A Late April Assessment Indicates Poor Long Rains and Low Gu Harvests for Somalia

Chris Funk, Pete Peterson, Peris Muchiri, Diego Pedreros, Greg Husak, Diriba Korecha, Gideon Galu, Laura Harrison, Will Turner, Marty Landsfeld and Shrad Shukla

This post examines conditions across East Africa at the close of April. As predicted by the  CHG, ICPAC, and a joint assessment by FEWS NET, WFP, FAO and JRC, exceptional warming in the West Pacific appears to have continued to produce subsidence and drying over East Africa.  At present (Figure 1), the FEWS NET food security outlook for Eastern Africa is very concerning, with a June-September outlook calling for IPC phase 3 (crisis) or 4 (emergency) across Kenya, South Sudan, southern Ethiopia and Somalia. The FEWS NET perspective seems largely congruent with the most recent seasonal assessment by the World Food Programme (here).

In Kenya, southern Ethiopia and Somalia June-September food security outcomes will be strongly influenced by rainfall in March and April, since most of the long rains tend to come in these months, and moist soils during this period support the establishment of healthy crops. As we will show below, poor March-April rainfall can be a good predictor of low crop production in Somalia.

FEWS NET Food Security Outlook for June-September 2017. Orange and red shades denote crisis and emergency conditions. Black shading in South Sudan indicates famine.
Figure 1. FEWS NET Food Security Outlook for June-September 2017. Orange and red shades denote crisis and emergency conditions. Black shading in South Sudan indicates famine.

We begin by looking at the observed March to late April rainfall performance using NOAA CPC ARC2 and CHIRPS rainfall fields enhanced with data provided by FAO SWALIM and the National Meteorological Agency of Ethiopia. All evidence indicates poor rainfall performance for much of the Greater Horn of Africa. We then examine the relationship between Somali Gu Sorghum crop production statistics and March-May rainfall. We find that March-April rains are by far the most important – and current March-April totals indicate very poor sorghum production totals for 2017. We conclude with a brief look at the current climate conditions and the performance of the NOAA GEFS weather forecasts.

March-April Rainfall Assessment

March-April 24th ARC2 anomalies (Figure 2) and March-April 20th ‘enhanced’ CHIRPS data, expressed as standardized precipitation index values (Figure 3) are in strong agreement that there has been wide-spread drought across almost all of Kenya, Somalia and Uganda as well as southern Ethiopia, eastern South Sudan, north-central Tanzania and western Yemen. Note that the units in these maps are different. It is useful to consider rainfall anomalies both in terms of absolute magnitude (Figure 2) and as standardized anomalies (Figure 3). In Figure 2 we note very large (100 mm) rainfall deficits across central Kenya, Uganda and in the SNNPR region of Ethiopia; these large deficits could be associated with large disruptions in key crop growing areas. In Figure 3 we see that the seasonal rainfall progress has been exceptionally dry, in a statistical sense (<-1 standard deviations) across most of the Horn.

NOAA CPC ARC2 precipitation totals from March 1st-April 24th 2017.
Figure 2. NOAA CPC ARC2 precipitation totals from March 1st-April 24th 2017.

In Somalia, where even normal rainfall totals are characteristically low, we find that our estimates indicate an exceptionally poor March-May season. While the results in Figure 2 do not indicate performance over the last dekad of April, CPC ARC2 totals for April 21, 22, 23 and 24 show almost no rainfall over Somalia.

CHIRPS Standardized Precipitation Index values from March 1st-April 20th 2017.
Figure 3. CHIRPS Standardized Precipitation Index values from March 1st-April 20th 2017.

For Somalia, it is important to realize that we have been able to incorporate a fairly dense network of gauge observations provided by FAO SWALIM. Figure 4 shows April 1st to April 20th enhanced CHIRPS rainfall totals. The numbers on this map show rainfall totals from either the SWALIM stations or WMO GTS observations. Across all of East Africa, very few regions appear to have received more than 60 mm of rain so far in April. From a crop perspective, this means that planting has been delayed across Kenya and Somalia, and crop growth is likely to be running substantially behind normal. For example, ARC2 data at Meru, in central Kenya, indicates a seasonal accumulation of ~120 mm, less than half of the normal 270 mm. Results in Baidoa (Bay Region Somalia), Dif in far eastern Kenya, and  Kibre Mengist in south-central Ethiopia are similar.

Enhanced CHIRPS precipitation for April 1 to April 20, 2017.
Figure 4. Enhanced CHIRPS precipitation for April 1 to April 20, 2017.

Seasonal Rainfall Ensembles

To examine likely outcomes for the total March-May season we have combined March 1 to April 20 CHIRPS rainfall totals and then examined the possible combinations of future rainfall by sequentially inserting one of the past 36 years (1981-2016) and then examining the associated distribution of seasonal rainfall totals. We begin by showing these results for the Bay Province of Somalia (Figure 5), which is currently facing food security crisis (i.e. just short of famine) conditions (see Figure 1). We start at a low seasonal total of 42 mm for Bay on April 20th – this low value and large deficit is primarily due to the low April rainfall totals, as shown in the SWALIM station data (Figure 4). To explore the remainder of the season, we sample the CHIRPS data using all prior seasons. Advancing one dekad by this approach gives us a seasonal total for the end of April of 90 mm, only 60% of long term average. As we will see below, this large March-April deficit will very likely be associated with large crop production deficits. Proceeding through the rest of May in this same fashion we arrive at a spread of possible outcomes ranging from near normal to very low, with an average outcome of 174 mm, 74% of the long term average. In the context of the past 20 years, this would be a 1-in-5 year drought (i.e. 20th percentile); 2011, 2001, 2008, and 1999 were a little drier.

Cumulative rainfall ensemble for Bay region of Somalia. Observed CHIRPS dekads are used from February through April 20th.
Figure 5. Cumulative rainfall ensemble for Bay region of Somalia. Observed CHIRPS dekads are used from February through April 20th.

Repeating this process for each pixel, we can assess the probability of March-June rainfall being less than 85% of the long term average (Figure 6) and less than 50% of the long term average (Figure 7). Figure 6 indicates that the regional as a whole is very likely at this point to end with below normal rainfall. The certainty of this outcome is much less in northern East Africa, although some Belg growing regions in the eastern highlands of Ethiopia and the northernmost parts of Somalia and Yemen are shown to have an 80% chance of below normal rains. Across southern Somalia, southern Ethiopia, all of Kenya and much of Tanzania a below normal outcome seems almost certain, given historical rainfall distributions.

Probability of March-May rainfall totals being below normal (less than 85% of average) based on historical rainfall distributions.
Figure 6. Probability of March-May rainfall totals being below normal (less than 85% of average) based on historical rainfall distributions.

Looking at areas likely to see catastrophic (<50% of normal) March-June outcomes, we see that such an outcome is very likely (>50% probability) across much of Kenya and near the Mandera triangle area at the intersection of Somalia, Ethiopia and Kenya. These are regions that have received low March-April 2017 rains (Figure 2 and 3) and have historically had short ‘long’ seasons – such that they now have low chances of anything but poor outcomes. We can see this in more detail by looking at cumulative rainfall totals for the Eastern (Figure 8) and Central (Figure 9) province of Kenya using the USGS Map Viewer. For Eastern province, seasonal rainfall totals have been extremely low (~140 mm), in line with 2010/11, and far below the typical seasonal total of ~310 mm. Historically, rainfall stops in this region at the end of April, hence we find a very high probability of very low rainfall (Figure 7).

Probability of March-May rainfall totals being extremely low (less than 50% of average) based on historical rainfall distributions.
Figure 7. Probability of March-May rainfall totals being extremely low (less than 50% of average) based on historical rainfall distributions.

For the densely populated, well observed Central Province of Kenya, we find that seasonal rainfall accumulations are the lowest in the 2001-2016 RFE2 period of record. The observed 257 mm is far below the average of 518 mm, and substantially lower than values in 2010-2011 at this time (330 mm).

Cumulative RFE2 rainfall totals for the Eastern province of Kenya. Data from https://earlywarning.usgs.gov/fews/mapviewer/index.php?region=af.
Figure 8. Cumulative RFE2 rainfall totals for the Eastern province of Kenya. Data from https://earlywarning.usgs.gov/fews/mapviewer/index.php?region=af.

In many of these arid land regions current assessments of water hole conditions indicate alert or near-dry conditions – at or near the end of the rainy seasons – it is very likely that conditions will soon get worse in these locations as evaporation takes its toll on surface water stores.

Cumulative RFE2 rainfall totals for the Eastern province of Kenya. Data from https://earlywarning.usgs.gov/fews/mapviewer/index.php?region=af
Figure 9. Cumulative RFE2 rainfall totals for the Eastern province of Kenya. Data from https://earlywarning.usgs.gov/fews/mapviewer/index.php?region=af

Assessing likely crop growing outcomes for Somalia’s Gu season

We next turn to Somalia’s Gu sorghum production outlook. This analysis is based on 1999-2016 Gu sorghum production for three key growing regions: Bay, Shabelle Dhexe and Shabelle Hoose. Our objective here is not to produce a precise crop production assessment for Somalia Gu production, but rather to highlight that the poor March-April rainfall totals, alone, are likely to produce serious reductions in crop production. Both the available production data and crop water requirement estimates from a simple crop model indicate that May rainfall will be unable to make up for the poor rainfall distribution in April. Both the crop production and CHIRPS rainfall data in Somalia are likely to be noisy. This analysis is intended to imply that a poor harvest is very likely – but not provide a precise quantitative Somali production forecast.

We began by totaling sorghum production from Bay, Shabelle Dhexe and Shabelle Hoose and related these totals to CHIRPS rainfall from March-May, March-April, and May.  We found an okay level of correspondence between crop production and March-May and March-April rainfall, with corresponding R2 values of 0.22 and 0.34. The correlation between sorghum production and May rainfall was actually weakly negative (-0.25), which helps explain why using March-April totals, rather than March-May totals, improved our predictive skill. The corresponding correlation between April rainfall and Gu sorghum production was fairly high (0.51). April is the key month for crops, according to the empirical data.

To generate a prediction of Gu production in Southern Somalia we regressed (Figure 10) March-April rains in Bay, Shabelle Dhexe and Shabelle Hoose against observed production anomalies (based on a 1999-2016 baseline). We then extracted the average April 1-20 rainfall total from our enhanced CHIRPS data set (27 mm) and assumed 25 mm for the last dekad of April. This latter value was a compromise between the April 21-25 observed ARC2 outcome (~0 mm) and the optimistic weather forecasts (discussed further in the next section). These assumptions and our regression indicate very low March-April rainfall totals and corresponding very poor level of crop performance (-50%), similar to previous recent drought years.

Figure 10. Scatterplot of southern Somalia sorghum production percent anomalies and March-April rainfall [mm].
Figure 10. Scatterplot of southern Somalia sorghum production percent anomalies and March-April rainfall [mm]. The circle marked in red is the production estimate for 2017, based on March-April rainfall.
To further corroborate these results we looked at the relationship between pixel-level Water Requirement Satisfaction Index (WRSI) end-of-season values and onset of rains dates for Bay (Figure 11). The WRSI is an index that shows crop water stress – a value of 100 means no water stress. Note that under normal conditions (with a start in the first or second dekad of April), Bay WRSI values are low (~50). This is a very marginal farming region. In the WRSI model the onset of rains triggers crop growth. It is calculated by identifying areas that receive at least 25 mm of rainfall in a dekad, followed by 20 more mm in the next 20 days. Since our enhanced CHIRPS data indicate average Bay rainfall totals of 12 and 15 mm for the first and second dekad of April, it seems unlikely that the region experienced onset conditions in those dekads. While this outcome was uncommon, we see a large decline in end-of-season WRSI. Our crop simulation results reinforce the critical nature of good early rains.  Crops require weeks of decent rainfall to emerge, put on green vegetation and then divert resources to build up grains. Even if southern Somalia receives torrential rain in the next several weeks it seems unlikely that conditions will be conducive to rainfed agriculture.

Distribution of end of season WRSI in Bay Somalia, stratified by onset date.
Figure 11. Distribution of end of season WRSI in Bay Somalia, stratified by onset date.

GEFS Forecasts, Current Climate Conditions and What We Know Now

GEFS 7 Day precipitation forecast [mm] April 25-May 2nd.
Figure 12. GEFS 7 Day precipitation forecast [mm] April 25-May 2nd.
We next briefly explore the skill of NOAA’s Global Ensemble Forecast System (GEFS) weather forecasts and discuss their current forecasts for the Horn. The issue we focus on here is the likelihood that current optimistic forecasts (Figure 12) for rainfall over Somalia will verify. These forecasts call for more than 80 mm of rain during the upcoming week. Such relief could definitely improve rangeland conditions and prospects for irrigated agriculture. CHG assessments of GEFS forecast skill (Figure 13) show some promising areas of high correlation, but not for the first half of May or the end of April. Thus while these forecast could prove accurate, the should probably be treated cautiously, since historically they have had fairly low correlations with observations over Somalia, and we have yet to see any rainfall totals approaching this magnitude appear across the region. On the other hand, it is certainly plausible that Somalia could see a few weeks of healthy rain before the season comes to a close. Such rains could improve rangeland conditions and water availability, but may not provide much relief to crop areas in Somalia. Based on the data analyzed here, reliable maps of observed rainfall (Figs. 2-4) provide a solid basis for predicting Gu agricultural outcomes, which look bleak for 2017. Central and Eastern Kenya and the Mandera triangle region also appear very likely to large precipitation deficits.

Correlation between GEFS 2-week forecasts and CHIRPS data for different initialization dates.
Figure 13. Correlation between GEFS 2-week forecasts and CHIRPS data for different initialization dates.

For Kenya, field reports indicate that the area planted with crops is less than 50% of normal in the southeastern lowlands. Central, eastern and coast Kenya has experienced a late onset of rains, only episodic precipitation, and a shortened growing season. At present, maize crops are only just emerging or are very young. Forecasts from the Kenya Met Department are for a normal ending time for the March-May season, suggesting that these crops are unlikely to have time to complete germination and grain filling. The poor March-May trends appear to be part of an ongoing drying trend (Figure 14) associated with warming in the Western Pacific. This trend has helped produce repetitive shocks, reducing household food security and resilience.

March-May rainfall trends based on CHIRPS data enhanced with a dense network of Kenya Met Department observations.
Figure 14. March-May rainfall trends based on CHIRPS data enhanced with a dense network of Kenya Met Department observations.

As an example, consider a time series of NDVI anomalies for the Coastal Province of Kenya (Figure 15). Since 2009, typical vegetation conditions have been below normal, with large drought events in 2009, 2010/11, 2012, and 2016/17. The current 2016/17 appears to be the worst event.

USGS eMODIS NDVI anomalies from the Coastal Province of Kenya.
Figure 15. USGS eMODIS NDVI anomalies from the Coastal Province of Kenya.

Real-time monitoring of current Belg season in Ethiopia

Diego Pedreros, Diriba Korecha, and Chris Funk

Background

Ethiopia experiences three climatic seasons, with high rainfall during two major rainy seasons. The country’s economy is largely agrarian, in which pure farming, mixed farming, and livestock herding (pastoralists) are common practice. Consistent increases in population, over-exploitation of natural resources such as natural forest and swampy lands for agriculture, and well as an alarming expansion of urbanization impose untenable burdens on Ethiopia’s social and economic strata. The agricultural sector in particular supports 85% of the population and thus is central to the livelihoods of the rural poor in Ethiopia (Conway et al. 2007; Deressa 2006). Current agricultural and herding practices in the country mainly rely on seasonal rainfall and water available in perennial rivers and dams; only a small fraction of Ethiopian agriculture is irrigated.  A significant decline in annual agricultural production has been observed during drought years (Lemi 2005).

It has been documented that food shortage and scarcity of water have led to local and nationwide famines, mainly due to complete or partial failures of short (Belg, February-May) and long (Kiremt, June-Sep) rainy seasons over various parts of Ethiopia (NMSA, 1996). The failure of seasonal rainfall is often caused by either misplacement or weakening of large-scale seasonal rain-producing systems. Stephanie et al. (2016) documented that droughts and famines, such as the socio-economic catastrophe of 2011, call attention to the need for reliable seasonal forecasts for rainfall in Ethiopia to allow for agricultural planning and drought preparations.

Drought-related famine is the result of several factors, where lack of rainfall is only the first (Webb et al. 1992). Famine, in itself, cannot be taken as evidence of drought, while it is also not possible to assess the role of societal conditions without knowledge of the extremeness of rainfall deficits (Viste et al, 2013). To address this quandary, some scholars (Funk et al. 2008; Williams and Funk 2011) have documented rainfall declines in southern and eastern Ethiopia, especially in the spring season.

Dry Belg seasons affect all of Ethiopia, causing the largest relative precipitation deficits in the south, where it is the main rainy season. The southern and southeastern lowlands have been drier than normal in every year from 1998 through 2010, with 2009 having the worst drought incidences. This description considers normal as being the average if rainfall over the years 1981-2010. For instance, Viste et al. (2013) noted that even though both the Belg and Kiremt seasons were dry in both 1984 and 2009, the large-scale patterns reflect the fact that in 1984 the Kiremt was one of the driest seasons, whereas the Belg was particularly dry in 2009. The core of the 2009 drought was located farther south, covering the Horn of Africa and the northern part of East Africa, where the February–May season is the main rainy season.

Belg as the main rainy season over south and southeast Ethiopia

While Kiremt is the main rainy season in many parts of Ethiopia, and Belg rains contribute about two-thirds of the annual rainfall for the southern and southeastern Ethiopia (Figure 1). For the Belg season, precipitation shows strong variability and is less reliable both from a temporal and spatial viewpoint, especially over the northern half of Ethiopia.

Percentage contribution of Belg rainfall for annual rainfall totals
Figure 1: Percentage contribution of Belg rainfall for annual rainfall totals

The Belg rains start falling over southern Ethiopian in February. During a wet year, rain usually starts around mid-January and continues without prolonged dry spells through February over Belg growing regions as well as Belg rain-benefiting regions of Ethiopia (Figure 2).

Spatial distribution of February mean rainfall climatology in mm
Figure 2: Spatial distribution of February mean rainfall climatology in mm

In March (Figure 3), Belg rains start to expand north and eastwards and cover the southwest to northeast regions of the Rift Valley. The western and eastern escarpment of the Rift Valley regions also receive rains. Much of the Southern Nations, Nationalities and People’s Region (SNNPR), the central and eastern half of Oromia, and the eastern Amhara regions usually receive more rainfall than other parts of the country.

Spatial distribution of March mean rainfall climatology in mm
Figure 3: Spatial distribution of March mean rainfall climatology in mm

Belg rains reach their peak in April (Figure 4), particularly over the regions where Belg is the main rainy season as well as secondary rainy season (over south-southeast, central, east and northeast Ethiopia). Aside from those regions, Belg rains extend eastward and covers the Somali region, where Belg is the main rainy season. In many cases, severe droughts happen when April rains fall short of their predicted climatological values.

Spatial distribution of April mean rainfall in mm
Figure 4: Spatial distribution of April mean rainfall in mm

May is the last month of Belg season (Figure 5), when rain starts to retreat/decline, slowing from eastern and southern sectors of Ethiopia. In contrast, Kiremt seasonal rains further expand west and northwards. Sometimes, Belg and Kiremt seasons merge during a fast transition from El Nino (Belg) to La Nina (Kiremt) or El Nino (Belg) to neutral (Kiremt) episodes.

Spatial distribution of May mean rainfall climatology in mm
Figure 5: Spatial distribution of May mean rainfall climatology in mm

Homogeneous regimes of Belg season

In Ethiopia, onset and cessation of seasonal rainfall vary considerably within a few kilometers distance due to altitudinal variations as well as orientation of mountain chains and their physical influence on atmospheric flow. In particular, diverse topography and strong seasonal variation over the country indicate the potential physical justifications to delineate rainfall patterns on various spatial scales. Based on existing evidence, rainfall seasonality, and above all by considering localized social and economic practices, we delineated the country into four homogeneous regimes. The characteristic of each homogeneous regime is mainly a reflection of their typical seasonal agro-climatic practices as well as the contribution and benefit of seasonal rainfall that prevails in each regime (Figure 6).

Homogeneous regimes of Belg season in Ethiopia
Figure 6: Homogeneous regimes of Belg season in Ethiopia

Social and economic description of homogeneous regimes

  • Purple area (Regime I): This regime receives 40-70% of the total annual rainfall during February, March, April, and May, with rainfall maxima occurring from late March to mid-May), a variety of grain crops (maize, sorghum, teff, barley), root crops (potatoes), pasture, and water storage are commonly practiced over various portions of the region. Most parts of this region are identified as pastoralist.
  • Green area (Regime II): Belg is the second rainy season and contributes 30-50% to annual rainfall totals. These regions usually receive less rainfall compared to southern Ethiopia despite the fact that they rely on Belg rains to produce short-cycled crops, root crops, land preparation for long-cycle crops, pasture, and water storage.
  • Blue area (Regime III): This regime receives up to 30% of their annual rainfall totals from Belg rainfall. Belg rains usually fall from March and continue without prolonged dry spells up to October/mid-November. Belg rains especially contribute to land preparation, planting/sowing of long-cycle (18 dekads) crops (e.g., maize, sorghum), and to contribute for extension of long rainy season that usually spans from March to November.
  • White area (Regime IV): Most parts of these regions receive less than 20% of annual rainfall totals from Belg rainfall because Kiremt is the major rainy season here, although it rains from mid-April/May to September/October.

 

The 2017 Belg Season up to the 2nd dekad of April

Up to the second dekad of April 2017 (to April 20th), the south eastern part of Ethiopia has received a low percentage of the expected rainfall.  See Figure 7.

Rainfall Percent anomalies (1st of February- 20th of April) for the 2017 season.
Figure 7: Rainfall Percent anomalies (1st of February- 20th of April) for the 2017 season.

On a regional level, we examined Regime 1 as shown on Figure 6.  At the end of dekad 2 of April 2017, this region overall had received below average rainfall. Figure 8 shows the seasonal rainfall accumulation for the entire Regime 1 region.  The black line shows how the overall rainfall for the region for the 2017 season deviates from the long term mean (thick red line). This season has received exceptionally low rainfall. The current total (~80 mm) is only 57% of the long term average (~140 mm). Only 3 out the 36 prior had lower seasonal totals, making this season a 10th percentile event – a one-in-ten-year drought if current conditions persist.

Seasonal cumulative rainfall for Regime 1 for every year since 1981 to 2017. Black line represents the development of the 2017 season, red thick line shows the long term mean.
Figure 8: Seasonal cumulative rainfall for Regime 1 for every year since 1981 to 2017. Black line represents the development of the 2017 season, red thick line shows the long term mean.

WRSI for grasslands

The low rainfall during the season in Regime 1 translates into low soil moisture availability for plants.  Figure 9 shows the WRSI for grasslands at the 2nd dekad of April.  By this dekad only a few areas have received enough rainfall for the WRSI model to start. We have been monitoring polygon 1 in southern Ethiopia as shown in Figure 9.  This polygon showed signs of stress, based on WRSI values, since the last dekad of March and exhibited no signs of recovery by the 2nd dekad of April.

The WRSI model show that just a few areas have received enough rainfall to support grasslands and some of these areas are already showing stress, as shown in Polygon 1.
Figure 9: The WRSI model show that just a few areas have received enough rainfall to support grasslands and some of these areas are already showing stress, as shown in Polygon 1.

Looking in more detail into Polygon 1, Figure 10 shows the seasonal cumulative rainfall for every year since 1981 using CHIRPS data. The black line represents the 2017 season, the thick red line shows the long term average, and the two black squares on the axis represent the 33rd and 67th percentiles. The plot shows that the season started as normal, but by the 1st dekad of March rainfall values started to decrease and have not recovered since.     The current total (~120 mm) is only 60% of the long term average (~200 mm), and we see that this season’s total to date is also associated with a one-in-ten-year drought, if the current dryness continues.

Spatially averaged seasonal cumulative rainfall for every year since 1981 for polygon 1. Black line depicts the development of the 2017 season to the 2nd dekad of April.
Figure 10: Spatially averaged seasonal cumulative rainfall for every year since 1981 for Polygon 1. Black line depicts the development of the 2017 season to the 2nd dekad of April.

Figure 11 shows the ensemble of potential outcomes for Polygon 1 based on the observed cumulative rainfall to the 2nd dekad of April, completing the season with historical CHIRPS data for each year. The red line shows the long term average, the black squares show the 33rd and 67th percentiles of the historical data while the red dot is the average of the ensemble. The green triangles represent 1 standard deviation (+) or (-) from the ensemble’s average. The outlook table on Figure 11 shows the probability at the end of the season.  This results indicates that there is a 94% probability that the seasonal accumulation for polygon1 is below normal.    Looking at the ±1 standard deviation range shown in Figure 11, the likely outcomes range from near normal to extremely dry. The most likely outcome, described by the mean of the ensemble, is a total of about 270 mm, or 77% of the long term average.

Potential outcomes based on the observed data to the 2nd dekad of April 2017 and completing the season with original data from each year. The side table show that there is a 94% chance that the seasonal total would be below normal.
Figure 11: Potential outcomes based on the observed data to the 2nd dekad of April 2017 and completing the season with original data from each year. The side table show that there is a 94% chance that the seasonal total would be below normal.

Conclusion

Low rainfall during the Belg 2017 season has been observed since the beginning of March.  These low values in rainfall have been affecting the availability of water for crops and grasslands primarily in southern and southeastern Ethiopia. This analysis focuses on a specific area in southern Ethiopia where the conditions are getting worse as the season progresses.  Even though there is more than a month to the end of the season, there is a 94% probability that the season ends below normal for this region. Substantial (75% of normal) rainfall deficits seem likely, and given the ensemble of historical outcomes, very low seasonal rainfall totals are quite possible.

References

Conway, D., L. Schipper, M. Yesuf, M. Kassie, A. Persechino, A., B. Kebede (2007). Reducing vulnerability in Ethiopia: addressing the issues of climate change: Integration of results from Phase I. Norwich: Overseas Development Group, University of East Anglia.

Deressa, T. T. (2006). Measuring the economic impact of climate change on Ethiopian agriculture: Ricardian approach. CEEPA Discussion Paper No. 21. Pretoria: University of Pretoria (http://econ.worldbank.org).

Lemi, A. (2005). Rainfall probability and agricultural yield in Ethiopia. Eastern Africa Social Science Research Review, 21(1): 57-96.

NMSA (1996), Climatic and agroclimatic resources of Ethiopia, Natl. Meteorol. Serv. Agency of Ethiopia, Meteorol. Res. Rep. Ser., 1(1), 1–137.

Webb P, Braun Jv, Yohannes Y (1992). Famine in Ethiopia: policy implications of coping failure at national and household levels. Research Reports, vol 92. International Food Policy Research Institute, Washington, D.C.

Funk C, Dettinger MD, Michaelsen JC, Verdin JP, Brown ME, Barlow M, Hoell A (2008). Warming of the Indian Ocean threatens eastern and southern African food security but could be mitigated by agricultural development. PNAS 105:11081–11087

Stephanie Gleixner, Noel Keenlyside, Ellen Viste,  Diriba Korecha (2016):  The El Niño effect on Ethiopian summer rainfall. Clim Dyn, DOI 10.1007/s00382-016-3421

Viste, E, Korecha, D. and Sorteberg, A. (2013): Recent drought and precipitation tendencies in Ethiopia. Theoretical & Applied Climatology. V. 112, p535-551

 

2017 Long/Gu Rains off to a slow start

Chris Funk, Pete Peterson, Diego Pedreros, Greg Husak and Laura Harrison

 

Introduction and context

As summarized in a recent UN Office for the Coordination of Human Affairs (UN OCHA) report, the drought and food security situation in East Africa is becoming increasingly fragile. Poor water access has helped lead to outbreaks of Acute Watery Diarrhea in Kenya, Somalia and Ethiopia, and cholera outbreaks in Somalia and Kenya. Two of the region’s biggest cereal exporters – Uganda and Tanzania face serious domestic challenges. In Uganda, thousands of immigrants continue to arrive from South Sudan, and fall armyworms may reduce Uganda’s four million metric ton maize output. According to UN OCHA, Ugandan maize prices have increased by 60%. Tanzania, typically the largest regional exporter has closed its borders for exports. Maize and bean prices have risen substantially in Kenya, but stabilized in Somalia in response to extensive humanitarian aid efforts. The FEWS NET March report for East Africa reports that “FEWS NET and FSNAU released joint statements on deteriorating food security in Somalia and the risk of Famine (IPC Phase 5) in a worst-case scenario in which the April to June 2017 Gu season performs very poorly, purchasing power declines to levels seen in 2010/11, and humanitarian assistance is unable to reach populations in need.” FEWS NET reports that there have been 10,571 cases of AWD/cholera since the beginning of 2017 – with half of these cases happening in Bay Province – which we focus on below. This report also identifies severe potential food insecurity in far eastern Ethiopia (Somali Region) and atypical levels of critical food insecurity across eastern and northern Kenya. More details for these countries can be found in the March reports for Somalia, Kenya, and Ethiopia. In general, these reports highlight high levels of concern for eastern Kenya, eastern Kenya, and southern Somalia, areas we believe likely to receive below normal March-May rainfall this year.

As emphasized by the most recent NOAA Climate Prediction Center Weather Hazards report – “Erratic and below-average rainfall continues in East Africa”.  Our objective with this posting is to a) highlight the severity of this drying over Eastern Kenya and Somalia, and b) emphasize the ongoing role that an overturning Walker Circulation intensification appears to be playing in supporting these arid conditions.
Current vegetation and climate conditions

We begin with a brief review of current vegetation and hydrologic conditions over East Africa. Poor October-December rains, warm temperatures, and low March precipitation totals in most areas have conspired to produce extensive vegetation stress across almost all of Kenya, southeastern Ethiopia, and Southern Somalia (Figure 1). Kenya appears to be hardest hit in terms of vegetation deficits.

Figure 1.Current (March 21-31 2017) percent anomalies from EROS eMODIS vegetation imagery. Source: earlywarning.usgs.gov.
Figure 1.Current (March 21-31 2017) percent anomalies from EROS eMODIS vegetation imagery. Source: earlywarning.usgs.gov.

These poor vegetation conditions are likely due to poor March rains

Figure 2. Current (March 6th 2017) water point conditions, based on estimated water availability. Source: https://earlywarning.usgs.gov/fews/waterpoint/index.php
Figure 2. Current (March 6th 2017) water point conditions, based on estimated water availability. Source: https://earlywarning.usgs.gov/fews/waterpoint/index.php

 

following a poor 2016 short rainy season. Current water availability, based on estimates of water levels at water holes across arid regions of East Africa (Figure 2), also appears to be very low (near-dry) in most of the locations being monitored. Time series of water availability, which can be explored here, indicate many areas with zero water at present in many areas, with the water holes typically drying up between December and March. These same regions – Kenya, southeastern Ethiopia and Somalia – also appear as warm or very warm in Land Surface Temperature images (Figure 3). These warm conditions are likely due to a combination of surface water stress, low vegetation, lower than normal cloud cover and subsiding air, which increases local air temperatures.

Figure 3. Standardized March 2014 MODIS Land Surface Temperature anomalies.
Figure 3. Standardized March 2014 MODIS Land Surface Temperature anomalies.

Poor March 2017 rainfall conditions

Due to the severity of the current crisis, the CHG has created a first draft CHIRPS products for March that incorporates a large number of additional gauges provided by FAO/SWALIM and the Ethiopian Meteorological Agency. Sixty-two and forty-five additional gauges were provided, respectively, by our colleagues in Somalia and Ethiopia. While these gauge observations are routinely incorporated in the ‘final’ CHIRPS product, this final product is typically only available three weeks after the end of the month. Figure 4 presents a provisional CHIRPS rainfall for March, superimposed with station values. The station values are presented as colored squares, marked with source identifiers.

What we see in this image is that the station data indicate that parts of Kenya and virtually all of Somalia received no rainfall in March. The background coloring shows the station-adjusted CHIRPS totals for March. These are very similar to the CPC ARC2 totals, except that the station data is more pessimistic over Southern Somalia. Both the ARC2 and satellite-only CHIRP estimates indicate about 25 mm of rain in some parts of this region. The station data, in contrast, indicate that extremely low rainfall totals are more likely. Note also that the gauge observations for typically rainy areas in western and central Kenya also exhibit very low rainfall.

Figure 4. March CHIRPS precipitation totals overlain with station rainfall observations (shown as squares). Stations were kindly provided by FAO SWALIM and the Ethiopian Meteorological Agency. The stations shown for Kenya are provided by the Kenya Meteorological Agency via the Global Telecommunication System.
Figure 4. March CHIRPS precipitation totals overlain with station rainfall observations (shown as squares). Stations were kindly provided by FAO SWALIM and the Ethiopian Meteorological Agency. The stations shown for Kenya are provided by the Kenya Meteorological Agency via the Global Telecommunication System.

We can also plot the March gauge and CHIRPS rainfall values as standardized anomalies (Figure 5). This map suggests that March has likely been much drier than normal over Somalia and parts of Kenya. While climatologically March is typically still quite dry in Somalia, the country appears to have received little relief in 2017.

Figure 5. March 2017 CHIRPS and station observations expressed as standardized anomalies. Stations were kindly provided by FAO SWALIM and the Ethiopian Meteorological Agency.
Figure 5. March 2017 CHIRPS and station observations expressed as standardized anomalies. Stations were kindly provided by FAO SWALIM and the Ethiopian Meteorological Agency.

Focus – Bay Region of Somalia

We next briefly place the current dryness in historic context for the Bay Region in Southern Somalia – an area facing severe IPC 4 (crisis) food security conditions (http://www.fews.net/east-africa/somalia).

 

 Figure 6. Potential cumulative rainfall curves for Bay Region, Somalia.

Figure 6. Potential cumulative rainfall curves for Bay Region, Somalia.


To explore potential outcomes for the coming Gu season, we plot (Figure 6) the observed accumulated rainfall for the Bay region through the 1st dekad of April, and then simulate potential future seasonal performance by completing the season using observed CHIRPS rainfall from previous years. We assume that the SWALIM data correctly indicate zero March rainfall for this region. Furthermore, given that GTS observations and CPC ARC2 satellite estimates around this region report little rainfall over the last 7 days, we also assume that the rainfall in this region for the 1st 10 days of April will be 20 millimeters, about half the long term average value for this region and dekad. The result of these assumptions is that it appears likely that the Bay Region will receive below normal March-May rains, just based on the current seasonal deficit and historic rainfall outcomes.

Focus on Eastern Kenya

We next briefly focus on the Eastern Province of Kenya. Parts of this area are heavily populated, and it shows up consistently as an area of high vegetation stress (Figure 1), high temperature anomalies (Figure 3) and low rainfall totals (Figures 4 and 5). Time series (Figure 7) of vegetation (normalized difference vegetation index values) for Eastern Province appear to be very low, and lower than during the 2010/2011 drought period. Note however, that we are just entering the period when NDVI in this area is expected to increase. Figure 8 shows the corresponding cumulative rainfall totals, based on NOAA CPC RFE2 rainfall estimates. This plot indicates conditions similar to 2010/11. During March, this region typically receives about ~34 mm of rainfall (about a third of the typical ~170 mm March-May total). This year the amount received in March and early April appears to be much lower than that amount (~10 mm).

Figure 7. Vegetation Health (NDVI) for Eastern Province of Kenya. Source: USGS Map Viewer.
Figure 7. Vegetation Health (NDVI) for Eastern Province of Kenya. Source: USGS Map Viewer.
Figure 8. Cumulative NOAA CPC Rainfall totals for Eastern Kenya.
Figure 8. Cumulative NOAA CPC Rainfall totals for Eastern Kenya.

Current climate conditions likely to indicate continued rainfall suppression

We now examine recent climate conditions, which appear likely to be associated with continued rainfall suppression, despite the fact that La Niña climate conditions have dissipated over the past few months.

What the most recent (April 3rd) NOAA CPC assessment shows is that equatorial Pacific sea surface temperatures (Figure 9) are currently exhibiting a ‘split personality’ with very warm temperatures in both the western and far eastern equatorial Pacific, and near neutral conditions in between, in the central Pacific. Since about July of 2016, we have seen relatively warm temperatures in the west Pacific and relatively cool temperatures in the central Pacific. This contrast in temperatures helps drive the Walker Circulation, producing enhanced rainfall near Indonesia and dry sinking air over eastern East Africa. We recently had a chance to discuss this with Steve Baragona, a science writer with Voice of America, who did a piece on the East African drought and climate change (link here). Steve was kind enough to share VOA’s animation describing the Walker Circulation’s impact on East Africa, which you can access here.

Figure 9. Evolution of equatorial sea surface departures. Source NOAA CPC: http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/lanina/enso_evolution-status-fcsts-web.ppt
Figure 9. Evolution of equatorial sea surface departures. Source NOAA CPC: http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/lanina/enso_evolution-status-fcsts-web.ppt

When the western Pacific warms relative to the central Pacific, and rainfall increases around Indonesia and decreases in the central equatorial Pacific, rainfall tends to decrease across eastern East Africa. FEWS NET research has suggested that these changes can help explain a shift towards drier long/Gu/Belg rains over East Africa. To help visualize these changes you can see an animation we prepared here at the CHG. The basic idea behind this animation is that when the west Pacific warms relative to the east Pacific and precipitation around Indonesia increases, the East African long rains tend to decrease. This pattern helps explain the very dry March conditions and may persist into April and May.

Figure 10. NOAA CPC OLR and near-surface wind anomalies. Source NOAA CPC: http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/lanina/enso_evolution-status-fcsts-web.ppt
Figure 10. NOAA CPC OLR and near-surface wind anomalies. Source NOAA CPC: http://www.cpc.ncep.noaa.gov/products/analysis_monitoring/lanina/enso_evolution-status-fcsts-web.ppt

Figure 10 shows another plot from the recent CPC ENSO assessment. The top panel shows outgoing longwave radiation (OLR) anomalies for March. Since clouds block OLR, decreased OLR around Indonesia and increased OLR around the dateline (180°E) indicates an enhanced Walker Circulation, with more/less rainfall near Indonesia and the dateline.

Between these areas of enhanced and reduced precipitation we find increased low level winds blowing into the area near Indonesia (Figure 10, bottom). These winds increase moisture convergence in that area while also acting to cool the central equatorial Pacific. Note that these OLR and wind responses are constrained to the western and central Pacific and are quite different than the typical ENSO response. This pattern appears similar to circulation anomalies associated with the West Pacific Warming Mode. This climate pattern is produced by a ‘Western V’ of warm sea surface temperatures beginning near Indonesia and stretching to the northeast and southeast – a pattern that appears very similar to sea surface temperature conditions in March of 2017 (Figure 11). This figure shows that March sea surface temperatures resemble a pattern commonly associated with drying over East Africa during March-June. This pattern could be exacerbated by La Niña conditions, but La Niña conditions are not necessary to produce dry East African rainfall outcomes. This ‘Western V’ pattern of warm sea surface temperatures enhances subsidence over the central Pacific, increases trade winds flowing into the region surrounding Indonesia, increases precipitation in that region, and decreases precipitation over Eastern East Africa.

Figure 11. Standardized March 2017 sea surface temperature anomalies. Based on NOAA Extended Reconstruction version 4 data and a 1981-2010 baseline.
Figure 11. Standardized March 2017 sea surface temperature anomalies. Based on NOAA Extended Reconstruction version 4 data and a 1981-2010 baseline.

This set of climate anomalies is occurring now and is likely to persist into April and possibly May.

This far into the long/Gu/Belg season, monitoring the strength of the Walker Circulation can help us anticipate how the rest of the season may play out. Unfortunately, what we are seeing is that we continue to observe a large scale pattern of ascending and descending air that acts to enhance the Walker Circulation and reduce eastern East African rainfall. Figure 12 shows March 26th to April 4th vertical velocity anomalies averaged along the equator. Purple shades show areas of ascent, yellow-orange shades indicate sinking air. While the pattern of rising and sinking air for 2017 (top) does not look quite as intense as it did in 2011 (bottom), it still does not look conducive to good long rains in eastern East Africa.

Figure 12. March 26th to April 4th vertical velocity anomalies. Source: NOAA Earth Systems Research Laboratory.
Figure 12. March 26th to April 4th vertical velocity anomalies. Source: NOAA Earth Systems Research Laboratory.

We can look more closely at the surface wind component of the Walker Circulation enhancement (Figure 10, bottom), by extracting and plotting reanalysis surface winds from the central equatorial Pacific (Figure 13). Negative values in this time series indicate increased wind flow towards Indonesia. Note that this time series shows an abrupt change in 1999 – when East Africa switched to a ‘new normal’ associated with much drier East African March-May rains. In March of 2017, the intensity of these winds were similar to 1999, 2000, 2008, 2011 and 2012 – La Niña-like East African drought years. In April-May of these potential analog years Global Precipitation Climatology Project (GPCP) precipitation estimates (Figure 14) showed drying over East Africa and a Walker Circulation enhancement similar to that observed in March 2017 climate anomalies.

Figure 13. Reanalysis east-west (zonal) winds over the equatorial central Pacific. Source: NOAA ESRL.
Figure 13. Reanalysis east-west (zonal) winds over the equatorial central Pacific. Source: NOAA ESRL.
Figure 14. Global Precipitation Climatology Project April-May rainfall anomalies for recent years with strong westward central Pacific winds in March. Source: NOAA ESRL.
Figure 14. Global Precipitation Climatology Project April-May rainfall anomalies for recent years with strong westward central Pacific winds in March. Source: NOAA ESRL.

Conclusion

Our analysis of March rainfall anomalies and recent large scale climate fields compel continued concerns for the 2017 Long/Gu rainy season, especially over central-eastern Kenya and southern Somalia, where the season appears to have had a very poor start. Newly available station data, provided by FAO/SWALIM, allow us make this assertion for Somalia with an unprecedented level of confidence. This poor March onset of rains appears related to a La Niña-like atmospheric response pattern, despite the current ENSO-neutral conditions. FEWS NET research has linked such a response to a warm ‘Western V’ SST pattern, and this SST pattern seems likely to have contributed to poor March rainfall, and a probable below-normal March-May outcome, if current conditions persist, which seems likely. For Somalia, this could lead to a third failed or poor growing season. For most the Horn, a poor March-May season will exacerbate the current severe drought crisis.