Official outlook comparison — FMAM 2026

Season 2026-02-01 to 2026-05-31 · areas: All Ethiopia; FMAM rainfall contribution domain; FMAM main-season rainfall domain (Belg / Gu / Ganna, >=40% of annual) · built from reviewed extractions only (rules v2, 2026-10-11). Research reconstruction; not an official EMI or ICPAC forecast. The first parts describe agreement between the outlooks; "Against CHIRPS observations" scores each of them against what was observed this season.

Key findings — All Ethiopia

ICPAC — All Ethiopia. Category agreement of 47.8% within the compared area, which covers 34% of the analysed national area. Predominant category in both: above normal. Periods differ (Mar 2026–May 2026 vs Feb 2026–May 2026).
EMI — whole zones. Platform area mean of local probabilities over each EMI zone, below / near / above: zone West: no clear category in one source, platform 32% / 33% / 35%; zone North-central: no clear category in one source, platform 34% / 32% / 34%; zone South-central: no clear category in one source, platform 39% / 32% / 29%; zone Southeast: no clear category in one source, platform 34% / 27% / 39%. Zone boundaries: the zone polygons digitized from the EMI figure (georeferenced by fitting the country outline).

Key findings — FMAM rainfall contribution domain

ICPAC — FMAM rainfall contribution domain. Category agreement of 45.4% within the compared area, which covers 39% of the analysed domain. Predominant category in both: above normal. Periods differ (Mar 2026–May 2026 vs Feb 2026–May 2026).
EMI — whole zones. Platform area mean of local probabilities over each EMI zone, below / near / above: zone West: no clear category in one source, platform 32% / 33% / 35% over the whole zone (76% of it in the domain); zone North-central: no clear category in one source, platform 34% / 32% / 34% over the whole zone (73% of it in the domain); zone South-central: no clear category in one source, platform 39% / 32% / 29%; zone Southeast: no clear category in one source, platform 34% / 27% / 39% over the whole zone (99% of it in the domain). Zone boundaries: the zone polygons digitized from the EMI figure (georeferenced by fitting the country outline).

Key findings — FMAM main-season rainfall domain (Belg / Gu / Ganna, >=40% of annual)

ICPAC — FMAM main-season rainfall domain (Belg / Gu / Ganna, >=40% of annual). Category agreement of 50.1% within the compared area, which covers 45% of the analysed domain. Predominant category in both: above normal. Periods differ (Mar 2026–May 2026 vs Feb 2026–May 2026).
EMI — whole zones. Platform area mean of local probabilities over each EMI zone, below / near / above: zone West: no clear category in one source, platform 32% / 33% / 35% over the whole zone (1% of it in the domain); zone North-central: no clear category in one source, platform 34% / 32% / 34% over the whole zone (16% of it in the domain); zone South-central: no clear category in one source, platform 39% / 32% / 29% over the whole zone (77% of it in the domain); zone Southeast: no clear category in one source, platform 34% / 27% / 39% over the whole zone (93% of it in the domain). Zone boundaries: the zone polygons digitized from the EMI figure (georeferenced by fitting the country outline).

Key findings — Robustness of the extraction

Robustness of the extraction. With the ICPAC map as digitized and shifted by one source pixel in each direction, the All Ethiopia category agreement stays between 45.8% and 48.4%. With the ICPAC map as digitized and shifted by one source pixel in each direction, the FMAM rainfall contribution domain category agreement stays between 43.6% and 45.9%. With the ICPAC map as digitized and shifted by one source pixel in each direction, the FMAM main-season rainfall domain (Belg / Gu / Ganna, >=40% of annual) category agreement stays between 48.2% and 50.9%. These are bounded sensitivity checks of the extraction, not statistical confidence intervals; exact EMI zone comparisons remain unavailable.

Comparison maps

EMI official figure: Bega 2025-26 Assessment and Belg2026 hydro met Bulletin, PDF page 22

ICPAC category agreement

Only where both outlooks show a favoured category.

Area	Compared area (share of the analysed area)	Agreement within it	Opposite within it
All Ethiopia	34%	47.8%	17.4%
FMAM rainfall contribution domain	39%	45.4%	19.4%
FMAM main-season rainfall domain (Belg / Gu / Ganna, >=40% of annual)	45%	50.1%	13.1%

EMI zones: printed values and the platform

Zone	Official below / near / above	Platform over the whole zone (region)
Zone West	25% / 30% / 45%	32% / 33% / 35% No clear category in one source
Zone North-central	10% / 30% / 60%	34% / 32% / 34% No clear category in one source
Zone South-central	15% / 30% / 55%	39% / 32% / 29% No clear category in one source
Zone Southeast	20% / 35% / 45%	34% / 27% / 39% No clear category in one source

Platform values are area means of local probabilities: over each EMI zone as drawn in EMI's figure for this outlook, digitized from that figure (approximate boundaries). Percentages are rounded to add up to 100%.

ICPAC: printed favoured category and interval vs the platform

Location	ICPAC (favoured category, printed interval)	Platform below / near / above	Relationship
Within zone West Partly overlaps the domain — 76% of sample cells	Above normal 40–50% 60% of shown cells; also Above normal 50–60% (36%), Near normal 33–40% (5%)	32% / 33% / 35%	No clear category in one source
Within zone North-central Partly overlaps the domain — 73% of sample cells	Above normal 40–50% 57% of shown cells; also Near normal 33–40% (33%), Below normal 40–50% (5%), Above normal 50–60% (5%), Above normal 33–40% (<1%); 4% of the sample grey (no forecast category shown)	34% / 32% / 34%	No clear category in one source
Within zone South-central Entire sample in the domain	Near normal 33–40% 48% of shown cells; also Above normal 40–50% (32%), Below normal 40–50% (20%)	39% / 32% / 29%	No clear category in one source
Within zone Southeast Partly overlaps the domain — 99% of sample cells	Above normal 40–50% 77% of shown cells; also Near normal 33–40% (19%), Below normal 40–50% (4%)	34% / 27% / 39%	No clear category in one source
All Ethiopia	Above normal (most often 40–50%) shown on 89% of the area; 11% grey (no forecast category shown)	34% / 31% / 35%	Same category (where both show a category)
FMAM rainfall contribution domain	Above normal (most often 40–50%) shown on >99% of the area; <1% grey (no forecast category shown)	35% / 31% / 34%	Same category (where both show a category)
FMAM main-season rainfall domain (Belg / Gu / Ganna, >=40% of annual)	Above normal (most often 40–50%) shown on 100% of the area; 0% grey (no forecast category shown)	35% / 30% / 35%	Same category (where both show a category)

ICPAC publishes only the favoured category and its probability interval (the other two categories are not published), for 2026-03-01 to 2026-05-31; the platform covers 2026-02-01 to 2026-05-31. Locations are the EMI arrow-tip boxes or the digitized EMI zones; platform values are area means of local probabilities over each location or area.

Against CHIRPS observations

Each outlook is scored against what CHIRPS observed for its own period; observed terciles from CHIRPS 1993–2025 per 0.25° cell. One season only.

Outlooks and observed terciles — Top: the favoured category of the platform, EMI and ICPAC (platform and EMI: untied leading probability of at least 40%; ICPAC: its printed dominant category, including 33–40% intervals; white: no favoured category under the applicable rule; hatched: no outlook); bottom: the tercile CHIRPS observed (same colours), for the season and for ICPAC's period.

All Ethiopia

Outlook	Period	Observations available	Outlook issued	Favoured category shown	Favoured category observed	Opposite observed	Chance	RPSS vs climatology
Platform — FMAM 2026	Feb–May 2026	100%	100%	37%	36%	34%	33%	-0.03
EMI — Belg (FMAM) 2026	Feb–May 2026	100%	88%	88%	49%	12%	33%	+0.14
Platform, on the EMI zones only	Feb–May 2026	100%	88%	34%	34%	36%	33%	-0.04
ICPAC — MAM 2026	Mar–May 2026	100%	89%	89% 25% of it on the 33–40% interval	53%	7%	33%	not computed

FMAM rainfall contribution domain

Outlook	Period	Observations available	Outlook issued	Favoured category shown	Favoured category observed	Opposite observed	Chance	RPSS vs climatology
Platform — FMAM 2026	Feb–May 2026	100%	100%	39%	32%	38%	33%	-0.05
EMI — Belg (FMAM) 2026	Feb–May 2026	100%	>99%	>99%	47%	13%	33%	+0.12
ICPAC — MAM 2026	Mar–May 2026	100%	>99%	>99% 27% of it on the 33–40% interval	50%	8%	33%	not computed

By EMI zone: favoured category observed (and RPSS)

Zone	Observed area share (below / near / above)	EMI	Platform	ICPAC
West	<1% / 12% / 88%	88% · RPSS +0.31	64% · RPSS +0.05	87%
North-central	2% / 52% / 46%	46% · RPSS +0.23	34% · RPSS -0.03	48%
South-central	16% / 34% / 50%	50% · RPSS +0.13	9% · RPSS -0.21	45%
Southeast	31% / 51% / 18%	18% · RPSS -0.16	33% · RPSS -0.01	30%

Outlooks against CHIRPS observations — Observed terciles and, for each outlook, where its favoured category was observed (blue), near normal vs an outer category (amber), and the opposite outer category (red).

Single season: descriptive scores for one year, not evidence of general skill.
Each outlook is scored against observations for its own target window; the observed tercile uses CHIRPS 1993-2025 terciles per 0.25° cell (official reference periods are not stated).
Favoured category: for the platform and EMI, the untied leading tercile where it reaches 40% (the map display rule); for ICPAC, the dominant category ICPAC printed, including its 33–40% intervals (the share resting on 33–40% is reported). "Chance" is the climatological probability of the favoured category over the same cells (about one in three).
Coverage: observation coverage is the share of the area with usable CHIRPS observations; outlook coverage the share of those cells where the provider issued an outlook (ICPAC grey areas and EMI's excluded dry areas have none); favoured-category coverage the share where its rule selects a category. Match percentages are within the favoured cells.
RPSS: ranked probability skill score against the CHIRPS climatology, on identical cells; ICPAC publishes only the favoured category and its interval, so no RPSS is computed for ICPAC.
EMI zone probabilities are applied uniformly over each digitized zone.

Detailed interpretation

The platform does not show a unique favoured category over EMI zone West under its applicable rule. EMI prints 25% / 30% / 45% (below / near / above). Over the whole zone (as digitized from the EMI figure), the platform area mean is 32% / 33% / 35%: no clear category in one source.

The platform does not show a unique favoured category over EMI zone North-central under its applicable rule. EMI prints 10% / 30% / 60% (below / near / above). Over the whole zone (as digitized from the EMI figure), the platform area mean is 34% / 32% / 34%: no clear category in one source.

The platform does not show a unique favoured category over EMI zone South-central under its applicable rule. EMI prints 15% / 30% / 55% (below / near / above). Over the whole zone (as digitized from the EMI figure), the platform area mean is 39% / 32% / 29%: no clear category in one source.

The platform does not show a unique favoured category over EMI zone Southeast under its applicable rule. EMI prints 20% / 35% / 45% (below / near / above). Over the whole zone (as digitized from the EMI figure), the platform area mean is 34% / 27% / 39%: no clear category in one source.

The EMI zones were digitized from the figure's fill colours and placed by fitting the country outline, so their boundaries are approximate; the figure states no reference period, so probability differences are not calculated.

The outlooks favour different categories over much of their shared displayed area in Ethiopia. ICPAC shows a favoured category on 89% of Ethiopia and no forecast category shown (the legend does not explain the grey areas) on 11%; where it shows one it favours above normal on 68%, most often at 40-50% and 33-40%. The platform favours above normal on 18% and below normal on 14% of Ethiopia. There is category agreement of 47.8% within the compared area, which covers 34% of the analysed national area; opposite categories (above vs below) on 17% of the compared area. ICPAC publishes only the favoured category and its probability interval, so probabilities are not differenced. The official outlook covers Mar 2026–May 2026 and the platform Feb 2026–May 2026 (Feb 2026 only in the platform), so this compares spatial tendencies, not probabilities for the same seasonal event.

The outlooks favour different categories over much of their shared displayed area in the FMAM rainfall contribution domain. ICPAC shows a favoured category on >99% of the FMAM rainfall contribution domain and no forecast category shown (the legend does not explain the grey areas) on <1%; where it shows one it favours above normal on 66%, most often at 40-50% and 33-40%. The platform favours above normal on 18% and below normal on 17% of the FMAM rainfall contribution domain. There is category agreement of 45.4% within the compared area, which covers 39% of the analysed domain; opposite categories (above vs below) on 19% of the compared area. ICPAC publishes only the favoured category and its probability interval, so probabilities are not differenced. The official outlook covers Mar 2026–May 2026 and the platform Feb 2026–May 2026 (Feb 2026 only in the platform), so this compares spatial tendencies, not probabilities for the same seasonal event.

The outlooks favour different categories over much of their shared displayed area in the FMAM main-season rainfall domain (Belg / Gu / Ganna, >=40% of annual). ICPAC shows a favoured category on 100% of the FMAM main-season rainfall domain (Belg / Gu / Ganna, >=40% of annual) and no forecast category shown (the legend does not explain the grey areas) on 0%; where it shows one it favours above normal on 51%, most often at 40-50% and 33-40%. The platform favours above normal on 23% and below normal on 18% of the FMAM main-season rainfall domain (Belg / Gu / Ganna, >=40% of annual). There is category agreement of 50.1% within the compared area, which covers 45% of the analysed domain; opposite categories (above vs below) on 13% of the compared area. ICPAC publishes only the favoured category and its probability interval, so probabilities are not differenced. The official outlook covers Mar 2026–May 2026 and the platform Feb 2026–May 2026 (Feb 2026 only in the platform), so this compares spatial tendencies, not probabilities for the same seasonal event.

Not calculated: rainfall-anomaly differences, because no official rainfall-amount anomaly product was found in the checked ICPAC and EMI products; same-event probability differences, because the target windows or spatial supports differ and ICPAC publishes only favoured-category intervals.

Single-season performance against CHIRPS observations is now evaluated (see Against CHIRPS observations). One season does not establish that any provider, or the platform, is generally more skilful.

Source: ICPAC — MAM 2026

Product	Rainfall forecast - March - May 2026 - IGAD Region
Target period	2026-03-01 to 2026-05-31
Issue date	not stated by the provider
Reference period	not stated
Retrieved	2026-10-09
Latest check	Checked 2026-10-09: new product, reviewed
Content hash (SHA-256)	a9d7f6106eb5cb61e2b7ac9aa3e15a43750615d0d55131e726d6c554738240fe
Extraction	validated
Review	validated by Yonas M, 2026-10-09
Source page	https://www.icpac.net/seasonal-forecast/Mar-May-2026/?region=1&resource_type=27
Original file	https://www.icpac.net/media/images/GHA-MAM-Rainfall.original.png

Wetter than normal conditions expected over most parts of Ethiopia, parts of northern and central and southern Somalia, northern Eritrea, South Sudan, central to southern Kenya, Uganda, Rwanda, Burundi, and Tanzania.
Normal to drier than normal conditions are expected over western and northern Kenya, western and eastern Uganda, south-western to central and north-eastern Ethiopia, Djibouti, western South Sudan and most parts of Somalia.
— ICPAC

Rainfall forecast - March - May 2026 - IGAD Region (original figure)

Source: EMI — Belg (FMAM) 2026

Product	Bega 2025-26 Assessment and Belg2026 hydro met Bulletin
Target period	2026-02-01 to 2026-05-31
Issue date	not stated by the provider
Reference period	not stated
Retrieved	2026-10-09
Latest check	Checked 2026-10-09: new product, reviewed
Content hash (SHA-256)	4a0d4eedcc7c108ab4a6f91c242adde43f46b4dad024e17a48b3b7164392d066
Extraction	validated
Review	validated by Yonas M, 2026-10-09
Source page	https://www.ethiomet.gov.et/products/seasonal-hydromet-bulletin/bega-2025-26-assessment-and-belg2026-hydro-met-bulletin/
Original file	https://www.ethiomet.gov.et/documents/625/Bega_2025_26_and_Belg_outlook2026E_Hydromet_bulletin_English_bulletin.pdf

Summery and advisory for coming Belg (FMAM), 2026 For Belg 2026, rainfall in the Belg-benefiting catchments is anticipated to be normal to above normal Rainfall condition. February is expected to remain mostly dry, while from March through May, the catchments are likely to experience better conditions, to increase the availability of surface water and groundwater recharge. Tercile probability on basin map for coming belg 2026 Season During the coming Belg 2026 season (typically February to May), meteorological forecasts indicate that Most of Belg Benefiting catchments are likely to receive above Normal rainfall amounts over Most of Genale dawa, and lower Reft vally and close to normal or above normal rainfall condition over Wabeshebele, Awash Afar Denakel, Ommo Gibe, Baro Akobo and Upper and Middle Abay Basins. This pattern has important implications for both surface water discharge (riv
— EMI

Bega 2025-26 Assessment and Belg2026 hydro met Bulletin, PDF page 22 (original figure)

What is compared, and what is not

Comparison	Where	Status	Reason
display official probabilities	West	available
favoured category relationship	West	unavailable	reviewed visual (arrow-tip neighbourhood)
zone mean probability	West	available	zone polygons digitized from the EMI figure
same event probability difference	West	unavailable	zone geometry requires alignment; official reference period unknown
display official probabilities	North-central	available
favoured category relationship	North-central	unavailable	reviewed visual (arrow-tip neighbourhood)
zone mean probability	North-central	available	zone polygons digitized from the EMI figure
same event probability difference	North-central	unavailable	zone geometry requires alignment; official reference period unknown
display official probabilities	South-central	available
favoured category relationship	South-central	unavailable	reviewed visual (arrow-tip neighbourhood)
zone mean probability	South-central	available	zone polygons digitized from the EMI figure
same event probability difference	South-central	unavailable	zone geometry requires alignment; official reference period unknown
display official probabilities	Southeast	available
favoured category relationship	Southeast	unavailable	reviewed visual (arrow-tip neighbourhood)
zone mean probability	Southeast	available	zone polygons digitized from the EMI figure
same event probability difference	Southeast	unavailable	zone geometry requires alignment; official reference period unknown
mapped category agreement	All Ethiopia	available	reviewed digitization
mapped category agreement	FMAM rainfall contribution domain	available	reviewed digitization
mapped category agreement	FMAM main-season rainfall domain (Belg / Gu / Ganna, >=40% of annual)	available	reviewed digitization
same event probability difference	—	unavailable	target window mismatch; official publishes only favoured category interval
rainfall anomaly difference	—	unavailable	official amount anomaly not found in checked products
forecast accuracy	—	available	single-season scores against CHIRPS observations (see Against CHIRPS observations)

Notes

Area means are area-weighted means of local grid-cell probabilities, not probabilities of area-mean rainfall.
Favoured category: for the platform and EMI, the untied leading tercile where it reaches 40% (the map display rule), otherwise no clear category; for ICPAC, the dominant category ICPAC printed, including its 33–40% intervals.
ICPAC covers Mar–May 2026 and the platform Feb–May 2026: a probability for one period cannot be built by averaging monthly probabilities of the other, so they are compared only as spatial tendencies.