In April 2021, a devastating wildfire tore through the iconic Table Mountain area of Cape Town, South Africa.1 Following a human-induced ignition on the morning of 18 April, worsening weather conditions led to increased fire spread that lasted until the afternoon of 20 April when the fire was eventually extinguished. The fire burned across more than 600 ha of wildland,2 with its incursion into urban areas resulting in widespread evacuations and several hospitalizations.3 Up to ZAR 1 billion (approximately USD 60 million) worth of damage to buildings and infrastructure was incurred by the University of Cape Town campus alone,3 and irreplaceable collections in its Jagger Library were destroyed. While summer wildfires are common in the Cape Town area, the rapid spread, spotting behavior, and unprecedented impacts of this fire so late in the fire season, which is usually considered to run from mid-November to mid-April (Forsyth and Bridgett 2004; Christ et al. 2022), raise important questions about the challenges in responding to changing fire regimes at the wildland–urban interface.
The first three weeks of April 2021 were abnormally warm and dry along South Africa’s west coast, at the southern tip of which Cape Town is situated. These conditions were highly conducive to wildfire ignition and spread. Previous work has demonstrated a link between extreme hydroclimatic events in the surroundings of Cape Town and anthropogenic climate change, most notably in an attribution study of the 2015–17 drought (Otto et al. 2018a). While such droughts are likely to enhance fire risks, a quantification of how climate change has altered the likelihood of extreme weather conducive to late-season fires is worthy of dedicated analysis. Here, we analyze the exceptional nature of the meteorological conditions that coincided with the April 2021 event. Using an established probabilistic methodology applied to fire weather extremes simulated by multiple large ensembles from the latest generation of climate models, we quantify the influence of rising global temperatures on the likelihood of such conditions.
Data and methods
First, to place the April 2021 event in the context of the regional fire regime, location and intensity data on historical fires (2001–21) are taken from the Moderate Resolution Imaging Spectroradiometer (MODIS; Giglio et al. 2016) via the Fire Information for Resource Management System (FIRMS). Our analysis of fire-conducive meteorology is based on the Canadian Fire Weather Index (FWI; Van Wagner 1987), which combines temperature, surface wind speed, relative humidity, and precipitation. FWI has been widely used in related fire analysis across the world (e.g., Krikken et al. 2021; Liu et al. 2022a,b) and forms the basis of GEFF-ERA5, the fire danger reanalysis based on the Global ECMWF Fire Forecast model and the ERA5 reanalysis (Vitolo et al. 2020), from which we derive historical FWI data for the period 1979–2021. The FWI value of 67.77 on 18 April 2021 is the highest recorded during autumn (March–May) in GEFF-ERA5. Our attribution analysis questions to what extent rising global temperature associated with anthropogenic climate change has altered the likelihood of a “2021-type event,” defined by the exceedance of the 18 April 2021 threshold by yearly maxima in autumn FWI. It is widely accepted that global mean temperature change since the late nineteenth century has been predominantly driven by anthropogenic forcings, with the influence of natural forcings very small by comparison (Hegerl et al. 2010; Bindoff et al. 2013; Philip et al. 2020; Ara Begum et al. 2022). Recent work has revealed positive trends in observed fire weather extremes (Jain et al. 2022) and fire weather maxima (Liu et al. 2022a) across much of southern Africa, although the extent of the observational record limits each analysis to just a few decades. Here, simulations of historical FWI are derived from six large ensembles (at least 10 members) from phase 6 of the Coupled Model Intercomparison Project (CMIP6; Eyring et al. 2016) for the period 1850–2014 (see supplemental material for details). As the extent of the April 2021 fire was relatively small, model output is taken for a single grid point closest to the fire’s approximate origin (33.92°S, 18.42°E). The meteorological and climatic diversity of the wider region (Conradie et al. 2022) means that including model output across a larger area is very likely to conflate spatially heterogeneous change signals not relevant to the event in question.
We apply a probabilistic statistical methodology based on a time-dependent generalized extreme value (GEV) distribution to each of the six CMIP6 model ensembles to quantify changes in the likelihood of extreme fire weather to rising global temperatures. This method has been widely used in the attribution of different extreme events (e.g., Schaller et al. 2014; Eden et al. 2016; van der Wiel et al. 2017; Eden et al. 2018; Otto et al. 2018b), including episodes of extreme fire weather (e.g., Krikken et al. 2021; Liu et al. 2022b). For each model, 165 yearly FWI maxima (1850–2014) across all corresponding ensemble members are fitted to a GEV distribution scaled with the 4-yr smoothed global mean surface temperature (GMST), under the assumption that the location parameter µ and the scale parameter σ have the same exponential dependency on GMST, while the “dispersion ratio” σ/µ and the shape parameter ξ remain constant (Philip et al. 2020; van Oldenborgh et al. 2021a).
We evaluate the FWI threshold associated with the April 2021 event for each CMIP6 model following a bias correction based on the ratio between the µ parameters of the stationary GEV fit and that fitted with FWI maxima from GEFF-ERA5. We then estimate the probability of this threshold being exceeded, first, in a “past” climate of 1880 (p0) and, second, in a “present” climate of 2021 (p1), both of which are defined by observed GMST (GISTEMP Team 2022; Lenssen et al. 2019). The probability ratio (PR) p1/p0 is used to express the overall change in likelihood. A 1,000-sample nonparametric bootstrap is used to estimate confidence intervals (CIs) for each model. Following a model evaluation and selection step based on the dispersion ratio of each model’s GEV fit, a final PR result is obtained by a multimodel weighted average (e.g., Eden et al. 2016; Philip et al. 2018).
Results
Between 2001 and 2021, fires frequently occurred across the Cape Floristic Region along South Africa’s southern and southwestern coastal margins. Fires during March–May occurred predominantly in the west of this region (Fig. 1a) and regularly exceeded a fire radiative power (FRP) of 900 MW (Fig. 1b). The majority of fires observed within 50 km of Cape Town occurred between December and March; far fewer fires are observed later than mid-March (Fig. 1c). Synoptic conditions during the week leading up to the 18 April 2021 were characterized by a quasi-stationary midtropospheric ridge over South Africa and dry, downslope easterly or northerly drainage winds along the west coast, known locally as “berg winds” (Fig. 1d), which contributed to the exceptional meteorological conditions. The approximate time of the fire’s spread coincided with temperatures over 33°C and very low relative humidity (Figs. 1e,f), in addition to the emergence of strong northwesterly winds (Fig. 1g). While, during the 2020–21 summer months, the FWI was generally above average, the absence of prolonged periods of extreme conditions and isolated daily FWI values as anomalous as that recorded on 18 April 2021 further illustrates the exceptionality of the event (Fig. 1h). FWI anomalies from the MAM climatology on 18 April 2021 were very positive (>40) along the west and south coasts, yielding FWI values around Cape Town usually seen in the arid western interior (Fig. 1i).
(a) Location and (b) intensity (FRP) of FIRMS-detected fires (2001–21). (c) Intra-annual timing and FRP of FIRMS-detected fires within the Western Cape province. Fires within 50 km of Cape Town are shown in red. (d) ERA5 mean 500-hPa geopotential height (contours) and surface winds (arrows) for 11–17 Apr 2021. (e) Temperature (°C), (f) relative humidity (%), and (g) wind speed (m s−1) and direction observed between 11 and 19 Apr 2021 at Cape Town WO. (h) Cape Town FWI between July 2020 and June 2021 from GEFF-ERA5 (line) and 1979–2021 monthly climatological quantiles (bars). (i) GEFF-ERA5 FWI anomalies on 18 Apr 2021 with respect to the 1979–2021 March–May climatology. The Western Cape province is shaded in (a), (b), and (d) and outlined in (h).
Citation: Bulletin of the American Meteorological Society 104, 1; 10.1175/BAMS-D-22-0204.1
An overall increase in the likelihood of a 2021-type event between 1880 and 2021 was found for all six CMIP6 models, with PR ranging from 1.2 (INM-CM5-0) to 4.1 (MPI-ESM1-2-HR) (Figs. 2a–f). The uncertainty ranges vary between models, and statistical significance is found only in CanESM5 (95% CI: 1.3–5.6; Fig. 2a) and MPI-ESM1-2-HR (95% CI: 1.6–29.5; Fig. 2f). These results complement the positive trends in observed extreme fire weather revealed in recent work (Jain et al. 2022; Liu et al. 2022a). In view of the intermodel differences, it is notable that the highest resolution model, MPI-ESM1-2-HR, is associated with the strongest trend, but it is unclear whether results are sensitive to model resolution.
(a)–(f) Gumbel plots for the six CMIP6 models, showing the GEV model fit scaled to the smoothed observed GMST (GISTEMP Team 2022; Lenssen et al. 2019) of 1880 (blue) and 2021 (red). Shaded areas represent the 95% CIs following nonparametric bootstrapping. The magenta lines represent the 2021-type event, scaled to the model distribution using bias correction. The blue (red) bars represent the 95% CIs for the return period of a 2021-type event in the climate of 1880 (2021). (g) PR estimates for the six CMIP6 models and the weighted average (for which CNRM-ESM2-1 is excluded). Bars show 95% CIs; central values are shown in bold.
Citation: Bulletin of the American Meteorological Society 104, 1; 10.1175/BAMS-D-22-0204.1
The small spatial extent of the April 2021 event, and the subsequent application of the method to a single model grid cell, results in a relatively large influence of internally driven natural variability on PR uncertainty (Kay et al. 2015). Combining results as part of a multimodel synthesis is a useful way to summarize and communicate overall findings when internal variability is large. Here, the synthesis is limited to those models that realistically represent FWI extremes, defined by the dispersion ratio of the GEV fit (see supplemental material). A weighted average is generated for the five models that meet the selection criteria, with weights for each model’s PR given by the inverse of the squared uncertainty. The uncertainty of the weighted average is approximated by adding the errors for each PR estimate in quadrature (e.g., Philip et al. 2018). The multimodel synthesis result suggests that the weighted average of the likelihood of the 2021-type event increased by a factor of 1.9 (95% CI: 1.2–3.1; Fig. 2h) between 1880 and 2021 as a result of rising global temperatures.
Conclusions
Our analysis aimed to quantify the impact of a changing climate on the extreme fire weather that coincided with the Cape Town wildfire on 18 April 2021. We applied an established statistical method to the outputs of six large ensembles from CMIP6 to estimate how the likelihood of the 2021-type conditions has been altered by anthropogenic climate change, here expressed as the change in global mean temperature since the late nineteenth century. Averaging the results from multiple models revealed a mean probability ratio of 1.9, i.e., an overall increase in likelihood of around 90%. Diagnosing discrepancies among different models of differing resolutions, particularly when the analysis is limited to a single model grid point, is challenging and a potential avenue for further study.
The results complement existing efforts to attribute hydroclimatological extremes around Cape Town, including droughts (e.g., Otto et al. 2018a; Zscheischler and Lehner 2022), and add to the growing set of attribution studies on wildfires and extreme fire weather in different parts of the world (e.g., Krikken et al. 2021; van Oldenborgh et al. 2021b; Liu et al. 2022b). Our analysis also highlights the importance of drawing findings from multiple models in pursuit of the most robust statement possible for a singular wildfire episode. The model-derived evidence of trends in fire weather extremes add to that drawn from observational analysis (Jain et al. 2022; Liu et al. 2022a), and the application of alternative modeling approaches and statistical methodologies is a potential pathway toward further building this evidence base (Otto et al. 2020).
Acknowledgments.
We acknowledge the use of data and/or imagery from NASA’s Fire Information for Resource Management System (FIRMS) (https://earthdata.nasa.gov/firms), part of NASA’s Earth Observing System Data and Information System (EOSDIS). Weather data from the Cape Town WO station were obtained from South African Weather Service SYNOP data.
References
Ara Begum, R., and Coauthors, 2022: Point of departure and key concepts. Climate Change 2022: Impacts, Adaptation, and Vulnerability, H.-O. Pörtner et al., Eds., Cambridge University Press, 121–196.
Bindoff, N. L., and Coauthors, 2013: Detection and attribution of climate change: From global to regional. Climate Change 2013: The Physical Science Basis, T. F. Stocker et al., Eds., Cambridge University Press, 867–952.
Christ, S., N. Schwarz, and R. Sliuzas, 2022: Wildland urban interface of the City of Cape Town 1990–2019. Geogr. Res., 60, 395–413, https://doi.org/10.1111/1745-5871.12535.
Conradie, W. S., P. Wolski, and B. C. Hewitson, 2022: Spatial heterogeneity in rain-bearing winds, seasonality and rainfall variability in southern Africa’s winter rainfall zone. Adv. Stat. Climatol. Meteor. Oceanogr., 8, 31–62, https://doi.org/10.5194/ascmo-8-31-2022.
Eden, J. M., K. Wolter, F. E. Otto, and G. J. van Oldenborgh, 2016: Multi-method attribution analysis of extreme precipitation in Boulder, Colorado. Environ. Res. Lett., 11, 124009, https://doi.org/10.1088/1748-9326/11/12/124009.
Eden, J. M., S. F. Kew, O. Bellprat, G. Lenderink, I. Manola, H. Omrani, and G. J. van Oldenborgh, 2018: Extreme precipitation in the Netherlands: An event attribution case study. Wea. Climate Extremes, 21, 90–101, https://doi.org/10.1016/j.wace.2018.07.003.
Eyring, V., S. Bony, G. A. Meehl, C. A. Senior, B. Stevens, R. J. Stouffer, and K. E. Taylor, 2016: Overview of the Coupled Model Intercomparison Project Phase 6 (CMIP6) experimental design and organization. Geosci. Model Dev., 9, 1937–1958, https://doi.org/10.5194/gmd-9-1937-2016.
Forsyth, G., and J. Bridgett, 2004: Table Mountain National Park Fire Management Plan. 85 pp., www.sanparks.org/docs/parks_table_mountain/library/fire_management.pdf.
Giglio, L., W. Schroeder, and C. O. Justice, 2016: The collection 6 MODIS active fire detection algorithm and fire products. Remote Sens. Environ., 178, 31–41, https://doi.org/10.1016/j.rse.2016.02.054.
GISTEMP Team, 2022: GISS Surface Temperature Analysis (GISTEMP), version 4. NASA Goddard Institute for Space Studies, accessed 10 August 2022, https://data.giss.nasa.gov/gistemp/.
Hegerl, G. C., O. Hoegh-Guldberg, G. Casassa, M. P. Hoerling, R. Kovats, C. Parmesan, D. W. Pierce, and P. A. Stott, 2010: Good practice guidance paper on detection and attribution related to anthropogenic climate change. Meeting Report of the Intergovernmental Panel on Climate Change Expert Meeting on Detection and Attribution of Anthropogenic Climate Change. T. F. Stocker et al., Eds., University of Bern, 8 pp., www.ipcc.ch/publication/ipcc-expert-meeting-on-detection-and-attribution-related-to-anthropogenic-climate-change/.
Jain, P., D. Castellanos-Acuna, S. C. P. Coogan, J. T. Abatzoglou, and M. D. Flannigan, 2022: Observed increases in extreme fire weather driven by atmospheric humidity and temperature. Nat. Climate Change, 12, 63–70, https://doi.org/10.1038/s41558-021-01224-1.
Kay, J. E., and Coauthors, 2015: The Community Earth System Model (CESM) large ensemble project: A community resource for studying climate change in the presence of internal climate variability. Bull. Amer. Meteor. Soc., 96, 1333–1349, https://doi.org/10.1175/BAMS-D-13-00255.1.
Krikken, F., F. Lehner, K. Haustein, I. Drobyshev, and G. J. van Oldenborgh, 2021: Attribution of the role of climate change in the forest fires in Sweden 2018. Nat. Hazards Earth Syst. Sci., 21, 2169–2179, https://doi.org/10.5194/nhess-21-2169-2021.
Lenssen, N., G. Schmidt, J. Hansen, M. Menne, A. Persin, R. Ruedy, and D. Zyss, 2019: Improvements in the GISTEMP uncertainty model. J. Geophys. Res. Atmos., 124, 6307–6326, https://doi.org/10.1029/2018JD029522.
Liu, Z., J. M. Eden, B. Dieppois, and M. Blackett, 2022a: A global view of observed changes in fire weather extremes: Uncertainties and attribution to climate change. Climatic Change, 173, 14, https://doi.org/10.1007/s10584-022-03409-9.
Liu, Z., J. M. Eden, B. Dieppois, I. Drobyshev, C. Gallo, and M. Blackett, 2022b: Were meteorological conditions related to the 2020 Siberia wildfires made more likely by anthropogenic climate change? [in “Explaining Extreme Events in 2020 from a Climate Perspective”] Bull. Amer. Meteor. Soc., 103 (3), S44–S49, https://doi.org/10.1175/BAMS-D-21-0168.1.
Otto, F. E. L., and Coauthors, 2018a: Anthropogenic influence on the drivers of the Western Cape drought 2015–2017. Environ. Res. Lett., 13, 124010, https://doi.org/10.1088/1748-9326/aae9f9.
Otto, F. E. L., K. van der Wiel, G. J. van Oldenborgh, S. Philip, S. F. Kew, P. Uhe, and H. Cullen, 2018b: Climate change increases the probability of heavy rains in Northern England/Southern Scotland like those of storm Desmond—A real-time event attribution revisited. Environ. Res. Lett., 13, 024006, https://doi.org/10.1088/1748-9326/aa9663.
Otto, F. E. L., and Coauthors, 2020: Toward an inventory of the impacts of human-induced climate change. Bull. Amer. Meteor. Soc., 101, E1972–E1979, https://doi.org/10.1175/BAMS-D-20-0027.1.
Philip, S., and Coauthors, 2018: Validation of a rapid attribution of the May/June 2016 flood-inducing precipitation in France to climate change. J. Hydrometeor., 19, 1881–1898, https://doi.org/10.1175/JHM-D-18-0074.1.
Philip, S., and Coauthors, 2020: A protocol for probabilistic extreme event attribution analyses. Adv. Stat. Climatol. Meteor. Oceanogr., 6, 177–203, https://doi.org/10.5194/ascmo-6-177-2020.
Schaller, N., F. E. L. Otto, G. J. van Oldenborgh, N. R. Massey, S. Sparrow, and M. R. Allen, 2014: The heavy precipitation event of May–June 2013 in the upper Danube and Elbe basins [in “Explaining Extreme Events of 2013 from a Climate Perspective”]. Bull. Amer. Meteor. Soc., 95 (9), S69–S72, https://doi.org/10.1175/1520-0477-95.9.S1.1.
van der Wiel, K., and Coauthors, 2017: Rapid attribution of the August 2016 flood-inducing extreme precipitation in south Louisiana to climate change. Hydrol. Earth Syst. Sci., 21, 897–921, https://doi.org/10.5194/hess-21-897-2017.
van Oldenborgh, G. J., and Coauthors, 2021a: Pathways and pitfalls in extreme event attribution. Climatic Change, 166, 13, https://doi.org/10.1007/s10584-021-03071-7.
van Oldenborgh, G. J., and Coauthors, 2021b: Attribution of the Australian bushfire risk to anthropogenic climate change. Nat. Hazards Earth Syst. Sci., 21, 941–960, https://doi.org/10.5194/nhess-21-941-2021.
Van Wagner, C. E., 1987: Development and Structure of the Canadian Forest Fire Weather Index System. Forestry Tech. Rep. 35, Canadian Forestry Service Headquarters, 37 pp., https://cfs.nrcan.gc.ca/publications?id=19927.
Vitolo, C., F. Di Giuseppe, C. Barnard, R. Coughlan, J. San-Miguel-Ayanz, G. Libertá, and B. Krzeminski, 2020: ERA5-based global meteorological wildfire danger maps. Sci. Data, 7, 216, https://doi.org/10.1038/s41597-020-0554-z.
Zscheischler, J., and F. Lehner, 2022: Attributing compound events to anthropogenic climate change. Bull. Amer. Meteor. Soc., 103, E936–E953, https://doi.org/10.1175/BAMS-D-21-0116.1.
