Below normal Long/Gu rains appears likely for eastern East Africa

Chris Funk 3/23/2017

In early November of 2016 we predicted that eastern Kenya and southern Somalia would likely experience a very poor October-December short rainy season and one of the poorest combined March-May (Gu) + October-December (Deyr) seasons in the last 20 years. Unfortunately, the March 2017 report from the Somalia Food Security and Nutrition Unit (FSNAU) indicates (full report here) that agricultural conditions in Somalia tracked closely with our November outlook.

The combined Gu/Deyr crop production levels were the worst on record, and the current 2016 local cereal supply situation is worse than in 2011 – when Somalia experienced a severe famine. Current humanitarian access, however, is much better than in 2011, and more than one million severely impacted people are receiving assistance in some of the most afflicted areas.  Recent FEWS NET and FSNAU alerts, however, have also underscored the high food security impacts that could accompany a third poor 2017 March-May Gu season.

Since December of 2016 this blog has maintained a negative outlook for the 2017 Gu season. Here, we update this assessment, based on the most up to date information available. We find that even though La Niña SST conditions have disappeared, a La Niña-like atmospheric response continues. Warm conditions in the Western Pacific continue to increase the chances of below normal rainfall in  eastern East Africa.

February 2017 CHIRP precipitation
Figure 1, February 2017 CHIRP precipitation

Before examining the data, however, it is worth considering the atmospheric conditions that, on average, keep eastern East Africa dry. Consider Figure 1, which shows February rainfall totals from the Climate Hazards Group Infrared Precipitation data set for South America, Africa and Indonesia. Why is eastern East Africa, an equatorial region located on the western rim of an ocean, so dry in comparison with the eastern seaboard of South America? The answer is the Walker Circulation. East Africa is dry because the very heavy rainfall around Indonesia, interacting with the Indian Monsoon winds, tends to produce dry warm sinking air over the eastern Horn of Africa.

The Walker Circulation has three main centers of action – one region of very heavy rainfall, centered around Indonesia – an area with very warm sea surface temperatures, and two equatorial dry regions to the west (over eastern East Africa and the Indian Ocean) and to the east (over the eastern Equatorial Pacific). When the ocean around Indonesia is warmer than the equatorial East Pacific, we tend to see a La Niña-like climate response that brings warm dry air down over eastern East Africa, increasing the risk of drought.

It is important to realize that we do not have to have a strong La Niña to have a severe East African March-May drought and La Niña-like atmospheric responses. For example, the signature 2011 drought was accompanied by mild (-0.5°C) La Niña conditions. Very dry conditions were also experienced in March-May of 2004 and 2009, years with neutral El Niño-Southern Oscillation conditions. These seasons, as well as ‘classic’ La Niña-type drought seasons in 1999, 2000 and 2008, share a common relationship – an enhanced Walker Circulation pattern contrasting increased moisture convergence, rainfall and ascending air near Indonesia with decreased moisture and dry descending air over eastern East Africa.

Below, we analyze current climate conditions and suggest that  persistence of the current  Niña-like climate response appears likely, at least for the early part of the 2017 March-May growing season. While considerable uncertainty remains, a third poor or failed crop growing season for eastern Kenya and Somalia appears to be likely, with below normal March-May rainfall the most probable outcome. Even a late start to the rainfall in these regions could have very serious food security implications. The ‘long’ rains in these areas are actually quite short, and crops depend heavily on April precipitation.

We present two lines of evidence converging on our below normal outlook: 1) a diagnostic analysis of current conditions and near-term weather forecasts, 2) a cross-validated statistical forecast based on the difference between March western Pacific and eastern Pacific sea surface temperatures (SSTs). Section 1 provides our diagnostic discussion. Section 2 provides our quantitative March-May forecast for a subset of eastern East Africa : a standardized rainfall outlook of -0.8Z, with an 80% confidence interval of -1.8Z to -0.2Z, and a 65% chance of below normal rainfall (rainfall less than -0.5Z).

Figure 2. Past thirty day OLR and near-surface wind anomalies from the Climate Prediction Center’s March 20th ENSO discussion.
Figure 2. Past thirty day OLR and near-surface wind anomalies from the Climate Prediction Center’s March 20th ENSO discussion.
  1. Observed conditions and near-term weather forecasts indicate below normal rains.

At this point (March 23rd), when the March-May rainy season typically commences in earnest, the observed climate conditions can be useful indicators of April rainfall performance.

Are we seeing conditions indicating an intensification of the Walker Circulation and/or a reduction in the onshore moisture transports into northern Tanzania, Kenya and southern Somalia?

The most recent (March 20th 2017) ENSO discussion from NOAA’s Climate Prediction Center identifies current neutral La Niña conditions. This same report, however, describes La Niña-like atmospheric responses, at least over the Indo-Pacific warm pool and eastern equatorial Pacific Ocean (near the dateline, 180°E/180°W). Looking at anomalies (differences from average) of Outgoing Longwave Radiation (OLR) data for the last thirty days (Figure 2), we see more OLR/less rainfall near the dateline and less OLR/more rainfall over the Warm Pool (110°E-150°E). OLR is blocked by clouds, so we see more cloudiness near the Maritime Continent and less cloudiness near the dateline — a modest enhancement of the Walker Circulation. Thus we are seeing a modestly La Niña-like atmospheric response despite neutral La Niña SST conditions. We are also seeing stronger near-surface trade winds along the equator between 160 and 180°E. These trade winds help bring moisture into the warm pool, feeding convergence and rising air near the Maritime Continent. This convergence and rising air can enhance the Walker Circulation and contribute to below normal rainfall over East Africa.

Figure 3. A longitude-time plot of OLR anomalies from the Climate Prediction Center’s March 20th ENSO discussion.
Figure 3. A longitude-time plot of OLR anomalies from the Climate Prediction Center’s March 20th ENSO discussion.

Figure 3 shows a longitude-time plot of OLR anomalies, also taken from the CPC ENSO assessment. The x-axis in this figure shows cloudiness anomalies (OLR) along the equator, with Africa and Indian Ocean being on the right, the Maritime Continent in Center, and the East Pacific on the right. Since early October we have seen a persistent pattern of cloudier conditions over the warm pool and less cloudier conditions over the dateline. This pattern helped produce the severe October-December drought over East Africa. There was a brief period of increased cloudiness and rainfall over East Africa in February. This respite was associated with the passage of a Madden-Julian Oscillation, and is not likely to extend into April. In March of 2017 we see a continued enhancement of the Walker Circulation, suggestive of a below normal outlook for April and perhaps May. At the bottom this panel (end of March) we see conditions similar October-November of 2016 – when the 2016 short/Deyr rains failed.

Figure 4. March 11-20 2017 CHIRPS standardized precipitation anomalies.
Figure 4. March 11-20 2017 CHIRPS standardized precipitation anomalies.

Recent rainfall conditions over East Africa have been dry, though we are still very early in the March-May season. Figure 4 shows the CHIRPS rainfall anomalies for March 11-20 2017, expressed as standardized anomalies (z-scores). Across northern Tanzania, Kenya, Somalia and Ethiopia we see widespread, modestly below normal precipitation. Note that the rains typically have not really started yet across most of this area. Using a different satellite (MODIS) we can also look at March 11-20 Land Surface Temperature (LST) anomalies (Figure 5). Again, we also see warm or very conditions across eastern East Africa. Later in this post we will suggest that dry sinking air (atmospheric subsidence) associated with a Walker Circulation enhancement may be contributing to these above normal temperatures.

Figure 5. March 11-20 2017 MODIS Land Surface Temperature standardized anomalies.
Figure 5. March 11-20 2017 MODIS Land Surface Temperature standardized anomalies.

The next several figures (Figures 6-8) show March 11-20 2017 NCEP/NCAR reanalysis data provided by NOAA’s Earth System Research Laboratory (ESRL). We use these figures to explore current atmospheric conditions, looking for potential drivers related to the observed dry and warm conditions in the Horn of Africa (Figures 4 and 5) .

Figure 6 shows March 11-20 2017 Total Precipitable Water anomalies. Blue areas in this image have less atmospheric water vapor. We see drier than normal conditions over east Africa where rainfall has been low (Figure 4) and LST has been warm (Figure 5). Not a great start to the rainy season.

Figure 6. March 11-20 2017 Total Precipitable Water anomalies from the NCEP/NCAR Reanalysis.
Figure 6. March 11-20 2017 Total Precipitable Water anomalies from the NCEP/NCAR Reanalysis.

At this time of the year, we should be seeing low level moisture transports from the Indian Ocean blowing just north of Madagascar and into Tanzania, Kenya and southern Somalia . Instead, we are seeing that low level wind anomalies are heading the other way (Figure 7) at a pretty rapid rate (4-6 ms-1). This reduction in onshore moisture transports may be helping to reduce water vapor levels over East Africa (Figure 6). In this figure we also see anomalous winds over the equatorial eastern Indian and Central Pacific that feed enhanced moisture convergence near the Maritime Continent – enhancing the Walker Circulation.

Figure 7. March 11-20 2017 Near-Surface Wind anomalies from the NCEP/NCAR Reanalysis.
Figure 7. March 11-20 2017 Near-Surface Wind anomalies from the NCEP/NCAR Reanalysis.

We conclude our analysis of reanalysis data by looking at a longitude-height plot of vertical velocities (Figure 8). Vertical velocities have been averaged along the equator and shown along the x-axis. Orange-red values in this plot indicate sinking air. Purple-blue shades denote rapidly ascending air. The y-axis represents elevation, with the top of the plot representing the top of the atmosphere. What we see in this figure is enhanced upward vertical motions near Indonesia (130-160°E) and enhanced downward motions over the western Indian Ocean and eastern East Africa (45-60°E). This subsidence may help explain the above normal temperatures we see in satellite data (Figure 5).

Figure 8. March 11-20 2017 vertical velocity anomalies from the NCEP/NCAR Reanalysis.
Figure 8. March 11-20 2017 vertical velocity anomalies from the NCEP/NCAR Reanalysis.

These reanalysis images just examine a brief ten-day snapshot of the climate, but do seem to indicate troubling patterns as we enter a critical growing season for Somalia. We can look out a little further out using Global Forecast System weather forecasts provided by NOAA’s Climate Prediction Center (CPC). Figure 9 shows the CPC’s March 31st-April 6th low level wind forecast. This appears similar to Figure 7, suggesting that continued reductions in onshore moisture flows may be likely, at least for early April. Once again we find large (4-5 ms-1)  changes in wind speeds just off the coast of East Africa.

Figure 9. CPC low level wind anomalies forecast for March 31-April 6th 2017.
Figure 9. CPC low level wind anomalies forecast for March 31-April 6th 2017. Image obtained from the CPC African desk.

Here at the Climate Hazards Group, funding provided by NASA SERVIR is allowing us (Marty Landsfeld and Pete Peterson) to develop CHIRPS-compatible downscaled GEFS precipitation forecasts. Figure 10 shows these forecasts for the last 11 days of March. Drought stricken areas of southern Somalia are not forecast to receive any relief in March. This may indicate that continued dry warm conditions are likely for the start of the growing season.

Figure 10. Downscaled (CHIRPS compatible) GFS rainfall forecast for March 21-31 2017. The area outlined in red will be examined in our forecast section.
Figure 10. Downscaled (CHIRPS compatible) GFS rainfall forecast for March 21-31 2017. The area outlined in red will be examined in our forecast section.
  1. Statistical March-May rainfall forecasts

We now present statistical March-May rainfall forecasts for the region identified in red in Figure 10. This area was selected because i) it received poor rainfall in October-December of 2016, and ii) prior research by FEWS NET scientists have linked March-May rainfall in this region to SSTs in the western Pacific and central Pacific. We will use these SSTs to predict March-May rainfall.

Figure 11. Correlation between 1997-20116 March NOAA Optimum Interpolation SSTs and March-May CHIRPS rainfall in eastern East Africa region shown in Figure 10. Purple boxes identify the equatorial western and central Pacific regions.
Figure 11. Correlation between 1997-20116 March NOAA Optimum Interpolation SSTs and March-May CHIRPS rainfall in eastern East Africa region shown in Figure 10. Purple boxes identify the equatorial western and central Pacific regions.

Figure 11 shows the correlation between March-May eastern East African rainfall and March SSTs. When the western Pacific is relatively warm and the central Pacific is relatively cool in March, precipitation in our study site tends to be below normal.

Figure 12. March 1 to March 23rd NOAA Optimum Interpolation SSTs, expressed as standardized anomalies based on a 1981-2010 baseline. Purple boxes identify the equatorial western and central Pacific regions.
Figure 12. March 1 to March 23rd NOAA Optimum Interpolation SSTs, expressed as standardized anomalies based on a 1981-2010 baseline. Purple boxes identify the equatorial western and central Pacific regions.

Figure 12 shows the observed March 1 to March 23 SSTs, expressed as standardized anomalies. We will use these SSTs to represent the month of March and derive a March-May rainfall forecast.  In these observations we find a strong difference between western Pacific and central Pacific SSTs – with relatively warm West Pacific and relatively cool central Pacific conditions. While the West Pacific has cooled in over the last few months, and we may be heading toward potential El Niño conditions later this year, moderate gradient conditions still persist in March. These conditions appear consistent with the observed Walker Circulation enhancement discussed in Figures 6-8.

In Figure 12 we also find a very strong pattern of warm and cold SST anomalies across the southern Indian Ocean. To the southeast of Madagascar we find very warm ocean conditions. To the northeast of this warm region we find very cool anomalies. This pattern appears associated with an anomalous sea level pressure gradient with low pressures in the southwestern Indian Ocean and higher pressures in the southwestern Indian Ocean. The sea level pressure gradients associated with these SSTs may be helping to drive the observed (Figure 7) and predicted (Figure 9) anomalous wind patterns just of the coast of East Africa, contributing to below normal moisture levels over East Africa (Figure 6).

Figure 13. Standardized difference between West Pacific and Central Pacific SST time series.
Figure 13. Standardized difference between West Pacific and Central Pacific SST time series.

FEWS NET scientists have published numerous papers linking the gradient between West Pacific and Central Pacific SSTs to dry East Africa conditions during March-May. Figure 13 shows these standardized March SSTs differences from 1982 through 2017. The strength of this gradient has increased after 1998 as the West Pacific warmed dramatically and we experienced relatively frequent La Niña conditions.

While the western-versus-central Pacific SST difference has weakened substantially over the past few months, as the West Pacific has cooled and the Central Pacific warmed, conditions in March 2017 still appear relatively unfavorable for eastern East Africa, with a standardized SST difference of +1Z. This gradient value is similar to 2012 and 2004, but substantially lower than strong drought years like 2000, 2008 or 2011 – when we also faced La Niña conditions in the eastern Pacific.

On the other hand, unlike 2011, we are facing the prospect of three poor Somali rainy seasons.  Crop simulations for East Africa indicate that most seasons in Somali poor fairly, so even mediocre 2017 rainfall could lead to poor crop production.

The SST difference time series shown in Figure 13 is fairly well correlated with 1997-2016 March-May rainfall over eastern East Africa (R=-0.7). We can use this relationship to make statistical forecasts of March-May rainfall based on observed March SSTs. Our paper on this topic is available here. To test this forecast strategy, we use take-one-away cross validation. We remove each year of data, re-estimate our regression coefficients, and compare our forecast results with the removed value. Figure 14 shows our cross-validated forecast results, along with our 2017 forecast (-0.8Z). Overall the cross-validated R2 is relatively low (R2=0.33), but all the recent droughts are forecast to be below normal – we have a good ‘hit rate’. One very wet year (2013) was predicted to be dry. The cross-validated standard error was 0.8Z, resulting in 80% confidence intervals of -1.8Z to -0.2Z, and a 65% chance of below normal rainfall (rainfall less than -0.5Z).

Figure 14. Standardized difference between West Pacific and Central Pacific SST time series.
Figure 14. Standardized difference between West Pacific and Central Pacific SST time series.

In conclusion, both our statistical forecast analysis and our review of recent climate diagnostics and weather forecasts indicate that below normal rains for eastern East Africa are likely. This may be especially true for the first part of the growing season. At present, observed east African moisture levels, precipitation, and land surface temperatures appear drier and hotter than usual. Warm SSTs to the east of Madagascar and Indonesia, combined with relatively cool SSTs in the southeast Indian Ocean and central Pacific, appear to be enhancing the Walker Circulation and increasing subsidence over East Africa (Figure 8), while also  reducing onshore moisture transports into East Africa (Figure 6, 7 and 9). It seems probable that these conditions will persist into April, contributing to below normal March-May rains in at least parts of eastern East Africa.

This pessimistic outlook is corroborated by our statistical forecast. This forecast indicates a substantial tilt in the odds towards below normal precipitation. While conditions may change rapidly if we transition into an El Niño-like climate state, a poor start to the March-May growing seasons in Southern Somalia and Eastern Kenya seems likely, and a third poor Somali growing season appears probable as well.

Please note that this outlook does not cover northern Somalia or most  of Ethiopia.