SEARCH

SEARCH BY CITATION

Keywords:

  • Sahel teleconnection;
  • interannual variability;
  • CFSv2 validation

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The North Atlantic influence on PS
  5. 3. Forecast skill for EA
  6. 4. Summary
  7. Acknowledgements
  8. References

This study presents new findings on the link between interannual variabilities of atmospheric circulations over the North Atlantic and precipitation over the African Sahel (PS). Our analysis shows a meridionally stratified circulation wave train resembling the East Atlantic (EA) mode, apparently connected to PS through Rossby wave dispersion in the middle troposphere originating from the North Atlantic. However, the Climate Forecast System version 2 fails to depict the EA and its PS impact. Because the EA explains 29% of variance of PS, this portion of PS variability is either missing in the seasonal forecast, or being made up by an alternative process.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The North Atlantic influence on PS
  5. 3. Forecast skill for EA
  6. 4. Summary
  7. Acknowledgements
  8. References

The African Sahel, a semi-arid region lying along the southern edge of the Sahara Desert, is characterized by large climate variability in summer precipitation on interannual and interdecadal timescales. The El Niño-Southern Oscillation (ENSO) is known as an important modulator of Sahel summer precipitation (PS) (e.g. Semazzi et al., 1988; Janicot et al., 1996; Joly and Voldoire, 2009). Other climatic forcing factors that modulate PS include sea surface temperature (SST) anomalies in the Indian Ocean (Bader and Latif, 2003) and the tropical Atlantic (Brandt et al., 2010), as well as the Atlantic Multi-decadal Oscillation (AMO) (Zhang and Delworth, 2006). These previous findings were based upon the concept of SST-driven, tropically confined teleconnections affecting PS. Changes in SSTs also can modulate planetary-scale circulation, such as modifying the strength of the Hadley cells and changing the position of the extratropical storm tracks (Brayshaw et al., 2008; Graff and LaCasce, 2012).

There are, however, some studies that hint at teleconnectional forcing of PS from the mid-latitudes. Focusing on the North Atlantic Oscillation (NAO; Hurrell et al., 2003) during summer months, Folland et al. (2009) found that temperature and precipitation in the Sahel are correlated with NAO. Chen and Wang (2007) noticed a connection between PS and an atmospheric short-wave train in the North Atlantic during ENSO-active years. In this study, we show a meridional wave train pattern that connects PS with a possible higher latitude influence, regardless of the state of ENSO. Our analysis indicates that such a North Atlantic wave train is linked to the so-called East Atlantic (EA) mode. The EA was first identified by Barnston and Livezey (1987) as a center of anomalous 700-mb geopotential height off the coast of Ireland accompanied by wave-like disturbances across Europe. We also report that this EA-PS linkage is missing in one of the major climate forecast models, hence representing missing variability of PS in seasonal climate prediction.

This study utilized the National Center for Environmental Prediction (NCEP)/National Corporation for Atmospheric Research (NCAR) Reanalysis I (Kalnay et al., 1996) for atmospheric data. Rain gauge observations compiled and gridded by the National Oceanic and Atmospheric Administration (NOAA) precipitation reconstruction over land (PREC/L; Chen et al., 2002) were used. Seasonal composites derived from monthly average reforecasts (or hindcast, 0–9 months) from the NCEP Climate Forecast System version 2 (CFSv2; Saha et al., 2013) were examined for the forecast skill of PS and EA. PS was defined as the average precipitation within 12–20°N and 15°W–30°E; hereafter the term PS and all other analyses are focused on the July to September (JAS) season. Precipitation and various climate indices used in the following analyses were normalized by standard deviation for ease of comparison, and all data were linearly detrended to remove any anthropogenic forcing while retaining decadal-scale variability such as the recovery of the Sahel drought (e.g. Fink et al., 2010; Wang and Gillies, 2011).

2. The North Atlantic influence on PS

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The North Atlantic influence on PS
  5. 3. Forecast skill for EA
  6. 4. Summary
  7. Acknowledgements
  8. References

2.1. Empirical evidence

The well-known ENSO influence on PS is illustrated in Figure 1(a) by the composites of streamfunction anomalies during ENSO-active years, i.e. La Niña minus El Niño years based on |normalized JAS Niño3.4 index| > 1.25  standard deviations (s.d.), for the period 1950–2010 (Niño3.4 index obtained from the NOAA Climate Prediction Center (CPC; http://www. cpc.ncep.noaa.gov/products/precip/CWlink/MJO/enso.shtml). These composites reflect the resultant wet conditions in the Sahel during ENSO-active years. There is a broad area of increased anticyclonic flow at 200 mb over the Mediterranean Sea, North Africa and the tropical Atlantic, with weaker cyclonic anomalies in the middle and lower troposphere over North Africa. During La Niña (El Niño) events, the circulation anomalies result in increased (decreased) strength of the tropical easterly jet (TEJ) and enhanced (suppressed) divergence aloft, which in turn enhances (suppresses) convection over the Sahel and West Africa (e.g. Fontaine et al., 1995). The anomaly at 600 mb over North Africa is important in that it represents a modulation of the African easterly Jet (AEJ), whose strength and position plays an important role in the development of African easterly waves (AEWs, Thorncroft and Hoskins, 1994; Chen, 2006; Nicholson and Grist, 2001).

image

Figure 1. Composited differences of JAS streamfunction over the years 1950–2010 during (a) La Niña years (normalized Nino3.4 index < −1.25 standard deviation, s.d., six members) minus El Niño years (Nino3.4 index > 1.25 s.d., seven members), and (b) wet years (normalized PS > 1.25 s.d., six members) minus dry years (normalized PS < −1.25 s.d., four members during ENSO-neutral years (normalized Nino3.4 index between −0.75 and 0.75 s.d.). Stippling indicates statistical significance of 95% per test.

Download figure to PowerPoint

Composite circulation anomalies for wet minus dry years (i.e. when |normalized PS| > 1.25 s.d.) in Figure 1(b), compiled during ENSO-neutral years (i.e. |normalized JAS Niño3.4 index| < 0.75 s.d.), show some similarities with the ENSO composite, but with intriguing differences. Both cases reveal cyclonic anomalies in the middle and lower troposphere, but the ENSO-neutral composite has a stronger cyclonic anomaly over Great Britain at all levels, especially at 200 mb. Second, there is a northwest–southeast oriented wave train extending from Greenland to Eurasia particularly at 200 mb, which includes the pronounced negative anomaly over the British Isles. At 850 mb the cyclonic anomaly over the Sahel implies increased monsoonal southwesterly flows, thereby enhancing low-level moisture flux from the Gulf of Guinea; this feature is present in both composites, but is stronger and more coherent in the ENSO-neutral ‘wet’ case.

The meridional wave train pattern revealed in Figure 1(b) resembles the spatial loading pattern of the EA (Barnston and Livezey, 1987). Based upon the EA index obtained from NOAA CPC (http://www.cpc.ncep.noaa.gov/data/teledoc/ea.shtml), the streamfunction anomaly composites during EA-active [(EA index > 1.25 s.d.)—(EA index < −1.25 s.d.)], ENSO-neutral years are shown in Figure 2(a), which depict the characteristic anomalous center west of Ireland and the wave train extending over Europe. Because the NCEP/NCAR reanalysis was produced from assimilation systems that are two decades old, we also compared against the newer, but shorter-duration NCEP/Department of Energy (DOE) Reanalysis II (Kanamitsu et al., 2002) and the CFS Reanalysis (Saha et al., 2010) during their respective time periods and found the results from both reanalyses to be equivalent (not shown).

image

Figure 2. (a) Streamfunction anomalies during positive EA index years (normalized EA index >1.25 s.d., four members) minus negative EA index years (normalized EA index >1.25 s.d., three members), during ENSO-neutral years (normalized Nino3.4 index between −0.75 and 0.75 s.d.). (b) Time series for normalized EA index (black) and PS (green). (c) Spatial correlation map between EA index and PREC/L precipitation. Red box indicates Sahel region.

Download figure to PowerPoint

By correlating PS with the EA index, we found that there is indeed a statistically significant relationship (Figure 2(b), r = 0.54, p < 0.01). Based upon their correlations with PS during the period 1950–2010 (i.e. r2), ENSO explains 11% of the variance of PS while EA explains 29%. Spatial correlations between EA and PREC/L precipitation (Figure 2(c)) depict a significant, positive relationship across the Sahel region (boxed area). This correlation map also reveals a north–south stratification that corresponds to the EA circulation pattern (Figure 2(a)).

2.2. Dynamical inference

Having demonstrated the empirical connection between EA and PS, we next explored the dynamic mechanism of this connection by analyzing the horizontal-component Rossby wave-activity flux (W) for our composites. We derived the formulation of W by following Takaya and Nakamura (2001),

  • display math

where U is the two-dimensional JAS mean flow (u,v) and ψ is the perturbation streamfunction, with subscripting representing partial derivatives. This W vector provides a measure of propagating Rossby wave energy. Calculation of this wave-activity flux is not dependent on spatial or temporal averaging and therefore is applicable for any particular time period (see Takaya and Nakamura, 2001 for details).

To capture and isolate the wave train that is characteristic of the EA, we conducted an empirical orthogonal function (EOF) analysis of 600-mb geopotential height over the North Atlantic, Europe and North Africa only (20–90 N, 100 W–60E), so as to disregard potentially confounding climate signals from the rest of the Northern Hemisphere. The 600 mb level was chosen to focus on the height at which the core of the AEJ is located. The resulting analysis yields an EA-like pattern for in the second mode (the first being NAO). The corresponding time series is a reasonable representation of the globally and 700-mb-based EA index (r = 0.58), and broadly captures the interannual and interdecadal variability of the EA (not shown). Using this EOF-based EA index when |normalized EOF time series | > 1.25 s.d. and during ENSO neutral years, the EA circulation patterns are reconstructed in Figure 3(a) with the geopotential height—i.e. with a regional weighting over the North Atlantic and Africa. When compared to the ENSO composite in Figure 3(b), the meridional wave train pattern in this EA composite is much more pronounced in the middle and lower troposphere.

image

Figure 3. JAS geopotential height anomalies (contours) and Rossbywave-activity flux vectors for (a) high regional EOF time series values (normalized PC > 1.25 s.d., three members) minus low regional EOF time series (normalized PC< −1.25 s.d., six members) during ENSO-neutral years (normalized Nino3.4 index between −0.75 and 0.75 s.d.) and (b) La Niña years (normalized Nino3.4 index < −1.25 s.d., six members) minus El Niño years (Nino3.4 index > 1.25 s.d., seven members).

Download figure to PowerPoint

During ENSO-active years (Figure 3(b)), Rossby wave-activity flux is confined to the upper troposphere and directs from North Africa towards Europe, implying the movement of energy from the tropics to the middle latitudes. By contrast, in EA-active years (Figure 3(a)—during ENSO-neutral years) the wave-activity flux at the upper troposphere (200 mb) is mainly oriented toward southeast but is confined to north of 30° N. However, in the middle troposphere (600 mb) the wave-activity flux penetrates into North Africa along the downstream portion of the wave train over the Mediterranean Sea. At lower troposphere (850 mb), the wave-activity flux becomes weaker and is mostly confined to the anomaly off the coast of Ireland. Here we propose that this W propagation is a manifestation of the forcing mechanism leading to the middle tropospheric circulation patterns associated with the Sahelian wet/dry anomalies (Figure 1(b)).

The wave-activity flux over North Africa at 600 mb has an implication for regional circulation anomalies. It appears that the mid-level anticyclone that stations itself over the Sahara Desert can be modulated by teleconnection emanating from higher latitudes, i.e. it oscillates in response to the wave train initiated during positive/negative EA years. This teleconnection may affect the position and/or intensity of the AEJ, which in turn regulates the activity of AEWs (Thorncroft and Hoskins, 1994; Chen, 2006). For instance, Wang and Gillies (2011) found that the post-1990 recovery of the Sahel drought occurred in association with the northward migration of both the AEJ core and the AEW activity, leading the summer rainband to move northward. Because AEWs form at both sides of the AEJ core (Reed et al., 1988; Chen, 2006), it is possible that this injection of Rossby wave energy weakens (strengthens) the middle-level anticyclone during high EA-index (low EA-index) years. This would then affect the AEJ, as well as the activity of AEWs, thereby influencing PS. Further investigation is needed to detail the dynamic processes of this EA–PS teleconnection.

3. Forecast skill for EA

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The North Atlantic influence on PS
  5. 3. Forecast skill for EA
  6. 4. Summary
  7. Acknowledgements
  8. References

An important question derived from the aforementioned finding of the EA–PS connection lies in its depiction in climate prediction, that is, how well do operational climate forecast models capture EA and its impact on PS? To examine this, we analyzed hindcast output from the CFSv2 to evaluate the model's performance. Following the methodology used by the CPC to calculate EA, we first conducted a rotated EOF analysis on summer (JAS) Northern Hemispheric 500-mb geopotential height of the CFSv2 at zero-month lead time. The first ten principal components are plotted in Figure 4. The loading patterns for some of the EOFs are similar to the EA pattern, but those EOFs did not have a corresponding time series that correlated significantly with either the EA index or PS (not shown). Apparently, CFSv2 was unable to reproduce EA through the EOF approach as was used in Barnston and Livezey (1987) and the CPC.

image

Figure 4. First ten leading modes of rotated EOF analysis on CFSv2 500-mb geopotential height.

Download figure to PowerPoint

Seeking a more direct comparison, we computed the correlations between the reanalysis 500-mb JAS geopotential height and the EA and ENSO indices, and then preformed the same correlations using the CFSv2 zero-month reforecast (Figure 5). The atmospheric response to teleconnectional forcing is substantially different between the reanalysis and CFSv2 reforecast. When considering ENSO forcing, reanalysis shows positive correlation between geopotential height in North Africa (Figure 5(a)), while the CFSv2 reforecast (Figure 5(c)) shows a much more local positive response over the Sahara Desert and a negative relationship over Central and South America—in disagreement with reanalysis. Similarly, spatial correlation of the EA index with reanalysis (Figure 5(b)) depicts the distinct wave train over Europe and the negative anomaly west of Ireland, while CFSv2 does not reproduce the wave train (Figure 5(d)) and, furthermore, reveals an opposite relationship over the North Atlantic when compared with reanalysis. The CFSv2 output of PS does not respond to EA (r < 0.1) in all forecast months. Therefore, the variance in PS explained by EA is either missing from the model, or the modeled PS is mistakenly being modified by other processes.

image

Figure 5. Spatial correlation patterns between NCEP-I 500-mb geopotential height and (a) Nino3.4 index, (b) EA index, and between CFSv2 reforecast (0-month) 500-mb geopotential height and (c) Nino3.4 index, (d) EA index.

Download figure to PowerPoint

4. Summary

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The North Atlantic influence on PS
  5. 3. Forecast skill for EA
  6. 4. Summary
  7. Acknowledgements
  8. References

Analysis of the Sahel rainfall anomalies (PS) during ENSO-neutral years shows that, in addition to its known connection with ENSO, PS is significantly and positively correlated with EA. During the EA-active seasons, fluxes of Rossby wave activity bring energy dispersed from the mid-latitudes into North Africa; this process is particularly pronounced at the middle troposphere where the core of the AEJ is located. However, the CFSv2 fails to capture the EA mode and its association with PS. On the other hand, the CFSv2 produces reasonable ENSO states as well as the associated PS variation up to 3 months.

CFSv2 has been improved in terms of forecasting ENSO, particularly SST evolutions in the Niño-3.4 region (e.g. Wu et al., 2009; Zhu et al., 2012). This is a promising step in forecasting the well-established connection between ENSO and PS. Correspondingly, modeled PS responds appropriately to ENSO forcing through the first few forecast months (r = 0.78 at 0-month forecast, r = 0.33 at 3-month forecast). However, there is a discrepancy in the variability of atmospheric circulations in the North Atlantic and North Africa between the CFSv2 and reanalysis data, in that EA is not captured at all, even at 0-month forecast. Therefore, seasonal prediction of PS will likely suffer from the inability of CFSv2 in depicting EA. The only remedy at this point in time is to either develop empirical method supplementing the EA-related prediction or engaging multi-model ensembles, similar to what has been developed for the difficult-to-forecast NAO (Orsolini et al., 2011, Jia et al., 2012).

Finally, because EA is predominately a winter mode (whereas this study only considers the summer season), further examination of model performance in the wintertime atmospheric variability could help determine if this deficiency of EA forecasting is seasonal in nature, or is intrinsic to CFSv2 regardless of season. In future work we also will further explore the transfer of mid-latitude energy to the African subtropical climate that is implied by the Rossby wave-activity flux.

Acknowledgements

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The North Atlantic influence on PS
  5. 3. Forecast skill for EA
  6. 4. Summary
  7. Acknowledgements
  8. References

This project is supported by Grant NNX13AC37G, USAID Grant EEM-A-00-10-00001 and the Utah State University Agricultural Experiment Station as paper number 8556. Editorial assistance by Jon Meyer and comments offered by three anonymous reviewers are appreciated.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. The North Atlantic influence on PS
  5. 3. Forecast skill for EA
  6. 4. Summary
  7. Acknowledgements
  8. References