Eastern East African March-to-May Rains Fail Again, Creating Severe Food Insecurity in Somalia

Eastern East African March-to-May Rains Fail Again, Creating Severe Food Insecurity in Somalia

Chris Funk, Juliet Way-Henthorne

This blog presents time series of March-to-May East African Standardized Precipitation Index (SPI) for eastern East Africa and Somalia. We also examine annual average maximum daily temperature data from the new CHIRTSmax data set. While the latter data only goes through 2016, it does provide a valuable assessment of the general temperature trends in these regions.

The intent is to place the region’s recent drought in a historic context, while also emphasizing the dangerous repetitive nature of these severe rainfall failures. According to FEWS NET, Drought, conflict, and macro-economic shocks in East Africa have placed millions of people in extreme food insecurity. In northern Somalia, the Famine Early Warning Systems Network anticipates emergency levels of food insecurity – conditions approaching potential famine – for some 2.2 million people due to consecutive poor rains in October-December of 2018 and March-May of 2019.  During the typical peak of the March-May season (April), exceptionally dry conditions prevailed over the northwest, northeast, and parts of the south of the country. Between January and April, the number of children admitted to health posts for treatment of severe malnutrition increased by 92%. The country faces substantial shortfalls in funding for humanitarian assistance, with only about 1.2 million out of a targeted 2.2 million currently receiving aid. This is in a situation where conditions are likely to deteriorate rapidly as stressors associated with the current dry season accumulate. 

Given this context, we begin by examining a map of March-May CHIRPS rainfall anomalies (Figure 1), calculated using a 1981-2010 baseline. As noted by many agencies, these rainfall totals were very poor over almost all of Kenya, most of eastern Ethiopia, and much of Somalia. While the most severely food insecure region is northern Somalia, here, we focus on time series of rainfall and temperature for all of Somalia and eastern East Africa (delineated with a light blue polygon in Figure 1), because i) the local food security is tied to regional food deficits, ii) the data in northern Somalia is somewhat uncertain, due to an almost complete lack of weather stations, and iii) the climatic patterns associated with these droughts tend to act on a regional spatial scale. 

Figure 1. March-to-May CHIRPS rainfall anomalies.

To produce long 1900-2019 time series of precipitation data, we use the 1900-2014 station-only Centennial Trends data set and the 1981-2019 CHIRPS precipitation archive. Time series of regional averages of the 1900-2014 and 1981-2019 Centennial Trends and CHIRPS data sets are highly correlated, with correlation values of approximately 0.9. This allows us to combine the two data sets to place the current drought in historic context. The combined time series use 1981-2019 CHIRPS values and bias-corrected 1900-1980 Centennial Trends rainfall data. The bias correction accounts for small mean differences between the two archives. Regional rainfall averages were calculated over the eastern East Africa region, outlined in light blue in Figure 1. This region has been used in numerous previous studies by the Climate Hazards Center. Regional averages were also calculated over Somalia, which was selected because of the very high levels of food insecurity in that country. These time series are expressed as SPI values, which have a mean of 0 and standard deviation of 1. 

Figure 2 shows the 1900-2019 March-May SPI time series for eastern East Africa. Since 1999, there have only been a few wet seasons: the very wet 2018, and 2010 and 2013. Conversely, 9 years had poor rainfall: 1999, 2000, 2001, 2004, 2008, 2009, 2011, 2017, and 2019. This means that the “new normal” in eastern East Africa has been a substantial or severe March-to-May drought about every other year. Such sequential rainfall deficits have had dangerous impacts, eroding resilience, economic reserves, and herd size and health.

Figure 2. Time series of eastern East Africa March-to-May SPI values.



Figure 3 shows a similar time series for Somalia. The 2016, 2017, and 2019 March-May Gu seasons were poor, as were the national averages for 2016 and 2018 during the October-December Deyr season. In Somaliland, the 2017 Deyr rains were also poor. Somalia has experienced poor Gu rains in 1999, 2000, 2001, 2004, 2008, 2009, 2011, 2012, 2017, and 2019—ten out of the last 21 years.

Averaging these time series using running 20-year averages allows us to visualize how the mean March-May rainfall conditions are changing over Somalia and East Africa (Figure 4). We see a large ~-1 standardized anomaly decline between the beginning of our data (the 1900-1919 time period), and the last 20-year time period (2000-2019). Dry has become the new normal in regions that struggle to produce enough food in the good years. Following the 1997/98 El Niño, the eastern Indian Ocean and Western Pacific saw dramatic increases in sea surface temperatures, and our research has suggested that this warming can describe the substantial post-1998 decline in East Africa precipitation. 

Figure 3. Somalia March-May Standardized Precipitation Index time series.

Figure 4. Eastern East Africa and Somalia SPI time series smoothed with running twenty-year averages.

We next briefly examine (Figure 5) annual averages from the new Climate Hazards InfraRed Temperatures with Stations Tmax product (CHIRTSmax).  The paper and data are available here. This product is uniquely suited to track temperatures in areas with poor station coverage (like Somalia). The Tmax values are based on the monthly averages of daily maximum temperatures. Here, we are examining the average of these 12 monthly values. While we are working on setting up operational updates for this data set, for now, the archive stops in 2016. What we can see (Figure 5), however, is a clear upward trend that has already increased average maximum temperatures from about 32.8 ° Celsius (91°F) to about 33.8 ° Celsius (92.8°F), just in the last 30 years or so. Very warm conditions have become substantially warmer.  These increased temperatures will exacerbate the impacts of droughts by increasing atmospheric water demand, damaging crops, and weakening the health of livestock. This means that when there is a drought, limited soil and plant moisture will return more quickly to the atmosphere through evaporation and transpiration. During grain filling, crops are likely to exhibit more temperature stress. And livestock herds will likely suffer from more health problems due to extremely warm temperatures. 

Figure 5. Average annual maximum daily temperatures for Somalia (red) and eastern East Africa (dark red), based on the new CHIRTSmax data set.


A Brief Analysis of March-May 2019 Climate Conditions

This blog concludes with a brief analysis of the observed March-May climate conditions. Sea surface temperature conditions in this season did not resemble those during the most recent drought years (not shown). Rather, mild El Niño-like conditions were present in the equatorial Pacific, accompanied by warm ocean waters in the southwest Indian Ocean. The latter likely contributed to the formation of cyclone Idai, a very dangerous cyclone which struck southern Africa in mid-March. Low-level geopotential height and wind anomalies (Figure 6) indicate a low-pressure anomaly in the southern Indian Ocean that helped draw moisture away from eastern Africa. Also notable were large positive high-pressure anomalies to the southeast of Australia, near 140°W and 45°S. This high-pressure cell helped drive anomalous winds and moisture transports westward into the equatorial west Pacific. 

Total precipitable anomalies measure the change in the total atmospheric moisture at a given location. Anomalies in total precipitable water track changes in moisture over time. What we see in March-May anomaly fields (Figure 7), is 1. a large reduction in moisture over eastern East Africa and the equatorial western Indian Ocean, and 2. a large increase in moisture near Indonesia. These moisture anomalies are due to the winds and pressure anomalies shown in Figure 5. Despite the absence of La Niña or strong Indian Ocean dipole conditions, conditions over the southern Indo-Pacific (Figure 6) set up a strong moisture dipole (Figure 7) that helped produce an eastern East African drought. 

Maps of reanalysis precipitation anomalies (Figure 8) reinforce the idea that even though the sea surface temperature conditions were different in 2019, the east African drought was once again associated with increased rainfall over the Maritime Continent region. The total precipitable water anomalies and precipitation anomalies over this region are striking in magnitude (~8 kg m-2 and 6 mm day-1) and quite similar to those that accompany typical La Niña-like conditions. The energy released by this precipitation is known to trigger subsidence and dry conditions to the west, over equatorial East Africa and the Indian Ocean. So while the sea surface conditions were different than those associated with La Niña-like droughts in 2011 or 2017 and most recent drought years, the equatorial rainfall dipole appears similar. In 2019, however, it appears likely to be driven by conditions in and over the southern Indian and southwestern Pacific Ocean basins. 

Figure 6. March-May low-level geopotential height anomalies and anomalous wind vectors.

Figure 7. Total Precipitable Water anomalies for March-May.

Figure 8. March-May Reanalysis rainfall anomalies.

Summary of GeoCOF Training, AGRHYMET 24-27 June 2019

Summary of GeoCOF Training, AGRHYMET 24-27 June 2019

Contributors:  Alkhalil Adoum, Tamuka Magadzire

Reinforcing partnership and capacity in seasonal forecasting in West Africa has been part of SERVIR/West Africa’s objective.  In collaboration with AGRHYMET and USGS/FEWS NET, GeoCOF training has been jointly planned and carried out to reinforce AGRHYMET and ACMAD climate scientists’ capacity in seasonal forecasting.

GeoCOF is a statistical software tool for seasonal forecasting developed by the FEWS NET/USGS activity in support of regional seasonal forecasting processes, with assistance from the SADC Climate Services Centre.  It uses multiple-linear regression modeling between climatic predictors (e.g. sea surface temperatures) and the predictant—usually seasonal rainfall totals. By implementing this training, SERVIR/WA and FEWS NET aim to improve the capacity of ACMAD and AGRHYMET climate scientists in seasonal forecasting.

The training took place at the AGRHYMET Regional Center from 24-27 June 2017, with about 15 participants from ACMAD, AGRHYMET, and FEWS NET.  The training was completed to the satisfaction of the participants, who were given attendance certificates. Looking toward the future, a social media communication group of participants was set up, and a number of proposals were put forward as a way to continue efforts to improve the accuracy of seasonal forecasts. These included:

  • regular technical information-sharing meetings involving AGRHYMET, ACMAD, FEWS NET, perhaps quarterly, covering themes such as short-term training on selected topics and sharing of results from forecast-related research and operational methodologies
  • Assess and validate the relative performance of various sources of satellite data (e.g. CHIRPS, RFE, TAMSAT, etc.) compared to rain gauge data
  • Investigate the use of alternative predictors regarding whether they can improve upon the accuracy of the SST-based forecasts
  • Propose a 2-3 day GeoCOF training (by the participants) of colleagues at AGRHYMET and ACMAD who did not manage to fully attend the original training


Our special thanks to SERVIR/West Africa for full financial support and AGRHYMET Regional Center for the logistic support.