Impacts and Drivers of Summer Wildfires in the Cape Peninsula: A Remote Sensing Approach

Abstract

Over the years, the Cape Peninsula has seen a rise in the number of fires that occur seasonally. This study aimed to investigate the extent of fire spread and associated damages during the 2023/2024 Cape Peninsula fire events. Remote sensing datasets from Sentinel-5P, Sentinel-2, Moderate Resolution Imaging Spectroradiometer (MODIS), and Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2) were used. Most of the fires on the northern side of the Cape Peninsula had a short burning span of between 6 and 12 h, but fires with a duration of 12–24 h were minimal. The northern area is composed of low forests and thickets as well as fynbos species, which were the primary fuel sources. Excessive amounts of carbon monoxide (CO) and black carbon (BC) emissions were observed. High speeds were observed during the period of the fires. This is one of the factors that led to the spread of the fire. Relative humidity at 60% was observed, indicating slightly dry conditions. Additionally, the Leaf Water Content Index (LWCI) indicated drier vegetation, enhancing fire susceptibility. High temperatures, low moisture and strong winds were the main drivers of the fire. The Normalized Burn Ratio (NBR) values for the targeted fires showed values close to −1, which signifies presence of a fire scar. The study can be of use to those in the fire management agencies and biodiversity conservation in the region.

1. Introduction

The rise in global temperature has led to an increase in wildfire activities [1,2]. Wildfires can be produced by natural or anthropogenic causes and are estimated to burn approximately 82 million ha per year worldwide [3]. Wildfires can have harsh environmental and societal impacts on urban settlements, agricultural lands, climate and ecosystems [4,5]. Conditions such as high temperatures and low precipitation can lead to drier-than-normal conditions which favour the start and spread of wildfires. During dry conditions, fuels for wildfire such as grasses and trees, can desiccate and thus become more susceptible to ignition and high flammability [6,7].

Although most wildfires are considered harmful, some plant communities depend both directly and indirectly on consistent fires in order to survive. Fire plays a crucial role in shaping and maintaining fynbos ecosystems, and many plant species within this habitat have evolved adaptations to survive and even thrive in fire-prone conditions. [8,9]. Fires can also kill diseases and insects that could otherwise destroy many plants [10]. On the other hand, fires can modify ecosystems by introducing invasive alien species which can colonise and drive plant diversity in that habitat [11,12,13].

The burning of vegetation is known to release toxic greenhouse gases into the atmosphere, thereby creating a hazardous environment [14,15]. Such gases include carbon dioxide, methane and nitrous oxide [16]. Other pollutants released into the atmosphere include black carbon, organic carbon and smoke. Due to the dispersion and transport of these pollutants, this has an impact on the budget of organic trace gases in the tropical marine atmosphere and in the remote troposphere [17]. Moreover, vegetation burning emissions are also a source of a number of halogenated compounds such as methyl chloride and methyl bromide that are long-lived enough to destroy ozone in the stratosphere [16].

Over the years there have been several devasting wildfires globally, which resulted in large amounts of pollutants injected into the atmosphere. California, in the United States of America (USA), has experienced many extensive fires. Keeley and Syphard [18] showed that in the year 2020, the most extensive wildfires were experienced in the history of California. Eck et al. [19] also showed that the California wildfires during the August-to-October-2020 period were exceptional in terms of the fire severity and area burned. Extremely dry conditions exacerbated by high temperatures were some of the drivers for the large fires. The Australian region also experienced extensive wildfires in the period of September 2019 to February 2020 [20,21]. The fires were responsible for 33 direct deaths, over 400 smoke-related premature deaths and the loss of over 3000 houses [22]. Jalaludin et al. [23] showed that the fire burned more than 12.6 million hectares and emitted about 430 tonnes of carbon dioxide into the atmosphere.

South Africa has also witnessed extreme wildfires in the past, but not to the level of the California and Australia fires discussed earlier. As an example, approximately 35,000 fires were reported during the 2008 fire season, with a majority of fires originating from open flames during waste, grass or bush burnings [24], but none of the fires were devasting. More recently, in June 2017, a large wildfire was reported to have burned about 150 km² around the town of Knysna in the Western Cape province [25]. The fire destroyed over 800 buildings and about 50 km² of forest plantations, and was responsible for seven deaths [24]. Drought conditions experienced during the winter season of 2017 were one of the factors that contributed to the spread of the fire [25,26]. In the weeks leading to the fire, hot and dry berg wind conditions were observed, which reduced already critically low plant-moisture and soil-moisture levels [26].

Since 2019, the WC fires have been more frequent and severe, raising a concern about the water and air quality, land condition, biodiversity, vegetation structure, and greenhouse gas emissions of the affected area [27]. More recently, large wildfires were observed near Cape Town between the 24th and 26th of December 2023. The wildfires started on the mountain slopes of Simon’s Town and swept across thousands of hectares in the area, resulting in a lot of residents being evacuated from their homes. Minimising the occurrence of these fires and reducing their impact on the environment requires a clear understanding of the drivers, and the potential consequences. However, obtaining such information is often challenging to acquire, especially during the fire, due to its hazardous nature.

Remote sensing (RS) technologies allow for acquisition of information even in inaccessible areas without any human contact, and are equipped with instruments that can detect burnt surface areas. When vegetation burns, it goes through physical and chemical changes. This process involves the changes in the temperature, charring of stems, and loss of chlorophyll and water, leading to alterations in the colour and texture of the burned area, and, consequently, a formation of a blackened surface and charcoal residues. These changes produce a distinct spectral reflectance of the burnt vegetation, with charcoal residues causing reduced absorption of red light and increased reflection of near-infrared (NIR) light. The thermal infrared (TIR) wavelength is sensitive to the rise in temperatures caused by a fire, enabling effective detection of emitted smoke. The loss of water from vegetation during fires induces water stress. The short-wave infrared (SWIR) region is responsive to vegetation’s water content, and burnt scars cause increased reflection in this region [28]. It is these alterations in the spectral signature of vegetation that enable the detection of post-fire scars and greenhouse emissions from satellite imagery.

Currently, RS of wildfire management focuses on the use of MODIS 4- and 11-micrometre radiances fire products, which have been proven to be accurate and effective for fire management [29]. Several sensors from the ESA Sentinel missions are equipped with the above-mentioned fire-sensitive wavelengths; for example, Sentinel-2, with an improved spatial resolution of 10 m to 20 m. The sensor is also equipped with three narrow bands in the red-edge position (REP) which is proven to be sensitive to changes in vegetation’s biochemicals and biophysical composition, demonstrating a potential for detection of burn-area scars. The use of Sentinel-2 datasets for mapping wildfire scars and quantifying fire severity have rarely been investigated [30]. It is against this background that this work explores the potential of the Sentinel-2 MSI to detect the post-fire scars, quantify the severity, and also monitor the emission of greenhouse gasses from wildfires in fynbos vegetation using Sentinel-5P datasets.

The objective of this study is, therefore, to characterise the extent and drivers of the wildfires that occurred in the Cape Peninsula during the 2023/2024 summer season by analysing climatic, environmental, and meteorological data. In this work, we provide insights into the factors that contributed to the wildfires and their impacts on the environment and local communities. Studying factors such as fire sources, fuels, and the environment can enable the use of measures aimed at controlling and preventing wildfires. Moreover, leveraging advanced technologies such as remote sensing can assist in forecasting forest fire occurrences. This information facilitates timely interventions, mitigating the impact of wildfires on biodiversity and human health. Ultimately, research into forest fires plays a crucial role in safeguarding ecosystems, resources, and human well-being.

2. Study Site

The study area covers approximately 475 km² and is located in the southwestern region of South Africa [31]. It is predominantly covered by fire-prone fynbos shrubs. Figure 1 displays a land-cover map depicting part of the Cape Peninsula. The Cape Peninsula experiences a Mediterranean climate characterized by warm, dry summers and cool, rainy winters. Winter precipitation is brought by frontal cyclones originating from circumpolar westerly winds, occurring typically once a week during the mid-winter months from June to August [32]. In summer, the climate is influenced by the South Atlantic High-Pressure Cell, resulting in southeasterly winds along the west coast. Anthropogenic climate change has altered weather patterns in this area, leading to increased conditions favourable for fires [32].

The topography of the Cape Peninsula features sandstone mountains, nutrient-poor soils, and diverse climatic conditions, contributing to its designation as a biodiversity hotspot. These environmental factors, coupled with the region’s infertile soils and fire intervals occurring every 4–40 years, make the Cape Peninsula highly susceptible to fires. The dominant vegetation types, such as fynbos and renosterveld, have evolved to depend on and thrive under these fire regimes, which are crucial for maintaining ecological balance and biodiversity in the area [31].

3. Data Acquisition and Methods

The description of the datasets used in this study is summarised in

Table 1. Summary of the parameters used in this study.

Data Source	Product Name	Spatial Resolution	Spectral Resolution	Date of Acquisition	Output Data
TROPOMI	CO	7.0 km × 3.5 km	267––2389 nm	12–25 December 2023	Spatial distribution map and Timeseries plot
MODIS	Burnt Area date and LST	500 m	36 bands (400–14,400 nm)	1 January 2023 to 30 April 2024	Spatial distribution map and Timeseries plot
AIRS	Relative Humidity	13.5 km	2378 channels (3.7–15.4 µm)	1 January 2023 to 30 April 2024	Timeseries plot
MERRA-2	Wind Speed	0.625° × 0.5°	N/A	1 January 2023 to 30 April 2024	Timeseries plot
OMI	UVAI	13 × 24 km2	0.42–0.63 nm	1 January 2023 to 30 April 2024	Timeseries plot
Sentinel-2	ΔNBR	10 m, 20 m, 60 m	13 bands (443–2190 nm)	Mean for Jan 2024; October 2023	Spatial distribution map
HYSPLIT model	5-day forward air-mass trajectories	-	-	25–29 December 2023	Map showing trajectory of air masses

. A brief description of the dataset and the methods are shown in Section 3.1 and Section 3.2.

3.1. Data Acquisition

3.1.1. Sentinel-5P/TROPOMI

Sentinel-5P employs the TROPOspheric Monitoring Instrument (TROPOMI) to conduct global measurements of atmospheric trace gases, aerosols, and cloud distribution. TROPOMI, a hyperspectral imaging spectrometer, captures Earth’s radiance in ultraviolet–visible (UV–VIS, 267–499 nm), near-infrared (NIR, 661–786 nm), and shortwave infrared (SWIR, 2300–2389 nm) wavelengths, achieving a ground pixel resolution as fine as 7.0 km × 3.5 km. With a swath width of 2600 km, it provides nearly daily coverage of the entire globe. TROPOMI can map various trace gases such as nitrogen dioxide, ozone, formaldehyde, sulfur dioxide, methane, and carbon monoxide (CO). More details about Sentinel-5P can be found in References [33,34,35]. This study specifically focuses on the CO data.

3.1.2. MODIS

The Moderate Resolution Imaging Spectroradiometer (MODIS) is installed on both the Terra and Aqua satellites [36]. It features a broad swath width of 2330 km and captures images of the entire Earth’s surface every 1 to 2 days. MODIS sensors are equipped to measure 36 spectral bands ranging from 0.405 to 14.385 µm, providing data at three spatial resolutions: 250 m, 500 m, and 1000 m [36]. Justice et al. [37] provide a comprehensive discussion of various MODIS products. This study specifically focuses on two MODIS products: Land Surface Temperature (LST) and Thermal Anomalies/Fire products. LST data are derived using two algorithms: the generalized split-window algorithm and the day/night algorithm, detailed by Zhao et al. [38]. The fire detection algorithm is further elaborated in the work of Giglio et al. [39].

3.1.3. AIRS

The Atmospheric Infrared Sounder (AIRS) is a hyperspectral instrument onboard NASA’s Aqua satellite, launched in 2002 for climate research on greenhouse gases and weather forecasting. It utilizes a scanning spectrometer with 2378 infrared (IR) channels distributed across wavelengths in the ranges of 3.74–4.61 µm, 6.20–8.22 µm, and 8.8–15.5 µm, crucial for profiling atmospheric temperature and relative humidity. Additionally, AIRS includes four visible (VIS) and near-infrared (NIR) channels spanning 0.40–0.94 μm, primarily used for cloud detection within the IR field of view. The instrument achieves an IR horizontal resolution of 13.5 km and vertical resolution of 1 km, while the spatial resolution for VIS/NIR is approximately 2.3 km. For more details on AIRS, refer to Aumann and Miller [40], Chahine et al. [41], and Menzel et al. [42]. This study specifically focuses on utilizing relative humidity data obtained from AIRS.

3.1.4. MERRA-2

The Modern-Era Retrospective analysis for Research and Applications, Version 2 (MERRA-2) integrates satellite observations of aerosols and models their interactions with other physical processes in the climate system. It replaces the original Global Modeling and Assimilation Office (GMAO) reanalysis, MERRA [43]. MERRA-2 is generated using version 5.12.4 of the Goddard Earth Observing System Data Assimilation System (GEOS DAS). Gridded data are available at a resolution of 0.625° longitude × 0.5° latitude across 72 sigma–pressure hybrid layers from the Earth’s surface up to 0.01 hPa [44]. For more information on MERRA-2, refer to Gelaro et al. [45], Buchard et al. [46], and Randles et al. [47]. This study utilizes wind speed data obtained from MERRA-2.

3.1.5. OMI

The Ozone Monitoring Instrument (OMI) was launched aboard the EOS-Aura satellite in July 2004. OMI observes backscattered solar radiance during the dayside part of each orbit and solar irradiance near the northern hemisphere terminator once daily, across three wavelength channels covering 270–500 nm (UV-1: 270–310 nm, UV-2: 310–365 nm, visible: 350–500 nm), with spectral resolutions ranging from 0.42 to 0.63 nm [48]. OMI data products are derived from the ratio of Earth radiance to solar irradiance and are provided in Level-2 orbit files containing measurements of trace gases such as O₃, NO₂, SO₂, and HCHO, as well as UV-absorbing aerosol and cloud properties. For a comprehensive overview of OMI, consult Levelt et al. [49].

3.1.6. HYSPLIT Model

The Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT) model is a versatile tool capable of computing simple air-parcel movements as well as conducting complex simulations of dispersion and deposition [50]. One of its primary uses involves performing back-trajectory analyses to determine the origins of air masses and establish relationships between emission sources and receptor locations [51]. The trajectory-frequency feature initiates trajectories from a specified location and altitude every 3, 6, or 12 h throughout the duration of the meteorological dataset. It then counts how often trajectories pass over each grid cell defined by the user. Trajectories are computed using vertical motion calculations, detailed at http://ready.arl.noaa.gov/HYSPLIT.php, accessed on 1 June 2024. The computation of trajectories introduces inherent uncertainties due to potential inaccuracies in input meteorological data and numerical methods used. To address uncertainties associated with individual trajectories, HYSPLIT can operate in ensemble mode, generating multiple trajectories from a single meteorological dataset [52].

3.2. Methods

Sentinel-2

Sentinel-2 consists of a pair of polar-orbiting satellites, Sentinel-2A and Sentinel-2B, positioned in the same sun-synchronous orbit with a 180° phase difference between them. The primary objective of Sentinel-2 is to monitor changes in land surface conditions. It features a broad swath width of 290 km and provides high revisit times: every 10 days at the equator with one satellite, and every 5 days with two satellites under cloud-free conditions, resulting in a revisit time of 2–3 days at mid-latitudes. Sentinel-2 is equipped with 13 bands covering the Visible and Near-Infrared (VNIR) region with 8 bands, and the Shortwave Infrared (SWIR) region with 2 bands, offering spatial resolutions of 10 m, 20 m, and 60 m. For further information on Sentinel-2, consult Spoto et al. [53] and Sudmanns et al. [54].

Two Sentinel-2 images from before (December 2019) and after (December 2023) the fire were acquired from the Google Earth Engine cloud processing software. The Leaf Water Content Index (LWCI) data were calculated for both images, using the formula

$$\mathrm{L}\mathrm{W}\mathrm{C}\mathrm{I}={\displaystyle \frac{{\mathrm{R}}_{\mathrm{N}\mathrm{I}\mathrm{R}}-{\mathrm{R}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}}}{{\mathrm{R}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+{\mathrm{R}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}}}}$$

(1)

where ${\mathrm{R}}_{\mathrm{N}\mathrm{I}\mathrm{R}}$ is the Near Infrared Band and ${\mathrm{R}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}}$ is the Short-Wave Infrared Band. The difference in leaf water content between the two years was determined by subtracting the 2019 Leaf Water Content Index (LWCI) values from the 2023 LWCI values. The LWCI index ranges from −1 to 1, where higher values indicate greater leaf-water content, suggesting healthier vegetation, while lower values indicate reduced water content, often due to stress or damage such as fire [55].

Mapping fire scars and assessing burn severity can be challenging, due to the absence of standardized measures [56]. Consequently, various techniques are employed to identify post-fire scars and evaluate severity, with many methods relying on vegetation response. Previous research has demonstrated that image differencing is an effective method for detecting post-fire scars [56,57]. Normalizing data improves the accuracy of fire-scar detection [58]. Although the Normalized Difference Vegetation Index (NDVI) has traditionally been used for fire-scar mapping between Near-Infrared (NIR) and visible (red) wavelengths, its accuracy can be affected by pre-fire vegetation density and greenness [27,56]. Moreover, atmospheric interference from smoke and dust during fires can diminish the accuracy of indices derived from visible wavelengths. Studies indicate that utilizing the Shortwave Infrared (SWIR) region enhances accuracy in mapping burn severity. In this study, the Normalized Burn Ratio (NBR) between NIR and SWIR regions is employed to detect recent fires in the Cape Peninsula area and assess burn severity based on NBR changes. The change in NBR (ΔNBR) is then utilized to quantify fire severity. NBR values range from –1 to 1, with values closer to –1 indicating recently burned areas or bare ground, and values approaching 1 indicating healthy vegetation presence. Values near zero signify non-burned areas.

$$\mathrm{N}\mathrm{B}\mathrm{R}=({\mathrm{R}}_{\mathrm{N}\mathrm{I}\mathrm{R}}-{\mathrm{R}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}})/({\mathrm{R}}_{\mathrm{N}\mathrm{I}\mathrm{R}}+{\mathrm{R}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}})$$

(2)

Then,

$$\mathrm{d}\mathrm{N}\mathrm{B}\mathrm{R}=\mathrm{p}\mathrm{r}\mathrm{e}\mathrm{f}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{N}\mathrm{B}\mathrm{R}-\mathrm{p}\mathrm{o}\mathrm{s}\mathrm{t}\mathrm{f}\mathrm{i}\mathrm{r}\mathrm{e}\mathrm{N}\mathrm{B}\mathrm{R}$$

(3)

where $\mathrm{d}\mathrm{N}\mathrm{B}\mathrm{R}$ is delta NBR, ${\mathrm{R}}_{\mathrm{N}\mathrm{I}\mathrm{R}}$ is the reflectance at the NIR and ${\mathrm{R}}_{\mathrm{S}\mathrm{W}\mathrm{I}\mathrm{R}}$ is the reflectance at SWIR. For this work we adopted the $\mathrm{d}\mathrm{N}\mathrm{B}\mathrm{R}$ interpretation values proposed by the USGS to classify levels of severity of the present fire; see

Table 2. Burn-severity levels obtained calculating dNBR, proposed by USGS.

Severity Level	dNBR Range (Scaled by 1000)
Enhanced regrowth, high (post-fire)	−500 to −250
Enhanced regrowth, low (post-fire)	−250 to −101
Unburned	−100 to 99
Low Severity	100 to 269
Moderate–low severity	270 to 439
Moderate–high severity	440 to 659
High severity	660 to 1300

for reference.

4. Results and Discussion

4.1. Fire Detection and Burnt Scars

Figure 2 shows images of the burned-area dates in the Cape Peninsula (Figure 2a) and the areas where fires were detected on 22 December 2023 (Figure 2b). Most of the fires on the northern side of the Cape Peninsula had a short burning span of between 6 and 12 h, but fires with a duration of 12–24 h were minimal. The northern area is composed of low forests and thickets, as well as fynbos species (see Appendix A). The low forests and thicket have predominantly short-lived, bird-dispersed diaspores which produce shade-tolerant seedlings that require long fire-free intervals for recruitment into fynbos [24]. This is one of the reasons for the short fire span. Fynbos species, on the other hand, depend on regular crown fires which stimulate recruitment from persistent soil- and canopy-stored seed banks of relatively short-lived species [59]. In the central part of the Cape Peninsula, several fires with a duration of 12–24 h are dominant. This area is dominated by low-shrubland fynbos species. Although fynbos requires burning every 12–15 years to survive, shorter or longer time intervals would eliminate many shrub species. A few fires with a duration of more than a day are also observed. These fires had an impact on the residential areas (built up area) (see Figure 1). Additionally, the fires are likely to have burnt the infrastructure in this area. Figure 3c shows the NBR index derived from Sentinel-2 imagery for the burnt area. This index was used to detect the presence of fire scars in the Cape Peninsula area. NBR values for the targeted fires are close to −1, which denotes the presence of a fire scar. The largest burnt area is calculated to be 32.93 km², accounting for 6.93% of the total area burnt. The fire scar can be used for (i) assessing fire hazard and risk, (ii) monitoring and reporting ecological impacts of fire, (iii) developing strategic and operational fire planning and management, and (iv) informing carbon abatement schemes and greenhouse emissions.

The severity of the fire was quantified using deltaNBR, shown in Figure 3d. The deltaNBR values observed from the detected scar range between 140 and 600. The largest portion of the fire scar had deltaNBR values ranging from 440 to 600, which, according to the burn-severity levels proposed by the USGS (see Table 2), indicate moderate to highly severe fires.

4.2. Emissions and Meteorological Drivers

Wildfires release a mix of gases and particulate matter into the atmosphere [60]. CO and BC are some of the pollutants emitted during the biomass burning processes [17]. They are by-products of incomplete combustion of carbon-containing fuels. Figure 3a,b show the spatial distribution of CO during and after the wildfire episodes in the Cape Peninsula. During the wildfire, see Figure 3a, low CO column-density values between 0.0199–0.0250 mol/m² are observed in the northern part of the Cape Peninsula. Moreover, moderate-to-high CO column-density values between 0.0277 and 0.0438 mol/m² are observed from the central to some southern parts of Cape Peninsula. Wildfires tend to increase the concentration of CO in the region. This is also an area dominated by fynbos. The results suggests that fynbos species release significant amounts of CO into the atmosphere. Figure 3b, on the other hand, shows low-to-moderate CO column-density values of 0.0199–0.0277 mol/m². This can be regarded as the background CO levels during non-fire periods. Agriculture, agro-processing, vehicle emissions, and domestic fuels for heating and cooking [16], are some of the daily activities that release CO and contribute to the background CO.

Figure 4a shows the BC concentration time-series for the January to December 2023 period. The blue rectangles represent the time when the wildfires occurred and the period with the highest BC emissions. Of interest to this study are peaks that occurred during the December 2023 and the January 2024 periods. BC concentrations of above 3.1 µg/m³ are observed during these periods. This result implies large emissions from large amounts of fuel burnt. This is contrary to the CO column density with large peaks in September 2023 and January 2024 (see Figure 4b). The results imply that the fuel burnt does not release a large amount of CO. The CO released could be mainly from the vegetation, and not from the built-up areas. The Angstrom Exponent (AE) is related to the aerosol particle size. Values less than 1.5 imply an optical dominance of coarse particles, while values greater than 1.5 imply dominance of fine particles. Figure 4c shows the time-series of AE for the period of January 2023 to January 2024. A mixture of fine particles (due to the smoke) and coarse particles (due to black carbon) are observed during the wildfire episodes. AE values of up to 1.8 are observed during the December 2023 period. This result confirms the release of pollutants into the atmosphere from the wildfires. Similar to the AE, the ultraviolet Aerosol Index (UVAI) also determines the presence of dust and smoke. The UVAI indicates the presence of elevated absorbing aerosols in the troposphere. Figure 4d shows a time-series of the UVAI for the period January 2023 to January 2024. The negative UVAI values indicate non-absorbing aerosols such as sulfate, while positive UVAI values indicate absorbing aerosols. The positive UVAI values include biomass-burning emissions from wildfires such as smoke. During December 2023 and January 2024, positive UVAI values are observed, correlating to the wildfires.

The meteorological parameters such as temperature, wind speed, precipitation and relative humidity are important in generating extreme fire behaviour [61,62]. Figure 5 shows the meteorological parameters that influenced the wildfires in the Cape Peninsula region. Figure 5a shows the land-surface-temperature (LST) timeseries for the period of January 2021 to January 2024. High temperatures above 27 °C are observed during the summer season. No unusual temperatures were observed during the December 2023 and January 2024 wildfire period. However, with a lack of precipitation, these temperatures have the capacity to desiccate the vegetation, leaving it vulnerable to fires. Additionally, low relative humidity also helps in drying out vegetation. Again, moderate relative humidity of about 60% does play a role in drying out the vegetation. Higher relative humidity above 80% usually prevents fire from starting and spreading more easily. Figure 5b shows that during the wildfire episodes, the relative humidity was within the range of 60%, contributing slightly to the dry vegetation. Figure 5c shows speeds from December 2023 to January 2024. Wind speeds above 14.5 m/s imply a greater chance of a fire spreading rapidly. These higher speeds were observed during the period of the fires. This is one of the factors that led to the spread of the fire. Figure 5d further shows a 5-day airmass forward trajectory from the source point. All the air mass from 500, 1500 and 2000 m showed that the air parcels travelled to Namibia. This is significant, because it is known that air masses transport pollutants [63,64]. The transport of these pollutants can have a negative impact on health, climate and the environment [65].

The Leaf Water Content Index (LWCI) is used as a representation of water content in vegetation. It is used to estimate changes in the water content in leaves [66]. Negative values of LWCI indicate dry leaves, while positive values indicate higher moisture content in the leaves. Figure 6 shows the LWCI in the Cape Peninsula for the period of 2019 and 2023 during the December periods. In the period 2019, positive values around the fire scar are observed, while negative values are observed in 2023; see Figure 6a,b, respectively. The values shown in Figure 6c clearly shows a decrease in LWCI for the year 2023. The vegetation became drier and more susceptible to the spread of wildfires.

5. Summary and Conclusions

Over the years, the Cape Peninsula has seen a rise in the number of fires that occur seasonally. The region is known to have hot and dry summer conditions which have led to uncontrollable wildfires. Emergency services are always on the alert to intervene when the blazing fires become intense and dangerous. The current wildfire under study is one of the worst of its kind in the region. Therefore, an effort was made to understand the drivers of these fires. Meteorological conditions are definitely a significant driver in the start and spread of the fires. High temperatures, low moisture and strong winds are the main drivers for the fire. It is also clear that fires also destroyed fynbos, which need to burn every 10–15 years to reproduce properly. A four-to-five-year repeated short interval between fires can eliminate dominant reseeding shrubs, and increased fire frequency favours resprouting species. The presented results provide valuable insights into the wildfire regime at the Cape Peninsula. The study can be useful to those in fire management and biodiversity conservation in the region. It is advisable to implement proactive measures such as controlled burns and ecosystem restoration efforts to help mitigate the risk of wildfires and promote the resilience of the local ecosystems.

Monitoring of wildfires is crucial, and remote sensing plays a critical role in observing the wildfire locations and characterising its emissions. Currently low-earth-orbiting (LEO) satellites are used for this purpose. However, one of the challenge is the long revisit time, which impedes the near real-time estimation of wildfire emissions. Furthermore, the passive sensors on these satellites are limited by obstacles such as clouds and thick smoke, i.e., when the imagery of the burnt area is required. The grand challenge is that most of these satellite sensors were not specifically designed for fire monitoring, which poses some limitations on their use. Nonetheless, a multi-sensor approach does give some insight into the wildfire. On the upside, Chen et al. [67] showed that the growing number of medium- and high-resolution satellites, such as Sentinel-2a and Sentinel-2b, Landsat 8 and Landsat 9, and commercial platforms like Planet Labs PlanetScope and DigitalGlobe WorldView, provide detailed imagery for detecting and mapping burnt areas with unprecedented accuracy. Moreover, the launch and operation of Sentinel 5P has be used to assess the impact of wildfires on the air quality and to monitor toxic gases. The monitoring of toxic gases facilitates a quick response to wildfire events. One of the tasks going forward is to develop a localised fire-forecasting system that predicts emissions of wildfires in that area using past fire data and climate variables.