. 2007). land use and the landscape in the country. Please cite this article in press as: Wisdom M.ARTICLE IN PRESS W. indicate that these eco-climatic zones are riskiest in terms of wildfires under such conditions and as such fire management and control should pay particular attention to these areas. Thus.2008.M. 2). This geographic distribution of the fires illustrates the spatial pattern of the current burning practices. MORDYA. natural vegetation (mainly grass) and homesteads were burnt (Dlamini. Csiszar et al. can often generate the hot. Fig. located in the Highveld. MODIS detected 71% and 11% of all active fires in the Highveld and Upper Middleveld. Characterization of the July 2007 Swaziland fire disaster using satellite remote sensing and GIS. the most affected eco-climatic zone was the Highveld. fires in Swaziland are generally used by people to facilitate pasture regeneration and in clearing vegetation for farming and settlements whilst commercial/industrial plantation forests are typically designed with networks of firebreaks that are annually burnt in June and July to provide clean belts around the plantation compartments (Dlamini. 2007. consisting mainly of plantation forests and grasslands. the eastern zones recording the least number of fires (Fig. Dlamini. During the period of the fires. 2005). doi:10. 1993).1016/j. The observations.

thereby minimising omission errors especially from MODIS.27) 98 (7. MODIS detected 76.35 ha while the probability of detection increases with fire size reaching 100% with fires greater than 15 ha (Costantini. Dlamini. Zavagli. i.76) 1358 MSG-SEVIRI 3529 (89. The prevalent strong winds most likely caused spotting.12) 37 (2. However. the false alarm rate decreases with increasing fire size reaching zero with fire sizes of the order of 0.91) 153 (11.89) 221 (5. While MSG-SEVIRI was still detecting an increase in fires on Fig.89 The ratios of MSG-SEVIRI/MODIS active fires for all the eco-climatic zones are shown in Table 1. doi:10. the fires in the country. possible that for numerous small fires detected as one fire pixel by MSG-SEVIRI fire product. Applied Geography (2009). only fires actively burning at the time of the polar-orbiting MODIS satellite overpass could be detected. Dlamini / Applied Geography xxx (2009) 1–9 Table 1 MODIS and MSG-SEVIRI fire hotspots by eco-climatic zone (percentage of total indicated in brackets).66 1.9 depending on the size and characteristics of the fire.63) 113 (2.88) 18 (0.e.M. especially MODIS. A notable signal of fire activity was first detected on 24 July followed by a dramatic upsurge on 27 July. However. duration and intensity of the intense fires. as Fig. Characterization of the July 2007 Swaziland fire disaster using satellite remote sensing and GIS. The MSG-SEVIRI/MODIS ratios also give credibility to a proposition that the size.06 2.10. Calle et al. to detect fires.22) 56 (4. in the three-day period of intense fire activity.15 0. which typically occurs during high speed winds exceeding 11 km/h and in crown fires within highly flammable vegetation such as plantation forests (Trollope & Potgieter.apgeog. however.44 1. 3.13 0. were very large and intense and so were their detection probability.73% of the July 2007 fires during the three-day period whilst MSG-SEVIRI detected 88.ARTICLE IN PRESS 6 W. Cisbani. Fire suppression efforts were first made on 24 July 2007 and hopes were elevated among the plantation forest companies that the fire had been successfully put out. is the main reason for the differences in detected fires. Van den Berg and Frost (2005) also observe MSG-SEVIRI/MODIS active fire ratios varying from 7 to 1.32 1. However. 1993).007 . MODIS and MSG-SEVIRI active fire counts for 1–31 July 2007 in Swaziland. Actually.2008.1016/j.46) 42 (1. The ratios indicate that MSG-SEVIRI detected more of the destructive and intense large fires of the Highveld and Middleveld. 3 indicates. It is. the fires were rekindled aided by strong winds and dry conditions resulting. the start of a new fire ahead of a main fire by an airborne firebrand (Trollope. and fire location on the topography are some of the factors that may have limited the ability of the satellites. whilst MODIS was also effective in detecting smaller fires in the Western Lowveld and Lubombo Plateau. Of particular note is the prevalence of thick plumes around the country which might have also hindered or limited the detection of fires beneath the smoke. The findings indicate that the MSG-SEVIRI detections were the fast-spreading fires of the Highveld and Upper Middleveld. particularly in the Highveld. MSG-SEVIRI Cloud or smoke cover. (2005) provide a very good comparative analysis of MSG and MODIS active fire performances and found false alarm rates of as high as 70% and as low as 10% within the hottest zones of fires.07) 3 (0. & Greco. 1985). MODIS detected these as separate fires due to its smaller spatial resolution. 2006).08) 3926 MSG/MODIS ratio 3. Eco-climatic zone Highveld Upper Middleveld Lower Middleveld Western Lowveld Eastern Lowveld Lubombo Plateau Total MODIS 963 (70. Accordingly.79% during the same period indicating high spread rates during this time. 3). Temporal distribution Temporal analysis reveals that a majority of the fires were detected from the period 27–29 July 2007 (Fig.72) 51 (3. Please cite this article in press as: Wisdom M. coupled with differences in sensor characteristics and algorithms.

Steele. Cummings. This could be attributed to the fact that most of the fires were intense and fast due to the windy conditions and could have been missed by MODIS due to its orbital characteristics. coupled with certain synoptic weather patterns. (2006) attribute the broad morning to afternoon peak to the peak fire season whilst the switch from afternoon to night time fires follows the rainfall reduction gradient and land cover type. terrestrial carbon sinks have also been observed to have a strong diurnal cycle (Kauffman. Long (2006). which is of particular interest for short-lived trace gases and aerosols in the lower troposphere. in a study of extreme fire events in Victoria (Australia). Similarly. The results. 2006. Van Wilgen & Scholes. increasing rapidly to an overnight maximum that is maintained until the late evening/early morning.2008. Surface meteorological conditions. The observed diurnal cycle patterns are comparable to the observations by Eck et al. Dlamini. 1993). found peak fire danger at 15:00. Diurnal fire behaviour information is also important for applications in support of fire suppression and for emergency preparedness purposes. the burning likely became long lasting and persistent through the night particularly in the Highveld. and Oshika (2000) also found a strong southern African diurnal fire cycle with a peak around 14:00 local time. Menzel.apgeog. have been shown to be important factors affecting fire occurrence probability (Long. 2005). 4. Another reason.007 . Rice. 4) data reveals a strong and interesting diurnal cycle with most fires flaring strongly from late morning (11:00) to the afternoon (17:00) and again at late night (22:00) to early morning (04:00) with an overall peak in fire activity centred around 14:00 local time. fire occurrence is delayed and confined during the warmer periods of the day. Giglio et al. Characterization of the July 2007 Swaziland fire disaster using satellite remote sensing and GIS. WRAP. Temperature has also an important role in diurnal fire behaviour because in the cooler higher altitudes (such as the Middleveld and Highveld). analogous to Beck and Armitage (2004) and Zoumas et al. 2005). thereby indicating that risk is high and suppression efforts should be increased during these times. Gregoire. 1998. 2003. Diurnal variation in MSG-SEVIRI active fire counts in Swaziland. Tressel. (2006) and Peck. The usefulness of satellite fire observation in studying temporal fire evolution is clearly demonstrated in these observations. MODIS identified the beginning of a gradual decline. Cheney (1981) and Zoumas et al. at which time fires tend to be burning in larger numbers and with greater intensity. Applied Geography (2009). doi:10.ARTICLE IN PRESS W. Prins. Diurnal-scale analysis of biomass burning is very important because in the tropics. These fires were aided by the prevalent and fire-favourable climate conditions of long and persistent drought experienced in the country in the past few years. & Stibig. are indicative of the response of fuel moisture content throughout the day typically reaching minimum in the afternoon. However.M. and Everson (1985) found that the moisture content of the upper grass layer declined from 33% in the morning to 11% in the afternoon in the montane grasslands of Natal (South Africa) thus influencing the diurnal fire activity patterns. Since July is near the peak of the fire season in the country (Dlamini. In this study. 1997). Low fire activity is observed in the mornings and in the early evening. typical of human managed ecosystems (Zoumas. (2003) during southern African biomass burning seasons. The diurnal pattern in this study is more or less similar for all the eco-climatic zones with minor differences that could be attributed to differences in fuel (vegetation and land cover) types and the possible influence of elevation and prevailing weather conditions. & Jaramillo.1016/j. the MSG-SEVIRI (Fig. 2006). (2006). Feltz. Smith. the behaviour of the Highveld Fig. Lee-Wagner. Eva. Please cite this article in press as: Wisdom M. The diurnal variations also indicate the influence of diurnally driven processes of boundary layer mixing or fast changing conditions such as frontal passages and windy conditions (Trollope. as stated in the previous section. & Ward.10. is the effect of the dense smoke plumes which could have obscured a majority of the fires mainly burning intensely within the plantation forests. Dlamini / Applied Geography xxx (2009) 1–9 7 28 July. This indicates that the fires were spreading faster and burning more intensely. The number of active fires decreased after 29 July 2007 after the large areas of vegetation (mainly plantation forests and grass) biomass and other material had been quickly consumed. Everson.

211–221). J. The diurnal fire cycle slightly varies between the eco-climatic zones since most of the large fires were concentrated in the Highveld and Upper Middleveld and also points to the influence of land cover/ land use. Ceccato. O’Neill. TRENDS in Ecology and Evolution. 2035.1997.2006. daytime thermal radiation from hot background surfaces and/or increased solar reflection can also saturate the 3. & W.. M..F. J. M. Di Bisceglie. D. Dubovik. & Verbyla. Xiong. Similarly. There is an observed distinctive spatial distribution pattern in the fire activity during the month of the Swaziland fire disaster. C. Tallahassee. Geophysical Research Letters.. The burning patterns presented in this study could be used as input information for the further analysis of current and future fire regimes and fire risk according to local and global change. R. Rupp.. Galdi. Proceedings of the 22nd tall timbers fire ecology conference: Fire in temperate. The high temporal resolution MSG-SEVIRI detected more fires in the Highveld and Upper Middleveld and reveals a distinct diurnal variation in fire activity where peak activity is observed mainly in the afternoon and late evening/early morning during when fires are most likely to burn. & Armitage.. (1997). Eck. Schafer. doi:10. Boreal. In R. S. In Proceedings of 2nd international conference on recent advances in space technologies (pp.O. doi:10.). In general. Raimadoya. Frost. de Groot (Eds. High aerosol optical depth biomass burning events: a comparison of optical properties for different source regions.2. H.. Elvidge. Zavagli.J.. California: Environmental Systems Research Institute. Beck.10. D. M. The capabilities of remote sensing data. Cheney. Dlamini. fire algorithms. T. P. C. P. M. In IEEE international conference on geoscience and remote sensing symposium. & Li. R. Dlamini. I.. Holben. S. 61–65. Fire behaviour. Stocks. Kasischke. S... Active fires from polar-orbiting satellites.. M. (2001). Cochrane. A. S. B.. R. P..7 mm may be a combination of both emitted thermal and a reflected solar radiation which may elevate the signal to produce false fire alarms (Gao. J... Land Use and Fires. M.1029/2003GL017861. and Understanding Trajectories of Change on the Earth’s Surface (pp.A. E. (2006). L. et al.2008.F. T. S. Moran. elevation and the prevalent weather characteristics in line with observation from the region ad other parts of the world.. InProceedings of the IEEE international geoscience and remote sensing symposium 2005. Brown.J. Fire situation in Swaziland. A contextual algorithm for AVHRR fire detection. Berlin: Springer-Verlag. Justice E. J. 24–26).. Applied Geography (2009). J. Mccrae. caution needs to be taken and considered with satellite observations because the daytime radiance signal at 3. Dlamini.L. Romo. D. Everson. Gill. represent a limited temporal sample due to the satellites’ restricted overpass frequency. S. B.D. Latest algorithms and scientific developments for forest fire detection and monitoring using MSG/SEVIRI and MODIS sensors. 151–175).A. Cisbani. B. C. M. R.. In Proceedings of the 5th SAFNet workshop: Towards meeting fire management challenges in Southern Africa (pp. Groves.. O. D. Mcguire. 2005). Galdi. the highest proportion of fire hotspots were detected in the Highveld and Middleveld. Flannigan.M. Fire and the Australian biota (pp. Justice. Bond. C. RAST. P. Please cite this article in press as: Wisdom M.P. Calle. P. Constant false alarm rate in fire detection for MODIS data.G. and Montane ecosystems (pp.apgeog. on the contrary. Costantini. This study also helps to understand active fire detection differences that can be explained by differences in orbits. J. C. S. O. Diurnal fine fuel moisture and FFMC characteristics at northern latitudes.. T. including the unique strengths and limitations. 1065. J.. Redlands.1109/IGARSS. Fire as a global ‘herbivore’: the ecology and evolution of flammable ecosystems. N. N.O. A review of the July 2007 Swaziland fire disaster using GIS and MODIS data. & Navarro. Global Biogeochemical Cycles. Dlamini / Applied Geography xxx (2009) 1–9 and Upper Middleveld fires produced a relatively uniform graph indicating little variation due to the intensity and size of the fires.. Conclusion and recommendations The usefulness. L. I. W.L. PositionIT. 2006 (pp. C. Conard.ARTICLE IN PRESS 8 W.007 . Giglio.). ArcGIS: Release 9. & Everson.G. E.. A. However.. Janetos. IGARSS. doi:10. Mustard. Kasischke. coupled with the diurnal fire cycle (Hyer. (1985). A. Turner & M.S. Gutman. 17. Smith. Characterization of the July 2007 Swaziland fire disaster using satellite remote sensing and GIS. Episcopo. 4153–4156). M. International Journal of Remote Sensing.. Despite the observed limitations. B. W.). T. IGARSS. & Keeley. & Merlet. B.. N. T. et al.1016/j.. L. A technique for automatic fire detection from geostationary optical sensors and its validation on MSG SEVIRI data. References Andreae. & Ullo. 1569–1572). A.. & Ullo. and fire pixel characteristics.. Mangochi. (1981). Skole. Reid. J. M. IGARSS. Emission of trace gases and aerosols from biomass burning.. A. (2005). Malawi: United Nations Environment Programme (UNEP)/Southern Africa Fire Network (SAFNet). Flasse. F. Downey. Monitoring. satellite monitoring of fires proves to be useful for a resource-constrained developing country such as Swaziland. Roy. 2006). P. (2005). Journal of the Grassland Society of Southern Africa. where a majority of the devastating fires raged through the country’s plantation forests thus indicating the high fire risk in these areas.608950. Characteristics of fire behaviour in the montane grasslands of Natal. 387–394.. Increasing observational and technological skill and experience at recognizing dangerous fire activity will offer significant lead time to anticipate future disasters and minimize losses and environmental impacts from such disasters.R.. 2007).D. Moclan. E. (2004). 20(7). Costantini. 30(20). & Greco. (September/October 2007). R. P. Cisbani. (2005). Noble (Eds. Cochrane (Eds. (9–13 August 2005). Florida: Tall Timbers Research Station... 955–966.C. F..H. such as demonstrated in this study. 8 (pp 5717–5720). doi:10. integration of multi-sensor data can benefit from high temporal resolution (MSG-SEVIRI) data and medium spatial resolution (MODIS) fire products to provide more information than either product could provide alone. Land Change Science: Observing. detecting shortduration fires and for resolving the diurnal behaviour of large fires.. A. & Ceccato. Episcopo.. & I. M. In: G. 118–123). Csiszar. 329-351). (1996). 2007. of remotely sensed active fire data from both polar-orbiting and geostationary satellites in understanding the geographic and temporal characteristics of fire is demonstrated in this study. (2003). M. (2006).. Flasse. 2.. Rindfuss.1109/IGARSS.... Casanova. 419–424.7 mm channel (Di Bisceglie. O. & Allen. In Proceedings of IEEE international geoscience and remote sensing (pp. F. The findings point to the geostationary sensor’s possible usefulness for evaluating diurnal fire evolution. Ultimately. (2004). 13–21. P. offer exceptional value for early warning and disaster assessment even for developing countries like Swaziland. In A. J. E. Canberra: Australian Academy of Science. Remote sensing and GIS tools to support vegetation fire management in developing countries. The geostationary MSG-SEVIRI sensor offers a great opportunity to develop a robust early warning system capable of timely identification of forest fires and to monitor them in real time thus minimising damage. Engstrom. 15. D. M. ESRI. W.

J.. Feltz. A study of MODIS fire detecting channel centered at 3. Remote Sensing of Environment.. Kasischke. Proceedings of the 29th ISRSE conference. -J. Journal of Geophysical Research. A. A. G. J.10. 244–262. Journal of the Grassland Society of Southern Africa. Giglio. & Vosloo. MODIS collection 4 active fire product user’s guide version 2. A research on fire automatic recognition using MODIS data. 73–76. D.. J.. & C. 47. An overview of GOES-8 diurnal fire and smoke results for SCAR-B and the 1995 fire season in South America. Biomass dynamics associated with deforestation. Satellite-based early warnings of fires: reducing fire flashovers on transmission lines. L. L. Gibbons. van Wilgen. (2004). (2005). C. (2006). & J. J. Remote Sensing of the Environment. M. Crosslink. J. S. (2006). A. Just. 107. & Kaufman. F.. W. (2005). Project No. 45–52. Giglio. L. Giglio.07. 2006firecongressproceedings/PDF/Athanassios%20Zoumas%20possible%20poster.ARTICLE IN PRESS W. Y. Trollope. Preparing for METEOSAT second generation: early fire detection in the Mediterranean area based on MSG. In F. J... 31821–31835.. Characterization of the July 2007 Swaziland fire disaster using satellite remote sensing and GIS.. Csiszar. G. Tors in Swaziland. J. 178-8. M. Justice. Wildland fire handbook of Sub-Saharan Africa (pp. & Price-Williams. & C. Johannesburg: Witwatersrand University Press. 6.). S. In: IEEE International Geoscience and Remote Sensing Symposium (IGARSS03). land use and climate. & Potgieter. Australian Meteorological Magazine. J.. A. Fire regime of the Kruger National Park for the period 1980–1992. The vegetation and fire regimes of southern hemisphere Africa. Fire emissions from C-3 and C-4 vegetation and their influence on interannual variability of atmospheric CO2 and delta(CO2)-C-13. G. J. S. Justice. (1997).. J. R. Continuous monitoring of forest fires in the Mediterranean area using MSG. 35.emmps. Fire in Southern African savannas.881716. J. B. (August 2005). 27–43). M. Prins. P. IGARSS. Bulletin of the American Meteorological Society. W. D. M.wsu. 2–8.. Remote Sensing of Environment. 72--78. Ahern. A. De Ronde. climate. Lavorel. Owens. The Geographical Journal.. Trollope. L. Andreae. 7. An introduction to Meteosat Second Generation (MSG). W.. B. H.). 571–581. 156–160. M. et al. Matson.. 273–282. J. 3175–3226. San-Miguel-Ayanz. M.. (1981). W. P. T. 36.. de Ronde (Eds. C. Current approaches to modelling the spread of wildland fire: a review. Photogrammetric Engineering and Remote Sensing. G. V. Goldammer.. 2006.95-mm. Field Doc. 3rd International Fire Ecology & Management Congress.. Goldammer. 83. E. doi:10. An enhanced contextual fire detection algorithm for MODIS. 176. Physiographic map of Swaziland. F.pdf>. D.2006.. Giglio. Giglio. 4. M. G. & Jaramillo. MORDYA. 33–53. et al. (2003). Kasibhatla.. J. C. Burned area forecasting using past burned area records and southern oscillation index for tropical Africa (1981–1999).umd. & Allen.1016/j. Steele. D. J.apgeog. Peck. Cummings.. ´ ´ ˜ Riano. Global Biogeochemical Cycles. W. (1993). 147. Zhou. E. S. Tjemkes. J. 90pp. Rota. 977–992. Natural Hazards. & Geldenhuys.. In J. Collatz.. H. (2005). 1311–1318. Air Sciences. Laneve. (2000).). Goldammer. Randerson. Kasibhatla. A multi temporal approach to fire detection using MSG data. Collatz. Hyer. (1985).. Ministry of regional development and youth affairs. 19(2). 83. Please cite this article in press as: Wisdom M. Justice (Eds. 111. Freiburg: Global Fire Monitoring Centre. B.. D. (1998). Forest Ecology and Management. <http://maps. (2003). and conversion to cattle pasture in a Mexican tropical dry forest. Schmetz. 17–22. San Diego. Long. (2004). WRAP. O. G.1029/2006JD007234.. T. & Ustin.2008. & Ollero. (2007). Kerkann. I. Journal of Geophysical Research..000. Van den Bergh. 222–245. (2006). & Ward. G.. P. F. J. 2502–2504. Hillger. Applied Geography (2009). F. P. Scale 1:250. W. Dlamini / Applied Geography xxx (2009) 1–9 9 Frost.geog. (2006).M.07.. J. Baron Martınez. Rice. doi:10.. Wang. L. L. 27–59). G. E. & Justice. Science.007 .. D. fire. Schmetz. S. (2003). J.. E. R. A. 303. Continental-scale partitioning of fire emissions ˜ during the 1997 to 2001 El Nino/La Nina period.3. J. Xiong. O. 1. (2002). Active fire detection for fire emergency management: potential and limitations for the operational use of remote sensing. W. H. C.. L. California <http://www. (2006). Van der Werf. 41 Mbabane. Remmelzwaal. 87. Accessed 25.. A.. Kelha.. E. 44(10). Morisette. Randerson. T. Van der Werf. FAO/UNDP/GOS land use planning for rational utilization of land and water resources Project SWA/89/001. Lobamba: Swaziland National Trust Commission.. Collatz. S. Laneve. The Hague: SPB Academic Publishing.. 145–170). E. S. P. 3–18. D01302.. Atmospheric Chemistry and Physics Discussions.... R.. Castronuovo. L. Gregoire. E.. D. Vulnerability of land systems to fire: interactions among humans... pp.pdf>. W.3. W. & Cadau. Y. Prins. R.... & Stibig.. V. L. G. The Swaziland national trust commission occasional paper no. S. 22. Van Wilgen.. doi:10. A climatology of extreme fire weather days in Victoria. G.. Buenos Aires. 12. Identification of subresolution high temperature sources using a thermal IR sensor. S... S.. Zoumas. J. C. The MODIS fire products. Korontzi. Ravail. France: IGARSS. 55. Menzel. & Oshika.. J.. 4. van der Werf. 2. Roy. Arellano. Lindesay (Eds. M. Flynn. O. G. (2007). D. 1–12. 48–51. (2001). J. -M. E.07. B. N. Jahjah. doi:10. Argentina. Trollope. Ecological and atmospheric perspectives (pp. X. Giglio.. Mitigation and Adaptation Strategies for Global Change. (1981).. M. (1993). Dlamini.. Tressel.. (25th 2007). In Proceedings of the IEEE international geoscience and remote sensing symposium (pp. 2002. C. M. & Feltz. G. H... (2002). Journal of Geophysical Research. Global distribution and seasonality of active fires as observed with the Terra and Aqua Moderate Resolution Imaging Spectroradiometer (MODIS) sensors. & Li.1029/2005JG000142.. J. doi:10. D. R. C. viii.-C. Kasibhatla. (1998).. W. L. L. J. D. et al.. J. T. M. Koedoe. L. & Marzoli. ESI Africa. 361–376. Pili. P. Interannual variability of global biomass burning emissions from 1997 to 2004. & Wang. A. A. et al. Inc.. J. (2007). 2. Flannigan. Eva.. S. M. Toulouse. O. IEEE Transactions on Geoscience and Remote Sensing. the atmosphere. L.. Global and regional wildfire monitoring: Current status and future plans (pp. 2761–2768. Fire behaviour. and ecosystems. 112. C. P. G. Development of 2000–2004 baseline period and 10 2018 projection year emission inventories. Perry. E. W. C. (2006). 1102–1104). J. Fire disaster-report. Randerson. Overview of current and future diurnal active fire monitoring using a suite of international geostationary satellites. doi:10. M.. Western Regional Air Partnership.1109/TGRS. J. Accessed 14. Still. Kauffman. Progress in Physical Geography. & Scholes. J. S..07. Effects of source temporal resolution on transport simulations of boreal fire emissions. Proceedings of the 2005 International Workshop on the Analysis of Multi-Temporal Remote Sensing Images. J. J. Gao. Lambin. C. Lee-Wagner. (1983)The Atlas of Swaziland. Descloitres. & Frost.1029/2004GB002366.1007/s11027-006-9046-5. G02016. & Dozier. Fire behaviour in the Kruger National Park. D. & Arellano. Goudie. O. November 13–17. (2007). J.. Diurnal fire information of tropical fires in relation to land cover.. In B.. S. F. L. Giglio. Civilian uses of surveillance satellites. & Scholes. D. The Times of Swaziland. Moreno Ruiz.

Sign up to vote on this title
UsefulNot useful