Indo-Pacific sea surface temperature influences on failed consecutive rainy seasons over eastern Africa

Clim Dyn DOI 10.1007/s00382-013-1991-6

Indo-Pacific sea surface temperature influences on failed consecutive rainy seasons over eastern Africa Andrew Hoell • Chris Funk

Received: 17 June 2013 / Accepted: 30 October 2013 Ó Springer-Verlag (outside the USA) 2013

Abstract Rainfall over eastern Africa (10°S–10°N; 35°E–50°E) is bimodal, with seasonal maxima during the ‘‘long rains’’ of March–April–May (MAM) and the ‘‘short rains’’ of October–November–December (OND). Below average precipitation during consecutive long and short rains seasons over eastern Africa can have devastating long-term impacts on water availability and agriculture. Here, we examine the forcing of drought during consecutive long and short rains seasons over eastern Africa by Indo-Pacific sea surface temperatures (SSTs). The forcing of eastern Africa precipitation and circulation by SSTs is tested using ten ensemble simulations of a global weather forecast model forced by 1950–2010 observed global SSTs. Since the 1980s, Indo-Pacific SSTs have forced more frequent droughts spanning consecutive long and short rains seasons over eastern Africa. The increased frequency of dry conditions is linked to warming SSTs over the Indowest Pacific and to a lesser degree to Pacific Decadal Variability. During MAM, long-term warming of tropical west Pacific SSTs from 1950–2010 has forced statistically significant precipitation reductions over eastern Africa. The warming west Pacific SSTs have forced changes in the regional lower tropospheric circulation by weakening the Somali Jet, which has reduced moisture and rainfall over the Horn of Africa. During OND, reductions in A. Hoell (&)
C. Funk Department of Geography, University of California, Santa Barbara, Santa Barbara, CA, USA e-mail: [email protected] C. Funk e-mail: [email protected] C. Funk Earth Resources Observation and Science Center, U.S. Geological Survey, Sioux Falls, SD, USA

precipitation over recent decades are oftentimes overshadowed by strong year-to-year precipitation variability forced by the Indian Ocean Dipole and the El Nin˜o– Southern Oscillation. Keywords Eastern Africa
Drought
West Pacific warming
ENSO
Indian Ocean Dipole
Pacific Decadal Variability

1 Introduction Precipitation over eastern Africa, defined here as the rectangular region bounded by the points 10°S–10°N and 35°E–50°E (Fig. 1a), is bimodal due to the movement of the the Intertropical Convergence Zone through the region (Nicholson 1996). Seasonal precipitation maxima over eastern Africa occur during the ‘‘long rains’’ of March– April–May (MAM) and the ‘‘short rains’’ of October– November–December (OND) (Fig. 1b). Precipitation during MAM and OND are critical to food security and water resources over eastern Africa (Washington and Downing 1999; Haile 2005). When rains fail during MAM or OND, the local populations endure significant hardships (Verdin et al. 2005). When rains fail during consecutive MAM and OND seasons, these hardships are exacerbated and have the potential to result in widespread famine, similar to what occurred during 2011 (FEWSNET 2011; Hillbruner and Moloney 2012). Rainfall over eastern Africa during the MAM and OND seasons are influenced by Indo-Pacific sea surface temperatures (SST). Precipitation over eastern Africa during OND is closely related with the Indian Ocean Dipole (IOD) (Abram et al. 2008) discovered by Saji et al (1999) and the El Nin˜o–Southern Oscillation (ENSO) (Ogallo 1988;

123

A. Hoell, C. Funk Fig. 1 a The eastern Africa domain, taken to be the land area bounded by the purple box (10°S–10°N and 35°E–50°E). b Average monthly observed precipitation for 1981–2010 spatially averaged over the eastern Africa domain, highlighting the long rains of March–April–May (red) and the short rains of October– November–December (green)

(a)

Hastenrath et al. 1993; Nicholson and Kim 1997; Nicholson and Selato 2000). Precipitation over eastern Africa during MAM is related to SST changes over the west Pacific (Williams and Funk 2011; Lyon and DeWitt 2012) and the southern Indian Ocean (Funk et al. 2008) in addition to weaker influences by ENSO (Indeje et al. 2000; Camberlin and Philippon 2002). A close examination of the SST influences on consecutive MAM and OND season droughts has yet to be performed. Here, we examine how Indo-Pacific SST force consecutive MAM and OND season droughts and whether the frequency of consecutive rainy season droughts have increased in recent decades. During OND, seasonal-to-interannual SST variations over the tropical Indo-Pacific Oceans associated with ENSO (Ogallo 1988; Hastenrath et al. 1993; Nicholson and Kim 1997; Nicholson and Selato 2000) and the IOD (Abram et al. 2008; Black et al. 2003) strongly influence eastern Africa rainfall. The positive phases of ENSO (El Nin˜o) and the IOD (Black et al. 2003) result in enhanced rainfall over eastern Africa while the negative phases of ENSO (La Nin˜a, Nicholson and Selato (2000) and the IOD result in diminished rainfall over eastern Africa. During La Nin˜a events, the anomalous lower tropospheric circulation over the Arabian Peninsula is cyclonic, which results in precipitation-reducing dry air advection over eastern Africa (Hoell et al. 2013) in addition to widespread anomalous descent associated with the Walker Circulation. During MAM, seasonal-to-interannual SST variations over the tropical Indo-Pacific Oceans (Indeje et al. 2000; Camberlin and Philippon, 2002) are less closely related with eastern Africa rainfall. However, declines in total precipitation over eastern Africa during MAM in the recent decade (Williams and Funk 2011; Lyon and DeWitt 2012) have occurred contemporaneously with SST increases over the southern Indian Ocean (Funk et al. 2008), a westward extension of the Indo-Pacific Warm Pool (Williams and Funk 2011) and the 1999–2010 averaged SST pattern of tropical Pacific SSTs (Lyon and DeWitt 2012). All of these studies concluded that during the recent decade regional tropical Indian Ocean Rim circulations during MAM were

123

(b)

modified, reducing moisture fluxes and precipitation over eastern Africa. While the studies of Funk et al (2008) and Williams and Funk (2011) observationally demonstrated the co-variation of eastern Africa precipitation decreases with Indo-west Pacific SST increases over the last decade, the modeling experiments of Lyon and DeWitt (2012) showed that the primary forcing of MAM eastern Africa precipitation decreases since 1998 originated in tropical Pacific SSTs. In the present study, we examine how drought during consecutive MAM and OND seasons are influenced by decadal and multi-decadal SST changes between 1950–2010 over the Pacific Ocean. Furthermore, we examine the interplay between the forcing of drought by Pacific SST on multidecadal time scales and seasonal-to-interannual time scales associated with ENSO and the IOD. Considerable effort has been invested to examine the Indo-Pacific SST influences on east African rainfall during the OND and MAM seasons individually. In this study, the influences of Indo-Pacific SST on eastern Africa drought spanning multiple MAM and OND seasons are tested using ten ensemble simulations of the Global Forecast System (GFS) forced by 1950–2010 observed global SST. Specifically, we examine whether the frequency of drought spanning consecutive MAM and OND seasons forced by SST have increased in recent decades and identify the modes of climate variability, and their interaction, that may be dynamically important. The focus on back-to-back droughts is motivated by a desire to support better drought preparedness and mitigation. In Sect. 3, we examine the seasonal SST-forced precipitation over eastern Africa during OND and MAM. We also investigate whether the frequency of consecutive drought seasons are increasing as forced by SST. In Sect. 4, we examine the decadal and multidecadal variations of global SST and eastern Africa precipitation. In Sect. 5, we examine the dynamical forcing of eastern Africa precipitation by individual Indo-Pacific SST modes, ranging from seasonal to multidecadal time scales. In Sect. 5, we provide a summary and brief discussion.

Indo-Pacific SST and consecutive failed eastern Africa rainy seasons

2 Data and methods 2.1 Observed climate The monthly observed precipitation climatology over eastern Africa (Fig. 1b) was drawn from monthly averages of the 1981–2010 Global Precipitation Climatology Project (GPCP) version 2.2 (Adler et al. 2003) on a fixed 2.5° 9 2.5° latitude–longitude grid. The GPCP dataset incorporates satellite-based precipitation estimates using infrared and microwave channels and global station data. Monthly observed SSTs for 1950–2010 were drawn from the Extended Reconstructed Sea Surface Temperature (ERSST) analysis version 3b (Smith et al. 2008) on a fixed 2.0° 9 2.0° latitude–longitude grid. ERSST version 3b is based upon the International Comprehensive Ocean– Atmosphere Data Set version 2.4 and does not incorporate satellite SST data, as satellite SST data introduce residual biases. SST with the seasonal cycle removed relative to the 1981–2010 period was analyzed. 2.2 Testing the forcing of precipitation by SST The SST influences on eastern Africa precipitation are tested using ten ensemble simulations of the GFS model driven by observed 1950–2010 SST. The average of all ten simulations was calculated prior to all analyses, with the intent of removing influences of internal atmospheric

Fig. 2 a SST-forced GFS simulation MAM and OND percent precipitation departure spatially averaged over eastern Africa (purple box in Fig. 1) for 1950–2010. The maximum number of consecutive dry MAM and OND seasons that occurred within a 5-year window centered around the labeled year for precipitation in a

variability. All departures, whether displayed as an anomaly or as a percent change, are calculated from the 1981–2010 climatology. 2.3 Calculation of statistical significance Statistical significance in all latitude–longitude plots are calculated using either a Student’s t test or a resampling approach. The Student’s t test is utilized to test for significance in regression and correlation diagrams. The resampling approach is utilized to test for significance of temporal averages of SST and precipitation. For the resampling approach, 10,000 composites were constructed from the average number of randomly selected monthly anomalies of the variable in question. The number of months used in each composite is reflected by the length of the analyzed time period. Statistical significance at each grid point was determined from the 10,000 composites. For example, we explain how statistical significance was computed for the average of MAM rainfall during 1950–1979 (Fig. 4a). Twenty-nine MAM seasons, corresponding to the length of the 1950–1978 period, were selected randomly without replacement from the possible 61 MAM seasons in the 1950–2010 record. The 29 random samples were averaged at all grid points, and this process was repeated to generate 10,000 composites. The 10,000 composites were sorted according to rank at all grid points and statistical significance was computed.

(a)

(b)

123

A. Hoell, C. Funk

3 SST-forced consecutive MAM and OND droughts The objective of this section is to examine whether the frequency of SST-forced droughts spanning consecutive MAM and OND seasons over eastern Africa (Fig. 1a) have increased in recent decades. MAM and OND precipitation departures for 1950–2010 are displayed in Fig. 2a. Prior to 1983, almost all MAM and OND seasons were wetter than average. After 1983, there was an abrupt change in the amount of precipitation that fell during MAM and OND, which resulted in a more even distribution between drought and pluvial seasons. As a result of this more even distribution between pluvials and droughts, there was a substantial increase in the frequency of droughts spanning consecutive MAM and OND seasons. The frequency of drought spanning consecutive MAM and OND seasons is assessed by counting the maximum number of consecutive dry MAM and OND seasons within a 5-year window of the plotted year (Fig. 2b)—2 years prior to the year, that year, and two years after the year. In any given year, the maximum possible overlapping drought seasons is 10. Here, dry seasons are defined as having below average precipitation. Prior to 1983, the only consecutive dry MAM and OND seasons occurred in 1959. There was an increase in the number of droughts spanning consecutive MAM and OND seasons in the 1980s. From OND 1983 to OND 1985, there were five consecutive OND and MAM dry seasons, which was followed by three consecutive dry seasons between MAM 1988 and MAM 1989. Fig. 3 Observed SST-forced GFS simulation percent precipitation departure during a MAM and b OND spatially averaged over eastern Africa (purple box in Fig. 1) for 1950–2010

The frequency of drought spanning consecutive MAM and OND seasons increased substantially during the late 1990s, lasting into the 2000s. The most persistent drought conditions occurred between OND 1998 and MAM 2002, during which there were eight consecutive dry MAM and OND seasons. During the mid- and late 2000s, there were two separate occurrences of three consecutive dry MAM and OND seasons. MAM and OND precipitation over eastern Africa for 1950–2010 is displayed in Fig. 3 in an attempt to determine how each season contributed to droughts spanning consecutive MAM and OND seasons. For MAM, precipitation during 1950–2010 over eastern Africa decreased considerably between the beginning and end of the period. While there were clear year-to-year variations in precipitation, the dominant feature throughout this period was the strong decreasing trend. Three precipitation periods are distinguishable during MAM, each receiving less precipitation than the previous period: 1950–1983, 1984–1998 and 1999–2010. Prior to 1983, MAM seasons were wetter than average. For 1984–1998, there was an even mixture of wet and dry MAM seasons. For 1999–2010, almost all MAM seasons were dry. For OND, precipitation during 1950–2010 over eastern Africa varied considerably on interannual time scales associated with ENSO events. For example, precipitation was reduced during the La Nin˜a events of the mid-1980s and around 2000 while precipitation was enhanced during the El Nin˜o events during the mid-1990s. Despite the

(a)

(b)

123

Indo-Pacific SST and consecutive failed eastern Africa rainy seasons Fig. 4 Observed SST-forced GFS simulation average percent precipitation departure for (top row) 1950–1978, (middle row) 1979–1998 and (bottom row) 1999–2010 during (left column) MAM and (right column) OND. The white contour denotes values significant to p \ 0.05

interannual variation of precipitation during OND, the precipitation time series can be separated into two periods: 1950–1978 and 1979–2010. Prior to 1978, OND seasons were predominantly wetter than average. For 1979–2010, there was a more even mixture of wet and dry OND seasons. The preceding analysis shows that the SST-forced precipitation over eastern Africa during both OND and MAM displays strong decreasing trends and considerable decadal to multidecadal variability throughout the 1950–2010 period. The long-term precipitation decreases shown here over eastern Africa forced by SST have been corroborated by the observational of analyses of Williams and Funk (2011) and Lyon and DeWitt (2012) during MAM for the more recent 1979–2010 period. The decreasing trend in eastern Africa precipitation forced by Indo-Pacific SST have contributed to the increased frequency of drought spanning consecutive MAM and OND seasons. Due to the strong coupling of OND precipitation over eastern Africa with ENSO and IOD, there is still considerable interannual variation in precipitation despite the overall trend, which can help to alleviate multi-season droughts during El Nin˜o

(a)

(d)

(b)

(e)

(c)

(f)

events (i.e. the mid-1990s), but exacerbate multi-season droughts during La Nin˜a events (i.e. around 2000).

4 Decadal to multi-decadal SST and precipitation variability Previously, we showed that the SST-forced precipitation over eastern Africa during both MAM and OND of 1950–2010 displays decadal and multidecadal variabilities. The decadal and multidecadal precipitation variability have resulted in an increased frequency of droughts spanning consecutive MAM and OND seasons in recent decades. Here, we examine the spatial patterns of SST-forced precipitation variability over the Indian Ocean Rim and the associated SST patterns on the decadal and multidecadal time scales. Figure 4 shows the average percent precipitation departure over the Indian Ocean Rim for 1950–1978, 1979–1998 and 1999–2010 during MAM and OND. The time periods were chosen because they best capture the decadal to multidecadal trifurcations in eastern Africa

123

A. Hoell, C. Funk Fig. 5 Observed average SST anomaly (°C) for a 1950–1978, b 1979–1998 and c 1999–2010. All plots are significant to p \ 0.05

(a)

(b)

(c)

precipitation displayed in Figs. 2a and 3. White contours denote values significant to p \ 0.05. During MAM, there were significant precipitation differences from the mean during the 1950–1978 and 1999–2010 periods over the Indian Ocean and Indian Ocean Rim, which highlights the strong trend in precipitation throughout the period. During 1950–1978 (Fig. 4a), precipitation was significantly greater than average over the Horn of Africa, extending northwest along the Red Sea and significantly less than average over the tropical central Indian Ocean. Over the Maritime Continent, precipitation was less than average, but these differences were not significant. During 1979–1998 (Fig. 4b), there were no significant precipitation differences, though the pattern closely resembles MAM 1950–1978 (Fig. 4a), but with smaller precipitation magnitudes. During 1999–2010 (Fig. 4c), precipitation was significantly less than average over eastern Africa, extending northeast across the Gulf of Aden into central-southwest Asia. Over the tropical central Indian Ocean, precipitation was significantly greater than average. Elsewhere over the tropical Indian Ocean,

123

precipitation was greater than average to the north and less than average to the south off of the west coast of Australia. During OND, there were significant precipitation differences from the mean during the 1950–1978 period over the Indian Ocean and Indian Ocean Rim. During 1950–1978 (Fig. 4d), precipitation was significantly greater than average over the the Horn of Africa, extending northward across the Red Sea and Arabian Penninsula and was significantly less than average over the tropical central Indian Ocean. OND 1950–1978 and MAM 1950–1978 share very similar average precipitation patterns over the central and western tropical Indian Ocean, eastern Africa and Middle East. During the 1979–1998 (Fig. 4d) and 1999–2010 (Fig. 4f) periods, precipitation departures were small and insignificant. Figure 5 shows the average SST anomalies for 1950–1978, 1979–1998 and 1999–2010, corresponding to the same time periods shown in the average precipitation plots of Fig. 4. We show the average of all calendar months during the aforementioned periods because they closely correspond to the SST patterns averaged during the

Indo-Pacific SST and consecutive failed eastern Africa rainy seasons

individual MAM and OND seasons (not shown). All shaded values are significant to p \ 0.05. The averaged decadal to multidecadal SST patterns of Fig. 5 resemble Pacific Decadal Variability, also known as the Pacific Decadal Oscillation [PDO, Mantua et al (1997)], and a warming of the global oceans throughout the 1950–2010. For 1950–1978 (Fig. 5a), the global oceans were significantly cooler than average, particularly over the Indian Ocean and tropical Pacific Ocean. SST over the north Pacific were significantly warmer than average pattern associated with the PDO. The cool SST signature of the PDO is not readily visible because it is dominated by the SST trend. For 1979–1998 (Fig. 5b), the strong trend in global SST is not readily visible because the average is taken in the middle of the period. However, a strong positive PDO is clear in the extratropical Pacific Ocean. For 1999–2010 (Fig. 5c), the global oceans were significantly warmer than average, especially over the Indo-west Pacific Ocean with weak cooling over the central tropical Pacific. This pattern is very similar to the ENSO-free longterm warming pattern shown by Compo and Sardeshmukh (2009) and Solomon and Newman (2012). Averages in SST-forced precipitation over the Indian Ocean Rim (Fig. 4) and SST (Fig. 5) for 1950–1978, 1979–1998 and 1999–2010 highlight a potential coupling between decadal to multidecadal SST variation and eastern

Fig. 6 Correlation of SSTforced GFS simulation percent precipitation departure spatially averaged over eastern Africa (purple box in Fig. 1) and observed SST for a MAM and b OND. The white contour denotes values significant to p \ 0.05

Africa precipitation. Warming SST over the Indo-west Pacific Oceans is associated with diminished precipitation over eastern Africa, particularly during MAM. Additionally, each of the time periods examined were characterized by SST patterns associated with the PDO. A negative PDO SST pattern occurred in the 1950–1978 and 1999–2010 periods (Fig. 5a, c) while a positive PDO SST pattern occurred in the 1979–1998 period (Fig. 5b).

5 Seasonal precipitation forcing by SST areas 5.1 Identification We identify SST areas likely responsible for forcing precipitation over eastern Africa during MAM and OND. SST areas are identified through the correlation of observed SST-forced precipitation spatially-averaged over eastern Africa (Fig. 1) and global SST for MAM and OND 1950–2010 (Fig. 6). During MAM, the correlation of observed SST-forced precipitation and global SST (Fig. 6a) shares similarities with the average SST pattern of 1999–2010 (Fig. 5c) over the Indo-west Pacific. Furthermore, this correlation pattern resembles the ENSO-free warming pattern of tropical Pacific SST (Compo and Sardeshmukh, 2009; Solomon and

(a)

(b)

123

A. Hoell, C. Funk

(a)

(b)

(c) (d)

(e)

(f)

Fig. 7 a ENSO time series as measured by observed SST averaged over the b Nin˜o3.4 region. c IOD time series as measured by the d Dipole Mode Index. e Observed west Pacific time series as measured by area-averaged SST over 15°S–15°N and 120°E–150°E, the boxed region in f

Newman 2012), also with the sign inverted. Specifically, there are three distinct areas of significant correlations with east African rainfall over the Indo-Pacific: (1) the entire Indo-west Pacific Warm Pool; (2) extratropical north Pacific and (3) extratropical south Pacific. Lyon and DeWitt (2012) found that SSTs over the Pacific Ocean during 1999–2010 (Fig. 5c) were responsible for forcing the recent precipitation declines over eastern Africa during MAM. Here, we show that the only region of SST over the Pacific significantly related with MAM precipitation is over the west Pacific (Fig. 6a). Therefore, in the following analyses, we focus on the influences of SST over the tropical west Pacific Ocean on eastern Africa rainfall. Tropical west Pacific SST is quantified through calculation of the monthly SST anomaly spatially-averaged over 15°S–15°N and 120°E–150°E (black box in Fig. 7f). During OND, the correlation of observed SST-forced precipitation and global SST (Fig. 6b) strongly resembles a canonical ENSO pattern over the tropical Pacific (Rasmusson and Carpenter 1982) and an inverted IOD pattern over the tropical Indian Ocean (Saji et al. 1999). Therefore, we focus on the influences ENSO and the IOD on east African rainfall. ENSO is quantified using the Nin˜o3.4 index of monthly SST anomaly spatially-average over 5°S– 5°N and 170°W–120°W (black box in Fig. 7b). The IOD is quantified by the Dipole Mode Index (Saji et al (1999);

123

black boxes in Fig. 7d), and is calculated as the difference in spatially-averaged monthly SST anomaly over the western Indian Ocean (10°S–10°N; 60°E–80°E) and eastern Indian Ocean (10°S–0°N; 90°E–110°E). 5.2 Temporal SST variability We describe the temporal variability of ENSO, IOD and west Pacific SST during 1950–2010. ENSO (Fig. 7a) varies semi-regularly every 2–7 years between its cold (La Nin˜a) and warm (El Nin˜o) phases. There is a small warming trend in the Nin˜o3.4 index used to measure ENSO, and that warming trend is neglected in this analysis. The IOD (Fig. 7c), which describes the west-to-east gradient across the Indian Ocean basin, varies irregularly on seasonal to interannual time scales. The variation of the total anomalous SST over the west Pacific is shown in Figs. 7e and 8a. Also, west Pacific SST increased steadily throughout 1950–1990, then sharply increased in the late 1990s. Within the increasing longterm trend, there appeared to be regular variability in west Pacific SST on interannual time scales (Wang et al. 1999). We separate the long-term trend from the interannual variability in west Pacific SST. The long-term trend was captured using the 5-year running mean of SST anomaly (Fig. 8b), and displays the steady increase through much of

Indo-Pacific SST and consecutive failed eastern Africa rainy seasons Fig. 8 a Observed west Pacific (WP) SST, which is measured by area-averaged SST over 15°S–15°N and 120°E–150°E, the boxed region in Fig. 7f. b Observed western Pacific SST trend, as captured by a 5-year running mean of the observed western Pacific SST time series in a. c Observed de-trended west Pacific SST, which is calculated as the difference of the western Pacific SST time series (a) and the western Pacific SST trend (b)

(a)

(b)

(c)

the period and a sharp increase since the late 1990s that persisted through the end of the period. The interannual variability was captured by subtracting the long-term trend from the total anomalous SST variability over the west Pacific (Fig. 8c). The interannual component of west Pacific SST and ENSO are inversely related at first glance, but their temporal variabilities differ in some instances, and the magnitudes of west Pacific SSTs are not necessarily related with the magnitude of east-central Pacific SSTs (Hoell and Funk 2013). 5.3 Precipitation forcing We examine the forcing of precipitation and circulation over the Indian Ocean Rim during MAM (Figs. 9, 10) and OND (Figs. 11, 12) by ENSO, the IOD, the long-term trend in west Pacific SST and the interannual component of west Pacific SST. The SST influences on precipitation are tested using ten ensemble simulations of the GFS model driven by observed 1950–2010 SST. The average of all ten

simulations was calculated prior to all analyses, with the intent of removing influences of internal atmospheric variability. 5.3.1 MAM ENSO forces widespread changes in the regional circulation (Figs. 9a, e, 10a), which significantly influences the north to south distribution of precipitation (Fig. 10b) over the Indian Ocean basin. Over southern Asia, ENSO forces a broad anticyclonic circulation at 200 hPa (Fig. 9a), highlighted by anomalous easterly flow over the Arabian Sea. Over eastern Africa, ENSO forces significant low-level circulation modifications, but these circulation modifications do not result in significant precipitation departures. The 850 hPa wind over the western Indian Ocean (Figs. 9b, 10a), near the Horn of Africa, is easterly and northeasterly during the negative phase of ENSO (La Nin˜a), which results in competing moisture flux contributions. Over coastal Somalia, the northeasterly flow provides anomalously dry air, which

123

A. Hoell, C. Funk Fig. 9 Regression of observed SST-forced GFS simulation (left column) 200 hPa wind and (right column) 850 hPa wind to standardized indices of ENSO, the IOD, west Pacific SST trend and detrended west Pacific SST during MAM in units of m s-1 °C-1 per standard deviation

results in slight precipitation decreases. Over coastal Tanzania and Kenya, easterly flow provides anomalously moist air, which results in slight precipitation increases just off of the eastern Africa coast. The IOD forces significant changes to the zonal wind over the Indian Ocean at 850 hPa (Figs. 9f, 10c), but not at 200 hPa (Fig. 9b), which results in significant east-to-west precipitation changes (Fig. 10d) over the Indian Ocean basin and surrounding continental areas. During negative IOD periods, weakened easterly winds across the Indian Ocean result in reduced water vapor transports toward east

123

(a)

(e)

(b)

(f)

(c)

(g)

(d)

(h)

Africa and the adjoining northwest rim of the Indian Ocean, reducing precipitation in this region. The trend component of west Pacific SST forces considerable changes to the tropospheric circulation over the Indian Ocean. The trend in west Pacific SST forces tropospheric circulations consistent with a baroclinic Rossby wave, with anticyclonic circulation at 200 hPa (Fig. 9c) and cyclonic circulation at 850 hPa over southern Asia and the northern Indian Ocean (Fig. 9g). The changes in the regional circulation associated with the trend component of west Pacific SST force significant changes in the wind field

Indo-Pacific SST and consecutive failed eastern Africa rainy seasons Fig. 10 Left column regression of observed SST-forced GFS simulation 850 hPa wind magnitude (shaded) and 850 hPa wind vectors to standardized indices of ENSO, the IOD, west Pacific SST trend and detrended west Pacific SST during MAM. The units of wind magnitude are m s-1 °C-1 per standard deviation and all shaded values are significant to p \ 0.05 according to a Student’s t test. The wind vectors have been normalized. Right column regression of observed SST-forced GFS simulation precipitation to standardized indices of ENSO, the IOD, west Pacific SST trend and detrended west Pacific SST during MAM in units of percent change °C-1 per standard deviation. The white contour denotes values significant to p \ 0.05

at 850 hPa (Fig. 10e) and precipitation (Fig. 10f) over eastern Africa. Easterly flow over the eastern African continent coupled with westerly flow over the western Indian Ocean north of Tanzania results in significant divergence in the 850 hPa wind field (not shown). These circulation modifications in the low-level wind field result in significant precipitation reductions throughout eastern

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

Africa and the Arabian Penninsula on the order of -21 to -28 percent per standard deviation of warming (Fig. 10f) of the west Pacific. The interannual component of west Pacific SST does not force large areas of statistically significant modifications to the 850 hPa wind (Fig. 10g) and precipitation (Fig. 10h) fields over the Indian Ocean basin.

123

A. Hoell, C. Funk Fig. 11 Regression of observed SST-forced GFS simulation (left column) 200 hPa wind and (right column) 850 hPa wind to standardized indices of ENSO, the IOD, west Pacific SST trend and detrended west Pacific SST during OND in units of m s-1 °C-1 per standard deviation

5.3.2 OND ENSO and IOD force very similar basin-wide circulation (Figs. 11, 12a, c) and precipitation (Fig. 12b, d) modifications over the Indian Ocean and surrounding continental areas during OND. Both ENSO and the IOD force tropospheric circulations consistent with a baroclinic Rossby wave, with anticyclonic circulation at 200 hPa and cyclonic circulation at 850 hPa over southern Asia and the northern Indian Ocean. Over the eastern Indian Ocean, lower tropospheric cyclonic circulation develops during the negative phase of ENSO (La Nin˜a) and the negative phase of the IOD, which drives strong westerly flow throughout the equatorial Indian Ocean. Over the northwest region of

123

(a)

(e)

(b)

(f)

(c)

(g)

(d)

(h)

the Indian Ocean Rim, the anomalous lower tropospheric circulation during the negative phases of ENSO (La Nin˜a) and the IOD originates over central-southwest Asia, blows across the Arabian Peninsula into eastern Africa and then eastward into the western Indian Ocean. The anomalous regional circulation reduces the amount of moisture, thereby reducing precipitation over eastern Africa, the Arabian Peninsula and central-southwest Asia. These circulation and precipitation modification patterns compare well with the observed climate during individual La Nin˜a events (Hoell et al. 2013). The interannual component of west Pacific SST drives similar, yet weaker, basin-wide circulation (Fig. 11d, h) and precipitation (Fig. 12h) modifications over the Indian

Indo-Pacific SST and consecutive failed eastern Africa rainy seasons Fig. 12 Left column regression of observed SST-forced GFS simulation 850 hPa wind magnitude (shaded) and 850 hPa wind vectors to standardized indices of ENSO, the IOD, west Pacific SST trend and detrended west Pacific SST during OND. The units of wind magnitude are m s-1 °C-1 per standard deviation and all shaded values are significant to p \ 0.05 according to a Student’s t test. The wind vectors have been normalized. Right column Regression of observed SST-forced GFS simulation precipitation to standardized indices of ENSO, the IOD, west Pacific SST trend and detrended west Pacific SST during OND in units of percent change °C-1 per standard deviation. The white contour denotes values significant to p \ 0.05

Ocean Rim to ENSO and IOD. Significant changes in the 850 hPa circulation over central-southwest Asia, the Arabian Peninsula and eastern Africa are related with the interannual component of west Pacific SST, and these changes drive significant precipitation modifications over the aforementioned regions. However, significant precipitation changes are not widespread over eastern Africa. While the trend component of west Pacific SST during OND forces anticylonic circulation over southern Asia at 200 hPa, the trend component is not associated with significant precipitation (Fig. 12f) and 850 hPa circulation modifications (Fig. 12e). While the trend component of

(a)

(b)

(c)

(d)

(e)

(f)

(g)

(h)

west Pacific SST is related with precipitation reductions over the Arabian Peninsula and central-southwest Asia, only a very small region experiences significant precipitation changes and a result of insignificant circulation modifications.

6 Summary and discussion Drought conditions during consecutive MAM and OND seasons over eastern Africa substantially increase the potential for widespread famine, an important contributor

123

A. Hoell, C. Funk

to the humanitarian crisis over the Horn of Africa and surrounding nations during 2011 (FEWSNET 2011; Hillbruner and Moloney 2012). In this light, we examined how the interplay of seasonal-to-interannual and multidecadal SST variation over the Indo-Pacific Ocean forced drought during consecutive MAM and OND seasons over eastern Africa for the period of 1950–2010. The forcing of eastern Africa precipitation by Indo-Pacific SST was tested using ten ensemble simulations of the GFS weather forecast model forced by 1950–2010 observed global SST. We examined the average of all ten simulations, which essentially removed the influences of internal atmospheric variabilities from our analyses. Indo-Pacific SST have forced more frequent droughts spanning consecutive MAM and OND seasons over eastern Africa since the 1980s (Fig. 2). Prior to 1980, SST throughout the tropical Indo-Pacific Ocean were significantly cooler than average (Fig. 5a), which forced significant precipitation surpluses relative to average during MAM and OND over central-southwest Asia, the Arabian Peninsula and eastern Africa (Fig. 4a). Over eastern Africa, consecutive dry MAM and OND seasons were very infrequent, only occurring twice near 1960 (Fig. 2b). Between 1980 and 1998, SSTs throughout the tropical Indo-Pacific Ocean were near average (Fig. 5b), which forced near average precipitation during MAM and OND over the Indian Ocean Rim (Fig. 4b). Since precipitation during MAM and OND varied between dry and pluvial conditions during this period, it appears that the warming trend in Indo-Pacific SST forced more frequent consecutive MAM and OND droughts. After 1998, SST throughout the tropical Indo-Pacific Ocean warmed rapidly (Fig. 5c), particularly over the tropical west Pacific (Fig. 8) and forced significant drying over eastern Africa, the Middle East and central Asia during MAM. For the 13 MAM between 1998–2010 (Fig. 3a), precipitation was below average during 10 seasons, and was the primary reason behind the increased trend in consecutive MAM and OND season droughts (Fig. 2b). During OND, the trend was for increased drying, but this drying trend was not statistically significant. The period-averaged SST prior to 1980, for 1980–1998 and after 1998 (Fig. 5) implies a possible relationship between Pacific Decadal SST variability and Indian Ocean Rim rainfall. During each of these periods, averaged SST over the North Pacific Ocean resemble patterns consistent with the PDO (Mantua et al. 1997), which occurred contemporaneously with significant decadal precipitation variability over portions of the Indian Ocean Rim. It is impossible from this analysis to conclude whether Pacific Decadal SST Variability is a driver of precipitation variability over the Indian Ocean Rim due to the dominant effects of the long-term tropical Indo-west Pacific

123

warming. However, the results presented here indicate the need to further investigate the the links between Pacific Decadal Variability and Indian Ocean Rim climate. Consistent with previous works, precipitation over eastern Africa is significantly related to ENSO (Ogallo 1988; Hastenrath et al. 1993; Nicholson and Kim 1997; Nicholson and Selato 2000), the IOD (Abram et al. 2008) and west Pacific SST during OND (Fig. 6a) and to Indowest Pacific Warm Pool SST (Williams and Funk 2011; Lyon and DeWitt 2012; Tierney et al. 2013) during MAM (Fig. 6b). The results presented here suggest three epochs. Between 1950–1978 and 1979–1998, the Indian Ocean warmed substantially (Fig. 5) and rainfall increased (decreased) over the ocean (East Africa) in a manner that may be consistent with Funk et al (2008). Between 1979–1998 and 1999–2011, warming in the western Pacific appears to be the primary forcing component (Lyon and DeWitt 2012). SST over the tropical west Pacific Ocean varies on two time scales, the first is a long-term warming trend and the second is an interannual component (Fig. 8a). We isolated the long-term warming trend of tropical west Pacific through a calculation of the 5-year running mean (Fig. 8b). The interannual component of tropical west Pacific SST (Fig. 8c) was obtained by subtracting the long-term west Pacific SST trend from the total west Pacific SST variability. We examined the dynamical forcing of eastern Africa precipitation during MAM and OND by the interannual and long-term trend components of west Pacific SST, the IOD and ENSO. Each of the aforementioned modes of SST variability influence Indian Ocean Rim circulations and precipitation similarly, but with varying levels of statistical significance. The regional lower tropospheric circulation is cyclonic over eastern Africa, the Middle East and centralsouthwest Asia, which draws moisture away from these regions thereby reducing precipitation (Figs. 10, 12). While the long-term MAM precipitation trend is decreasing, and has contributed to more frequent consecutive MAM and OND season droughts, the IOD and ENSO can still overshadow these long-term trends and either alleviate or enhance consecutive droughts. This study, along with the recent studies of Funk et al. (2008), Williams and Funk (2011), Lyon and DeWitt (2012) and Tierney et al. (2013), highlight the importance of the Indo-west Pacific in forcing eastern Africa drying during MAM. While the aforementioned studies have identified important causal relationships, the sensitivity of Indian Ocean Rim climate to a warming Indo-west Pacific Warm Pool remains unknown. As a community, we have yet to concretely quantify how changes in the pattern and magnitude of Indo-west Pacific warming influence the circulation and precipitation over the Indian Ocean Rim.

Indo-Pacific SST and consecutive failed eastern Africa rainy seasons

Furthermore, we have yet to fully understand the relative influences of the eastern Indian Ocean and the west Pacific Ocean SST to decadal precipitation variability over eastern Africa during MAM. To better understand the sensitivity of Indian Ocean Rim climate to changes over the west Pacific, we need to perform more atmospheric modeling experiments with a variety of climate models forced by realistic SST patterns with various SST magnitudes. The future changes of tropical Pacific SST and the subsequent forcing of Indian Ocean Rim precipitation is also very important. Unfortunately, the current generation of coupled climate models have projected tropical IndoPacific SST to change in a different way (Yeh et al. 2012) than the observed long-term trend identified by recent studies (Compo and Sardeshmukh 2009; Solomon and Newman 2012). The current generation of climate models indicate a uniform warming of Indo-Pacific SST while the observed long-term trend in the absence of ENSO prefers warming over the west Pacific and slight cooling over the central Pacific. Therefore, the coupled model projections of Indian Ocean Rim circulations and rainfall may be considerably different than what may actually occur (L/’Heureux et al. 2013). Acknowledgments The authors would like to thank two anonymous reviewers whose comments and suggestions helped to improve the manuscript. This research builds upon a multi-year research project carried out under a U.S. Agency for International Development-funded Famine Early Warnings System Network agreement with the U.S. Geological Survey.

References Abram NJ, Gagan MK, Cole JE, Hantoro WS, Mudelsee M (2008) Recent intensification of tropical climate variability in the indian ocean. Nat Geosci 1(12):849–853. doi:10.1038/ngeo357 Adler RF, Huffman GJ, Chang A, Ferraro R, Xie PP, Janowiak J, Rudolf B, Schneider U, Curtis S, Bolvin D, Gruber A, Susskind J, Arkin P, Nelkin E (2003) The version-2 global precipitation climatology project (GPCP) monthly precipitation analysis (1979 present). J Hydrometeorol 4(6):1147–1167. doi:10.1175/ 1525-7541(2003)004\1147:TVGPCP[2.0.CO;2 Black E, Slingo J, Sperber KR (2003) An observational study of the relationship between excessively strong short rains in coastal east africa and indian ocean SST. Mon Weather Rev 131(1):74–94 doi:10.1175/1520-0493(2003)131\0074:AOSOTR[ 2.0.CO;2 Camberlin P, Philippon N (2002) The east african March - May rainy season: Associated atmospheric dynamics and predictability over the 1968–97 period. J Clim 15(9):1002–1019. doi:10.1175/15200442(2002)015\1002:TEAMMR[2.0.CO;2 Compo GP, Sardeshmukh PD (2009) Removing ENSO-Related variations from the climate record. J Clim 23(8):1957–1978. doi:10.1175/2009JCLI2735.1 FEWSNET (2011) Past year one of the driest on record in the eastern horn, famine early warning system network report. Tech. rep., U.S. Agency for International Development, Washington, DC

Funk C, Dettinger MD, Michaelsen JC, Verdin JP, Brown ME, Barlow M, Hoell A (2008) Warming of the indian ocean threatens eastern and southern african food security but could be mitigated by agricultural development. Proc Natl Acad Sci 105(32):11081– 11086. http://www.pnas.org/content/105/32/11081.abstract Haile M (2005) Weather patterns, food security and humanitarian response in sub-saharan africa. Philos Trans R Soc B Biol Sci 360(1463):2169–2182. http://rstb.royalsocietypublishing.org/ content/360/1463/2169.abstract Hastenrath S, Nicklis A, Greischar L (1993) Atmospheric-hydrospheric mechanisms of climate anomalies in the western equatorial indian ocean. J Geophys Res 98(C11):20219–20235. doi:10.1029/93JC02330 Hillbruner C, Moloney G (2012) When early warning is not enough lessons learned from the 2011 somalia famine. Special Issue on the Somalia Famine of 2011–2012. Glob Food Secur 1(1):20–28 doi:10.1016/j.gfs.2012.08.001 Hoell A, Funk C (2013) The ENSO-related west pacific sea surface temperature gradient. J Clim. doi:10.1175/JCLI-D-12-00344.1 Hoell A, Funk C, Barlow M (2013) The regional forcing of northern hemisphere drought during recent warm tropical west Pacific Ocean la nina events. Clim Dyn. doi:10.1007/s00382-0131799-4 Indeje M, Semazzi FH, Ogallo LJ (2000) ENSO signals in east african rainfall seasons. Int J Climatol 20(1):19–46. doi:10.1002/ (SICI)1097-0088(200001)20:1\19::AID-JOC449[3.0.CO;2-0 L/’Heureux ML, Lee S, Lyon B (2013) Recent multidecadal strengthening of the walker circulation across the tropical pacific. Nat Clim Change. doi:10.1038/nclimate1840 Lyon B, DeWitt DG (2012) A recent and abrupt decline in the east african long rains. Geophys Res Lett 39(2):L02702. doi:10.1029/ 2011GL050337 Mantua NJ, Hare SR, Zhang Y, Wallace JM, Francis RC (1997) A pacific interdecadal climate oscillation with impacts on salmon production. Bull Am Meteorol Soc 78(6):1069–1079. doi:10. 1175/1520-0477(1997)078\1069:APICOW[2.0.CO;2 Nicholson S, Selato J (2000) The influence of La Nina on african rainfall. Int J Climatol 20(14):1761–1776. doi:10.1002/10970088(20001130)20:14\1761::AID-JOC580[3.0.CO;2-W Nicholson SE, Kim J (1997) The relationship of the El nino-southern oscillation to african rainfall. Int J Climatol 17(2):117–135. doi:10.1002/(SICI)1097-0088(199702)17:2\117::AID-JOC84[ 3.0.CO;2-O Ogallo LJ (1988) Relationships between seasonal rainfall in east africa and the southern oscillation. J Climatol 8(1):31–43. doi:10.1002/joc.3370080104 Rasmusson EM, Carpenter TH (1982) Variations in tropical sea surface temperature and surface wind fields associated with the southern Oscillation/El nino. Mon Weather Rev 110(5):354–384. doi:10.1175/1520-0493(1982)110\0354:VITSST[2.0.CO;2 Saji NH, Goswami BN, Vinayachandran PN, Yamagata T (1999) A dipole mode in the tropical indian ocean. Nature 401(6751): 360–363 Smith TM, Reynolds RW, Peterson TC, Lawrimore J (2008) Improvements to NOAAs historical merged land–ocean surface temperature analysis (1880–2006). J Clim 21(10):2283–2296. doi:10.1175/2007JCLI2100.1 Solomon A, Newman M (2012) Reconciling disparate twentiethcentury Indo-Pacific ocean temperature trends in the instrumental record. Nat Clim Change 2(9):691–699. doi:10.1038/ nclimate1591 Tierney JE, Smerdon JE, Anchukaitis KJ, Seager R (2013) Multidecadal variability in east African hydroclimate controlled by the Indian ocean. Nature 493(7432):389–392. doi:10.1038/ nature11785

123

A. Hoell, C. Funk Verdin J, Funk C, Senay G, Choularton R (2005) Climate science and famine early warning. Philos Trans R Soc B Biol Sci 360(1463): 2155–2168. http://rstb.royalsocietypublishing.org/content/360/ 1463/2155 Wang C, Weisberg RH, Virmani JI (1999) Western pacific interannual variability associated with the El nio-southern oscillation. J Geophys Res 104(C3):5131–5149. doi:10.1029/1998JC900090 Washington R, Downing TE (1999) Seasonal forecasting of african rainfall: prediction, responses and household food security. Geogr J 165(3):255–274. doi:10.2307/3060442. ArticleType:

123

research-article / Full publication date: Nov., 1999 / Copyright 1999 The Royal Geographical Society (with the Institute of British Geographers) Williams A, Funk C (2011) A westward extension of the warm pool leads to a westward extension of the walker circulation, drying eastern africa. Clim Dyn 37(11–12):2417–2435. doi:10.1007/ s00382-010-0984-y Yeh SW, Ham YG, Lee JY (2012) Changes in the tropical pacific SST trend from CMIP3 to CMIP5 and its implication of ENSO*. J Clim 25(21):7764–7771. doi:10.1175/JCLI-D-12-00304.1