Seasonal Climate

Predictability in Ethiopia:
Review of best predictor sets for sub-
seasonal to seasonal forecasting
Working Paper No. 301

CGIAR Research Program on Climate Change,

Agriculture and Food Security (CCAFS)

Robel Takele
Kindie Tesfaye
Pierre C. Sibiry Traore
Seasonal Climate
Predictability in Ethiopia:
Review of best predictor sets for sub-
seasonal to seasonal forecasting

Working Paper No. 301

CGIAR Research Program on Climate Change,

Agriculture and Food Security (CCAFS)

Robel Takele
Kindie Tesfaye
Pierre C. Sibiry Traore

1
Correct citation:
Takele R., Tesfaye K., Traore P.C.S., 2020. Seasonal Climate Predictability in Ethiopia. Identifying
best predictor sets for sub-seasonal to seasonal forecasting. CCAFS Working Paper no. 301.
Wageningen, the Netherlands: CGIAR Research Program on Climate Change, Agriculture and Food
Security (CCAFS). Available online at: www.ccafs.cgiar.org

Titles in this Working Paper series aim to disseminate interim climate change, agriculture and food
security research and practices and stimulate feedback from the scientific community.

The CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS) is a
strategic partnership of CGIAR and Future Earth, led by the International Center for Tropical
Agriculture (CIAT). The Program is carried out with funding by CGIAR Fund Donors, Australia
(ACIAR), Ireland (Irish Aid), Netherlands (Ministry of Foreign Affairs), New Zealand Ministry of
Foreign Affairs & Trade; Switzerland (SDC); Thailand; The UK Government (UK Aid); USA
(USAID); The European Union (EU); and with technical support from The International Fund for
Agricultural Development (IFAD). For more information, please visit https://ccafs.cgiar.org/donors.

Contact:
CCAFS Program Management Unit, Wageningen University & Research, Lumen building,
Droevendaalsesteeg 3a, 6708 PB Wageningen, the Netherlands. Email: [email protected]

Creative Commons License

This Working Paper is licensed under a Creative Commons Attribution – NonCommercial 4.0
International License.

© 2020 CGIAR Research Program on Climate Change, Agriculture and Food Security (CCAFS).
CCAFS Working Paper no. 301

Photos:

DISCLAIMER:
This Working Paper has been prepared as an output for the CASCAID-II project on Capacitating
African Stakeholders with Climate Advisories and Insurance Development under the CCAFS Program
Flagship # 4 on Climate Services and Safety Nets and has not been peer reviewed. Any opinions stated
herein are those of the author(s) and do not necessarily reflect the policies or opinions of CCAFS,
donor agencies, or partners. All images remain the sole property of their source and may not be used
for any purpose without written permission of the source.

2
Abstract

As elsewhere in the Tropics, the climate of Ethiopia is highly variable. Capturing its
variability has been a major challenge for climate models and tools. Understanding
teleconnections and predictors is, therefore, an important step towards improving the
skill of seasonal and intra-seasonal climate forecasts, derivative products such as
seasonal yield predictions, and climate services in general. This report presents a review
of existing knowledge on teleconnections, climate predictability and seasonal to intra-
seasonal climate forecasting advances and challenges for Ethiopia. Literature reviewed
indicates an association between the seasonal climate of Ethiopia and seas surface
temperature (SST) forcings over the Atlantic, Indian Oceans and, to a greater extent,
over the equatorial Pacific along with associated atmospheric circulations. The main
(Kiremt) season’s climate is strongly influenced by teleconnections with SST anomalies
and the El Nino Southern Oscillation (ENSO) in the Nino-3.4 region of the equatorial
Pacific and can yield moderate skill forecasts with 1 to 2 month lead time, while the
Indian Ocean Dipole (IOD) has relatively stronger influence on the climates of the dry
season (Bega) and small rains season (Belg). Best climate predictors and prediction skill
therefore vary for the different seasons of Ethiopia. The procedures and methods used
by the National Meteorology Agency (NMA) of Ethiopia to forecast seasonal and intra-
seasonal climates and their pros and cons are discussed.

Keywords

Seasonal climate predictability; Forecasting skill; Teleconnections; Ethiopia.

3
About the authors
Robel Takele, Agro-Meteorologist, Ethiopian Institute of Agricultural Research (EIAR), P.O.
Box 2003, Addis Ababa, Ethiopia. [email protected]

Kindie Tesfaye, Senior Scientist – Crop Modeling, CIMMYT, c/o ILRI, P.O. Box 5689, Addis
Ababa, Ethiopia. [email protected]

Pierre C. Sibiry Traore, Principal In-Business Researcher, ICRISAT, c/o agCelerant-Senegal,

P.O. Box 25026, Dakar-Fann, Senegal. [email protected]

4
5
Contents

Introduction .................................................................................................................... 9
The Seasonal Climate of Ethiopia................................................................................ 10
Potential Predictability of Ethiopia’s Seasonal Climate .............................................. 13
Predictability of Kiremt and Belg Season Climate .................................................. 13
Predictability of Bega Season Climate .................................................................... 15
Recent Developments on Climate Predictability ..................................................... 16
Operational Seasonal Climate Forecasting at NMA .................................................... 18
The Analogue Method ............................................................................................. 21
The Climate Predictability Tool (CPT).................................................................... 22
The Geospatial Climate Outlook Forecasting Tool (GeoCOF) ............................... 23
Conclusion/recommendations ...................................................................................... 25
References .................................................................................................................... 29
Glossary ....................................................................................................................... 31

6
Acronyms

AGCM Atmospheric General Circulation Model

CCA Canonical Correlation Analysis

CHIRPS Climate Hazards Group Infrared Precipitation with Stations

CPC Climate Prediction Center

CPT Climate Predictability Tool

ENSO El Nino Southern Oscillation

ERSST Extended Reconstructed Sea Surface Temperature

GCM General/Global Circulation Model

GeoCOF Geospatial Climate Outlook Forecasting Tool

IOD Indian Ocean Dipole

ITCZ Inter Tropical Convergence Zone

JAS July, August and September

JJAS June, July, August and September

MAM March, April and May

MLR Multiple Linear Regression

LDA Linear Discriminant Analysis

NMA National Meteorological Agency

NOAA National Oceanic and Atmospheric Administration

MME Multi-Model Ensemble

RCOF Regional Climate Outlook Forum

SLP Seal Level Pressure

SOI Southern Oscillation Index

SON September, October and November

SST Sea Surface Temperature

7
SSTA Sea Surface Temperature Anomaly

RSCZ Red Sea Convergence Zone

TEJ Tropical Easterly Jet

WMO World Meteorological Organization

8
Introduction

Mankind recognized millennia ago the importance of climate variability for the
sustenance of life, whether that variability was expressed in the form of droughts,
floods, heat, cold, or wind. Coping strategies developed to handle the consequences of
climate variability helped ensure the survival of mankind, although the historic record
indicates that not all civilizations successfully overcame past challenges imposed by
long-term droughts, extensive flooding, and the like. Much has changed in the modern
era, with coping strategies such as migration, invasion and appropriation frequently
constrained by international boundaries and laws. Included amongst the technological
advances that have led to increased resilience against climate variability are remarkable
achievements in the understanding, monitoring and prediction of climate variability
itself (Troccoli, 2008).
In principle, modern seasonal to inter-annual predictions are an answer to the needs of
many whose activities are influenced in some manner by climate variability, whether
in terms of creating profit through the marketing of an appropriate range of goods, or
for critical decisions regarding agriculture and food security. From a practical
perspective, there is only one reason for undertaking research and development to
advance seasonal climate predictions and for investing in the infrastructure to produce
and deliver them. That reason is to assist whatever decision processes are of concern to
those who might make use of them. For that matter, prediction information must be
reliable to be admissible into the decision processes of recipients (WMO, 2002).
The idea that the climate may be predictable at seasonal timescales may seem
counterintuitive, given that weather does not appear to be predictable with much
accuracy beyond a few days. At seasonal scale, the errors become so large that there is
no longer a forecast, but an accidental resemblance between the forecast and the
observed conditions. However, it is possible to provide information, based on a
different source of predictability (Hansen et al., 2011). Changes in the earth’s surface,
particularly sea surface temperatures (SST), can influence the atmosphere. Since ocean
temperatures tend to change slowly relative to the atmosphere because of their high heat
capacity, knowing the current state of the oceans may provide some degree of
predictability on seasonal time scales. Thus, while it is harder to forecast the weather in
the Tropics, it tends to be easier to predict the seasonal conditions, although

9
predictability at seasonal timescales is highly dependent on location and on the time of
the year (Troccoli, 2008).
It is important to note that average circulation in the tropics is, on the one hand, a major
characteristic of the atmospheric variability and, on the other hand, strongly influenced
by the large scale organized convection (notably the Walker-Hadley divergent
circulation), the latter being strongly controlled by the evolutions of the conditions of
oceanic and continental surfaces. That particularly explains why seasonal forecasting
has enhanced scores and skills in tropical areas compared to the mid-latitude regions.
In addition to solar forcing, the principal source of energy for the climatic system, one
can distinguish between continental and oceanic forcings. SSTs are particularly used in
both coupled and forced numerical models, but also in the majority of statistical models.
This information coming from the oceanic surface allows us to get reasonable forecasts
up to a lead time of 4 months. Beyond, information from coupled Ocean-Atmosphere
dynamics is typically required, using general circulation models (GCM) including sub-
surface information. The best known of these forcings is related to the El Niño Southern
Oscillation (ENSO), with planetary consequences (Clarke, 2008).
Ethiopia, like many parts of the Tropics, is prone to extreme climate events such as
droughts and floods. In an effort to minimize the negative impacts of extreme climate
events, and to better exploit opportunities offered by climate, in 1980 the Government
of Ethiopia established the National Meteorological Services Agency (currently,
National Meteorological Agency; NMA). Today, NMA uses a statistical method based
on analogue multivariate ENSO index years. The outputs of this method are
probabilistic categorical forecasts of regional Ethiopian rainfall (Korecha and
Sorteberg, 2013). The goal of this review is to review and discuss current knowledge
concerning predictability and predictions of Ethiopian seasonal climate. In this review,
we focus on assessing the current practices, methods and tools used to produce seasonal
climate forecasts by NMA.

The Seasonal Climate of Ethiopia

Meteorologically, a season is a period when an air mass characterized by homogeneity

in temperature, humidity, wind patterns, rainfall, etc., influences a region (Désalmand,
1998). Ethiopia’s rainfall climatology is mainly determined by seasonal changes in

10
large-scale circulation, part of which involve the latitudinal movement of the
intertropical convergence zone (ITCZ) as happens throughout the larger Sahel region
from Sudan to Senegal (Nicholson 1989). In Ethiopia, seasons are unique and are
classified mainly based on rainfall and its distribution.
“Bega” is a generally dry season that covers the period from October to January. During
this season, Ethiopia is predominately influenced by both warm and dry air masses
originated from the Saharan Anticyclone as well as cool and dry air masses originated
from the Siberia and Arabian Anticyclones. However, associated with occasional
eastward movements of mid-latitude depressions, the Arabian high can displace and
establish itself over the Arabian Sea. In such situations, the interaction between the
southeast warm moist and mid latitude cool dry air masses, coupled with mid-
tropospheric deep troughs and with the Subtropical Jet (STJ) can produce substantial
and untimely rains.
The “Belg” season refers to a small rainy season that covers the period from Mid-
February to Mid-May. This season coincides with the dominance of the Arabian high
as it moves toward the North Arabian Sea. In this season a thermal low develops over
Sudan, an anticyclone forms over the Arabian Sea, and resulting easterly winds bring
considerable amounts of moisture to the region under consideration. Sometimes the
Northern part of the region is under the influence of the warm and dry Sahara and
Arabian air masses. In general, the weather systems that are responsible for rainfall
activities during the belg season are:
▪ The development of high pressure over the Arabian Sea;
▪ The generation and propagation of disturbances, sometimes coupled with Easterly
Waves;
▪ The interaction with mid- latitude depressions accompanied by trough and STJ, the
tropical disturbances and ITCZ; and
▪ The occasional development of Red Sea Convergence Zone (RSCZ), which
produces substantial rainfall in the Northeast parts of the region.

“Kiremt” refers to the main rainy season that covers the period from June to September.
Airflow during this season is dominated by a zone of convergence in low-pressure
systems accompanied by the oscillatory Inter-tropical convergence zone (here after
ITCZ) extending from West Africa towards India through Ethiopia or North of it. The
major rain-producing components during kiremt are:

11
▪ The northward Migration of ITCZ following the sun’s movement;
▪ The development and persistence of the Arabian and Sudan thermal lows along 20o
latitude;
▪ The development of the Quasi-high-pressure system over the Southern Atlantic and
Indian Ocean
▪ The development and persistence of the Indian sub-continent’s depression and the
associated Monsoon trough; and
▪ The development of the tropical Easterly jet stream (TEJ) and its persistence for
distribution and intensity of rain.

The National Oceanic and Atmospheric Administration (NOAA) defines the ENSO
state (e.g., El Niño or La Niña) as a departure of magnitude 0.5°C or more from normal
SST in the Niño 3.4 region, lasting for at least five running three-month periods over
the tropical Equatorial Pacific Ocean. The main ENSO signal is found during the
northern summer (Camberlin, 2009) at which time a negative correlation is found with
the Niño 3.4 index, depicting lower than normal rainfall in the years of higher sea-
surface temperatures (SST) in the eastern equatorial Pacific (i.e., El Niño years). It is
argued that each ENSO state (El Niño, neutral or La Niña) has its own influence on the
rainy or dry season.
The physical connection between these changes in the atmospheric general circulation
and ENSO is complicated and not well understood. Other factors such as southern
Atlantic and Indian ocean SSTs also influence the rain-bearing systems in Ethiopia. Not
all ENSO events correlate directly with drought in Ethiopia. The 1982-83 El Niño, for
example, did not cause failure during the (kiremt) main rainy season strictu senso
although one of the worst droughts in Ethiopian history occurred over 1983-84 (NMSA,
1996). Despite this complexity, Ethiopian researchers have developed a system for
identifying when an El Niño event is likely to produce climatic variations in Ethiopia,
and for forecasting ENSO-induced climate anomalies. The NMA of Ethiopia, based on
criteria that define particular types of ENSO events, has concluded that negative SST
anomalies are strongly associated with below-normal rainfall during belg; positive SST
anomalies are often correlated with good rainfall during belg, while the effects are
opposite for the main rainy season (kiremt).
ENSO events are always associated with the overturning of the Walker cell, such that
the descending limb of the cell rather than the ascending limb sits over Africa.

12
Atmospheric dynamics responsible appear to be causing cold air outbreaks from the
Siberian High area. To the east of the Tibetan Plateau, over the western Pacific between
Japan and Philippines at about 13E a quasi-stationary upper level westerly long wave
trough with a length around 500km is maintained during bega. Through this trough, the
short-wave disturbance in the form of cyclonic vortices travels. During ENSO the effect
of this to start instance period of eastward moving of extra tropical systems across the
Mediterranean areas and its trough move toward the equator at 40 oE. This indirectly
makes warm and moist air to be pumped to the North and central part of the Ethiopia
resulting in warmer and wetter condition than normal. In addition, during this season
the Red Sea convergence zone (RSCZ) intensifies during ENSO years. This abnormal
rain occurs during the harvesting season and may reduce annual crop yield product both
in terms of harvest volume and quality. El Niño affects the bega season in such a way
that it makes the season wetter than normal. Hence, during bega, El Niño (positive
SSTAs) is correlated with normal or above normal rainfall.
The impact of La Niña over the bega season is negative. La Niña makes the bega season
extremely dry. Depressed atmospheric water vapor hampers cloud formation, and due
to enhanced nocturnal emissions (night-time long wave radiation) minimum
temperatures tend to decrease, resulting in frost conditions over some highland areas.

Potential Predictability of Ethiopia’s Seasonal Climate

Predictability of Kiremt and Belg Season Climate

Given that the atmosphere is predominantly heated from the earth’s surface rather than
directly from the sun and given that the atmosphere receives its moisture from the
earth’s surface, changes in the earth’s surface, particularly the SST distribution, can
influence the atmosphere. There are significant interconnections between the surface
temperature of the oceans and the associated atmospheric circulation due to heat
transfers from oceans to the atmosphere. Any significant departure of the earth’s surface
from its normal conditions can disrupt weather patterns over a prolonged period. These
disruptions are likely to be strongest in the Tropics where sea surface temperatures are
warmest. Since ocean temperatures tend to change slowly relative to the atmosphere
because of their high heat capacity, knowing the current state of the oceans may provide
some degree of predictability of how weather patterns may be disrupted. The basis of

13
this burgeoning industry is that slowly varying components of the geosystem, most
significantly SST across tropical ocean basins can impart a ‘memory’ to the atmosphere
in the vicinity of any such long-lived anomalies. To the extent that this ‘memory’ can
be transmitted to parts of the globe remote from the originating sea surface temperature
anomalies (SSTA), meteorologists refer to this phenomenon as teleconnections
(Rosenzweig and Hillel, 2008).
The inter-annual variability of the seasonal mean in the tropics is mainly determined by
slowly varying components such as SST, albedo, sea ice, and soil moisture. Model
experiments showed that the tropical circulation and precipitation are strongly
determined by the underlying SST with very little sensitivity to the changes in the initial
conditions of the atmosphere. The ENSO is the best-known example of a slowly varying
phenomenon that results from ocean‐atmosphere interaction in the tropical Pacific. The
prediction of conditions associated with ENSO has also seen more success in seasonal
prediction by the climate forecasting community.
Since Ethiopia is located within the Tropics, seasonal predictability is higher compared
to mid-latitude countries. The Ethiopian seasonal climate shows significant year-to-
year variations. One of the important mechanisms that control this year-to-year
variability is the ENSO state. Thus, the ENSO condition and the associated SST pattern
is the primary source of predictability so far. The impact of ENSO on Ethiopian
seasonal climate and its potential for predictability is widely documented. Sir Gilbert
Walker was the first to indicate indirectly the presence of a link between the Southern
Oscillation and rainfall variability in parts of Ethiopia. In his calculations of the
Southern Oscillation, one of the variables he used was the Nile flood level, whose major
water source is the Ethiopian Highlands (Walker and Bliss, 1932). The first
experimental seasonal forecast based on El Niño effects was made in 1987 by NMA.

Previous studies have identified the role of remote SST forcings over the Atlantic,
Indian, and to a greater extent, equatorial Pacific Oceans and the associated atmospheric
circulation over the Ethiopian seasonal climate (Table 1). The Kiremt seasonal climate
variability can be strongly influenced through teleconnection patterns originated by
SST anomalies in the Nino-3.4 region of the equatorial Pacific. For instance, (Gissila
et al., 2004) used empirical methods to predict Kiremt rains, using the relationship
between SST data for March, April and May in the Indian and Pacific Oceans. The most

14
extensive study of the predictability of the Kiremt rains is that of Korecha and Barnston
(2007). This study investigates the strength of linear relationship between all Ethiopian
Kiremt rainfall and the Niño-3.4 SST index derived from Extended Reconstructed Sea
Surface Temperature version 2 (ERSSTv2). The association of summer rainfall with
ENSO in early pre-summer months (January–April) is weak and increases as the ENSO
state approaches the beginning of the rainfall season. The correlation is moderate (-
0.59) for the May Niño-3.4 SST, suggestive of some predictability based solely on the
May ENSO state.

The polarity of the ENSO teleconnection depends on the season in question with
positive SST anomalies in the eastern Pacific (El Niño) being associated with rainfall
deficits in the Kiremt season and excess rainfall in the Belg season. The ENSO signal
on observed Belg rainfall anomalies over southern Ethiopia seems to be weaker than
for Kiremt. (Diro et al., 2008) identifies various regions of SST anomalies over Atlantic,
Indian and Pacific Ocean to predict the Belg rains (Annex Fig. 7.1). Because of the
spatial variation across Ethiopia in both interannual variability and the annual cycle,
Diro et al. (2008) identified five homogeneous rainfall zones within Ethiopia and
produced separate forecast models for each zone. Both multiple linear regression and
linear discriminant analysis were applied to four sets of predictors. It was also shown
that the models had the most skill in the southern and eastern parts of Ethiopia and that
the extreme years were more reliably forecasted than the average years.

Predictability of Bega Season Climate

Interannual variability of Bega rains over south-eastern Ethiopia is dominated by large-

scale changes in the Indian Ocean and its coupled atmosphere with a clear link to the
Indian Ocean dipole (IOD). Bega rainfall over south-eastern parts of Ethiopia is
increased during positive IOD events. Bahaga et al. (2015) explored the predictability
of Bega rains in their Atmospheric General Circulation Model (AGCM) ensemble
experiments and found a substantial potential prediction skill associated with East
African short rains, given the predictability of SST anomalies over the Indian Ocean.
Besides, a temporally limited potential for the dynamical predictability of short rain has
been indicated using a coupled ocean–atmosphere general circulation model (Behera et
al., 2006). They showed that the July–August signal of the IOD in the SST dipole mode
index has a high prediction skill for the variability in Bega rains.

15
Recent Developments on Climate Predictability
More often, preceding month SSTA and ENSO state are the key operational predictors
in Ethiopia. Yet, Nicholson (2015, 2014) argues that atmospheric variables (zonal and
meridional wind and vertical motion) generally provide higher forecast skill compared
to surface variables (SST and season level pressure, SLP) on shorter lead forecasts (1-
month lead time). Hence, ENSO and IOD provide less forecast skill than atmospheric
variables associated with them. Surface variables become somewhat more important
for 2-month lead-time, longer-lead forecasts.
A well improved ensemble-based multiple linear regression technique is developed to
assess the predictability of regional and national Kiremt rainfall anomalies and local
monthly rainfall totals for Ethiopia (Segele et al., 2015). The ensemble prediction
approach captures potential predictive signals in regional circulations and global SSTs,
two to three months in advance of the Kiremt season. This ensemble features an
improvement in terms of skill and usability compared to previous studies.

16
Table 1. Sources of predictability for the seasonal climate of Ethiopia
Prediction method
No. Predictand Predictors Location Reference
and skill
Kiremt rainfall ▪ Tropical western Indian Ocean (10 °S - 10
totals over Sea Surface °N, 50 - 70 °E) and (10 °S -equator, 90 - 110 Multiple linear regression Gissila et al.
1
homogeneous Temperature °E). [r =0.6] (2004)
zones ▪ The Niño-3.4 (5°S - 5 °N, 120 - 170 °W).
▪ The difference of May minus February to
March SSTs over the south Atlantic (30° - Multiple linear regression
All Ethiopian
Sea Surface 40°S, 15° - 30°W). and Canonical Korecha and
2 Average Kiremt
Temperature ▪ The difference of May minus the February to Correlation Analysis Barnston (2007)
rainfall totals
March Niño-3.4 SST [r=0.73]
▪ May Niño-3.4 SST
Belg rainfall Multiple
totals over Sea Surface linear regression (MLR) Diro et al.
3 See Annex Fig.7.1 for details
homogeneous Temperature and linear discriminant (2008)
zones analysis (LDA)
Multiple
Atmospheric and Nicholson
4 JAS rainfall See Annex Table 7.1 for details linear regression (Annex
Oceanic Indices (2015, 2014)
Table 7.2 and 7.4)
Multiple
Atmospheric and Nicholson
5 MAM rainfall See Annex Table 7.1 for details linear regression (Annex
Oceanic Indices (2015, 2014)
Table 7.3)
September- Coupled
October- operational Multi-model ensemble Bahaga et al.
6 1-month lead IOD index
November seasonal forecast (MME) (2016, 2015)
(SON) rains models (GCMs)

17
Operational Seasonal Climate Forecasting at NMA

Ethiopia started issuing seasonal forecasts in 1987, ten years before the first RCOF. The
NMA issues seasonal forecasts 3 times a year targeting the Bega, Belg and Kiremt
seasons, 1–2 weeks prior to the normal onset date of each season. During early stages,
the methods of forecasting used by the NMA are based on analogue, trend analysis
(short-term trends of SST), statistical assessments, and teleconnections (Bekele, 1997).
However recently the agency adopted a consensus seasonal climate outlook based on
guidance from various prediction methods and tools. The seasonal climate outlook
includes forecast information beyond seasonal rainfall totals, such as outlooks on the
start, end and duration of the rainfall season. In addition, the agency forecasts several
additional agriculturally important variables such as moisture status, water requirement
satisfaction index and vegetation condition. The agency also issues outlooks on climate
suitability for malaria transmission (Connor et al., n.d.).

Although there are various attempts to include a number of methods and tools to
produce the seasonal climate outlook, the main method includes a combination of the
analogue method and regression based statistical forecasting tools. The procedures of
producing consensus-based seasonal climate outlooks at NMA include:
A. Diagnosis: involves examining temporal evolution and current status of oceanic
and atmospheric synoptic scale regional and global meteorological features;
B. Prognosis: involves examining outlooks produced by global forecasting centers of
the status of oceanic and atmospheric meteorological features on synoptic, regional
and global scales (Figure 1).

18
Figure 1. Schematic of the
analytical steps in the
preparation of seasonal
rainfall forecasts by the
National Meteorological
Agency of Ethiopia (after
Korecha and Sorteberg,
2013)

19
NMA uses indices of sea surface temperatures (SSTs) over the tropical Pacific Ocean,
the Southern Oscillation Index (SOI), the Multivariate ENSO Index (MEI: Wolter and
Timlin, 1998) and the ENSO outlook obtained from NOAA/CPC. Historical and
current Niño 3.4 SSTs (located in the central equatorial tropical Pacific Ocean) are
used to select years with an ENSO evolution similar to the current year.

NMA’s seasonal rainfall forecasts are then prepared as a probability of the regional

seasonal rainfall being below, near, and above the climatological normal for eight

homogeneous rainfall regions (Figure 2). The classification is based on typical rain-

producing systems affecting the region and on spatial and temporal response of each

respective region to major atmospheric and oceanic circulation systems. Although some

authors (Diro et al., 2010, 2008; Gissila et al., 2004; Tsidu, 2012) have proposed

modifications to the NMA homogeneous rainfall regions, NMA still uses the originally

defined eight rainfall zones for the preparation of seasonal forecasts. The tercile rainfall

categories, which are more commonly known as the probabilities, refer to the likelihood

that the region averaged rainfall will be below, near, or above average as the anomalies

in seasonal (4-month) rainfall are often large in geographical scale.

Figure 2. Homogeneous rainfall regions currently used for the preparation of seasonal
rainfall forecasts in Ethiopia.

20
The Analogue Method
The analogue forecast methodology is one of the major techniques operationally used
by NMA to produce seasonal forecasts. This technique involve an analysis of historical
data in search of previous periods that resemble the immediate past period, and
predicting the following season’s rainfall anomalies on the basis of what happened on
those previous occasions. In identifying the predictors for Ethiopia rainy seasons,
previous research guides the selection of the most appropriate predictors from the
historical archives. Various teleconnection patterns are linked to Indian, Atlantic, and
Pacific Oceans, where they produce different climatic anomalies in various parts of
Ethiopia (Gissila et al., 2004; Korecha and Barnston, 2007; Segele and Lamb, 2005;
Shanko and Camberlin, 1998) . ENSO-indices have been well identified as the potential
preseason indicators and thus have become the basis for analogue forecasting in
Ethiopia. ENSO indices are being retained year round, but allowing these indices to be
weighted differently from season to season as well as from region to region, depending
on the direct linkage between regional rainfall pattern and SST anomalies.
The analogue approach requires the selection of 3 to 5 analogues years with an ENSO
evolution similar to the target year, by comparing historical and target years’ SSTAs in
the Niño 3.4 region. The seasonal rainfall forecast for the target season is then prepared
based on rainfall observed in these analogue years. The seasonal rainfall total of each
station is expressed as a seasonal rainfall percentile and used to calculate tercile
categories (0–33, below; 34–66, near, and 67–100%, above the climatological normal)
for each homogeneous rainfall region.
The major advantage of the analogue forecasting technique currently used in NMA is
that the climate information consists of past observations, so the implications can be
readily connected with decision models and can be utilized in situations where the
computing facility is very weak. Because of its conceptual simplicity, already identified
various teleconnection patterns can be easily related to the observed climate to produce
the seasonal forecast. This provides a unique opportunity for NMA to issue timely early
warnings on the adverse effect of climatic anomalies within a reasonable lead-time.
Although the analogue method is simple and computationally efficient, it is limited by
observational data and mostly depend on linear relationships. Non-stationarity due to
decadal/multi-decadal climate variability is thus not considered (Goddard et al., 2010).
Changes in observed mean, variability, and trend due to the changing climate and the
associated unprecedented situations severely limit the usefulness of analogue
approaches.

21
The Climate Predictability Tool (CPT)
The Climate Predictability Tool (CPT), developed by the International Research
Institute on Climate and Society, provides a package for constructing a seasonal climate
forecast model, performing model validation, and producing forecasts given updated
data (Mason, 2011). Although the tool is specifically tailored for these applications, it
can be used in more general settings to perform canonical correlation analysis (CCA),
principal components regression (PCR), or multiple linear regression (MLR) on any
data, and for any application. The Climate Predictability Tool (CPT) is an easy to use
package for making tailored and downscaled seasonal climate forecasts. CPT is
designed to produce statistical forecasts of seasonal climate using either the output from
a GCM, or empirical predictors.
The underlying goal in using the CPT has been to address the widespread creation and
communication of quality-controlled seasonal climate forecasts that address specific
needs of different user groups. There are two main approaches to generating seasonal
forecasts: using large-scale models of the global atmosphere, known as general
circulation models (GCMs), or using a statistical approach to relate seasonal climate to
changes in sea-surface temperatures, such as those associated with El Niño, or to other
predictors. In the former case, predictions are made for large-areas, and are often not
very relevant for specific locations. In addition, because of the coarse scale at which the
GCMs operate, the geography in the models is often distorted, and so geographical
locations can be displaced. These GCM outputs therefore need to be adjusted so that
they can be applied at the local level. The CPT tool is designed to perform both forms
of prediction, namely downscaling of GCM output, and purely statistical predictions.

22
Figure 3. Methodological framework to produce seasonal forecasts using CPT

The Geospatial Climate Outlook Forecasting Tool (GeoCOF)

The Geospatial Climate Outlook Forecasting Tool (GeoCOF) is a statistical tool,
developed by FEWS NET for producing probabilistic seasonal forecasts for rainfall and
other climatic parameters. GeoCOF facilitates multiple-linear regression modeling,
identification of potential predictors (e.g. regional sea surface temperatures) and
forecast model skill assessments. Graphical outputs are designed to support regional
climate outlook forums, such as through presentation of forecasts using tercile
categories and maps (Magadzire et al., 2016).
GeoCOF uses observed SST data to forecast rainfall in a given homogeneous region.
At the beginning, the datasets are summed up over the forecast period of interest, for
example, monthly data can be summed up over Jun-July-August-September (JJAS)

23
period in order to generate a time-series of JJAS totals. The JJAS totals are then
converted to a standard normal variable (Z-score). The predefined homogeneous
rainfall zones are used to summarize the predictand (rainfall) variable, allowing the
rainfall for each zone to be correlated to potential predictors, especially oceanic regions
that display a high correlation between their SSTs and rainfall. This allows the
correlations between rainfall and SST to be analyzed, and ultimately for a regression-
based forecast model to be developed for each climate zone.
The SST data aggregated from the oceanic regions that are well correlated with rainfall
is selected. In doing so, very small regions should be avoided as these are not robust in
analysis. Oceanic regions whose correlation with the rainfall zone of interest is difficult
to explain should also be avoided as predictors. Different combinations of potential
predictors are then tested using a number of statistical tests to identify the best statistical
models that can be used to predict the climate variable of interest.
A linear regression model is produced using the NOAA Extended Reconstructed Sea
Surface Temperature (ERSST) v4 monthly lagged SST values as predictor and the
Climate Hazards Group Infrared Precipitation with Stations (CHIRPS) dataset as
predictand data. The regression model can be developed using either a single month
SST, as available in the GeoCLIM ERSST database, or using multiple months, such as
3-month averaged SSTs. The established regression model is trained using the data
from 1981 to 2010. The regression model is refined with the remainder of the historical
rainfall and SST fields using the error variance of the cross-validated forecasts over the
training period. Once a satisfactory model is identified, the current observed SST value
of the predictor then substituted into the regression equation to make a forecast.

24
Conclusion/recommendations

This review provides an overview of the current state of knowledge and operations with
regard to seasonal climate forecasting in Ethiopia. Most of the studies reviewed agree
that potential predictability with moderate skill is achievable by exploiting the
equatorial Pacific SSTAs, particularly in the Niño-3.4 region and the associated ENSO
state. For this oceanic region, the ENSO predictive skill is higher during May (the
preceding month) for the Kiremt season. This is the approach followed by NMA.
SSTAs over the Atlantic Ocean (Gulf of Guinea), the Indian Ocean, and the associated
IOD could provide additional sources of predictability, particularly for the Bega and
Belg seasons. Clearly, strategic and tactical decisions could benefit from seasonal
predictions on longer time scales, such as 2 to 3 month lead forecasts which may be
improved using multivariate indices combining oceanic and atmospheric predictors.
Some recent studies proposed indeed that atmospheric variables such as low-level and
upper level winds could provide greater predictability than surface variables. The merit
of this idea stems from the association between patterns of tropospheric convergence,
divergence and vertical motion (which provide the direct forcing) and the dominant
stability mechanisms in the region. Finally, the demonstration of the predictability of
climate by CGCMs using multi-model ensemble approaches brings optimism for better
operational forecasts, notably for the seasonal prediction of high frequency variables
and particular events. For the time being, the Nino-3.4 SST and ENSO state for the
Kiremt season, supplemented by the IOD (for the Bega and Belg seasons) provide
reliable moderate skill at 1 to 2 months lead time for the Ethiopian seasonal climate.

25
Appendix

Fig. 7.1. Location of SST predictors for Belg season, as described by (Diro et al.,
2008): (a) Zone I; (b) Zone II; (c) Zone III; (d) Zone IV; and(e) Zone V; (f)
homogeneous rainfall zones for the Belg season.

(f

26
Table 7.1. Location of predictors for JAS (top) and MAM (bottom); as developed by
(Nicholson, 2015) for 1-and 2-month lead times.

27
Table 7.2. Correlation between predictors and summer (Kiremt) rainfall regions for 1-
and 2-month lead times

Table 7.3. February predictors for the Belg (MAM) rainfall regions

Table 7.4. May predictors for the summer (JAS) rainfall regions

Source: Nicholson (2015)

28
References

Bahaga, T.K., Kucharski, F., Tsidu, G.M., Yang, H., 2016. Assessment of prediction
and predictability of short rains over equatorial East Africa using a multi-model
ensemble. Theor. Appl. Climatol. 123, 637–649. https://doi.org/10.1007/s00704-
014-1370-1
Bahaga, T.K., Mengistu Tsidu, G., Kucharski, F., Diro, G.T., 2015. Potential
predictability of the sea-surface temperature forced equatorial east african short
rains interannual variability in the 20th century. Q. J. R. Meteorol. Soc. 141, 16–
26. https://doi.org/10.1002/qj.2338
Behera, S.K., Luo, J.J., Masson, S., Delecluse, P., Gualdi, S., Navarra, A., Yamagata,
T., 2006. Erratum: Paramount impact of the Indian ocean dipole on the East
African short rains: A CGCM study (Journal of Climate vol. 18 (21) (4514-
4530)). J. Clim. https://doi.org/10.1175/JCLI9018.1
Bekele, F., 1997. Ethiopian Use of ENSO Information in Its Seasonal Forecasts.
Internet J. African Stud. # 2,
https://www.africabib.org/rec.php?RID=P00007395&DB=p
Camberlin, P., 2009. Nile Basin Climates BT - The Nile: Origin, Environments,
Limnology and Human Use. Monogr. Biol. Springer 307–333.
https://doi.org/10.1007/978-1-4020-9726-3_16
Clarke, A.J., 2008. An Introduction to the Dynamics of El Niño and the Southern
Oscillation. 1stdition, Academic Press.
Connor, S.J., Dinku, T., Wolde-georgis, T., n.d. A Collaborative Epidemic Early
Warning & Response Initiative in Ethiopia .
Désalmand, F., 1998. Meteorology today : An introduction to weather, Climate, and
the environment, Ninth Edit. ed, La Météorologie. Brooks/Cole, Cengage
Learning, Belmont, CA 94002 USA. https://doi.org/10.4267/2042/54533
Diro, G.T., Black, E., Grimes, D.I.F., 2008. Seasonal forecasting of Ethiopian spring
rains, in: Meteorological Applications. pp. 73–83. https://doi.org/10.1002/met.63
Diro, G.T., Canada, E., Catherine, E., Black, L., 2010. Teleconnections between
Ethiopian summer rainfall and sea surface temperature : Part I-observation and
modelling Teleconnections between Ethiopian summer rainfall and sea surface
temperature : part I — observation and modelling.
https://doi.org/10.1007/s00382-010-0837-8
Gissila, T., Black, E., Grimes, D.I.F., Slingo, J.M., 2004. Seasonal forecasting of the
Ethiopian summer rains. Int. J. Climatol. 24, 1345–1358.
https://doi.org/10.1002/joc.1078
Goddard, L., Aitchellouche, Y., Baethgen, W., Dettinger, M., Graham, R., Hayman,

29
 P., 2010. Providing Seasonal-to-Interannual Climate Information for Risk
Management and Decision-making 1, 81–101.
https://doi.org/10.1016/j.proenv.2010.09.007
Hansen, J.W., Mason, S.J., Sun, L., Tall, A., 2011. Review of seasonal climate
forecasting for agriculture in sub-Saharan Africa. Exp. Agric. 47, 205–240.
https://doi.org/10.1017/S0014479710000876
Korecha, D., Barnston, A.G., 2007. Predictability of June-September rainfall in
Ethiopia. Mon. Weather Rev. 135, 628–650.
https://doi.org/10.1175/MWR3304.1
Korecha, D., Sorteberg, A., 2013. Validation of operational seasonal rainfall forecast
in Ethiopia. Water Resour. Res. 49, 7681–7697.
https://doi.org/10.1002/2013WR013760
Magadzire T, Harrison L, Galu G, N.F., 2016. Seasonal Forecasting Using the
GeoCOF Software.GeopCOF Version 2.1 Manual, FEWSN NET, UCSB
Cliamte hazard Group, and SADC Climate Services Center.
Mason, S.J., 2011. Seasonal Forecasting Using the Climate Predictability Tool (CPT).
In: Annual Climate Diagnostics and Prediction Workshop. Science and
Technology Infusion Climate Bulletin , NOAA’s National Weather Service 36th
NOAA Annual Climate Diagnostics and Prediction Workshop Fort Worth, TX,
3-6 October 2011. Available at:
https://www.nws.noaa.gov/ost/climate/STIP/36CDPW/36cdpw-smason.pdf
Nicholson, S.E., 2015. The predictability of rainfall over the Greater Horn of Africa.
Part II: Prediction of monthly rainfall during the long rains. J. Hydrometeorol.
16, 2001–2012. https://doi.org/10.1175/JHM-D-14-0138.1
Nicholson, S.E., 2014. The predictability of rainfall over the greater horn of Africa.
Part I: Prediction of seasonal rainfall. J. Hydrometeorol. 15, 1011–1027.
https://doi.org/10.1175/JHM-D-13-062.1
NMSA, 1996. Climatic and agroclimatic resources of Ethiopia. National Meteorology
Service Agency (NMSA), Meteorological Research Report Series, Vol. 1, No. 1,
Addis Ababa, Ethiopia.
Rosenzweig, C., Hillel, D., 2008. Climate Variability and the Global Harvest
Impacts:Impacts of El Niño and Other Oscillations on Agroecosystems, Oxford
University Press, Inc. https://doi.org/10.1017/CBO9781107415324.004
Segele, Z.T., Lamb, P.J., 2005. Characterization and variability of Kiremt rainy
season over Ethiopia. Meteorol. Atmos. Phys. 89, 153–180.
https://doi.org/10.1007/s00703-005-0127-x
Segele, Z.T., Richman, M.B., Leslie, L.M., Lamb, P.J., 2015. Seasonal-to-interannual
variability of ethiopia/horn of Africa monsoon. Part II: Statistical multimodel

30
 ensemble rainfall predictions. J. Clim. 28, 3511–3536.
https://doi.org/10.1175/JCLI-D-14-00476.1
Shanko, D., Camberlin, P., 1998. The effects of the southwest Indian ocean tropical
cyclones on Ethiopian drought. Int. J. Climatol. 18, 1373–1388.
https://doi.org/10.1002/(SICI)1097-0088(1998100)18:12<1373::AID-
JOC313>3.0.CO;2-K
Troccoli, A., 2008. Seasonal Climate: Forecasting and Managing Risk, Seasonal
Climate: Forecasting and Managing Risk. https://doi.org/10.1007/978-1-4020-
6992-5
Tsidu, G.M., 2012. High-resolution monthly rainfall database for Ethiopia:
Homogenization, reconstruction, and gridding. J. Clim. 25, 8422–8443.
https://doi.org/10.1175/JCLI-D-12-00027.1
WMO, 2002. Applications of Climate Forecasts for Agriculture. Proceedings of an
Expert Group Meeting for Regional Association I (Africa), held 9-13 December
2002 in Banjul, Gambia, AGM-7, WCAC-1, WMO/TD-No. 1223.
Wolter, K., Timlin, M.S., 1998. Measuring the strength of ENSO events: How does
1997/98 rank? Weather 53, 315–324. https://doi.org/10.1002/j.1477-
8696.1998.tb06408.x

31
Science for a food-secure future