Forecast Update: East Africa Likely to Experience Six Droughts in a Row

This post was originally published on the University of California, Santa Barbara’s Climate Hazards Center (CHC) Blog on September 19, 2022. This post was written by Chris Funk, Gideon Galu, Diriba Korecha, Laura Harrison, Weston Anderson, Andrew Hoell, Kim Slinski and Greg Husak.
Summary
Reeling from three consecutive droughts capped by the worst March-April-May (MAM) rains in 73 years, famine looms in southern Somalia, and millions of people in eastern and northern Kenya, southern and southeastern Ethiopia and Somalia are facing crisis to emergency levels of food insecurity. Tailored forecasts, based on September sea surface temperature (SST) forecasts, indicate a “perfect ocean for drought” for the upcoming October-November-December (OND) rainy season, with a strong Indian Ocean Dipole (IOD) combined with a strong West Pacific Gradient (WPG), leading to a very high (90%) chance of moderate-to-severe OND rainfall deficits, with a most-likely outcome similar to signature historically severe and widespread drought years, like 2010 and 1998. For MAM 2023, September forecasts of the Pacific Western V Gradient (WVG) are surprisingly skillful (coefficient of determination (R2) of 0.78), and indicate a concerning La Niña-like pattern similar to conditions during many recent dry MAM rainy seasons. Using these forecasts to identify a large set of analogs, we conclude that there is a 62% chance of a below-normal MAM rainy season. The Eastern Horn of Africa is, therefore, very likely to experience a poor fifth rainy season in OND, and likely to undergo a sixth drought in 2023. Adverse hydrological and agricultural impacts would further degrade the precarious food security, water resources and health situation in this region, with adverse effects extending into early 2024.
The Alarming Reality: Four Unprecedented Sequential Droughts with Two More Droughts Likely
A relentless and increasingly intense series of droughts has impacted communities in semiarid and arid eastern East Africa (EEA). Coming on the heels of three poor 2020-2021 rainy seasons, 2022 brought the worst MAM long rains in 73 years. Resilience to these consecutive droughts’ impacts has been severely eroded by high food and fuel prices and conflict. Many families have been forced to go hungry, leave their homes, liquidate their assets and break away from typical coping strategies in a desperate search for relief. In Somalia, the situation is harrowing. Experts stress the critical need for a scaled-up, large-scale humanitarian response to reduce hunger-related deaths and to battle acute malnutrition. Without this, famine conditions will likely emerge in the Bay region of southern Somalia by the end of 2022. Southern and eastern Ethiopia and northern and eastern Kenya’s pastoral and marginal agricultural communities have also faced high staple food prices, poor agricultural production and very poor livestock production conditions, and are experiencing crisis to emergency levels of food insecurity. Acute malnutrition is increasingly harming the vulnerable, and the latest measurements in Kenya are striking. Acute malnutrition is compromising the health of more than 880,000 Kenyan children between six months and five years old and more than 155,000 new and expecting mothers. In southern and southeastern Ethiopia, where acute malnutrition and hunger-related mortality is also occurring, thousands of households lost their entire herds during the droughts, as part of the reported 3.8 million livestock deaths during the relentless, multiyear drought, as of late July.
An aspect of serious concern, from a climate standpoint, is that models are again indicating that similar adverse, large-scale climate conditions will be present for the next two main rainfall seasons in this region. Here, this post provides “tailored” forecasts for OND 2022 and MAM 2023. Growing out of the Famine Early Warning Systems Network (FEWS NET) research focused on understanding sequential OND/MAM droughts, similar tailored forecasts developed explicitly for EEA (Ethiopia, Kenya and Somalia east and south of 8°N and 38°E) successfully anticipated the OND 2016/MAM 2017, OND 2020/MAM 2021 and OND 2021/MAM 2022 sequences. Given this proven performance, reliable forecasts of drought probability and intensity can highlight when there are elevated risks for drought, and these estimates, in turn, can inform decision-making. A detailed discussion of why these tailored forecasts work so well for the MAM growing season is available here.
Last month, we presented a tailored forecast for OND 2022, and discussed it in detail here. The updated September OND 2022 outlook (described below) is similarly pessimistic, identifying a regional average Standardized Precipitation Index (SPI) value of -1.3 ± 0.8 radar reflectivity (Z), with a 90% chance of below-normal rainfall (i.e., SPI < -0.44Z). This pessimistic forecast is consistent with precipitation forecasts from the Intergovernmental Authority on Development’s (IGAD) Climate Prediction and Applications Centre (ICPAC) and other climate modeling centers. Hydrologic forecasts provided by the FEWS NET Land Data Assimilation System Forecast (FLDAS-Forecast) also indicate high probabilities of below-normal soil moisture and streamflow, in combination with predicted hotter-than-normal air temperature (land surface temperature (LST)), with significant implications for water insecurity and environmental degradation, in these arid and semiarid lands of the eastern Horn.

Here, as in September of 2020 and 2021, we present a long-lead forecast for the MAM WVG, and use it to identify a large set of 13 analogs. These analogs were years with similar SST forecasts. Based on these analogs, we conclude that a moderately below-normal MAM 2023 rainfall season (SPI MAM = -0.8Z) is likely, and that there is a 62% chance of below-normal rains (i.e., SPI < -0.44Z). Taken together, these results suggest that a fifth and sixth drought are, respectively, very likely during OND 2022 and likely during MAM 2023 (Figure 1). This would mean that an unprecedented four-season drought sequence would be extended by two more consecutive dry seasons, shown in Figure 1.
Throughout this post, we focus on EEA (Ethiopia, Kenya and Somalia; located east and south of 38°E, 8°N). We use the terms dry/normal/wet seasons to refer to 1-in-3 years below-normal, normal and above-normal rainfall outcomes. We also refer to dry seasons as “droughts” to avoid repetition. In general, the EEA region is so semiarid or arid that below-normal rainy seasons produce severe impacts on crops, rangelands and water resources. These forecasts are based on SST forecasts from the September update of the North American Multi-Model Ensemble (NMME). Data are standardized over a 1981-2022 baseline.
Outlook for OND 2022
Here we give a brief update for OND 2022, before focusing on key new information about the outlook for MAM 2023. These forecasts are based on September NMME SST forecasts. Please see our August 2022 post for additional details and earlier forecasts. These are quite similar to the September outlook, corresponding to active and stable large-scale climate drivers. The following links show the September NMME forecasts for the Indian and Pacific Oceans (SST standardized anomaly map) and a time series of the WPG index (time-series plot). The forecast map shows a strong west-to-east SST gradient in the tropical Indian Ocean (a negative IOD event) and a strong east-to-west SST gradient in the tropical Pacific Ocean (a strong WPG event) during OND 2022. The IOD measures the difference between the west and east Indian Ocean. The WPG measures the difference between the east and west Pacific. Negative IOD conditions are associated with cool western Indian Ocean and warm eastern Indian Ocean SST conditions.
The time series of OND West Pacific, east Pacific (NINO3.4) and WPG values underscore the likely, yet exceptional, three-year persistence of strong La Niña-related gradients in the Pacific Ocean. This plot emphasizes the unique, and dangerous, sequence of three negative NINO3.4 values (~ -1Z) in 2020, 2021 and 2022. It also shows that forecasts for equatorial West Pacific SSTs in these years were also extremely warm (~ +2Z). Our research has documented climate change enhancements to such extremely equatorial West Pacific SSTs, and shown how these very warm SSTs are associated with climate disruptions that reduce EEA OND rains. These three years of extremely warm OND SSTs combined with three years of cool NINO3.4 SSTs to produce three years of strong WPG gradients (~ -2Z), such as those in OND 2016, 2020 and 2021. Assuming the OND 2022 forecast verifies, which is very likely at this point, this will be the first time in the observational record that such a sequence of three strong-gradient La Niñas has occurred.

When both the IOD and WPG are negative, as will likely be the case in OND 2022, the overall Indo-Pacific SST gradients provide a “perfect ocean for drought” for the EEA OND rainy season.
Figure 2 summarizes the likely state of the Indian Ocean (left) and Pacific Ocean (right), as characterized by the IOD and WPG and based on the September forecast. Similar to last month, the models predict that a moderate-to-strong IOD event is highly likely during OND 2022. The horizontal red bar in the left panel of Figure 2 describes the 80% confidence interval associated with the OND 2022 IOD forecast. Based on this, there is a 90% chance of the IOD being less than -0.5Z. The models also continue to predict a strong negative WPG during OND 2022 (Figure 2, right panel). This outlook combines a strong probability of La Niña conditions in the eastern Pacific, as anticipated by the National Oceanic and Atmospheric Administration (NOAA), with a high probability of exceptionally warm equatorial West Pacific SSTs (time series plot available here). The 80% confidence interval suggests that there is a 90% chance of a WPG value of less than -1Z; therefore, there is a 90% chance of a strong negative WPG event. Historical forecasts predicted this state of things during seven other years, and of these, below-normal OND EEA rainfall was the outcome in six. During the single outlier, 2017, the OND EEA rainfall total was normal, but that season did not also contend with negative IOD conditions.

We next use a simple regression, based on the IOD and WPG forecasts, to produce an EEA OND SPI forecast with 80% confidence intervals. The same approach was used with observed SSTs in 2016, and NMME SST forecasts in 2020 and 2021. One elegant aspect of this approach is that it automatically assigns a stronger weight to the forecast WPG than to the IOD in longer lead time iterations, and then as lead time to the OND season decreases, a comparatively stronger weight is assigned to the IOD forecast. The WPG has a larger weight in the June and July forecasts, associated with higher skill in Pacific/El Niño-Southern Oscillation (ENSO) predictions, while the IOD has larger weight in the August and September forecasts, associated with improved model skill for the IOD forecast and because the IOD is a better predictor of EEA SPI values. Based on the September NMME, the forecast model equation is SPI OND = 0.09 + 0.65*IOD + 0.36*WPG (R2 = 0.72). The model is giving more weight to the IOD, but it should be noted that WPG and IOD covary, and many years (like 1998, 2010, 2016 and 2021) exhibited both negative IOD and WPG forecasts. Figure 3 shows the forecast scatterplot, with the 2022 prediction (SPI OND = -1.3Z) and 80% confidence intervals shown in red (-2.1, -0.5). The model indicates that the most-likely outcome is a very intense drought, similar to 1998 or 2010, with a 90% chance of SPI outcomes less than -0.7Z. The orange dots indicate likely analogs. In addition to years with similar regression estimates (1996, 1998, 2010, 2016, 2020 and 2021), we have also included 2005. This year (2005) exhibited a very strong, negative IOD event, which was not anticipated by the NMME. The negative IOD forecast for 2022 suggests that 2005 should be included as an analog because of the strong, negative IOD in 2005.
While no forecast is certain, Figures 2 and 3 indicate a high degree of predictability in the IOD, WPG and EEA SPI time series. The overall outlook is extremely pessimistic, with a 90% chance of negative IOD, WPG and EEA SPI outcomes in OND 2022. The closest forecast values are some of the driest OND seasons on record. Furthermore, we are entering the OND 2022 season with depleted soil moisture levels, herd resilience and surface water stores; even normal rains would unlikely to be enough to support full recovery from such conditions.
A Pessimistic Outlook for MAM 2023
When considering the outlook for the next MAM rainfall season, it is important to consider two related issues. The first is that there has been a very well-documented decline in MAM EEA rains, which is associated with a large, post-1997 warming in the western Pacific, as discussed most recently by FEWS NET in a long 2018 paper in the Quarterly Journal of the Royal Meteorological Society (QJRMS) and a short Bulletin of the American Meteorological Society (BAMS) paper, “Examining the Potential Contributions of Extreme “Western V” Sea Surface Temperatures to the 2017 March-June East African Drought” (Explaining Extreme Events (EEE)). In the EEA, between 1999-2022, there have been 11 dry seasons in 24 years (assuming a 1981-2022 baseline). That provides a background probability of a 46% chance for below-normal rains. Adding to these concerns is the fact that, following post-1997 La Niñas, the probability of EEA MAM dry seasons increases to ~75% (see Figure 2C in the multiagency alert). As discussed in our recent research (QJRMS, EEE) and used in our successful MAM forecasts for the 2017, 2021 and 2022 seasons, the decline in the long rains, and the opportunity to predict many of the events that have created this decline, links to human-induced warming in the “Western V” region (EEE), and especially the Western North Pacific (QJRMS). As the Pacific Ocean warms and La Niña events increase the trade winds, more oceanic heat is pushed into the western Pacific. Hence, following recent La Niñas, we see very predictable warm Western V and Western North Pacific SSTs, and these warm waters can drive La Niña-like circulation anomalies, even after a La Niña fades (as in 2017). These La Niña-like SST gradients enhance the probability of EEA MAM drought, even if ENSO conditions return to a neutral state.

These insights are especially important because climate models do a very poor job of predicting the MAM EEA rains. A more detailed explanation of our inductive, tailored forecasting approach is available here.
Tailored forecasts for the EEA MAM season begin with composites of observed, standardized SSTs for all below-normal 1981-2022 seasons (Figure 4). This is what the Indo-Pacific looked like during 1981-2022 dry seasons (composite mean): 1984, 1992, 1999, 2000, 2001, 2004, 2008, 2009, 2011, 2017, 2019, 2021 and 2022.
To gain perspective on the EEA MAM 2023 season, we can compare Figure 4 with predicted MAM 2023 SSTs (Figure 5). What we see in Figure 5 is that some very warm SSTs (> +1.5Z) are expected in the extra-tropical Western V areas, which are also strongly associated in the observed dry season composites (Figure 4). Equatorial west Pacific SSTs are expected to be modestly warm, and eastern Pacific NINO3.4 SSTs are near neutral. As we will describe below, the overall gradient is concerning, but not as alarming as similar September forecasts made for MAM 2022 and 2021, because in 2022 and 2021, the predictions included more La Niña-like equatorial SST gradients, with cool NINO3.4 SSTs and warmer equatorial west Pacific SSTs. Neutral NINO3.4 conditions, however, do not mean neutral EEA MAM rainy season outlooks. CHC research has linked Western V (EEE) and Western North Pacific (QJRMS) SSTs to elevated risks of drought.


The link between Western V SST forecasts and the frequency of EEA MAM dry seasons is apparent in Figure 6, which shows predicted and observed MAM Western V values. First, it is important to note that even at an eight-month lead, Western V SSTs are highly predictable, with an R2 value of 0.76. This predictability arises from a strong warming trend, combined with the influence of La Niña. Many of the warmest MAM Western V forecasts and observations follow recent (post-1997) OND La Niña conditions (shown with green crosses in Figure 6). Heat builds up in the Pacific, and when there is a La Niña, increased Pacific trade winds push heat into the extra-tropics, where it can linger even after La Niña fades. The 80% confidence intervals shown in Figure 6 indicate that there is a 90% chance of very warm Western V SSTs (> +1.3Z). When such forecasts have been made in the past, we see a strong propensity for below-normal MAM outcomes. We also see, however, some very wet seasons, like 2018 and 2020.
While modeling studies suggest that warm Western V SSTs, when combined with neutral ENSO conditions, can double the chance of below normal EEA MAM rains (EEE), changes in the equatorial east Pacific SSTs will also influence EEA rains, hence we can use the standardized difference between the NINO3.4 and Western V region to represent large-scale temperature gradients in the Pacific Ocean. We refer to this as the WVG.

Figure 7 shows the observed and predicted WVG values for MAM. The WVG is also very predictable (R2 = 0.78). Furthermore, many strong, negative WVG events follow OND La Niña events. This strong link arises from two factors. First, if La Niña conditions exist in boreal fall, there is a very low chance of there being MAM El Niño conditions. According to the Oceanic Nino Index (ONI), this has never happened during the past 73 years. Furthermore, we see frequent and predictable, warm Western V SSTs following recent La Niñas (Figure 6). Hence, in September, the models can anticipate strong, negative WVG events. In 2023, the models anticipate neutral NINO3.4 conditions combined with warm Western V SSTs, leading to a -1Z WVG forecast, and a 90% chance of WVG values being less than -0.4Z.
We can use this upper bound of the WVG forecast (-0.4Z) to define a large set of 13 analogs: 1989, 1996, 1999, 2000, 2008, 2009, 2011, 2012, 2017, 2018, 2020, 2021 and 2022. Two of these analogs had very wet EEA MAM seasons, two were normal and nine were below-normal. The median SPI of these years is -0.8Z, and bootstrapping 80% confidence intervals from this sample correspond to a range of -1.4Z to +0.6Z. These assumptions are used in Figure 1. Bootstrapping can also be used to estimate that there is a 62% chance of below-normal MAM rains. This estimate seems reasonable given that, 1) the post-1998 probability of dry seasons is already 46%, and 2) the latest SST forecasts indicate a high probability of warm Western V SSTs (Figure 6) and strong, negative WVG gradients (Figure 7). Several negative WVG seasons were wet (1989 and 2018), and no forecast is certain, but a pessimistic outlook for MAM 2023 appears warranted.
Conclusion: An Unprecedented String of Five or Six Dry Seasons Appears Likely

Repeated climate shocks are extremely difficult for poor households to cope with; as their fragile resources dwindle, they are increasingly exposed to food access and availability crises. For EEA, the past six years have been exceptionally volatile. Beginning with a La Niña-related drought in OND of 2016, the region has experienced eight below-normal rainy seasons, broken by extreme rains in 2018, 2019 and 2020, followed by a severe desert locust outbreak (Figure 1). Only OND 2017 and 2018 have been normal seasons. Most (but not all) of these extreme rainy seasons have been tied to either La Niña and/or IOD events, and combinations of extremely warm Indo-Pacific sea surface temperatures and naturally cool waters. These gradients drive extreme EEA rainfall in predictable ways. This is especially true for OND EEA when negative IOD and La Niña events co-occur, and the results presented here suggest that a fifth EEA drought is extremely likely, with very high (>80%) probabilities of below-normal October-November or OND rains (Figure 8A). This map also emphasizes how widespread the OND 2022 drought is likely to be. In mid-September, climate observations can reinforce this pessimistic outlook. Low-level wind and precipitable water anomalies over the Indian Ocean, and longitude-height transects of vertical velocities display the classic signature of a poor OND rainy season, with east-west gradients in moisture and subsiding dry air over the eastern Horn. This anomaly pattern impacts the EEA in a very robust way, and one of the most reliable areas of impact is in central and southern Somalia, where famine conditions will likely emerge unless relief efforts are expanded dramatically. NASA forecasts of EEA streamflow and root zone soil moisture are extremely low for October and November. And a sixth below-normal season appears to be likely in MAM 2023, because SST forecasts (Figure 5) resemble conditions during dry EEA MAM rainy seasons (Figure 4). Even at eight-month lead times, warm Western V SSTs (Figure 6) and strong, negative WVGs (Figure 7) are highly predictable, with R2 values of 0.76 and 0.78. In the past, when these conditions have been predicted in September, we have seen a high frequency (> 60%) of below-normal March-April or MAM rainy seasons (Figure 8B). Strong, negative WVG conditions can cause the Indo-Pacific circulation to respond in a La Niña-like way, even after a La Niña event ends, as in MAM 2017. Responding effectively to the current 2022 food security crisis will help cope with the exceptional risks posed by a potential sixth dry season in MAM 2023.
Related Resources
UCSB Climate Hazards Center
Famine Early Warning Systems Network