Uses and limits of thermal indices: the case of Sahel

Résumé En

Our main goal here is to analyse extreme heat waves (HWs) in the Sahel (13°N-18°N; 16°W-30°E), using different thermal indices. In the ACASIS project funded bythe French “Agence Nationale de la Recherche”, HWs analyses are characterized forthe first time during the hottest season in the Sahel, using the Global Summary of theDay (GSOD) synoptic observations during April-May-June 1973-2013. Such extremehigh temperatures are usually defined by 3 criteria: 1/ Their low probability ofoccurrence: less than the 10% of the days. The use of absolute thresholds,associated with heat budget and physiological impacts, could be an alternative(Seneviratne et al., 2012). 2/ Their intensity, i.e. their amplitude in terms of standarddeviations from the local climatology (Goubanova, 2007), or otherwise excesses overabsolute intensity thresholds. 3/ Their severity: extreme events causing socioeconomicor human losses, the notion of impact is thus of major importance for theirdefinition (Beniston et al., 2007).According to prior IPCC assessments (TAR, AR4 and SREX), it is very likely thatincreased maximum temperatures and enhanced probabilities of hot days occurrenceswill occur at the global scale. There is also medium confidence that warm spells/heatweave frequencies, lengths or intensities will increase in many regions (Cubasch et al.,2013). Because of its latitude, Sahel is especially concerned, with diurnaltemperatures often exceeding 40°C in boreal spring and, to a lesser extent, autumn.AR4 concluded that global land-surface air temperature (LSAT) had increased over theinstrumental period of record, with the warming rate approximately double thatreported over the oceans since 1979. AR5 confirms previous estimates: Global LSATincreased: 0.1005°C ± 0.01925°C per decade over 1901-2012 period while the risereaches 0.262°C ± 0.05°C per decade over 1979-2012 time period, according to theobservations of CRUTEM4.1.1.0, GHCNv3.2.0, GISS and Berkeley dataset average(Hartman et al., 2013). Changes in the length of the spells are observed and they arelikely to increase but this needs to be investigated further over Sahel where severeheat can have damaging consequences for societies.For instance, in Niamey (Niger) in April 2010 whereas Tmax reaching 47°C (Ringard etal., 2015), increased morbidity and mortality rates were observed, especially amongthe elderly and young children. Moreover, in some places, HWs can be associated withincreased rates of atmospheric pollution.Terminology is large: a “ hot spell” implies the hottest temperature during an extremeof temperature with a small duration (a few hours), small extension (around thestation) and a high frequency; a “warm spell” should be less frequent and covers alarger area; a “heat wave” (HW) exceeds thresholds of temperature most frequentlyand covers at least a whole region (~ 500-1000 km²). These thresholds can beabsolute (T° fixed locally) or relative, exceeding 5 °C as local standards for examplefor Expert Team on Climate Change Detection and Indices (ETCCDI, 2013). ACMAD(African Centre of Meteorological Applications for Development) classifies the extremetemperatures following five high-impact weather events; they broadcasts a mail alertwhen Tmax> 40°C. In ACASIS, Sahelian HW is defined (Rome et al. 2015) as aperiod of at least 3 consecutive days of extremely high (above the local 90thpercentile) daily heat index. The Heat Index (HI) (Steadman 1979, 1994),combining temperature and relative humidity, appears as the most appropriate index,suitable for tropical climate, which takes into account the human-perceived equivalenttemperature. The role of humidity is rather small in boreal spring across the Sahel,when high incoming solar radiation is combined with extremely dry soils, preventingcooling effect associated with latent heat flux. As for global scales, results show aclear warming trend over the last three decades, with a clear trend for HW events tobecome more frequent, last longer, cover larger areas and reach higher intensities(see also Oueslati et al. 2015). We then observe a decrease in cases of “Caution” HIand an increase in cases of “Danger” and “Extreme Danger” HI values.