Diagnosis of the Causes of the Rain Flooding in June in the West Africa Coastal Area

Rain flooding during June on the West Africa coastal area is analyzed by using the 95th and 75th percentiles, which represent extreme and intense rainfall events respectively. Thus, the contribution of these events that reaches around 50% shows their impact on the rainfall in June. Atmospheric and oceanic factors influence the rain flooding. Indeed, the extreme events are associated with easterly waves propagating from 20˚E, while those of intense events are initiated around 5˚E. The impact of oceanic conditions exhibits the warming of the equatorial rail and the Atlantic cold tongue, the warming of the whole ocean basin and a north-south dipole of SST anomalies. The West African monsoon that reaches Abidjan corresponds to a low-level atmospheric flow, whose upward motion extends in latitude from the ocean to the continent. An increase of disturbance contributes to enhancing these events. This is confirmed by the inflow on to the continent of oceanic moisture coming from the ventilation by evaporation of warm water. In addition, the coupled ocean- atmosphere simulations are one of the best candidates that could help to better explain these dramatic events. This study is useful because of showing solutions that could help in adoption of policies for the risks management related to these events.

Abidjan (4.013˚W; 5.3˚N) is the economic capital of Côte d'Ivoire. It is the country's largest metropolis with an area of 2119 km 2 , 10 municipalities and 4.395 million inhabitants. This metropolis concentrates on the main economic and industrial activities of the country. In the last two decades, it has suffered disastrous social and economic consequences due to the rainfall intensification during the rainy season, and particularly in June which is the core of this season. For example, the ONU-HABITAT [4] report indicated that floods and landslides due to heavy rains were the main risks of natural disasters leading to loss of life and destruction of homes, especially those most at risk. This is in agreement with the "Société d'Exploitation et de Développement Aéroportuaire, Aéronautique et Météorologique" (hereafter SODEXAM) rain gauges measurements, which mentioned an exceptional rainfall amount (302.3 mm) during the night of June 18 th to June 19 th 2018, only in the municipality of Cocody (Table 1). Indeed, this amount represented about half of the rainfall in June. Therefore, many damages were observed such as around 18 deaths and 120 injuries, a flood and a landslide in the district of Abidjan (https://news.abidjan.net/h/639829.html). The frequent floods observed in recent times are therefore due to the exceptional quality of rainfall. However, ONU-HABITAT [4] also points out the occupation of roads and facilities of public interest, the city extension that does not match with the objectives of the urban planning and master plan for development which is never updating.
The disastrous consequences of these rainfall events require an understanding of their interannual variability and the influence of ocean and atmospheric conditions. Several studies have notably suggested a relationship between Atlantic Ocean conditions and rainfall in West Africa [2] [5] [6] [7], and relationship between both oceanic and atmospheric conditions and extreme rainfall [8]. Ta et al. [8] showed, for example, a different influence of oceanic and atmospheric conditions in the tropical Atlantic on the extreme rainfall in each of the climate zones of West Africa. Lafore et al. [9] also studied the oceanic and atmospheric conditions during an extreme rainfall event that occurred in 2009 in Ouagadougou (Burkina Faso) in the Sahel. Thus, they showed that the sea surface temperature (SST) anomalies observed in the Atlantic cold tongue, in the Atlantic Most of the studies that link these dramatic rainfall events to oceanic and atmospheric conditions were generally focused on all of West Africa [10], or on the Sahel [9] [11], but few on the Gulf of Guinea coast [12]. The main purpose of this present work is to better understand the oceanic and atmospheric conditions that can influence the occurrence of these events with disastrous consequences for the economy and social life of this booming city on the West African coast. Such study is important to propose solutions that could allow the adoption of adequate policies for the management of risks related to these events, and particularly to civil protection and urban planning. Section 2 presents the data and methods used in this study. Section 3 discusses the results by analyzing observed rainfall events and their relationships with oceanic and atmospheric conditions in the tropical Atlantic. Finally, a conclusion is given in the last section.

Data
June is the core of the rainy season in southern Côte d'Ivoire, especially in Abidjan. It is the month in which the city experiences heavy flooding. The history of these rain flooding events and the damages that occurred in specific municipalities of Abidjan were recorded by SODEXAM for each month of June during the 2007-2018 period. During this period, SODEXAM observed 13 events throughout the city of Abidjan in eight years (2007,2008,2009,2010,2011,2014,2017 and 2018) ( Table 1). This corresponds to about 1.6 events per year. Table 1 includes the municipality affected by the rain flooding event, the rainfall amount recorded by a rain gauge at a specific station, the types of disasters that occurred as a consequence of the event such as the human and material damages. When looking at the amount (Table 1), the rainfall concerns only specific measuring points and not the whole city of Abidjan. Moreover, some measurements are not accessible because they have not been carried out or they are expensive. To capture the whole study area, daily data from the Global Precipitation Climatology Project-GPCP [13] were used with 1˚ × 1˚ resolution. These data are derived from a combination of in-situ data and microwave measurements, and are suitable for an analysis of the temporal and spatial evolution of rainfall. Indeed, a comparative analysis of 10 satellite products over the Sahelian region of West Africa [14] shows that GPCP data are better than others. The GPCP data are more representative of longer duration and better reproduce extreme events with a behavior close to the reference data [14]. In comparison, the TRMM 3B42 and GSMaP-MVK satellite products have a large percentage of low rainfall event grid points of less than 5 mm with a maximum around 20 mm.
According to Roca et al. [15], satellite data differ also considerably when they compared 10-day rainfall cumulative from rain gauges and estimates from satellite products. The results highlighted that, for the estimated errors in satellite, GPCP have a better spatial distribution of rainfall with lower errors in the central Sahel, and in the coastal region of West Africa [14]. For this study, GPCP daily rainfall for each month of June from 2007 to 2018 (~12 years, i.e. ~360 values) was extracted for Abidjan.
The analysis of the causes of these rain flooding events is carried out by using daily data of sea surface temperature (SST), vertical and horizontal winds and specific humidity. Sea surface conditions are derived from Reynolds data [16] over the tropical Atlantic and are extracted from the IRI/LDEO Climate Data Library (http://iridl.ldeo.columbia.edu/). These data are in a regular grid of 1˚ × 1˚ for each month of June from 2007 to 2018. Vertical and horizontal winds, and specific humidity are extracted from the National Center for Environmental Prediction-National Center for Atmospheric Research (NCEP-NCAR) reanalysis for the same period [17]. These last data are in a 2.5˚ × 2.5˚ grid, with 17 pressure levels from 1000 hPa to 10 hPa for horizontal and vertical winds, and eight pressure levels from 1000 hPa to 300 hPa for specific humidity.
To assess the impact of moisture transport from the lower and upper layers of the atmosphere [18] on the occurrence of these rain flooding events, the integrated vertical moisture flux transport is calculated. The zonal and meridian components of the integrated vertical moisture flux transport are derived from the specific humidity, and the zonal and meridian components of the horizontal wind on both 1000 -700 hPa and 700 -300 hPa layers.

Methods
The interannual variability of the rainfall events that caused flooding is studied by using the 75th and 95th percentiles. Indeed, the Expert Team on Climate Change Detection and Index Study and the users of climate data have defined extreme precipitation using the 75th, 95th and 99th percentiles [19]. So, the 95th and 75th percentiles will be used to represent extreme and intense rainfall events respectively. Both thresholds are widely used in various studies at global, regional and local scales [20] [21] [22] [23].
These two thresholds are calculated on all the values (~360 values) for June during the 2007-2018 period. These estimates will allow evaluating the contribution of rainfall above these thresholds and which could or could not cause flooding. In addition, the methodology of Cao et al. [23] is used to avoid overlapping of the studied events. It stipulates that a unit of ±5 is applied to the percentile threshold. In our case, if the value of a selected event with the 75th percentile (hereafter intense event) falls within the range defined for the 95th percentile (hereafter extreme event), then it is classified as an extreme event. Similarly, if an event belongs to the range defined for the 75th percentile or is greater than it, but does not fall within the range defined for the 95th percentile, then it is classified as an intense event.
According to Burpee [24] and Zhang et al. [25], Côte d'Ivoire is located in zones C2 and D bordering the tropical Atlantic (see Figure 4 in [26]). These areas are influenced by the monsoon flow from the ocean. Diawara et al. [26] noted that this influence could vary from year-to-year. The impact of the West African monsoon on rain flooding events is studied by using the methodology described by Diawara et al. [26] and Ta et al. [8]. The June meridian wind reanalysis from NCEP-NCAR [17]    An upward trend of rainfall is observed from 2013 to 2018, after a downward trend from 2007 to 2012. Rainfall above the 75th percentile shows a decreasing trend, while that above the 95th percentile has an increasing trend. This could mean a resurgence and an intensification of extreme events over the city of Abidjan, as pointed out by several authors [8] for the West African coast.

Analysis of Rainfall Events
During the 2007-2018 period (~12 years, or 360 values), 25 events are extreme, while 57 events are intense. So, there are approximately 2 events per year on average (respectively 5 per year) which are above the 95th (respectively 75th) percentile; that means 2.5 times more intense events than extreme events. The contribution of extreme rainfall amounts (respectively intense) represents ~22.23% (respectively ~26.36%) of the cumulative rainfall amounts in June during the 2007-2018 period. The sum of these two contributions (~48.59%, i.e. almost 50%) highlights the importance of intense and extreme events in the June rainfall at Abidjan. Particularly, the low number of extreme events (~2 per year) above the 95th percentile is likely to produce about ~1/5 of the June rainfall. This indicates at least a large part of the rainfall of this month is concentrated in the occurrence of a few rainy events which can have disastrous consequences for this city.
The next paragraphs and sections now focus on the rainfall events recorded by SODEXAM as being rain flooding events that caused material and human damage (see Table 1). Figure 2 presents the time series and daily anomalies of June rainfall in Abidjan during the years when flooding events were recorded by SODEXAM. Anomalies are calculated as the difference between the rainfall on a calendar day in June and the annual average for that day over the entire 2007-2018 period. The contribution (in percentage) of a given event occurring in a given month, relative to the cumulative precipitation for that month is represented, as well as the 75th and 95th percentile thresholds calculated over the entire 2007-2018 period.
Extreme events contribution ranges between 11.18% (June 19, 2018) and 29.11% (June 29, 2008) of the monthly cumulative, while that of intense events is lesser than 7% (Figure 2). When considering these values and the number of extreme events per year in June (~2 events), it seems that the total contribution of extreme events causing flooding does not exceed ~30%. For example, the sum of the contributions of the extreme events (~29.36%) is the largest in 2014 because only this year presents two extreme events. Similarly, this figure shows that events classified as extreme have anomalies greater than +10 mm, while those classified as intense have anomalies less than 10 mm. This result is in agreement with the works of Liebman et al. [30] and Kouadio et al. [31] who used this anomaly threshold for heavy rains over Northeast Brazil. In the case of intense events, the anomalies are either negative or slightly positive and in most cases below 5 mm.  (~300 hPa depth) on average, that correspond to a low-level atmospheric flow.

Impact of the Monsoon Depth
This observation is in agreement with Burpee [24] and Zhang et al. [25] who showed that the monsoon flow from Atlantic Ocean, which influences Côte d'Ivoire, was particularly concentrated in the lower atmospheric layers between 1000 hPa and 500 hPa levels. The mean atmospheric level (~700 hPa) corresponds to that generally used for the study of the easterly waves [32] [33].

Relationship with Easterly Waves Propagation
In this section, the relationship between easterly waves and these rain flooding events that caused material and human damages is analyzed. This study is carried out because eight of the 13 events (~62%) occur when the moisture flow is located at the lower layers of the atmosphere.   . Indeed, those events above the 95th percentile are generated by the easterly waves propagating from Eastern Africa towards 20˚E, close to the Joss Plate [8]. This area is known as an initiation zone for convective systems crossing West Africa. Likewise, events above the 75th percentile are associated with easterly waves initiated around 5˚E, i.e. closer to Abidjan. As previously observed, extreme events exhibit contribution greater than 10%, while that of intense events do not exceed 7%. Let us also note that easterly waves are more active in August-September than in June-July [33]. That is in agreement with the lower number of rain flooding events in June related to these atmospheric disturbances. The remaining events could be influenced by other atmospheric processes, as for instance cold fronts from the southern tropical Atlantic and the Madden-Julian oscillation, not considered in this study. It would be interesting to extend, in future works, this study to all events, whether or not they create flooding, to clearly identify the processes involved in their occurrence.

Ocean and Atmospheric Conditions Associated with Rain Flooding Events
The impact of oceanic and atmospheric conditions on the rain flooding events that occurred in Abidjan is analyzed. This impact is studied by plotting the June SST daily anomalies for the 2007-2018 period, the daily specific humidity, the vertical wind, the meridian gradient of absolute vorticity at 700 hPa and the vertically integrated moisture flux on the 1000 -700 hPa and 700 -300 hPa columns.  To study the particularities that could exist for each event, daily anomaly patterns for four days before extreme or the intense rainfall event (Figure 6(a)) and for the onset day of this given event (Figure 6(b)) are produced. This time lag highlights the persistence of the thermal structures observed in both figures. The ocean warming observed four days before the onset day in Figure 6(a) enhances during the onset day in Figure 6 [34]. This abnormal warming could lead to strong oceanic evaporation causing increased instability of the lower monsoon layers over the sea and the coastline, and to higher humidity gradient creating a maximum mass advection. This is consistent with Caniaux et al. [35] who noticed that the Atlantic cold tongue plays a key role in West African monsoon variability.
The second structure illustrates a global warming of the whole ocean basin during the events of June 11 and June 24, 2017, June 29, 2008 and June 24, 2010.
Finally, the third structure indicates a north-south dipole of SST anomalies. This dipole shows a warming (respectively cooling) in the southern (respectively northern) basin in 2014 and 2009 and an inversion in 2011. This structure is consistent with the monthly composite structure of the anomalies seen in the previous paragraph (not shown). Likewise, the observed north-south dipole of SST anomalies corresponds to the second mode of interannual variability in the tropical Atlantic Ocean [36]. It results in a North-South oscillation whose axis of division lies at the ITCZ level. These different results provide the occurrence of these events is associated with particular patterns of tropical Atlantic SST anomalies. Other oceanic parameters, not considered here, could also influence these extreme events, as for instance the wind stress, the heat content and the ocean heat flux. It would be interesting to go further in future oceanic variables analysis to better understand the processes involved in their occurrence.
The evolution of upward motion in the atmosphere is also analyzed ( Figure   7). The altitude-latitude composite diagram of the vertical velocity averaged over 10˚W -10˚E (Figure 7, left) and its corresponding anomaly pattern (Figure 7  hancing rainfall during these periods. Li et al. [37] and Ta et al. [8] explained this phenomenon by greater convergence of oceanic and continental excess moisture input into the lower troposphere, which could influence monsoon circulation and affect rainfall. Figure 8 displays the composite pattern of the meridian gradient of absolute vorticity at 700 hPa, and the vertically integrated moisture flux and specific humidity anomalies performed in the 1000 hPa -700 hPa and 700 hPa -300 hPa columns. Negative values of the meridian gradient are located around 10˚N over West Africa. These negative values lie between the positive values at the north of the Gulf of Guinea at around 0˚N -5˚N, and on the continent at around 20˚N. It should also be noted that the ITCZ and its ground track, which is the intertropical front (FIT) on the continent, is located at this period around 5˚N, i.e. between the first positive vorticity structure in the south around 5˚N and the negative vorticity structure. This observation is corroborated by the results of Ferreira and Schubert mentioning that convection at the ITCZ creates such an inversion of the signs [38]. This implies an increase in disturbance indicating that these events follow, at this level, the Charney-Stern [29] criterion of barotropic instability. On the other hand, this instability could contribute to the intensification of these rain flooding events in Abidjan. In addition, this zone of instability  Figure 8. Latitude-longitude composite daily values of integrated moisture flux (×10 −4 g/kg/s, vectors) and anomalies of specific humidity calculated from 1000 hPa to 700 hPa (top) and 700 hPa to 300 hPa (bottom). Absolute vorticity at 700 hPa (×10 −11 s/m, contours) is also plotted. forms in the moisture streamlines due to the monsoon surge [39], which is a characteristic of the types of disturbed weather. This parameter also provides, between 700 hPa and 300 hPa, a strong moisture correspond to the intensification of the African easterly jet during these events [32]. These two observations thus result in a combined action of the influence of both oceanic and continental zones in the occurrence of the events.

Conclusions
This study analyses rain flooding events in June on the coastal area of West Africa, particularly in Abidjan, which is the economic capital of Côte d'Ivoire.
The June period corresponds to the core of the great rainy season over West Africa coast during which most of the dramatic rainfall events occur. This work also relates these events to oceanic and atmospheric conditions that could influence them. The variability of the rainfall events is studied by using the 95th and 75th percentiles, which represent extreme and intense rainfall events respectively.
The trend of the cumulative intense rainfall amount decreases, while that of extreme rainfall amount increases. The sum of their contributions reaches almost 50% and shows the importance of these two events in the June rainfall.
Particularly, the two extreme events per year are likely to produce about ~1/5 of the June rainfall.
The West African monsoon that reaches Abidjan in June during these events corresponds to a low-level atmospheric flow. The atmospheric depth increasing for certain years along the coast of West Africa could lead to greater penetration of monsoon flow over this region, and thus influence the intensity of extreme rainfall amounts. A proportion of 31% (third) of events is associated with easterly wave propagation.
The impact of oceanic and atmospheric conditions on the rain flooding events is also analyzed by using different sets of data. Daily SST anomalies provide three different structures. The first one concerns the warming of the equatorial rail and the Atlantic cold tongue during these events. The second pattern is related to a warming of the whole ocean basin. Finally, the third structure indicates a north-south dipole of SST anomalies. Further analyses are also in progress to understand the atmospheric processes which contribute to the occurrence of some events not relating to easterly waves.
Such a study is important to propose solutions that could allow the adoption of adequate policies for the management of risks related to these events, and particularly to civil protection and urban planning.