4.1 Events with the highest rainfall totals: The top 100
Figure 1a shows the location of the top 100 12Z – 12Z 24-h maximum rainfall rate events for the analysis region, while Table 1 provides information for each event. Top 100 events occur infrequently over the northern Sahel within the analysis region, with the vast majority occurring south of 14°N. There are 3 regions where events cluster over the southern Sahel. They are over southwestern Mali (~ 11°N, 8°W; 24 events), central Burkina Faso (~ 13°N, 2°W; 10 events), and north-central Nigeria (~12°N, 8°E; 17 events). Each of these regions has a strong connection with topography. Most events over Mali occur on the eastern slopes of the Mandingue Plateau, over Burkina Faso in close proximity to the North Mossi Plateau, and over north-central Nigeria to the northern slopes of the Jos Plateau. Elevation changes associated with these topographic features are modest, usually only 200 – 400 meters, compared to other regional features such as the Aïr Mountains of northern Niger (18°N, 9°E) and the Cameroon Highlands (6°N, 11°E). However, they are apparently large enough to consistently trigger heavy rainfall amounts. For the 9 most intense events, with rainfall totals greater than 200 mm, 8 occur south of 14°N. The one event north of 14°N occurred southeast of the Aïr Mountains of north-central Niger in the Damergou Gap, and it is the highest maximum total 24-h event identified in this study.
The MCSs associated with the top 100 rainfall events are back-tracked to find their genesis locations, shown in Fig. 1b. 94% originate between 10°N and 20°N. The longitudinal distribution is more uniform, with 29 generated west of 0°E, 35 between 0°E - 10°E, 17 between 10°E - 15°E, and 19 east of 15°E primarily over central and southern Chad and the Darfur Mountains of the Sudan. Thus, the distance between the genesis region and the location of maximum rainfall (Fig. 1a) can be large. Only 25 of the 100 events form locally, defined here as when storm genesis occurs within 100 km of the maximum 24-h rainfall location. Of the other 75 events, 44 form between 100 – 1000 km away from the maximum rainfall location, 24 form 1000 – 2000 km away, and 7 form over 2000 km away.
While topography is known to be associated with MCS generation, less than half of the 100 selected storms original over orographic features, including the Ahaggar Mountains of southern Algeria (3 storms), the Aïr Mountains of Niger (7 storms), and the Marra Mountains in the Darfur region of Sudan (8 storms). The most active storm genesis location is in the Lake Chad depression where 15 of the top 100 events originate (Fig. 1b). Whether and how Lake Chad contributes to storm development is under investigation in a separate study (Zhao et al. 2022).
In regards to the three event cluster regions mentioned earlier, for the 24 Mali event genesis locations, 5 form locally within 100 km of the maximum location, 10 form to the east over Burkina Faso, while the remaining 9 form more remotely at distances greater than 900 km from the Mandingue Plateau of Mali. For the Burkina Faso events over the North Mossi Plateau, 1 event forms locally while 3 originate to the north and east over the Ahaggar and Aïr Mountains. The remaining 6 events have origins east of Burkina Faso, primarily over the Jos Plateau region of Nigeria. For the north-central Nigeria events, 3 form locally, while 3 form east of 20°E. Most of the MCSs associated with extreme rainfall events over this region (11 events) originate in the vicinity of Lake Chad.
Table 1 also provides information on the duration and maximum rainfall intensity of each event. Seven of the longer events, that is, events that exceed 1 standard deviation over the mean, are among the 25 events with highest rainfall, with 4 in the top 10. In contrast, only 2 of the shorter duration events are among the top 25, and none are among the top 10. For peak rainfall intensity, 6 of the highest maximum intensity events are in the top 25, including 3 among the top 10. Conversely, only 2 of the lower peak intensity are in the top 25, and both are long duration events.
Figure 2a shows how rainfall totals are distributed for the top 100 events listed in Table 1. The wettest has a maximum of 291 mm with totals decreasing exponentially, falling below 200 mm by event 10, 175 mm by event 26, and 150 mm by event 78. It is 7.5 times more likely to have a maximum rainfall total between 150 – 200 mm than greater than 200 mm.
Figure 2b displays the seasonality of the top 100 events occur. 83 of the 100 events occur during the peak of the summer monsoon season, with 24 in July and 59 in August. There are only a few events in May, June, September, and October, and none from November to April.
Figure 3 shows the 24-h areal size of the top 100 events for various rainfall total thresholds. The bars are color coded to indicate event rainfall totals at 25 mm intervals from 100 mm – 200 mm. The majority of the top 100 events deliver 24-h rainfall totals exceeding 100 mm over an area of around 4000 km2 or less. The heavy-rain (>100 mm) areas of 12 of the top 100 events are greater than 20,000 km2, with 6 of these events ranked in the top 10 (Table 1). This indicates that the highest 24-h precipitation totals, that is, the events that lie on the steepest part of the distribution curve shown in Fig. 2a, are more likely to have a larger areal coverage.
Composites formed from the top 100 events are compared with climatological conditions to better understand the environmental conditions associated with these extreme cases. Fig. 4a shows the climatological precipitable water and low-level 925 hPa winds at 12Z, while Fig.4b shows the composite anomalies. Climatologically, precipitable water is greater than 45 mm south of 14°N, indicative of the deep moist tropical air that is typically transported into West Africa via the low-level southwesterly monsoon flow during the boreal summer. North of 14°N precipitable water decreases to about 30 mm by 20°N, reflective of the intrusion of drier, Saharan air into the northern Sahel. Most of the extreme events (Fig. 1a) occur where the climatological precipitable water is greater than 40 mm.
Extreme events are associated with a significant and widespread increase in precipitable water over the entire Sahel (Fig. 4b). In the analysis region, precipitable water is greater by 1 – 3 mm during these events, and even more so over the northern Chad Bodélé (> 4 mm) upstream of the analysis region. The increase over Chad is associated with a significant increase in the low-level southwesterly flow into the Bodélé. There is also a significant change in the low-level flow in the analysis region over Burkina Faso. Here the anomalous low-level flow is primarily zonal, indicating a more westerly low-level monsoon flow over Burkina Faso and from Ghana to Nigeria.
Fig. 4c shows the climatological 12Z wind difference between 600 – 925 hPa. There is strong zonal vertical wind shear over the West African Sahel with magnitudes greater than 14 m s-1 extending from the Senegal coast eastward to Chad. The anomalies at 12Z (Fig. 4d) indicate that, outside of a few isolated areas including the Bodélé of Chad and Côte d’Ivoire, pronounced differences in vertical wind shear are not associated with extreme rainfall totals in the composite. This result is independent of the compositing hour selected, as the 12Z anomaly pattern in Fig. 4d is consistent with those at other hours (not shown). Examination of the individual events comprising the composite indicates that the wind shear strength is anomalously weak in 46 events and anomalously strong in 54 events. (see last column of Table 1). If composites are formed individually using only events from one cluster area, namely, the Mali, Burkina Faso, and Nigeria region clusters, the conclusion is the same.
We conclude that extreme rainfall events in the analysis region are associated with a robust and significant increase in the atmospheric moisture content, not only over the region but over the entire Sahel. We do not find a connection to vertical wind shear.
4.2 Events with the most extreme rainfall totals: The top 10
A sub-sample of the top 100 population, namely, the 10 events that deliver the highest rainfall totals see Table 1) are next examined to gain a better understanding about the development of these storms, and extreme storms in general over the West African Sahel. Figure 5 shows the distribution of 24-h accumulated precipitation greater than 50 mm for each of these events. The red circles denote the location of the maximum rainfall for each case. There is no obvious similarity among the events, and they range from localized events (CASE01 and CASE08) to widespread events (CASE02 and CASE06).
Figure 5 also speaks to the potential impact of each event, especially in terms of flooding potential due to the total amount of accumulated precipitation over the 24-h period. For example, CASE01 which is associated with the highest maximum 24-h rainfall total (Table 1), has a limited are of heavy rain over central Niger with a more widespread area of 50 – 100 mm of rainfall extending from central Niger into northern Nigeria. Thus, it could be inferred from Fig. 5a that flash flooding for this case would be more likely localized over central Niger. In contrast, CASE02 and CASE07 are associated with widespread 24-h rainfall totals exceeding 150 mm over the Niger River basin in southern Mali that likely not only impact communities and cities including Bamako immediately after the event, but also the downstream Niger River communities in the days that follow.
FEWSNET reports, CRED EM-DAT disaster reports, OCHA Reliefweb, and DREF bulletin reports along with any available extreme rainfall studies are examined to attempt to confirm via ground-truth the impacts of the top 10 events. Results are summarized in Table 2, and indicate that 5 of the events (CASE01, CASE03, CASE04, CASE06, and CASE08) are directly associated with impactful flooding. For CASE02, CASE05, CASE07, and CASE09, available reports are vague on the exact timing of the flooding, and/or reports suggest that multiple events over many days led to the flooding. For CASE10, no ground-truth evidence is found, so it is unclear how impactful it is. Overall, the ground-truth results are encouraging, and give us confidence that we have identified impactful rainfall events.
Fig. 6 shows the diurnal time series of rainfall for the maximum 24-h total precipitation location (the red circles in Fig. 5) for the top 10 events. The gray lines in each panel denote the time of the maximum rainfall rate intensity for each event. Event types range from relatively shorter duration/higher peak rainfall intensity cases such as CASE01, CASE04, CASE08, and CASE10, to longer duration/weaker peak rainfall intensity cases such as CASE02, CASE03, CASE06, and CASE09. Longer duration/weaker peak intensity events occur primarily after 00Z, while the shorter duration/higher peak intensity events tend to occur in the late afternoon and evening.
To put into context how representative the above findings are for the larger population, all 100 events are also categorized by diurnal cycle. We calculate the mean and standard deviation for event duration and event peak intensity (Table 1), then select cases in which both the duration and peak intensity fall outside the ± 0.5 standard deviation level. This yields a list of events for the two diurnal types as well as the opposite categories, namely the longer duration/higher peak intensity and shorter duration/ weaker peak intensity events (Table 3). We find that there are 14 longer duration/weaker peak intensity and 14 shorter duration/ higher peak intensity events out of the top 100 cases. These types are approximately 3 times more likely to occur than the longer duration/higher peak intensity and shorter duration/weaker peak intensity types. Furthermore, the selected longer duration/weaker peak intensity events yield a larger 100 mm rain shield area (Fig. 3) and 7 are associated with an area greater than 10,000 km2 compared to 1 event at this size for the shorter duration/higher peak intensity type.
Fig. 7a shows the diurnal cycles of rainfall for the 14 longer duration/weaker peak intensity events (gray lines) listed in Table 3 along with the composite means (red line), while Fig. 7b shows their locations. While rainfall occurs prior to 00Z in some events, there is a clear preference for peak rainfall rates between 00Z to 03Z. 11 events (78%) occur over Burkina Faso and Mali and the remaining 3 events occur north of the Jos Plateau of Nigeria.
Fig. 7c shows the diurnal cycle of rainfall for the 14 shorter duration/higher peak intensity cases. The time of peak rainfall intensity is more variable for this type, so the composite mean is relatively constant over the diurnal cycle. Event locations (Fig. 7d) are evenly distributed longitudinally, with 6 cases east of 2°E and 8 cases west of 2°E. There is no clear association between the timing of peak rainfall intensity and the location for these events.
In summary, this analysis confirms that the top 10 events are representative of the top 100 population. This gives us confidence that the results focused on the top 10 events are applicable not only to the individual storms being analyzed, but also to a more robust population of storms.
Figure 8 shows the trajectories of the convective cores for the 10 events prior to (green) and following (blue) the time of peak rainfall intensity. As mentioned in section 3 the convective core is defined using a relatively high rainfall rate threshold (100 mm day-1). Rain shields of this magnitude vary in duration among the cases, so some trajectories in Fig. 8 have more points than other events. For example, CASE03 (Fig. 8c), CASE06 (Fig. 8f), and CASE09 (Fig. 8i) have relatively few tracking points and they tend to be spaced closer together indicating a shorter lived and slower moving convective core for these longer duration/ weaker peak intensity events. For CASE01 (Fig. 8a), CASE04 (Fig. 8d), and CASE10 (Fig. 8j) there are more tracking points covering a larger area, indicating the strong convective core is longer lasting and, in most cases, faster moving for the shorter duration/ higher peak intensity events. However, not all events conform to this characterization. For example, CASE02 is a long duration/weak peak intensity type, but the convective core is relatively long lasting and covers a large area. Likewise, CASE08 is a short duration/high peak intensity type, but the convective core is shorter lived as the storm is more limited in size compared to other top 10 events (Fig. 3).
Fig. 8 indicates that some storms change direction when the storm is in close proximity to the location of maximum rainfall rate (red circle). Examples include CASE01 (southwesterly to a south-southwesterly track shift), CASE05 (west-southwesterly to a southwesterly track shift), and CASE10 (west-southwesterly to a south-southwesterly track shift). The physical processes responsible for directional change could be influential in producing heavy rainfall totals for some events, for example, through a reorganization in the convection in response to environmental condition changes. Future work can explore this possibility and the implications for storm intensification.
One way to understand how a storm reorganizes over time is to examine how the physical size of the storm and the size of the convective core change. Fig. 9 shows the timing of the storm rain shield area development for each case. Two areas are shown. One, denoted by the black line, is the size of the rain shield defined by the 5 mm day-1 rainfall rate threshold. This represents an estimate of the area of the entire MCS that includes the convective and stratiform components of the system. The other area, indicated by the blue line, is an estimate of the intense convective core. It is defined by the 500 mm day-1 rainfall rate threshold. The gray line denotes the time of peak rainfall rate at the storm’s most intense location.
Fig. 9 indicates that events range in size from generally less than 200,000 km2 (CASE03, CASE08, and CASE09), to mid-size between 200,000 – 400,000 km2 (CASE01, CASE04, CASE05, and CASE10), to greater than 400,000 km2 (CASE02, CASE06, and CASE07). No apparent relationship is found between the areal size of the MCS rain shield and the longer duration/weaker peak intensity events as two cases are relatively large in size (CASE02 and CASE06), and two events are small in size (CASE03 and CASE09). In contrast, 3 of the 4 shorter duration/higher peak intensity events are mid-sized (CASE01, CASE04, and CASE10), while the fourth case, CASE07, eventually exceeds 400,000 km2.
The convective cores range in size from less than 10,000 km2 for CASE01, CASE03, CASE08, CASE09, and CASE10, to over 20,000 km2 for CASE02, CASE04, CASE05, and CASE06. When peak rainfall intensity occurs, there is a relative peak in the size of the convective core for most cases. This indicates that the convective core grows in size around the time of peak intensity. In most instances, when the intense convective core size is growing, the MCS temporarily contracts in size. In some cases, such as CASE01 (Fig. 10a), CASE02 (Fig. 10b), and CASE06 (Fig. 10f), this happens multiple times over the storm’s lifespan, signifying a potential cycle of convective core growth and decay that is generally opposite that of the larger MCS system. Furthermore, the most intense rainfall rate does not necessarily correspond to when the convective core size is largest.
4.3 Environmental conditions associated with the top 10 heavy rainfall events
Figure 10a shows ERA5 700 hPa geopotential height, 700 hPa wind, and precipitable water anomalies, while Fig. 10b shows the anomalous vertical wind shear for CASE01 for a representative time prior to the peak rainfall intensity. A strong mid-level disturbance was positioned east of the maximum rainfall location along the Niger/Chad border prior to the CASE01 rainfall maximum. Anomalous cyclonic flow enhanced atmospheric moisture loading over the central Sahel in support of the high rainfall totals over eastern Niger. There is a concentrated core of anomalously high meridional vertical shear, exceeding than 12 m s-1 extending northward to northern Niger, that is associated with the flow around the disturbance. South of 15°N there is a broad region of anomalously weak zonal shear associated with a weakening of the mid-level easterly flow.
CASE04 (Figs. 10c and d) and CASE08 (Fig. 10e and f) have different synoptic conditions and anomalous vertical wind shear patterns. Both occur over north-central Nigeria with mid-level anomalous circulation that is weaker than for CASE01, with a disturbance to the east near 15°N and 20°E. The environment is anomalously wet in the vicinity and upstream of the disturbance, with precipitable water anomalies 5 – 10 mm greater than normal. This loading of atmospheric moisture content is much less than the 20 – 25 mm observed for CASE01. Vertical wind shear is only modestly stronger for these cases, generally less than 3 m s-1, and confined to a narrow region along the Niger/Nigeria border. More influential for these two cases is the close proximity of the Sahel/tropical Africa dryline boundary denoted by the red-dashed line. Mid-level flow is anticyclonic over Niger and relatively drier Sahelian air extends equatorward into southern Niger, setting up a scenario with enhanced low-level convergence south of the dryline frontal boundary over northern Nigeria. The importance of the dryline for the development of convection in this region is consistent with results from other studies (Vizy and Cook 2018, 2019a).
CASE10 (Fig. 10g and h) is associated with two mid-level disturbances near southwestern Mali. The first is centered at 4°W and 11°N. It lies in the southern African wave storm track, with cyclonic flow that enhances the atmospheric moisture transport over southern and western Burkina Faso. The second disturbance is stronger and located west of the first disturbance at 6°W and 20°N in the northern African wave storm track. Cyclonic flow associated with this disturbance promotes increased equatorward flow of dry western Sahel air ahead of the trailing disturbance, shifting the dryline frontal boundary southward closer to 15°N while enhancing the vertical zonal wind shear westward to the coast by up to 10 m s-1. Combined, these two disturbances enhance the low to mid-level convergence over southwestern Mali.
Hovmöller plots of the 700 hPa relative vorticity and precipitable water anomalies averaged between 10°N - 15°N for the 4 cases are shown in Fig. 11. The circles in each panel denote the longitudinal position and time of the peak rainfall intensity at the maximum rainfall site. Propagation speeds estimated from the 700 hPa relative vorticity plots for relevant disturbances are noted.
CASE01 (Fig. 11a) is associated with a relatively slow-moving disturbance (5.5 m s-1) that is trailing a faster moving disturbance (9.5 m s-1). While the anomalous precipitable water (Fig. 11b) is primarily high in the 10-day window shown, the first disturbance is influential in further increases in the atmospheric moisture.
CASE04 (Fig. 11c) is also associated with two disturbances, but there is more time between the passage of these systems and both disturbances are propagating faster compared to CASE01. The maximum rainfall site occurs west of the second disturbance when the anticyclonic flow is transitioning to cyclonic flow and the atmospheric moisture content begins to increase (Fig. 11d).
CASE08 (Fig. 11e) is associated with a disturbance propagating at 7.1 m s-1. While the peak rainfall intensity occurs when the circulation is only starting to become cyclonic, the environment is relatively moist even when the flow is slightly anticyclonic between 18 – 21 July 2001 (Fig. 11f).
Finally, CASE14 is associated with a weak disturbance that originates on 17 July 2004 around 15°E, and propagates at 9.8 m s-1 (Fig. 11g). Around the time of maximum rainfall intensity, the propagation speed slows to 3.7 m s-1 where the disturbance eventually merges into the second disturbance that forms a couple days later. This slowdown in the first disturbance occurs when the system encounters drier (Fig. 11h) anticyclonic flow over far western West Africa. It is not until the stronger, second disturbance reaches the area that the storm sheds the West African coast into the eastern North Atlantic.
Figure 12 shows ERA5 700 hPa geopotential height, wind, precipitable water, and vertical wind shear anomalies for select times of the 4 longer duration/weaker peak intensity rainfall events. CASE02 (Fig. 12a and b), CASE03 (Fig. 12c and d), and CASE06 (Fig. 12e and f) are associated with strong disturbances in the southern AEW storm track, while the disturbance for CASE09 (Fig. 12g and h) is weaker than the other three events. All three events are associated with positive precipitable water anomalies in the vicinity of the disturbance. CASE03 and CASE06 are associated with anomalous westerly/west-southwesterly flow, and CASE02 and CASE09 are associated with anomalous northwesterly flow and drier air ahead of the disturbance similar to CASE10 (Fig. 10g). The Sahel/tropical Africa dryline boundary is influential only for CASE09 due to its location farther north.
CASE02 (Fig. 12b) is associated with enhanced meridional vertical wind shear greater than 10 m s-1 ahead of the disturbance, while CASE03 (Fig. 12d) is associated with a strong negative anomaly in the vertical zonal wind shear. The locations of CASE06 (Fig. 12f) and CASE09 (Fig. 12h) are located at the boundary between anomalously strong meridional vertical shear to the west and anomalously weak zonal vertical shear to the east. The former is associated with coastal ridging that is weakening as it leaves the West African coast, while the latter is associated with changes in the low- to mid-level flow along the southern flank of the approaching disturbance
Figure 13 shows Hovmöller plots of the 700 hPa relative vorticity and precipitable water anomalies for the longer duration/weaker peak intensity events. CASE02 (Fig. 13a) is associated with a disturbance that slows over western West Africa, followed by a stronger disturbance propagating at 9 m s-1 a couple of days later. The anticyclonic flow is relatively dry (Fig. 13b) until the second system approaches and wrap around moisture builds over the region. CASE03 occurs in a particularly active period with multiple disturbances propagating across the Sahel (Fig. 13c) associated with an anomalously moist environment west of 10°E over the 10-day period shown (Fig. 13d). The disturbance most closely associated with the event originates around 15°E and propagates at a speed of 7.4 m s-1 to about 4°W when it slows to 4.6 m s-1. The peak intensity of rainfall coincides with this slowdown. CASE06 is associated with two relatively strong disturbances (Fig. 13e) with the first moving faster than the second, helping to build atmospheric moisture over West Africa prior to the second system (Fig. 13f). The event is closely associated with the second disturbance which is relatively slow moving (5.5 m s-1). CASE09 (Fig. 13g) is associated with a short-lived disturbance that is moving westward into an environment that is primarily anticyclonic and anomalously dry (Fig. 13h). The environmental conditions associated with this case are similar to those for CASE10.
Finally, Fig. 14 shows results for CASE05 and CASE07. These two cases are associated with particularly strong disturbances with 700 hPa geopotential height anomalies less than -40 m, precipitable water anomalies greater than 15 mm, and strong vertical wind shear prior to the peak intensity (Figs. 14a - d). Ahead of the disturbances the environment is drier than normal, particularly for CASE05. The Sahel/tropical Africa dryline boundary is in close proximity to the maximum rainfall location for CASE05, but not CASE07. Figs. 14e – h show the 700 hPa relative vorticity and anomalous precipitable water Hovmöller plots for CASE05 and CASE07. Both are associated with fairly strong disturbances propagating westward into an anomalously dry environment, with the anticyclonic flow preceding the disturbance being stronger for CASE05 than CASE07.
Overall, a variety of factors are associated with the development of extreme rainfall events over the region. Based on the analysis of the top 10 events, three atmospheric conditions are identified here as being particularly important. They are the importance of moisture loading of the atmosphere prior to the event, interaction of the developing storm in the wake of a region of anticyclonic flow, and interaction of the storm in the wake of a region of anticyclonic flow and the Sahel/tropical Africa dryline boundary.