APP下载

Tanzanian rainfall responses to El Niño and positive Indian Ocean Dipole events during 1951—2015

2021-11-25DicksonMbigiZiniuXiao

Dickson Mbigi , Ziniu Xiao

a State Key Laboratory of Numerical Modeling for Atmospheric Sciences and Geophysical Fluid Dynamics, Institute of Atmospheric Physics, Chinese Academy of Sciences,Beijing, China

b University of Chinese Academy of Sciences, Beijing, China

c Department of Physics, University of Dar es Salaam, Dar es Salaam, Tanzania

Keywords:Rainfall Indian Ocean Dipole El Niño Tanzania

ABSTRACT The relative impacts of Indian and Pacific Ocean processes on Tanzanian rainfall was evaluated using composite and correlation analyses. It was found that the seasonal responses of rainfall to positive Indian Ocean Dipole(pIOD) and El Niño events are substantial from September—October—November (SON) to December—January—February (DJF), whereas the Indian Ocean Dipole (IOD) exerts more control than El Niño—Southern Oscillation(ENSO) in both seasons. The associated relationship with the sea surface temperature (SST) and large-scale atmospheric circulations revealed distinct features. For the pure pIOD years, there is above-normal rainfall over the entire country. A strong rainfall condition is evident over the Lake Victoria basin and coastal and northeastern highland parts of the country during SON, while areas of the central and southern highlands exhibit substantial rains during DJF. For the pure El-Niño events, Tanzania has suffered from insignificant, weak, and non-coherent rainfall conditions during SON. However, a contrasting insignificant rainfall signature is found between the northern and southern parts of the country during the subsequent DJF season. For the co-occurrence of pIOD and El Niño, significant, excessive rainfall conditions are restricted to over the northern coast and northeastern areas of the country during SON, consistent with the rainfall pattern for pIOD. A weak, positive rainfall condition is observed over the entire country in the following season of DJF. Generally, in terms of Tanzanian rainfall,the IOD/ENSO variability and the associated impacts can be explained by the anomalous SST and circulation anomalies.

1. Introduction

Tanzania, East Africa’s most populated country ( Population Reference Bureau, 2013 ), is located within 12 o S —1 o N and 28—42 o E . The climatological rainfall pattern is mainly influenced by the seasonal migration of the Intertropical Convergence Zone (ITCZ). The ITCZ is largely responsible for the two rainy seasons during October—November—December (OND) and March—April—May (MAM) in the northern coastal and northeastern regions of Tanzania, and one rainy season during November—April in the central and southern parts of the country( Fig. 1 (a)). This study was motivated by the important role of seasonal rainfall variability and evolution in influencing socioeconomic activities. Extreme climate events such as floods and droughts are common over the East African region. Recent observations have shown that the OND season has been struck by frequent flooding ( Chang’a et al., 2017 ),while the DJF and MAM seasons have been plagued by long-lasting drought conditions ( Harrison et al., 2019 ), leading to severe economic impacts. These impacts can be minimized by a good understanding of the physical processes that cause such events. Therefore, there is a need to better understand the seasonal dynamics of rainfall over Tanzania with large-scale general circulation processes that affect our climate,particularly over the Indian and Pacific Oceans.

East African rainfall displays large interannual variations. This condition has been associated with extreme climate occurrences with farreaching impacts on society. Some of the extreme rainfall conditions have been observed to persist for several months or years. The persistent nature of these conditions is linked to both the Indian Ocean Dipole (IOD) and El Niño—Southern Oscillation (ENSO). The potential roles of the IOD and ENSO in East African rainfall has been reported statistically in past studies ( Clark et al., 2003 ; Saji and Yamagata, 2003 ;Behera et al., 2005 ; Black, 2005 ; Bahaga et al., 2019 ; Chang’a et al.,2020 ; Lüdecke et al., 2021 ). Clark et al. (2003) suggested that ENSO variability is another forcing mechanism of rainfall variability in parts of the region. Furthermore, Black (2005) reported that the association between the East African rains and ENSO is a manifestation of a dynamic link between the IOD and ENSO. However, Saji and Yamagata (2003) revealed that the pure ENSO influence is weaker in parts of East Africa,consistent with Bahaga et al. (2019) . Lüdecke et al. (2021) , on the other hand, reported that the IOD signal appears prominent in Tanzanian short rains compared to that of ENSO, in line with Behera et al. (2005) . A study by Chang’a et al. (2020) revealed that the 2019 extreme rainfall condition over East Africa was coupled with the positive phase of the IOD (pIOD), while that of 1997 was associated with the cooccurrence of pIOD and El Niño (hereafter referred to as combined events).

Fig. 1. (a) Time—latitude Hovmöller diagram showing the mean monthly Tanzanian rainfall. Spatial patterns of SVD1 for (b) SON Indo-Pacific SST and (c) annual Tanzanian rainfall presented as heterogeneous correlation maps. Stippling denotes significance at the 99% confidence level. (d) SVD1 time series of the EC for SST(green line) and rainfall (red line), and temporal series of SON-DMI (blue line) and NDJ-NINO3.4 index (yellow line).

Most studies have investigated the individual effects of ENSO and IOD events on Tanzanian rainfall in a particular season without considering the phases of these events in a year. In this study, we investigated the individual and combined influences of ENSO and IOD on the seasonal evolution of Tanzanian rainfall and explored the possible atmospheric circulation structure associated with the IOD/ENSO sea surface temperature (SST) pattern to gain a better understanding of SST—rainfall links.

2. Data and methods

2.1. Data

The rainfall data were sourced from the Climate Research Unit( Harris et al., 2020 ), while the SST and wind data were obtained from the National Centers for Environmental Prediction—National Center for Atmospheric Research reanalysis ( Kalnay et al., 1996 ) (see Text S1 for further details).

2.2. Methods

The Dipole Mode Index (DMI) was calculated according to Saji et al. (1999) , while the NINO3.4 index was obtained following Trenberth (1997) . A non-simultaneous threshold value of the DMI(NINO3.4) index exceeding + 0.5 was adopted to classify pure pIOD (El Niño) events, while a combined event was identified when both events satisfied the threshold criterion. Based on these criteria, the identified events are presented in Table 1 .

Table 1 Classification of IOD and ENSO events, 1951—2015.

Furthermore, singular value decomposition (SVD) ( Bretherton et al.,1992 ) analysis was performed to understand the features of Tanzanian rainfall associated with Indo-Pacific SST in coupled modes. Composite analysis was performed to investigate the influence of IOD/ENSO and associated atmospheric circulation anomalies on rainfall. A partial correlation technique was used to assess the independent impact of climate modes on Tanzanian rainfall. As it is known that both the IOD and ENSO exhibit significant decadal variability, the use of a high-pass filter was adopted to remove lower-frequency signals. The correlation significance level was tested based on the effective degrees of freedom (see Text S2 for more details).

3. Covariability of rainfall and SST

The first SVD (SVD1) mode accounts for about 88% of the square covariance. The SST spatial pattern shows variability over the eastern and western Indian Ocean oppositely, resembling the pIOD mode and SST variability over the central-eastern Pacific Ocean, a pattern characteristic of El Niño ( Fig. 1 (b)). The corresponding rainfall over Tanzania exhibits positive correlations over the whole country, with the significant correlation being restricted to over the northern coast, Lake Victoria Basin (LVB), and northeastern highlands ( Fig. 1 (c)), consistent with previous work ( Owiti, 2008 ).

Furthermore, we assessed the coupling strength of the association between Tanzanian rainfall and Indo-Pacific SST ( Fig. 1 (d)). The Indian(Pacific) Ocean SST variability is represented by the DMI (NINO3.4) index. The expansion coefficient (EC) of rainfall and SST for the SVD1 exhibit a consistent pattern associated with a significant correlation value of 0.43. The rainfall EC shows a significant correlation with the DMI(

r

≅0.62) and NINO3.4 index (

r

≅0.45). These results indicate that the variability of Tanzanian rainfall is closely associated with IOD and ENSO phases in the Indian and Pacific Oceans.

4. Seasonal rainfall variation characteristics related to SST

In this part, we label the period in which the IOD/ENSO develops and matures as period(0) and the period before and after period(0) as period( − 1) and period(1), respectively (Fig. S1). The rainfall is negatively correlated with the DMI from October( − 1) to January(0), followed by a positive correlation during April(0) to May(0), and then a significant positive correlation persists from September(0) to February(1). On the other hand, rainfall and NINO3.4 index exhibit a positive correlation from March(0) to May(0), followed by a negative correlation in July(0), which changes sign and becomes a significant positive correlation during September(0) to February(1). Generally, a relatively low association between rainfall with the SON-DMI and NDJ-NINO3.4 index is observed around MAM(0), while a significantly high connection is concentrated from September(0) to February(1). The IOD can explain about 61% (37%), while ENSO can explain about 19% (13%), of the variability in rainfall during the SON (DJF) season.

5. Impact of the IOD and ENSO on Tanzanian rainfall

5.1. Pure pIOD

During SON(0) there is large-scale warming centred in the central Indian Ocean (CIO), spreading over the western tropical Indian Ocean(WTIO), but robust cooling in the southeastern tropical Indian Ocean(SETIO; Fig. 2 (a, c)). In this regard, an elongated warm SST is seen to stretch in a southwesterly orientation toward Australia from the WTIO.The anomalous rainfall is dominated by positive values over the entire country, although there appear to be more and significant rainfall records over the coastal, northeastern, LVB, and northwestern parts of the country ( Fig. 2 (e, g)). The anomalous 850-hPa wind fields are dominated by cyclonic and anticyclonic circulations over the north and south of the CIO, respectively ( Fig. 2 (g)). These circulation patterns are associated with significant easterly wind flow across the equatorial Indian Ocean, between 60and 95E , towards the western Indian Ocean, bringing moisture into the study region. There is moisture influx from anticyclonic easterly flow across the country from the southern Indian Ocean,with anticyclones centred to the northeast of Zambia. Consistently, the anomalous 200-hPa wind fields are dominated by significant and strong anticyclonic easterly flow over the northern parts of the country from the Indian Ocean, with the anticyclones centred to the south around the Mozambique Channel ( Fig. 2 (i)). The condition of strong tropospheric upper-level divergence is favourable for wet conditions over the northern parts of the country. During DJF(1) the observed location of the warm SST over the WTIO during SON(0) has shifted toward the central and southern Indian Ocean, while the cool SST persists over the SETIO and additionally over the Mozambique Channel ( Fig. 2 (b, d)).This indicates that, during a pure pIOD, the SETIO pole of the IOD is the robust centre in a changing season from SON(0) to DJF(1). A significant high rainfall condition is located over southwestern and southern parts of the country ( Fig. 2 (f, h)). The observed rainfall pattern is associated with a southwesterly wind anomaly from the South Atlantic Ocean ( Fig. 2 (h)). The moisture influx is also from the potential trough,extending from west to east of Madagascar Island, over the south Indian Ocean. In this regard, significant cyclonic southwesterly flow is observed to bring moisture across the southern parts of the country. In line with the 850-hPa wind orientation, the upper-level wind pattern indicates that the southern parts of the country are characterized by strong, albeit insignificant, anticyclonic southwesterly flow, extending from parts of central Angola to northern parts of Tanzania ( Fig. 2 (j)). The observed circulation orientations indicate that wet conditions over southern Tanzania can be associated with strong tropospheric upper-level divergent winds.

5.2. Pure El Niño

The pure El Niño during SON(0) is accompanied by a warmer and strong SST signature in the Arabian Sea and to the Indian Ocean coast of Indonesia, while the cooler waters are trapped offthe western Australian coastline ( Fig. 3 (a, c)). The rainfall response ( Fig. 3 (e, g)) is associated with the insignificant weak and non-coherent condition over Tanzania.The anomalous 850-hPa wind pattern is dominated by an inflow of variable and insignificant southwesterly flow from the South Atlantic Ocean and northeasterly flow from the Arabian Sea region. The weak convergence across the country suppresses moisture convection and hence dry conditions. Consistently, the upper troposphere over the country is dominated by anomalous cyclonic westerly flow, oriented along the Arabian Sea and the Mozambique Channel, indicating convergence of winds favouring dry conditions ( Fig. 3 (i)). For DJF(1), most of the Indian Ocean is warmer, with a small tongue of cold SST trapped offthe western Australian coastline ( Fig. 3 (b, d)). Although weak and insignificant, the observed rainfall anomalies exhibit positive (negative) rainfall anomalies over northern (southern) areas of the country ( Fig. 3 (f, h)).The observed rainfall pattern is associated with southeasterly and easterly flow from the easterly equatorial south Indian Ocean winds, mainly through the latitude of 10 o S ( Fig. 3 (h)). The southeasterly flow passes through most of the Indian Ocean towards northern parts of the country, while the easterly flow is traversing through Madagascar Island towards southern parts of the country. The former wind orientation can result in wet conditions, while the latter can lead to dry conditions over the respective regions. Northerly and southerly divergent winds dominate the anomalous upper-level circulation wind field over the country( Fig. 3 (j)). The latter winds pass through the northern parts of the country, indicating potentially wet conditions.

Fig. 2. Composite of the relevant data anomalies during SON(0) (left-hand column) and DJF(1) (right-hand column) for pure pIOD years. (a, b) SST; (g, h) rainfall(shade) and winds (vectors) at 850hPa; (i, j) 200-hPa winds. The partial correlation between seasonal data anomalies and DMI after partialing out the NINO3.4 influence for SON(0) (left-hand column) and DJF(1) (right-hand column). (c, d) SST; (e, f) rainfall. Stippling and red vectors indicate significance at the 95%confidence level.

5.3. Combined events

Fig. 3. Composite of the relevant data anomalies during SON(0) (left-hand column) and DJF(1) (right-hand column) for pure El Niño years. (a, b) SST; (g, h)rainfall (shade) and winds (vectors) at 850hPa; (i, j) 200-hPa winds. The partial correlation between seasonal data anomalies and NINO3.4 after partialing out the DMI influence for SON(0) (left-hand column) and DJF(1) (right-hand column). (c, d) SST; (e, f) rainfall. Stippling and red vectors indicate significance at the 95%confidence level.

The SST anomalies show that the western and eastern parts are in an antiphase signature during SON(0), reflecting a typical pIOD mode( Fig. 4 (a)). An elongated warm SST is seen to stretch in a southwesterly direction toward Australia from the WTIO. Significantly high rainfall amounts are restricted to over the northern coast and northeastern highlands of the country ( Fig. 4 (c)), resembling the boreal autumn rainfall pattern over Tanzania ( Owiti, 2008 ). The 850-hPa wind anomaly is dominated by strong and significant anticyclonic easterly flow, centred on the equatorial southern CIO, associated with easterly oriented significant winds from the eastern to the western equatorial Indian Ocean( Fig. 4 (c)). The observed dominance of easterly zonal flow towards the off-East-African coast indicates the diminished normal westerly flow over the equatorial Indian Ocean, contributing to more advection of moisture towards the study region. Consistently, there is strong anticyclonic upper-level easterly flow trapped around the southern Tanzanian coast associated with strong divergence of winds, which indicates wet conditions over the country ( Fig. 4 (e)). These circulation anomalies indicate that high rainfall over the northern coast and northeastern Tanzania are associated with substantial upper-level divergence of winds. During DJF(1), the widespread warm SST anomalies ( Fig. 4 (b)) imply a coherent monopole pattern over the tropical Indian Ocean. The anomalous rainfall is associated with weak and insignificant above-normal amounts that appear to be uniform over the entire country ( Fig. 4 (d)). The associated wind anomalies exhibit northeasterly flow from the north Indian Ocean towards the Tanzanian coast and then diverge, becoming northwesterly flowing to the southern Indian Ocean. This pattern indicates less moisture advection towards the country, leading to little rainfall.The potential moisture inflow over Tanzania, albeit weak and insignificant, is also from the southeasterly flow from Madagascar through the Mozambique Channel. For the 200-hPa winds ( Fig. 4 (f)), the centre of upper-level anticyclonic easterly flow over Tanzania has shifted towards the equatorial southern Indian Ocean (northeastern Madagascar Island).As such, the anticyclonic easterly flow blows mainly as northerly and southerly divergent winds offthe country, with weak and variable flow across the country. This orientation is insufficient for enhanced wet conditions.

Fig. 4. Composite of the relevant data anomalies during SON(0) (left-hand column) and DJF(1) (right-hand column) for combined events. (a, b) SST; (c, d) rainfall(shade) and winds (vectors) at 850-hPa; (e, f) 200-hPa winds. Stippling and red vectors denote significance at the 95% confidence level.

6. Summary and discussion

Studies (e.g. Chang’a et al., 2020 ) have associated the relative influences of the IOD and ENSO on Tanzanian and generally East African rainfall to a particular season, with little attention paid to seasonal and subseasonal rainfall evolution over a year. This study examined the seasonal evolution of Tanzanian rainfall anomalies in response to the individual and combined SON-pIOD and NDJ-El Niño. Results indicated that the change in Tanzanian rainfall is significantly sensitive to varying pIOD and El Niño from September to February of the following year, whereas the IOD exerts more control than ENSO, consistent with Clark et al. (2003) and Lüdecke et al. (2021) . However,Lüdecke et al. (2021) found significant IOD associations in October and November. This difference might be related to the different periods of data used in studies. The case studies of individual and combined events revealed distinct features, as briefly summarized below:

For pure pIOD, the Indian Ocean shows central-centred warming in the west and cooling in the east during SON(0). The associated circulation pattern emanates from the equatorial and southern Indian Ocean. These patterns are suggested to bring substantial rainfall over the coastal, northeastern, LVB and northwestern parts of the country,consistent with Behera et al. (2005) and Bahaga et al. (2019) . In the subsequent DJF(1) season, the warming over the western Indian Ocean is restricted to around the central and southern Indian Ocean and cooling to over the SETIO and Mozambique Channel. The moisture inflow over the study area originates from the Atlantic and Indian Oceans. Substantial rainfall amounts are centred over the southern parts of the country,consistent with Harrison et al. (2019) and Mapande and Reason (2005) .

During SON(0), the pure El Niño regions of SST influence in the eastern tropical Indian Ocean are confined to the Indonesian ( Meyers, 1996 )and Australian coastline, while over the western tropical Indian Ocean are apparent in the Arabian Sea. The circulation pattern is dominated by an inflow of weak and variable winds from the South Atlantic Ocean and the Arabian Sea region, leading to insignificant, noncoherent rainfall conditions. A similar rainfall orientation was reported by Behera et al. (2005) and Bahaga et al. (2019) . In the following season of DJF(1), El Niño events are accompanied by a general warming of the Indian Ocean ( Meyers, 1996 ), with a small tongue of cold SST trapped offthe western Australian coastline. The circulation anomalies emanate from the easterly equatorial southern Indian Ocean flow, mainly through the latitude of 10S , resulting in non-significant opposing rainfall conditions.

For combined events, the IOD and ENSO have significant western and eastern tropical Indian Ocean responses during SON(0). The results further showed that the centre of positive SST anomalies observed for pure pIOD has shifted to the WTIO, resulting in a typical IOD mode.Strong anticyclonic easterly flow, centred on the southern Indian Ocean,is evident. Correspondingly, significant wet conditions are restricted to over the northern coastal and northeastern highland areas of the country. Like pure El Niño, a coherent monopole pattern of SST over the entire tropical Indian Ocean is apparent for combined events during DJF(1). The eastward shifts of the anticyclonic centre over the southern Indian Ocean observed in SON(0) characterizes relative weak easterly flow during DJF(1). The pattern suggests weak rainfall conditions over the country.

Funding

The first author appreciates the financial support from the Chinese Academy of Sciences "Belt and Road" Masters Fellowship Programme.

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi: 10.1016/j.aosl.2021.100093 .


登录APP查看全文