A westward extension of the warm pool leads to a westward extension of the Walker circulation, drying eastern Africa
- First Online:
- Received:
- Accepted:
DOI: 10.1007/s00382-010-0984-y
- Cite this article as:
- Williams, A.P. & Funk, C. Clim Dyn (2011) 37: 2417. doi:10.1007/s00382-010-0984-y
- 136 Citations
- 3.6k Downloads
Abstract
Observations and simulations link anthropogenic greenhouse and aerosol emissions with rapidly increasing Indian Ocean sea surface temperatures (SSTs). Over the past 60 years, the Indian Ocean warmed two to three times faster than the central tropical Pacific, extending the tropical warm pool to the west by ~40° longitude (>4,000 km). This propensity toward rapid warming in the Indian Ocean has been the dominant mode of interannual variability among SSTs throughout the tropical Indian and Pacific Oceans (55°E–140°W) since at least 1948, explaining more variance than anomalies associated with the El Niño-Southern Oscillation (ENSO). In the atmosphere, the primary mode of variability has been a corresponding trend toward greatly increased convection and precipitation over the tropical Indian Ocean. The temperature and rainfall increases in this region have produced a westward extension of the western, ascending branch of the atmospheric Walker circulation. Diabatic heating due to increased mid-tropospheric water vapor condensation elicits a westward atmospheric response that sends an easterly flow of dry air aloft toward eastern Africa. In recent decades (1980–2009), this response has suppressed convection over tropical eastern Africa, decreasing precipitation during the ‘long-rains’ season of March–June. This trend toward drought contrasts with projections of increased rainfall in eastern Africa and more ‘El Niño-like’ conditions globally by the Intergovernmental Panel on Climate Change. Increased Indian Ocean SSTs appear likely to continue to strongly modulate the Warm Pool circulation, reducing precipitation in eastern Africa, regardless of whether the projected trend in ENSO is realized. These results have important food security implications, informing agricultural development, environmental conservation, and water resource planning.
Keywords
East Africa Indian Ocean Precipitation Drought Tropical warm pool Climate change1 Introduction
An estimated 17.5 million people are food insecure in Kenya, Ethiopia, and Somalia, and the US government has spent over $1.1 billion on food aid in these countries since 2009.1 Food balance modeling suggests that this insecurity stems (in part) from stagnating agricultural development, population growth, and recent drought (Funk et al. 2008; Funk and Brown 2009); which has been linked to human-caused warming in the Indian Ocean (Funk et al. 2005, 2008; Verdin et al. 2005). This warming appears to have had a large impact on eastern African rainfall from March to June (MAMJ, Funk et al. 2008; Funk and Verdin 2009). This season is known as the ‘long rains’ in Kenya and the ‘Belg’ rains in Ethiopia. In this paper, we show that a suite of observational datasets indicate a westward extension of the tropical warm pool into the Indian Ocean during MAMJ; this extension appears to be extending the zonally overturning atmospheric Walker circulation in a westward direction. While there appear to be many factors that govern interannual variability in east African long-rains precipitation, convective activity during MAMJ has steadily declined in eastern Africa for the past 30 years as the convective branch of the Walker circulation has become more active over the Indian Ocean.
Diagnoses of the reasons behind the declining long-rains precipitation have been elusive because the factors governing the interannual variability of long-rains precipitation totals are complicated and not well understood (Ogallo 1988; Camberlin and Philippon 2002; Pohl and Camberlin 2006b). While short-rains (October–December) precipitation is strongly influenced by the ENSO phenomenon (Ogallo 1988; Hastenrath et al. 1993; Nicholson and Kim 1997; Nicholson and Selato 2000) and the Indian Ocean Dipole (IOD, Saji et al. 1999; Abram et al. 2008), long-rains precipitation does not correlate well with ENSO or IOD indices (Pohl and Camberlin 2006b). Correlations are also weak between long rains and SST in any one region (Ogallo 1988; Rowell et al. 1994).
The poor relationships with SST patterns suggest that interannual variability in long-season precipitation must be associated with internal atmospheric variability (Camberlin and Philippon 2002). In search of an atmospheric mechanism influencing interannual variability of long rains, Pohl and Camberlin (2006a, b) implicated the Madden-Julian Oscillation (MJO, Madden and Julian 1971, 1972, 1994), where wet conditions during the early part of the long rains are associated with high-amplitude MJO conditions. Key characteristics of high-amplitude MJO conditions during MAMJ are westerly wind anomalies throughout the atmospheric column above tropical Africa and overturning circulation anomalies over the tropical Indian Ocean with westerly anomalies in the upper troposphere, downward anomalies over the Maritime Continent, and easterly anomalies in the lower troposphere. These conditions presumably allow for increased westerly moisture transports across the interior of Africa, somewhat increased easterly transports from the Indian Ocean, and reduced subsidence over eastern Africa (Pohl and Camberlin 2006a). Curiously, however, the statistical correlation between MJO amplitude and long-rains precipitation lost considerable strength after the 1980s (Pohl and Camberlin 2006b).
More recently, Funk et al. (2008) suggested that long-established interannual relationships between large-scale climate variability and long-rains precipitation are being altered due to warming within the south central Indian Ocean. Funk et al. suggest that warming near the western edge of the tropical Pacific-Indian Ocean warm pool tends to increase deep atmospheric convection over the Indian Ocean. This increased convection over the Indian Ocean is associated with increased subsidence and reduced rainfall across eastern Africa. Variations in temperature and convection in the Indian Ocean region modulate, and are modulated by, the tropical Pacific Walker Cell, which is characterized by low-level easterly wind flow from the relatively cool eastern equatorial Pacific toward the warmer western equatorial Pacific and return flow within the upper troposphere. Gill (1980, 1982) shows that this very large circulation can be considered as a steady-state solution to the shallow water wave equations when forced with a diabatic heating (latent heat release) term centered over the warm pool. This solution has two components, a long eastward Kelvin wave response, associated with the easterly Pacific tradewinds, and a shorter Rossby wave response that brings dry air down across eastern Africa. Indeed, the interannual analysis by Pohl and Camberlin (2006b) found that, from 1979 to 2001, anomalously high convective activity over the center of the warm-pool region (Maritime Continent) was associated with reduced convective activity over eastern Africa during the long-rains season. Thus, a change in the behavior of the western, ascending branch of the Walker cell would presumably initiate a change in the mean vertical atmospheric movement over eastern Africa.
At present, there is great interest and debate regarding how continued global warming would impact the Pacific Walker Cell, and its principal interannual variation (ENSO). One reason for uncertainty is the fact that anthropogenic changes in the Pacific climate will be intermingled with natural internal climate variations. These shifts vary on ~20-years time scales, and a large change before and after the 1970s has been well-documented (Graham 1994; Zhang et al. 1997; Kirtman and Schopf 1998). All models used in the 21st century climate model experiments performed for the Intergovernmental Panel on Climate Change (IPCC) 4th Assessment Report (AR4) predict that tropospheric warming will lead to a weakening of tropical circulation (Held and Soden 2006; Vecchi and Soden 2007). In short, this is because relatively rapid warming in the middle and upper troposphere are projected to cause tropospheric water vapor content to increase substantially more than precipitation within the tropics. This imbalance is expected to slow the overall vertical mass-movement of the tropical troposphere (Knutson and Manabe 1995; Held and Soden 2006).
Models generally project that decreased vertical circulation within the tropical atmosphere should weaken the tropical Walker circulation (Knutson and Manabe 1995; Held and Soden 2006; Vecchi et al. 2006; Lu et al. 2007). A weakened Walker circulation is projected to resemble a more “El Niño-like” climate with decreased convection over the western tropical Pacific and decreased subsidence over the eastern tropical Pacific and eastern tropical Africa (Vecchi and Soden 2007). Notably, the weakened, more “El Niño-like” Walker circulation only describes the mean state of the projected Walker circulation, not the proportion of El Niño events compared to La Niña events (Solomon et al. 2007; Vecchi and Soden 2007). It is also important to note that the IPCC AR4 presents a mixed view on changes in ENSO (Christiansen et al. 2007; Meehl et al. 2007), suggesting that substantial uncertainty remains in how precipitation is likely to change in locations where precipitation is impacted by the state of the ENSO cycle.
Forecasts of decreased tropical Walker circulation are at odds with several observational studies that noted trends indicating an intensified Walker circulation during the latter half of the 20th century. For example, Minobe (2005) found increased convection over the Maritime Continent and increased downward vertical anomalies over the central equatorial Pacific Ocean from 1948 through 2002 in NCEP/NCAR reanalysis wind divergence data. Compo and Sardeshmukh (2010) removed the impact of ENSO from the HadISST record of sea surface temperature (SST) anomalies and found that the remaining (ENSO-unrelated) component of the observed trend from 1949 to 2006 is above 1.2°C/century throughout much of the warm-pool region, and below −0.4°C/century in much of the cold-tongue region that extends from the South American equatorial Pacific coast westward along the equator. Consistent with this study, Cane et al. (1997) observed 20th century cooling in the cold-tongue region and a strengthening of the zonal SST gradient between the cold tongue and the warm pool. These studies suggest that in the absence of ENSO variability, there has been a multi-decade intensification of surface easterlies across the equatorial Pacific Ocean, in agreement with an intensified Walker circulation, as suggested by the modeling study presented by Clement et al. (1996). Considering more recent time periods, Chen et al. (2002) analyzed satellite imagery from 1985 through 2000 and found that long-wave radiation emitted to space decreased in areas associated with tropical Hadley and Walker convection, and long-wave emissions increased in areas associated with subsidence. Chen et al. invoked an intensification of Hadley and Walker circulations, concluding that outgoing terrestrial radiation increased where there was decreased cloud cover and conversely, decreased where there was increased cloud cover. Quan et al. (2004) corroborated these results using a modeling approach; when forced with observed monthly SSTs from 1950 through 1999, the ECHAM-3 GCM simulated increased precipitation over the Indian Ocean, and no change in precipitation over the equatorial central Pacific Ocean. All these studies suggest the Walker circulation is not weakening.
In this study, we use principal component analysis (PCA) of a suite of climate datasets to provide evidence for a westward extension, not a weakening, of the tropical Walker circulation during east Africa’s long rains season. We propose that this westward extension is the result of an extension of the western edge of the tropical warm pool in the Indian Ocean. We contrast this observational trend with Walker-circulation trends simulated by an ensemble of GCMs. Finally, we suggest that the westward extension of Walker circulation has likely contributed to increased subsidence and decreased MAMJ rainfall in eastern Africa. This result is contrary to the AR4 findings, for which 18 of 21 models evaluated for the IPCC AR4 agree upon increased precipitation in eastern Africa (Christiansen et al. 2007). The IPCC AR4 therefore states that the increased precipitation is ‘likely’. Presenting a countervailing view, Funk and Brown (2009) showed that when constrained by observed SSTs from 1980 through 2000, rainfall estimates by an ensemble of 10 GCMs used for the IPCC AR4 were actually anti-correlated with observed long-season rainfall totals in eastern Africa. Nonetheless, the projections of increased rainfall in eastern tropical Africa are taken seriously by food aid agencies, and hundreds of millions of dollars in development aid for agriculture, environmental conservation, and water resource planning are predicated on such uncertain IPCC projections. If the climate continues to tilt toward an intensified Walker circulation, or a westward extension of its western branch, rainfall should continue to decrease in the most food insecure region of the world.
2 Materials and methods
The purpose of this study is to marshal numerous data sources to evaluate two important scientific questions: what drives interannual variations in the Indo-Pacific warm pool region, and how do these variations impact rainfall in eastern Africa? We pursue the first question via (1) a principal component analysis of Indo-Pacific reanalysis fields, followed by (2) a corresponding evaluation of additional climate observations and climate change simulations. Having identified a dominant mode of warming-related Indo-Pacific climate variation, we then proceed to link this mode with station-based observations of declining rainfall in eastern Africa, and variations of the 500 hPa Meridional Height Gradient (MHG) across the Indian Ocean.
2.1 Principal component analysis of reanalysis fields
PCA is a useful way to identify climate processes and trends that are coherent among many datasets (von Storch and Zwiers 1999). We used PCA of tropical reanalysis data to identify and summarize the two primary modes of interannual MAMJ variability in the tropical tropospheric circulation overlying the tropical (20°S–20°N) Indian and Pacific Oceans (55°E–140°W). Extensive sensitivity testing indicated that the results of the PCA were not strongly impacted by the location of the eastern boundary of this region (i.e., the amount of the central and eastern tropical Pacific included). Four variables were analyzed: 2 m air temperature, precipitation totals, zonal wind velocity, and vertical velocity (omega). Temperature and precipitation datasets were surface grids between 20°S and 20°N, 55°E and 140°W. Zonal and vertical velocity datasets were also between 55°E and 140°W, but these data were three dimensional with 17 and 12 geopotential height levels, respectively. We converted these to two dimensional vertical profile fields by averaging values between 10°S and 10°N.
We converted records of temperature, zonal wind velocity, and vertical velocity into z-scores (by subtracting the mean and dividing by the standard deviation), and records of precipitation totals to anomalies. Sensitivity tests indicated that the results of the PCA were not impacted substantially by the use of 2-m temperature over SST, nor by the use of precipitation anomalies rather than z-scores. They were also not impacted substantially by the inclusion or exclusion of precipitation data. This indicates that the results of the PCA are robust despite the notoriously high uncertainty of precipitation estimates during the pre-satellite era (pre-1979). The PCA was then done in two steps. In the first step, we calculated four independent PCAs based respectively on the air temperature, zonal wind, vertical velocity, and rainfall data. For each variable, we then identified the three principal component (PC) time series sharing the most variability with the original data. These three time-series captured 42–68% of the variance among the four variables. We then combined these three time series from each variable into one matrix (12 variables by 62 years) and weighted the mean and variance of each time series by the fraction of variability that it shared with its original data. We conducted a second PCA on these 12 variance adjusted PC time series. This produced 12 new PC time series, each representing some component of the large-scale, multi-variate climate of the tropical Indian and Pacific Ocean area.
For each variable, we calculated a correlation map showing the spatial relationships between the original climate data and each of the first two new PC time series. We interpreted these correlation maps subjectively and determined how each of the first two PC time series related to well known large-scale climate phenomena.
We tested the robustness of the results of the PCA by repeating this procedure using three alternate climate datasets and comparing the resultant PC time series to those calculated using the NCEP/NCAR reanalysis data. In one repetition, we used gridded reanalysis data for the same four variables, produced by the European Centre for Medium-Range Weather Forecasts (ECMWF) ERA-40 project (accessed online at http://www.data.ecmwf.int/data/). This dataset only extended from 1958 through 2002. In the second repetition of the PCA, we used radiosonde data from 1979 through 2009 from 19 countries that are located within the tropical Indian-Pacific Ocean area of 20°S–20°N, 55°E–140°W. In this analysis, only surface temperature and zonal wind-profile data were used. See the appendix for further description of the radiosonde data. Finally, for a more direct comparison between the radiosonde and NCEP/NCAR data, we repeated the PCA on NCEP/NCAR data using only the same variables and temporal coverage as were used in the radiosonde analysis. The three climate datasets used in this analysis are not independent, as radiosonde measurements are used to constrain the climate models used to produce both reanalysis datasets.
To further test that the PC time series represent true climate processes and not artifacts of reanalysis and/or radiosonde data, we considered five new global precipitation products that are based upon satellite and/or station data. These products are the (1) Global Precipitation Climatology Project (GPCP) version 2.1 combined precipitation dataset (Huffman et al. 1997, 2009; Adler et al. 2003), (2) NCEP-DOE Reanalysis 2 (NCEP II, Kanamitsu et al. 2002), (3) CPC Merged Analysis of Precipitation (CMAP, Xie and Arkin 1997), (4) CPC Precipitation Reconstruction (CPC PREC, ftp.cpc.ncep.noaa.gov/precip/50yr/land_ocean), and (5) NOAA merged precipitation reconstruction (MergPr, Smith et al. 2010).
2.2 Comparison of reanalysis and modeled climate data
A primary goal of this study was to identify whether variability in the global mean surface temperature has impacted the Walker-like (zonal) circulation of the tropical atmosphere overlying the Indian and Pacific Oceans. Because changes in the Walker circulation should be accompanied by changes in the hydrological cycle throughout the tropics (Diaz and Markgraf 1992; Webster 1998), we wished to determine whether observed changes in the Walker circulation are represented among the 21 GCMs used by the IPCC AR4 to develop projections of regional changes in rainfall. To do this, we compared observational records of MAMJ climate data to 84 runs of modeled 20th century climate data produced by 21 GCMs as part of phase 3 of the Coupled Model Intercomparison Project (CMIP3) 20th century climate experiment (20c3 m) for the IPCC AR4. These models are described by Randall et al. (2007) and listed in the legend of Fig. 5. All model data were provided by the World Climate Research Programme Multi-Model Database (https://www.esg.llnl.gov:8443/about/overviewPage.do).
Model runs within the CMIP3 20c3m were driven entirely by external forcing such as solar insolation and the greenhouse effect. This means that the timing of high-frequency variability events (e.g., ENSO cycles) within 20c3m records do not necessarily agree with the observational record. All model runs do, however, simulate a 20th century increase in global mean tropospheric temperature. Within each modeled and observational record, we evaluated how specific characteristics of the MAMJ Indian-Pacific Walker circulation are related to global mean surface temperature. We then compared the modeled and observed relationships between MAMJ Walker circulation and global mean temperature.
Reanalysis variables evaluated, the geographic regions that they were evaluated in, and the anomalies expected in the case of intensified Walker circulation
Variable | Region | Region boundaries | Walker anomaly |
|---|---|---|---|
1. 2 m MAMJ temperature | 1 | 10°N–10°S, 180°E–140°W | Cool |
2 | 10°N–10°S, 55°E–120°E | Warm | |
2. Zonal wind velocity | 1 | 10°N–10°S, 90–180°E, 400–100 hPa | Positive |
2 | 10°N–10°S, 90–180°E, 1,000–750 hPa | Negative | |
3. Vertical velocity | 1 | 10°N–10°S, 180°E–140°W, 850–200 hPa | Downward |
2 | 0°S–15°S, 60°E–90°E, 850–200 hPa | Upward | |
4. Precipitation total | 1 | 10°N–10°S, 180°E–140°W | Low |
2 | 0°S–15°S, 60°E–90°E | High |
For temperature, we compared models to several gridded interpolations of observational records. These were the NCEP/NCAR reanalysis of 2 m temperature, NOAA extended SST reanalysis, Kaplan SST reanalysis, GISS surface temperature analysis, ECMWF ERA-40, and Hadley Centre’s CRU Air and Marine Temperature Anomalies (HADCRUT3) dataset. For zonal wind velocity, vertical velocity, and precipitation, we compared models to the NCEP/NCAR reanalysis and ECMWF ERA-40 datasets. All observational data besides the ECMWF ERA-40 covered the years 1948 through 2009. The ECMFW ERA-40 ranged from 1958 through 2002. All observational data besides the ECMWF ERA-40 were provided by the same source as the NCEP/NCAR reanalysis: the NOAA Earth System Research Laboratory (ESRL) Physical Sciences Division (PSD), Boulder, Colorado, USA, from their website (http://www.esrl.noaa.gov/psd).
To reduce the impact of interannual ENSO variability on Walker circulation and global temperatures, we worked with time series smoothed with a 5-years running mean. This isolated lower-frequency covariant within the datasets related to increasing global temperatures. For each modeled and observational dataset, we regressed smoothed data against the corresponding smoothed MAMJ mean global surface temperature (the NASA GISS land/sea air temperature estimates). We then calculated regional averages of the slope coefficients for each variable and region listed above. The temporal coverage among models and observational records varied between datasets, as did the long-term means and variability within the data. We therefore standardized all model and observational data and re-expressed them to have the same means and standard deviations as the NCEP/NCAR reanalysis data.
2.3 High quality interpolated precipitation data for eastern Africa
Because reanalysis data do not correlate well with observed precipitation in this region (Funk and Verdin 2003), we use interpolation of a unique set of observational precipitation records as the primary basis for evaluating trends in eastern Africa (Funk et al. 2003, 2007; Funk and Michaelsen 2004; Funk and Verdin 2009). The Climate Hazard Group (CHG) at the University of California, Santa Barbara is closely involved in famine early warning in eastern Africa, and has developed a dense quality controlled gauge dataset for the region, with over 350 stations in Kenya and Ethiopia. Many of these rain gauge observations were purchased from the Ethiopian Meteorological Agency and are not publically available. These stations were combined with gauges from neighboring countries, and quality controlled by hand via visual comparisons with neighbors. The data were then translated into MAMJ totals. Gamma distributions were also fit to each station and used to screen for abnormal values. Means were calculated over the 1960–1989 period and seasonal percent anomalies were estimated by dividing by means. Percent anomalies were then interpolated using kriging, and the data reconverted to millimeters by multiplying against a high resolution climatology (Funk and Verdin 2009). The interpolated precipitation dataset used in this study has 0.1° spatial resolution and is hereon referred to as CHG-CLIM. CHG-CLIM improves upon the estimates of east African rainfall made by the Global Precipitation Climatology Centre (GPCC) thanks to the substantially higher number of gauges and higher spatial resolution. The spatial resolution is especially important in the Ethiopian Highlands, where a very complex topography causes high heterogeneity in precipitation totals. The high spatial heterogeneity can be problematic because high correlation cannot always be expected among precipitation records from relatively nearby stations. Notably, Dinku et al. (2008) used many of the same Ethiopian Highland station datasets as are used in this study and found very strong agreement with several global gridded precipitation datasets. This contributes to our confidence in the CHG-CLIM product.
2.4 Interannual long-rains variability via meridional height gradient
We used rotated PCA on the standardized CHG-CLIM dataset and identified seven regions in Kenya and Ethiopia where interannual MAMJ precipitation variability is spatially coherent. The number of regions was determined using the Monte Carlo rule-N method (Preisendorfer et al. 1981). Within each region we calculated the mean standardized precipitation record and converted these standardized values to percent of the 30-years mean from 1950 through 1979. We then compared these CHG-CLIM regional records to those produced by the GPCC and the GPCP to demonstrate that all three datasets are in agreement regarding Kenyan and Ethiopian rainfall declines in recent decades.
For evaluation of the large-scale climate processes associated with drought during the long rains, we calculated a single standardized precipitation time series for Kenya and Ethiopia. For this calculation, we excluded all areas where (1) there have not been substantial declines over the past three decades, (2) MAMJ is not the dominant precipitation season, and/or (3) rainfall gauge data are sparse.
We next examined NCEP/NCAR reanalysis moisture transports during MAMJ within the lower troposphere (1,000–700 hPa) on a global scale, focusing on anomalies during the wettest and driest 20% of years in Kenya and Ethiopia. We also calculated how moisture transports and specific humidity have related to the primary modes of climate variability identified in the PCA. We compared the spatial structure of the correlation fields to the spatial patterns during wet and dry years. We complimented this analysis with an analysis of how the PC time series correlate with vertical and zonal wind velocities looking across tropical/northern Africa and the tropical Indian and western Pacific Oceans.
Finally, we evaluated how the long-established interannual relationship between MHG and long-rains precipitation may be impacted by the westward extension of the Walker circulation. We hypothesized that the westward extension of the Walker circulation has begun to suppress long-rains precipitation during both high and low MHG years. In this analysis, we considered GPCC and CHG-CLIM precipitation datasets to evaluate long-rains precipitation in Africa because of their long temporal coverage. We also considered MAMJ SSTs (NOAA extended dataset) to compare multi-decadal trends in SSTs compared to the interannual variability associated with MHG.
3 Results
3.1 Principal component analyses
a time series of PC 1 (black) and mean March–June GISS annual global temperature anomaly (red). Lower 4 panels: Correlation maps for regression of annual March–June climate on the PC 1 time series from (a). Climate variables are b mean zonal wind profile between 20°S and 20°N, c mean vertical velocity (omega) profile between 20°S and 20°N, d mean March–June temperature 2 m above sea level, and e NCEP/NCAR total precipitation. Values in c represent correlation with sinking motion, so cool colors represent correlations with rising motion, warm colors represent correlations with sinking motion
a time series of PC 2 (black) and March–June mean Niño-4 temperature anomaly (red, inverted). Lower 4 panels are the same as in Fig. 1, but show correlation with PC 2
Black and green lines represent a PC 1 and b PC 2 calculated using March-June NCEP/NCAR reanalysis data from 1948 to 2009 (also shown in Figs. 1 and 2) and ECMWF reanalysis from 1958 to 2002, respectively. Redlines are the first 2 PC time series calculated using March–June radiosonde 1,000 hPa temperature and profiles of zonal wind velocity from 1979 to 2009. Blue lines are the first 2 PC time series calculated using these same variables from the NCEP/NCAR reanalysis from 1979 to 2009. For PC time series calculated from 1979 to 2009, PC 2 appears in (a) and PC 1 appears in (b). R values indicate the correlation between the line of matching color and the black line
a Change in March–June SST (NOAA extended reanalysis) versus change in global mean March–June surface temperature (GISS dataset). Slope coefficients are 7-years running mean SSTs regressed on 7-years running mean global temperature from 1900 to 2009. A slope of 1 indicates SST warming at the same pace as the global mean. Analysis was limited to regions where mean 1951–1980 March–June SST was >24°C. Light and dark grey boxes bound the areas represented in (b) and (c–d), respectively. b Expansion of the warm pool from 1948 to 2009. Contours represent 5-years running 28.75°C isotherms. Warmer colored isotherms represent more recent 5-years periods. c and d Hovmöller plots of 5-years running mean March–June SST and 500 hPa vertical velocity (omega), respectively, between 10°S and 10°N. Black contours represent the 28.75°C isotherm in (c) and neutrality in (d). Negative values in (d) represent upward motion anomalies
3.2 Comparison of reanalysis to simulated and observed climate data
Modeled and observed March–June relationships between global mean temperature and variables related to the tropical Indian-Pacific Ocean Walker Circulation. Relationships with global mean temperature are measured here as unit change per degree change in mean global temperature. Relationships were evaluated using data smoothed with a 5-years running mean to dampen the impact of high-frequency ENSO variability. Model and observed datasets cover various time scales within the period from 1948 to 2009. Panela contrasts surface temperatures in the tropical Indian (10°N–10°S, 55°E–120°E) and central Pacific (10°N–10°S, 180°E–140°W). Panelb contrasts zonal wind velocity in the lower and upper troposphere in the Indian-Pacific warm-pool area (10°N–10°S, 90°E–180°E). Panelc contrasts vertical velocity (omega) from 850 to 200 mb above the southern Indian Ocean (0°S–15°S, 60°E–90°E) and central Pacific (10°N–10°S, 180°E–150°W). Paneld contrasts total precipitation in the same regions represented in panel (c). Circles model means. Vertical and horizontal lines running through circles of the same color: minimum and maximum of multiple runs. Stars Reanalysis and reconstructed data. Grey dashed boxes Unit change per °C given a correlation with global temperature of 0.3. Dashed diagonal lines in panels (a) and (d) are 1-to-1 lines
Among the other variables evaluated, there is little GCM consensus on how conditions might have been impacted as global temperatures rose in the 20th century (Fig. 5b–d). Interestingly, the NCEP/NCAR and ECMWF reanalyses also disagree as to how zonal wind, vertical velocity, and precipitation have varied over the past half century over the sparsely instrumented central Pacific Ocean (Fig. 5b–d). Over the Indian Ocean, however, these reanalyses agree that convection and rainfall have increased in step with SSTs since the mid-1900s. While the NCEP/NCAR reanalysis suggest intensification of upper-level westerly anomalies above the eastern Indian Ocean and western Pacific Ocean, the ECMWF does not indicate a significant change in upper-level (or lower) zonal wind velocities in this region (Fig. 5b).
Given the disagreement among modeled and reanalysis datasets in how global temperatures relate to the overturning Walker circulation, it was important to validate that the PC time series presented in this paper represent true climate processes and not artifacts of the reanalysis parameterization and/or inputs. In particular, we were interested in whether other precipitation datasets reflect the positive relationship between PC1 and precipitation over the Indian Ocean that is apparent in the NCEP/NCAR and ECMWF reanalyses. This is important because we believe trends in precipitation over the tropical Indian Ocean should impact trends in east African long rains. To examine this question we examined the correlation between PC1 and Indian Ocean Precipitation (0°S–15°S, 60°E–90°E) for each of the of the five validation precipitation datasets listed in Sect. 2.1. Considering 1979 through 2008 (the satellite era, during which we have relatively high confidence in ocean precipitation retrievals), PC1 correlated positively (0.41 < r < 0.66) and significantly (10−6 < p < 0.03) with all five data sets. In contrast, there was disagreement among the validation datasets regarding relationships with PC1 in the central Pacific Ocean, as was the case for NCEP/NCAR and ECMWF. PC2 was very well represented among the five validation datasets, with clear correlation patterns associated with ENSO (positive correlations over the Maritime Continent and negative correlations in the central Pacific).
It should be noted that the magnitude of the increases in latent heating shown by the various Indian Ocean rainfall observations (~30–50 cm °C−1) is thermodynamically substantial. A 40 cm change in seasonal precipitation is equivalent to a 97 W m−2 flux, associated with a total of 3.58 × 1014 W across the associated region (0°S–15°S, 60°E–90°E). This is roughly 10% of the total latent heating across the entire tropical Indian Pacific (15°S–15°NS, 60°E–80°W). It seems likely, therefore, that the ~40° geographic extension of the western border of the warm pool (Fig. 4b, c) has also influenced the warm pool thermodynamic circulation.
In summary, observational records and reanalysis products all show a dominant mode of climate variability in the tropical warm-pool area that is tightly coupled to global temperatures. Observational and reanalysis data also agree that the Indian Ocean’s temperature response to global warming appears to be roughly three times that of the central Pacific’s (Fig. 5a). The GCMs also show a greater rate of warming in the Indian versus Pacific, but overestimated warming in the central and eastern Pacific, causing the simulated gradients between Indian and Pacific SSTs to grow more slowly. Despite the fact that the observed gradients increased faster than simulated gradients, both observed and simulated records concur that the Indian Ocean has warmed faster than the central Pacific. This suggests, in turn, that the differential heating rates may continue for some time to come. Another consensus among the observations and most models is the increase in rainfall across the Indian Ocean (Fig. 5d). The ECMWF agrees with the NCAR/NCEP reanalysis, suggesting that 1°C of global warming will be associated with increased convection and an increase of more than 50 cm of rainfall over the Indian Ocean during MAMJ. It seems safe to conclude that models and observations all concur on a positive correlation among global temperatures, Indian Ocean SSTs, and Indian Ocean rainfall.
What is uncertain is the atmospheric response to the warming signal over the central tropical Pacific. The NCEP/NCAR reanalysis shows a distinct overturning circulation, with increased ascent over the Indian Ocean coupled with subsidence across the central Pacific. This response, however, is not shown in the ECMWF reanalysis. Further, the CMIP3 models simulate a weak association between global temperatures and increased convection over both the central Pacific and Indian Oceans (Fig. 5c). The tendency for models to simulate an increase in convection over the central Pacific contributes to the popular interpretation that continued warming will lead to a more ‘El Nino-like’ climate. The strength of the correlations between modeled central Pacific vertical velocities and global temperatures, however, are weak (less than 0.4), so it is unclear how atmospheric circulation over the central Pacific will respond to continued warming and how these circulation changes will impact precipitation in regions where precipitation has been historically correlated with variability in ENSO indices.
3.3 Precipitation and circulation impacts of PC1 and PC2
a and c MAMJ regional, 5-years running average precipitation within Ethiopia and Kenya relative to the 1950–1979 mean. Blacklines CHG-CLIM dataset, red lines GPCC, green lines GPCP. GPCC and GPCP records were linearly fit to the CHG-CLIM record. The numbered regions correspond to the regions outlined in (b). b Change in CHG-CLIM precipitation in the last 31 years relative to the 1950–1979 mean, calculated by linear regression. The seven regions were determined using a rotated principal components analysis. We did not consider locations where mean MAMJ precipitation <100 mm (grey areas) in any of these calculations
MAMJ water–vapor transports (arrows) and precipitable water (background) in the lower troposphere (1,000–700 hPa). a Mean anomaly during the driest 12 MAMJ seasons from 1948 to 2009. b Correlation with PC1. c Correlation with PC2. MAMJ rainfall records were calculated using regions 6 and 7 in Fig. 6b
While the mean wind velocity and humidity data provided by the NCEP/NCAR reanalysis are too low in spatial resolution to accurately discern the complex paths of air parcels before reaching eastern Africa (Gatebe et al. 1999), the anomaly map in Fig. 7a suggests that the tropical Atlantic Ocean and Congo Basin may be important sources of water vapor for eastern Africa during MAMJ, consistent with Segele et al. (2009). During dry seasons, Walker circulation over the tropical Indian Ocean is particularly strong and the descending western branch of the cell appears to suppress moisture transports from the Atlantic, interior of Africa, and Indian Ocean (Fig. 7a). During wet seasons, moisture more freely converges upon eastern Africa and a weakened or eliminated descending branch of the Indian Ocean Walker circulation allows for increased convection and precipitation (data not shown).
The strength of the Indian Ocean Walker circulation is impacted by the ENSO cycle. The warm ENSO phase leads to a weaker Walker circulation and is weakly associated with enhanced long-rains precipitation in Kenya and Ethiopia. While correlations between long-rains precipitation and ENSO indices are weak, there is a strong association between low-level moisture convergence over the tropical warm pool and low-level divergence and dry conditions over eastern Africa (Fig. 7a). This contributes to the La Niña-like pattern in the map of dry anomalies in Fig. 7a, as can be seen by comparing Fig. 7a to a map of correlation between PC2 and vapor-transports, where the length of arrows depict the strength of correlation (Fig. 7c). Interestingly, ENSO indices have slightly trended toward a warmer phase in recent decades, but east African vapor transports and rainfall have declined. Figure 7b suggests that the decline of vapor transports into eastern Africa, despite the slight tilt toward warm-phase conditions, may be associated with rapid warming of the tropical Indian Ocean and a westward extension of the ascending branch of the Walker circulation (increasing PC1).
Vertical profiles showing how tropical zonal and vertical circulation correlate with (a) PC1 and (b) PC2. a Zonal and vertical (omega) velocities were averaged across latitudes from 0°N to 20°N at longitudes west of, and including, 45°E. Velocities east of 45°E were averaged across latitudes from 20°S to 0°N. b Velocities were averaged across latitudes from 10°S to 10°N at all longitudes. Correlations with vertical velocity were multiplied by −1 so that upward arrows represent velocities away from the surface. Dark grey areas in the foreground represent average MAMJ surface pressure. Thick grey arrows in the background represent the zone of maximum correlation between the mode of climate variability and ascending atmospheric motion. The bold “H” in the background is for “high pressure” and represents the location of low-level divergence acting as the western source of return flow to the zone of convection
3.4 Relationship between PC1 and interannual variability
As was shown in Figs. 7b and 8a, PC1 manifests itself over tropical Africa and the Indian Ocean in a way that should impact the long rains. Using the correlation fields displayed in Fig. 8 as PC loadings, PC1 can be recalculated using zonal and vertical velocities over equatorial Africa and the Indian Ocean out to 120°E. The recalculated PC1 reflects how the circulation trends associated with the original PC1 impact climate across tropical Africa and the Indian Ocean. A 31-years running correlation between the recalculated PC1 and long-rains precipitation reveals a progressively more negative relationship over the past six decades, growing from r ≈ 0 to r ≈ −0.45 (not shown). This suggests that the manifestations of PC1 over Africa and the Indian Ocean are becoming increasingly influential on long-rains precipitation. Because these processes appear to have only a secondary influence on long rains totals, we were interested in identifying the primary atmospheric circulation factors related to long-rains precipitation and then determining whether the multi-decadal processes associated with PC1 have influenced these interannual relationships.
a Correlation between 500-hPa geopotential height and long-rains precipitation. Rectangles bound the regions where geopotential heights are used to calculate the meridional height gradient (MHG). MHG values are highest when 500-hPa heights are low in the blue rectangles and high in the red rectangles. b Standardized time series of long-rains precipitation, MHG, and MJO amplitude. c Background shows correlation between the MHG and specific humidity integrated from 1,000 to 700 hPa. Arrows show directional correlation between MHG and vapor transport velocities. d Vertical profiles of correlation of the MHG with zonal and vertical (omega) wind velocities. Zonal and vertical velocities (omega) were averaged across latitudes from 10°S to 10°N. Correlations with vertical velocity were multiplied by −1 so that upward arrows represent upward velocities. Dark grey areas in the foreground represent average MAMJ surface pressure (land)
Impacts of PC1 and MHG on long-rains precipitation (long-term linear trend was removed from MHG to better isolate the interannual nature of variability). a Scatter plot of 5-years running standardized MHG versus standardized PC1. Arrows indicate direction in time. Colors of circles represent the rank of 5-years running long-rains precipitation totals, calculated with CHG-CLIM. Each quadrant shows a unique class of season based upon MHG and PC1. bBox plot of long-rains precipitation totals in five classes of seasons. Neutral seasons are seasons when MHG and/or PC1 were within 1/8 of a standard deviation from the mean. In all other classes, both variables were greater than 1/8 of a standard deviation from the mean. Maps (c–f) show a comparison of climate (SST and precipitation) anomalies during each of the four non-neutral classes. In each set of three maps, top map NOAA extended SST, bottom-left map GPCC precipitation, bottom-right map CHG-CLIM precipitation
4 Discussion and conclusions
The PCA of reanalysis climate data indicated that the primary mode of variability within the tropical Indian Ocean has been strongly associated with increased global mean surface temperature and increased SST, convection, and rainfall in and above the Indian Ocean during MAMJ. These results are consistent with those of Goswami and Thomas (2000) and Minobe (2005). Both of these studies applied a PCA to global reanalysis data extending back to the late 1940s and also found that the first PC time series represents a long-term trend toward upward vertical velocity anomalies over the Indian Ocean and downward anomalies over the central equatorial Pacific Ocean. Notably, the long-term increase observed by Goswami and Thomas reversed back to neutral in the mid 1990s, shortly before their climate reanalysis time series ended in 1996. In contrast, the time series used in the Minobe study extended through 2002 and the positive trend of PC1 was continuous throughout the record. Corroboration of this post-1940s trend toward anomalously rapid warming in the warm pool and Indian Ocean can be seen in the Indian-Pacific SST trends, especially once the ENSO signal has been removed (see Figure 9 in Compo and Sardeshmukh 2010).
While there is generally high confidence in temperature observations for most places on earth, there is much less confidence in the historical variability of precipitation, particularly over oceans. The validation test using five alternate reconstructions of global precipitation confirmed that indeed PC1 did correlate positively and significantly with Indian Ocean precipitation during the post-1979 satellite era. This offers confidence that PC1 represents a significant and large-scale trend in not only temperature, but also hydrologic processes throughout the Indian Ocean region. As an interesting note, a follow-up analysis revealed that the first PC of GPCP precipitation (1979–2009) within the original PCA region correlates well (r = 0.75) with the first PC time series calculated from 1948 to 2009 2-m temperature data, and poorly with the second temperature PC (r = 0.28). This was unexpected because, as Fig. 3 demonstrates, other datasets limited to 1979 onward are more representative of PC2 than they are of PC1 (as is the case with the CMAP precipitation dataset). Among the five validation records, GPCP is arguably the most trusted over tropical ocean regions (Yin et al. 2004), further emphasizing that convective activity and precipitation over the tropical Indian Ocean are very much linked with the surface temperature dynamics associated with PC1.
Mean atmospheric vertical velocity during MAMJ between 700 and 400 hPa for a 1948–1978 and b 1979–2009. Cool shades indicate upward movement, warm shades indicate subsidence
The westward extension of MAMJ Walker circulation appears to be partly driven by disproportionately rapid warming in the central and western Indian Ocean, which has effectively expanded the warm-pool region westward (Fig. 4). The westward expansion of the warm pool appears to be a primary signature of increased global temperatures. Four SST datasets, two reanalysis products, and 21 CMIP3 models all indicate that Indian Ocean SSTs are tightly coupled with global temperatures during MAMJ (Fig. 5a); as global temperatures have increased since 1948, Indian Ocean SSTs have increased about 1.5 times as rapidly. These modeled and observational datasets also agree that SSTs in the tropical western and central Pacific are more weakly correlated with global temperatures during MAMJ, although models tend to over predict this relationship. Interestingly, regardless of whether they are constrained by actual 20th century SSTs, GCMs converge upon the projection of an increased Indian-Pacific SST gradient when they incorporate increased greenhouse forcing. This suggests that the increased SST gradient is anthropogenically driven (Barnett et al. 2005; Hoerling et al. 2010).
While the 21 GCMs evaluated in this study simulate a tilt toward more rapid warming in the Indian Ocean, they lack the substantial atmospheric Walker response above the Indian Ocean that is apparent in the reanalysis (Fig. 5b–d). Discrepancies between GCM simulations and reanalysis data within the tropics have been extensively documented (e.g., Santer et al. 2005, 2008; Soden et al. 2005; Karl et al. 2006; Mitas and Clement 2006; Douglass et al. 2008). Documented discrepancies generally regard trends in tropical lapse rates. While GCMs predict tropical lapse rates to decrease toward the moist adiabat (increased stability, weakened Hadley and Walker circulations, more ‘El Niño-like’), many observational records suggest the opposite. Recently, however, Santer et al. (2008) showed that, averaged throughout the tropics, the discrepancy between observed and modeled lapse rates is not statistically significant.
In the current study, we do not refute the modeled trends in mean lapse rate throughout the tropics, but we do challenge the idea that tropical climate is becoming more El Niño-like as the globe warms, at least over the Indian Ocean and eastern Africa during the MAMJ season. We suggest that, during this season, GCMs have generally overestimated the strength of the relationship between global mean temperatures and SSTs in the tropical central Pacific by a factor of two to three (Fig. 5a), thereby underestimating the SSTs contrast between the Indian Ocean and the central Pacific. As the SST gradient between the Indian Ocean and central Pacific has strengthened in step with global mean temperature, moisture convergence and convection have increased over the Indian Ocean. We suggest that this response to increasing global temperature is neither El Niño- nor La Niña-like, as is evidenced by the statistical independence of PC1, representing the westward expansion of the Walker circulation, and PC2, representing ENSO variability. Supporting this hypothesis, fossil corals in the tropical Pacific Ocean suggest that the mean state of ENSO and global temperatures have been uncorrelated for at least the past millennium (Cobb et al. 2003). Should global mean temperatures continue to rise beyond the realm of historical interannual variability, we expect that the gradient in SSTs between the warm tropical Indian Ocean and the relatively cooler Pacific Ocean will continue to increase in the near-term future. Du and Xie (2008) show that the differential warming rate is driven by the SSTs themselves: a given amount of warming causes relatively more evaporation from the already warmer Indian Ocean surface due to the exponential relationship between temperature and saturation vapor pressure. This local feedback mechanism likely accounts for the considerable coherence among CMIP3 estimates of increased air temperatures (Fig. 5a). The differential rates of evaporation lead to a more rapid increase in atmospheric water vapor over the warmer ocean surface, leading to enhanced greenhouse forcing and more warming. The specific mechanism of the warming feedback described by Du and Xie involves large increases in surface evaporation; these large increases in evaporation probably contribute to the coherent increases in rainfall over the tropical Indian Ocean (Funk et al. 2008). Further, greenhouse forcing should be more influential to the west of the cold-tongue region because forcing within the cold-tongue region is diluted by strong upwelling (Clement et al. 1996; Sun and Liu 1996; Cane et al. 1997; Seager and Murtugudde 1997). The disparate west versus east warming pattern could be exacerbated by a large-scale overturning circulation (Hoerling et al. 2010), advecting moisture into the Warm Pool region, and enhancing the water vapor feedback while simultaneously cooling the eastern Pacific via enhanced equatorial upwelling. Such an interpretation could be consistent with the observed trends (Fig. 4a).
Correlation between GPCP MAMJ rainfall totals and PC1 from 1979 to 2009
In summary, global surface temperatures are strongly correlated with Indian Ocean SSTs during MAMJ. As temperatures have increased, convection and precipitation above the Indian Ocean have increased and long-rains precipitation in eastern Africa has declined. This decline has been particularly strong on the eastern flank of the Ethiopian Highlands in Ethiopia and across central Kenya. Model simulations, driven by diabatic forcing over the southern Indian Ocean, suggest that increased convection and precipitation over the Indian Ocean are associated with a downwind reduction of low-level moisture transports toward eastern Africa (Funk et al. 2008). A similar process has been implicated previously as a possible cause of drought in the African Sahel (Giannini et al. 2003). While most GCMs simulate a shift toward a more ‘El Niño-like’ climate in tropical eastern Africa in response to increasing global surface temperatures, we show that there is no evidence of increased El Niño-like conditions (somewhat increased rainfall) for MAMJ in eastern Africa during the period of observed anthropogenic warming of the last 40 years. This is at odds with the IPCC AR4 predictions of increased precipitation in tropical eastern Africa as a ‘likely’ response to anthropogenic global warming. While important regional variations such as the MHG will still produce wet seasons, we warn that long-rains precipitation has steadily declined over the past three decades. As this trend is associated with globally increasing temperatures, we estimate that even a GCM-projected mild shift toward more El Niño-like conditions would still be accompanied by an increased frequency of drought conditions in tropical eastern Africa due to reduced MAMJ rainfall. These findings are significant for adaptation planning for the region. Drier, rather than wetter conditions in the century ahead appear likely. The anthropogenic Indian Ocean warming response appears to be one of the most consistent (Hoerling et al. 2004; Barnett et al. 2005) and well understood (Du and Xie 2008) responses to greenhouse gas emissions. This anthropogenic warming appears to have already significantly altered the earth’s largest circulation feature and impacted its most food insecure inhabitants.
http://www.usaid.gov/our_work/humanitarian_assistance/ffp/wherewework.html. Estimates based upon data from July 2010.
Acknowledgments
This research was supported by the U.S. Agency for International Development Famine Early Warning System Network under U.S. Geological Survey Cooperative Agreement #G09AC00001 and the National Aeronautics and Space Administration under Precipitation Science Grant #NNX07AG266. We thank these agencies for their support. We also thank and acknowledge the CMIP3 modeling groups for providing data for analysis, and the World Climate Research Programme for collecting and archiving the model output. Thanks to Mingyue Chen and Tom Smith for help accessing global gridded precipitation reconstructions. Comments by Henry Diaz, Charles Jones, Joel Michaelsen, James Verdin, James Rowland, Kyle Cavanaugh, Greg Husak, Laura Harrison, Michael Marshall, Brian Osborn, and two anonymous reviewers greatly improved the quality of the manuscript.
Open Access
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.