← ENSO country deep dives

ENSO deep dive · Occupied Palestinian Territory

Gaza and the West Bank

What the 2026 El Niño means for the 2026/27 winter in Gaza and the West Bank, including East Jerusalem, how far to trust it, and what winter weather does there  ·  GPCC 1891–2025, ERA5 1950–2026, IMERG 1998–2026, Beer Sheva gauge 1921–2016 (Gaza), Jerusalem gauge 1908–2015 (West Bank)  ·  Niño3.4 (NOAA)

Summary by section

1. What the forecasts say. A very strong El Niño is under way. NOAA gives El Niño 100 per cent odds through January–March and a 75 per cent chance that October–December will be the strongest since 1950, weakening through the winter. Every multi-model forecast checked (C3S, ECMWF SEAS5, NMME, IRI, the WMO Lead Centre) favours above-normal rain over the southern Levant in October–December (40–70 per cent for the wet tercile), with weaker odds in November–January and no wet signal for December–February, where one system (NMME) leans dry. All favour a warmer-than-normal winter. ECMWF's October forecast (SEAS5, published on 5 October) keeps the wet start: October–December is at about a 9-year wet return period over both Gaza and the West Bank, with high skill. The later windows are only slightly wet, with moderate skill for November–January and December–February and low skill after.

2. How much El Niño matters. Since 1979, clearly, and the same in both places. Gauge-based records and ERA5 correlate at +0.55 to +0.67 with winter Niño3.4 over Gaza and +0.57 to +0.59 over the West Bank, and 9 of the 14 El Niño winters fell in each area's wettest third. That moves the chance of a wettest-third winter from about one in three to about two in three, with a wide range on 14 cases. But the same records show no positive link before 1979, and the strong El Niños, the relevant analogues for this year, have been mixed: in Gaza three of six were wet, one normal and two normal to dry; in the West Bank none was in the driest third in GPCC or at Jerusalem, though other records put single winters there. The two wettest, 1982/83 and 1991/92, both followed major volcanic eruptions. El Niño shifts the odds towards a wet winter, mostly in its first half; it does not forecast one.

3. What changes in an El Niño winter. Mostly the number of rain days: about ten more per winter than in a La Niña winter in Gaza, and 10 to 14 more in the West Bank, and, in most records, more days of 10 mm or more. The biggest single storms are only loosely tied to El Niño, and cold nights not at all.

4. What winter weather does. In Gaza, every winter since October 2023 has flooded tents, including the very dry 2024/25. Damage followed ordinary rain, high tides with almost no rain, wind and cold nights. Last winter the Shelter Cluster counted more than 42,000 tents damaged in Storm Byron alone, 71 per cent of households were flooded at least once, and at least 11 children died of hypothermia (Ministry of Health in Gaza). In the West Bank, weather does far less harm and is poorly counted. The worst winters before October 2023 brought floods (January 2013) and snow (Storm Alexa, December 2013); since October 2023 the UN has quantified one West Bank weather event, when at least 66 households in 18 herding communities lost tents and animal shelters or had them damaged on 28–30 December 2025.

5. What this winter could bring. In Gaza, about half of winters bring a storm with a day of 50 mm or more, as Byron had, El Niño or not. Last winter's alert-based counts total about 80,000–100,000 household-impacts, about 58,500 tents and shelters damaged and 35–36 deaths reported, and they understate it: a household survey suggests about 250,000 households were flooded at least once. Exposure is higher this winter: 451 sites holding about 464,000 people are rated high flood risk. Counts per storm are tens to hundreds of times higher than before October 2023, a ratio that mixes how people are sheltered with what is counted and how closely. Scaled to the 1.7 million people now in displacement sites, last winter's counts imply about 140,000–230,000 people-impacts with no storm of 50 mm or more, 400,000–530,000 with one and 670,000–830,000 with two: planning scenarios from one winter, with alert counts as floors. In the West Bank, a wet first half is the most likely El Niño outcome (a storm of 20 mm or more came before December in 7 of 8 El Niño winters), and the most exposed are Bedouin and herding communities in Area C, about 34,000 people still displaced from the Jenin, Tulkarm and Nur Shams camps, and farmers whose greenhouses take the wind, frost and flood damage. In both places the wettest week of a typical winter brings a fifth to a quarter of the year's rain; Byron's week brought about half. None of this is a forecast.

6. Farming in the West Bank. The three farming zones cannot be told apart in their El Niño rain signal; they differ in what they grow. Rainfed wheat and barley were above trend in 7 of 9 crop years after El Niño winters and 3 of 11 after La Niña winters, and the four worst harvests since 1994 all followed La Niña winters: El Niño has made a bad cereal year less likely rather than a bumper one more likely. Spring greenness after El Niño winters was above normal in 8 of 8 years in the Central Highlands (several only slightly) and 6 of 8 in the other zones. Olives follow their own two-year bearing cycle and, since 2023, access, and show no clear El Niño signal. This winter feeds the 2027 cereal harvest, spring grazing and the autumn 2027 olive harvest.

About the catalogue row

The survey's “Palestine” row was computed on the Natural Earth polygon for the West Bank only, so it describes the West Bank (its ERA5 value was OND +0.56) and has never described Gaza. Its second citation is garbled: the paper is Price, Stone, Huppert, Rajagopalan & Alpert (1998) in Geophysical Research Letters, not a Journal of Climate paper. And Mariotti et al. (2002) found the El Niño autumn wet signal in the western Mediterranean, with negative anomalies in the east in their reanalysis. The grades here rest on this page's records and on the Israel-specific literature.

1. What the forecasts say for this winter

Latest Niño3.4 in NOAA PSL's current series: +1.89 °C for August 2026 (three-month mean +1.73).

Notes and method

The historical analysis on this page uses the series' pinned Niño3.4 (ERSST v5 basis, about 0.2 °C cooler than the current ERSST v6 series), and HadISST only where a record starts before 1950.

NOAA's September outlook gives El Niño 100 per cent odds for every season through January–March 2027. It has the event peaking in October–December, with a 75 per cent chance of exceeding +2.5 °C on its relative Niño3.4 index, which no event since 1950 has done. The coming winter is likely to be a stronger El Niño than any the relationships below were fitted on.

More detail

The median relative index at the peak is +2.67 °C. It then weakens to about +1.8 °C by January–March. On that same relative index, the October–December and November–January medians are above every El Niño since 1950. All 22 models in the IRI plume stay “very strong” through December–February, and IRI's traditional Niño3.4 (on a different sea-surface temperature dataset) was +2.52 °C in August and about +3.0 °C in mid-September; the highest month in the series this page uses is +2.72 °C (November 2015). A positive Indian Ocean Dipole is developing and is expected to fade by early 2027.

The rainfall forecasts agree on a shape, not only a sign: wet odds for October–December, weaker for November–January, and no wet signal for December–February, where NMME leans dry (40–60 per cent). That is the part of the season the El Niño signal has historically favoured (section 2), and the part where SEAS5 has skill. The forecasts also agree the winter will be warm, which makes snow on the West Bank hills less likely but does not remove cold nights.

More detail

In Gaza, the newborn hypothermia deaths of December 2024 happened on nights of about 9 °C. SEAS5 and the C3S multi-model mean lean towards a negative North Atlantic Oscillation in late winter. Most Levant studies find that pattern is not a wet one for Israel and the Occupied Palestinian Territory (Nissen et al. 2010; Sandler et al. 2024), and the forecast pressure anomaly over the Levant is near zero.

ECMWF's October forecast (SEAS5, published on 5 October) is the first of the October issuances. It keeps the wet start: October–December is at about a 9-year wet return period over both Gaza and the West Bank, with high skill. Like for like, that is a little wetter than a month ago for Gaza (about 7 years) and less extreme for the West Bank (23 years). The rest of the winter, slightly dry or near normal in the September issuance, is now slightly wet, and February–April is wetter again; skill is moderate for November–January and December–February and low after. It also keeps the warm signal: ECMWF's charts give a 70–100 per cent chance that November–January and December–February fall in the warmest third of 1993–2016 winters over the eastern Mediterranean; part of that is the warming since those years. The other centres' October forecasts are not yet out, so the table below is still the September one, including its ECMWF row.

Tercile probabilities for the wet (AN) or dry (BN) category over the southern Levant, read from the legend classes of each centre's published maps.

Notes and method

Gaza is smaller than every model's grid cell and the West Bank's hill country is not much larger, so these are southern-Levant values for both. “No signal” means no category above 40%. Compiled 30 September 2026. The October issuances (NOAA on 8 October, C3S around 10–13 October) will extend the horizon to January–March, and the Mediterranean Climate Outlook Forum (MedCOF) normally issues its winter consensus in late November.

SEAS5 for Gaza: skill and the October 2026 forecast

ECMWF's October 2026 forecast (SEAS5, published on 5 October) keeps a wet start to the season over Gaza. October–December is at a 9-year wet return period with high skill (r = 0.60), and November–January at about 5 years with moderate skill. The later windows also lean wet (about 4 to 8 years), with moderate skill for December–February (0.32) and low skill after it. SEAS5's wettest October–December hindcasts are El Niño winters, so this is the El Niño signal the model carries, not an independent source of information.

More detail

December–February and January–March, slightly dry in the September issuance, are now at about 4 years each, and February–April at about 8 years (skill 0.11–0.13 after December–February). Like for like, the September issuance had October–December at about 7 years. The raw ensemble-mean files for a box over Gaza show the same shape more modestly (table below): October–December is the 9th-wettest of 46 October issuances since 1981 and 11 per cent above the hindcast mean; November–January to January–March are 3–5 per cent above it, which is close to normal, and February–April is 11 per cent above. The return periods rank higher than the raw values because the app's method takes a long-term trend out of the hindcasts first. SEAS5's three wettest October–December hindcasts from October are 1986, 2018 and 1994, all El Niño winters.

What the Oct 2026 issuance forecasts. Windows at or beyond the app's 3-year return-period threshold (in-season windows excluded): OND wet 9 yr (high skill), NDJ wet 5 yr (moderate skill), DJF wet 4 yr (moderate skill), JFM wet 4 yr (low skill), FMA wet 8 yr (low skill). Under the app's rule, an alert needs the return period and at least moderate skill. Until the pixel product is out this is provisional: OND, NDJ meet both conditions on either measure of skill; DJF (pixel-median r 0.32, area-mean r 0.29) sits at the moderate-skill boundary.

Source of the October 2026 return periods

The seas5-skill app's pixel product, which this page normally reads, is recomputed the day after a new forecast lands, so the return periods here come from the team's raster statistics for Gaza as one area (the ensemble mean averaged over the admin-1 polygon, in the team database), ranked against the October hindcasts of 1981–2025 with the app's own method. Skill is still the median pixel correlation from the app's cube, which depends on the hindcasts and not on the new forecast. Windows that had already started at issuance are left blank: they blend in observed months, and the latest ERA5 month arrives a day after the forecast.

Each cell: median pixel r, its bin, the share of the zone's cells at moderate-or-better skill, and the return period of the Oct 2026 forecast anomaly for the area mean (dry = forecast below its hindcast median).

Notes and method

Skill source: skill_stats_grid_detrended.nc (seas5-skill, DEV blob), the same cube behind the app's pixel skill map. Median of pixel correlations, not the correlation of the area mean, so it is a conservative summary for a coherent area.

Cross-check from the raw ensemble-mean files (box 31.0–31.8°N, 34.0–34.8°E), Oct issuances 1981–2026: where the 2026 forecast ranks among all 46 (1 = wettest), its ratio to the hindcast mean, and the three wettest hindcast years.

The same forecast, month by month

By month, the forecast is above its hindcast mean in October (25 per cent, the 11th-wettest of 46 October forecasts), November (7 per cent) and December (10 per cent), at its mean in January and February, and well above it in March (28 per cent, the 4th-wettest) and April (18 per cent). The October signal is much smaller than a month ago, when the September issuance had October 79 per cent above its mean. Skill for single months is high for October (r = 0.54), moderate for November and low from December on (0.06–0.23), so read the monthly split as a guide to where the three-month signal sits, not as a forecast for individual months.

SEAS5 ensemble-mean totals are smoother than any single year, so compare the forecast with the hindcast mean and rank, not with observed rainfall. ERA5 mean: the three cells over Gaza, same years, for scale.

SEAS5 for the West Bank: skill and the October 2026 forecast

Over the West Bank, ECMWF's October 2026 forecast (SEAS5, published on 5 October) still shows a wet start to the season, though a less extreme one than a month ago. October–December is at a 9-year wet return period with high skill (r = 0.62); the September issuance had it at 23 years. The later windows also lean wet (about 3.5 to 7 years), with moderate skill at best. SEAS5's wettest October–December hindcasts for the area are El Niño winters, so this is the El Niño signal the model carries.

More detail

November–January is at about 7 years with moderate skill, December–February at about 3.5 years with moderate skill (0.33), January–March at about 4 years with low skill, and February–April at about 6 years with skill just at the moderate threshold (0.31). The raw ensemble-mean files show the same shape more modestly: October–December is the 8th-wettest of 46 October forecasts since 1981 and 12 per cent above the hindcast mean, and the later windows are 1–8 per cent above it. The return periods rank higher than the raw values because the app's method takes a long-term trend out of the hindcasts first. SEAS5's wettest October–December hindcasts for the area are 1986, 2018 and 2002, all El Niño winters. (The skill table counts 17 cells, every 0.25° cell touching the West Bank; the ERA5 series uses the 16 that cover more than 0.5 per cent of it.)

What the Oct 2026 issuance forecasts. Windows at or beyond the app's 3-year return-period threshold (in-season windows excluded): OND wet 9 yr (high skill), NDJ wet 7 yr (moderate skill), DJF wet 4 yr (moderate skill), JFM wet 4 yr (low skill), FMA wet 6 yr (moderate skill). Under the app's rule, an alert needs the return period and at least moderate skill. Until the pixel product is out this is provisional: OND, NDJ, DJF meet both conditions on either measure of skill; FMA (pixel-median r 0.31, area-mean r 0.27) sits at the moderate-skill boundary.

Source of the October 2026 return periods

The seas5-skill app's pixel product, which this page normally reads, is recomputed the day after a new forecast lands, so the return periods here come from the team's raster statistics for the West Bank as one area (the ensemble mean averaged over the admin-1 polygon, in the team database), ranked against the October hindcasts of 1981–2025 with the app's own method. Skill is still the median pixel correlation from the app's cube, which depends on the hindcasts and not on the new forecast. Windows that had already started at issuance are left blank: they blend in observed months, and the latest ERA5 month arrives a day after the forecast.

Each cell: median pixel r, its bin, the share of the zone's cells at moderate-or-better skill, and the return period of the Oct 2026 forecast anomaly for the area mean (dry = forecast below its hindcast median).

Notes and method

Skill source: skill_stats_grid_detrended.nc (seas5-skill, DEV blob), the same cube behind the app's pixel skill map. Median of pixel correlations, not the correlation of the area mean, so it is a conservative summary for a coherent area.

Cross-check from the raw ensemble-mean files (box 31.2–32.6°N, 34.8–35.6°E), Oct issuances 1981–2026: where the 2026 forecast ranks among all 46 (1 = wettest), its ratio to the hindcast mean, and the three wettest hindcast years.

The same forecast, month by month

By month, the forecast is above its hindcast mean in October (29 per cent, the 11th-wettest of 46 October forecasts), November (13 per cent) and December (7 per cent), close to it in January and February, and well above it in March (23 per cent, the 2nd-wettest) and April (15 per cent). Skill for single months is high for October (r = 0.57), moderate for November and December, and low after that, so read the monthly split as a guide to where the three-month signal sits.

SEAS5 ensemble-mean totals are smoother than any single year, so compare the forecast with the hindcast mean and rank, not with observed rainfall. ERA5 mean: the 16 cells over the West Bank, same years, for scale.

2. How much El Niño matters

In this part: The link switched on in the late 1970s · What El Niño winters have looked like since 1979 · Every strong El Niño winter since 1950 · Which part of the winter

Gaza's rain falls between October and April, most of it from December to February, almost all of it from Mediterranean cyclones. Gaza is 365 km²: three ERA5 cells and a corner of one GPCC cell. So this section uses four rainfall records: two gauge-based (the GPCC analysis, and the Beer Sheva gauge, which is one of GPCC's inputs), a reanalysis that assimilates no rain gauges (ERA5), and a satellite product (IMERG). It asks two questions: does El Niño make Gaza's winter wetter, and has it always done so?

The link switched on in the late 1970s

Since the late 1970s, yes; before that, no. Since 1979 the correlation between the October–April total and December–February Niño3.4 is +0.55 in GPCC (to 2024/25), +0.59 in ERA5 (to 2025/26) and +0.67 at the Beer Sheva gauge (to 2015/16). Each is significant at p < 0.001 and unchanged by detrending, and each holds at +0.45 to +0.62 without 1982/83 and 1991/92. Before 1979 the same records give −0.06 (GPCC, from 1891), −0.26 (Beer Sheva, from 1921; nominally significant, p = 0.045) and −0.26 (ERA5, from 1950). Over each full record the correlation is weak (+0.13 to +0.29). All three running correlations climb out of the noise band in the early 1980s and stay out. The switch is in the gauge records as well as the reanalysis, so it is not an artefact of ERA5.

The literature found the same break. Price et al. (1998) saw significant correlations only in the last 25 years of their Israeli record, and Alpert et al. (2005) found no connection before the 1970s. A 2026 Israel Meteorological Service technical report (not peer-reviewed) finds r ≈ 0.50–0.55 over the last 50 years, strongest in the northern Negev and the Shephelah next to Gaza, and about 11–14 per cent more rain per degree of Niño3.4. Similar switches have been found for Arabian winter rainfall (Kang et al. 2015) and Euro-Mediterranean winter storm frequency (Kamil et al. 2017: −0.24 in 1950–79, +0.44 in 1987–2016). Reviews call the eastern Mediterranean ENSO influence “seasonally dependent, small, and intermittent” (Dayan et al. 2015).

What this means for use: the relationship has held for 45 winters, the whole period in which seasonal forecasts have existed. But it is not a law of nature, and it could weaken again. IMERG (1998–) agrees on the sign but is weaker (r = +0.31, not significant over 28 winters). Its seasonal totals track the gauges less well (r = 0.72 with GPCC) than ERA5 does (0.91).

Is the change real? A Fisher z-test for the difference between the correlations before and after 1979: ERA5 z = 3.8, p < 0.001; GPCC z = 3.6, p < 0.001; Beer Sheva gauge z = 5.0, p < 0.001.

Notes and method

The 1979 break was not chosen blind: it is where the literature places the change (Price et al. 1998; Alpert et al. 2005) and the start of the satellite era. p-values for the period after the break are conditional on choosing it; the full-record rows show what an unsplit analysis gives.

What El Niño winters have looked like since 1979

Since 1979, 9 of the 14 El Niño winters in the GPCC record fell in the wettest third and 2 in the driest; in ERA5, 9 and none. La Niña winters ran the other way: 2 of 19 in the wettest third and 10 in the driest. Before 1979, El Niño winters were as likely to be dry as wet (8 in the wettest third and 11 in the driest, of 25), and La Niña winters were if anything wet (11 of 26 in the wettest third, 5 in the driest).

Terciles are computed within each period, so each period is judged against its own climate.

Notes and method

Phase from December–February Niño3.4 (±0.5 °C): HadISST before 1979, the pinned ERSST v5 series from 1979. On HadISST, 16 winters since 1979 count as El Niño rather than 14, and 11 of them were in the GPCC wettest third.

Every strong El Niño winter since 1950

Strength does not buy certainty, and the literature agrees that larger El Niños do not reliably bring larger rainfall anomalies here (Mariotti et al. 2002; Alpert et al. 2005). Of the six strong El Niños since 1979:

• three were wet: 1982/83 and 1991/92, the two wettest winters since 1979, and 2015/16;
• one was normal (1997/98);
• two were normal to dry, depending on the record: 2009/10, and 2023/24, the first winter after October 2023.

1982/83 and 1991/92 both followed major volcanic eruptions (El Chichón, Pinatubo), and 1991/92 was unusually cold across the Middle East. That makes them doubtful analogues for a winter every forecast expects to be warm. It is a caution about analogues, not a reason to discount them: we found no study attributing their rainfall to the eruptions. The two strong events before 1979 (1957/58, 1972/73) were normal to dry. The El Niño winters that delivered most reliably were moderate ones (1986/87, 1987/88, 1994/95, 2002/03).

Niño3.4 ≥ +1.5 °C in December–February on the pinned series. Percentiles are within 1979–2025 for winters from 1979, within 1950–1978 before.

Which part of the winter

The concurrent signal runs through the whole season, strongest early: r = 0.35–0.52 for every three-month window from October–December to January–March, in both GPCC and ERA5. What matters for a forecast now is the August–October Niño3.4, known before the season starts. It predicts the season total nearly as well (r = 0.45 in GPCC, 0.50 in ERA5), the windows from October–December to December–February at 0.32–0.46, and January–March more weakly (0.28–0.29, borderline significant). That matches SEAS5, whose skill fades after December, and the literature, which puts the eastern Mediterranean ENSO signal in autumn and early winter (Hochman & Gildor 2025; Mariotti et al. 2005).

1979–2025. Concurrent = Niño3.4 averaged over the same three months (December–February for October–April). August–October Niño3.4 is the value known when the season starts.

In this part: The link switched on in the late 1970s · What El Niño winters have looked like since 1979 · Every strong El Niño winter since 1950 · Which part of the winter

The West Bank's rain falls between October and April, mostly from December to February. The hills rise to 700–1,000 m and drop steeply east to the Jordan Valley, about 300 m below sea level, and rainfall more than halves from the western slopes to the valley. This section uses four records:

• the GPCC gauge analysis;
• the Jerusalem rain gauge (St Anne to spring 1948, Jerusalem Central from autumn 1948 to 2015), which is one of GPCC's inputs;
• ERA5 over the 16 cells covering the West Bank, a reanalysis that assimilates no rain gauges;
• IMERG.

The link switched on in the late 1970s

Since the late 1970s, yes; before that, no. Since 1979 the correlation between the October–April total and December–February Niño3.4 is +0.57 in GPCC (to 2024/25), +0.59 at the Jerusalem gauge (to 2013/14) and +0.59 in ERA5 (to 2025/26). Each is significant at p < 0.001 and unchanged by detrending. Before 1979 the same records give −0.03 (GPCC, from 1891), −0.16 (Jerusalem, from 1909) and −0.21 (ERA5, from 1950). Over each full record the correlation is weak (+0.16 to +0.30). As in Gaza, the switch shows in the gauge records as well as the reanalysis.

The literature found the same break. Price et al. (1998) saw significant correlations only in the last 25 years of their Israeli record, and Alpert et al. (2005) found no connection before the 1970s. A 2026 Israel Meteorological Service technical report (not peer-reviewed) finds r ≈ 0.50–0.55 over the last 50 years, strongest in the hills around Jerusalem, the Shephelah and the northern Negev. IMERG (1998–) agrees on the sign but is weaker (r = +0.30, not significant over 28 winters).

Is the change real? A Fisher z-test for the difference between the correlations before and after 1979: ERA5 z = 3.6, p < 0.001; GPCC z = 3.6, p < 0.001; Jerusalem gauge z = 3.9, p < 0.001.

Notes and method

The 1979 break was not chosen blind: it is where the literature places the change (Price et al. 1998; Alpert et al. 2005) and the start of the satellite era. p-values for the period after the break are conditional on choosing it; the full-record rows show what an unsplit analysis gives.

What El Niño winters have looked like since 1979

Since 1979, 9 of the 14 El Niño winters in GPCC fell in the West Bank's wettest third and none in the driest; in ERA5, 9 and one. La Niña winters ran the other way: 2 of 19 in the wettest third and 11 in the driest. Before 1979 El Niño winters were spread evenly (9 wettest, 7 driest of 25).

Terciles are computed within each period, so each period is judged against its own climate.

Notes and method

Phase from December–February Niño3.4 (±0.5 °C): HadISST before 1979, the pinned ERSST v5 series from 1979. On HadISST, 16 winters since 1979 count as El Niño rather than 14, and 11 of them were in the GPCC wettest third.

Every strong El Niño winter since 1950

None of the six strong El Niños since 1979 was in the driest third in GPCC or at the Jerusalem gauge, a stronger result than in Gaza. But other records put single winters there, and the two wettest strong events followed volcanic eruptions.

More detail

Two of the other records disagree for single winters: ERA5 puts 2009/10 just inside its driest third (32nd percentile), and IMERG puts 2015/16, the strongest event in the record, there too (21st percentile). The Israel Meteorological Service says a dry 2026/27 “like 2015/16” cannot be ruled out. The two wettest strong events both followed major volcanic eruptions: 1982/83, and 1991/92, Jerusalem's wettest winter on record, with heavy snow in February 1992. That makes them doubtful analogues for a winter forecast to be warm. 1997/98 was wet in GPCC and ERA5 but normal at Jerusalem; 2009/10, 2015/16 and 2023/24 were near normal in GPCC. The two strong events before 1979 were dry or normal.

Niño3.4 ≥ +1.5 °C in December–February on the pinned series. Percentiles are within 1979–2025 for winters from 1979, within 1950–1978 before.

Which part of the winter

The concurrent signal runs through the whole season and is strongest in October–December (r = 0.47–0.54). The August–October Niño3.4, known before the season starts, predicts the season total nearly as well (r = 0.49 in GPCC, 0.54 in ERA5). Unlike in Gaza, it still says something about January–March (0.33–0.37).

1979–2025. Concurrent = Niño3.4 averaged over the same three months (December–February for October–April). August–October Niño3.4 is the value known when the season starts.

The signal is not an artefact of either area. Across the southern Levant (the Israeli coastal plain, the hills of the West Bank, western Jordan and southern Lebanon), cells show the same positive correlation. It is most coherent in October–December (97 per cent of analysable cells at r ≥ 0.3, though only 71 cells are analysable; the rest are too dry in that window) and November–January (76 per cent of 144), and patchier in December–February and January–March. As in the survey, each cell keeps the best of four lags (0–3 months), which flatters r somewhat; the composite maps use concurrent December–February Niño3.4 and no lag search. Over the 14 El Niño winters of 1981/82–2025/26, every analysed cell in the region had a positive mean anomaly.

3. What changes in an El Niño winter, and what does not

A wetter El Niño winter is mostly a winter with more rainy days, not bigger storms. In all three records the number of rain days rises with Niño3.4 (r = 0.55–0.65): ERA5 has 54 in an average El Niño winter against 42 in a La Niña winter, IMERG 41 against 33, and the Beer Sheva gauge 33 against 22. The heaviest days are only loosely tied (correlations of 0.16–0.34 for days of 20 mm or more and for the season's wettest day), and cold nights not at all.

More detail

Days of 10 mm or more rise clearly at Beer Sheva (7.8 against 4.4) and in ERA5, but not in IMERG. Of the heaviest-day measures, only ERA5's wettest day clears p < 0.05. The Israel Meteorological Service likewise finds no significant El Niño link to daily or storm maximum rainfall, but about 15 per cent more days with a nearby low in El Niño than in La Niña winters. Cold nights show no link (Beer Sheva, r = +0.08). Windy days may be somewhat more frequent (12.8 against 9.3 days in the windiest 5 per cent of ERA5's winter days, r = +0.34, p = 0.02); with 17 tests in the table, treat that as suggestive.

The distinction matters for people in tents. The storms that tear down shelters and flood whole sites, like Byron in December 2025, can come in any winter. What El Niño adds is more wet days around them. Our reading, not tested here, is that this is what keeps ground saturated, stormwater basins full and bedding wet between storms.

Means per winter, by ENSO phase. The records differ in level: ERA5 spreads rain over a 25 km cell and understates heavy days; IMERG is satellite-only; Beer Sheva is drier and 46 km inland. Compare phases within a record, not across records.

A wetter El Niño winter here is a winter with more rain days and more moderate-to-heavy days. Rain days rise with Niño3.4 in all three records (r = 0.51–0.56): ERA5 has 65 in an average El Niño winter against 50 in a La Niña winter, IMERG 58 against 46, and the Jerusalem gauge 48 against 38. The single wettest day shows no consistent link, and nor do cold nights (ERA5).

More detail

Days of 10 mm or more rise in ERA5 and at Jerusalem (20 against 14). At Jerusalem so do days of 20 mm or more (10 against 7, r = +0.41). Windy days are a little more frequent (r = +0.31). More wet days should mean wetter ground, so that a given storm runs off faster into the wadis; that is a reasonable expectation, not something tested here.

Means per winter, by ENSO phase. The records differ in level: ERA5 spreads rain over 25 km cells and understates heavy days; IMERG is satellite-only; Jerusalem is one hill-town gauge. Compare phases within a record, not across records.

Cold snaps show no link to El Niño in either area. Since 1979/80 the number of cold nights in a winter does not follow winter Niño3.4: Gaza, nights of 8 °C or colder, r = +0.06 (p = 0.67); West Bank, nights of 3 °C or colder, r = +0.04 (p = 0.80). Nor do the winter's coldest night, the number of cold snaps of two nights or more, or the night count two degrees either side of each threshold (correlations between -0.09 and +0.13). El Niño is therefore no guide to cold snaps. The forecasts' warm signal is a separate matter: warmer winters have had fewer cold nights (r = -0.86 for Gaza and -0.64 for the West Bank between a winter's cold nights and its December–February mean night temperature). Cold deaths, though, have followed shelter more than temperature (section 5).

4. What winter weather does

As of September 2026 about 1.7 million people live in roughly 1,600 displacement sites, and 65 per cent of households live in tents or makeshift shelters. Last winter 71 per cent of households were flooded at least once, and 83 per cent of those in tents (Shelter Cluster). The inter-agency flood-exposure outlook for 2026/27 puts 989 of 1,764 assessed sites in medium- or high-risk areas, and 451 sites holding about 464,000 people in high-risk areas (IOM, Site Management Cluster and Shelter Cluster, September 2026; OCHA). That exposure map uses terrain and past flooding, not rainfall. No published analysis links rainfall amounts to tent flooding in Gaza, and no rain gauge inside Gaza reports internationally. The table below is a first pass at that link: every dated winter impact in UN reporting since October 2023, with the weather over Gaza on those days.

Three things stand out.

• The rain that floods tents is ordinary. In 2023/24, tent flooding was reported on days when neither product had more than about 16 mm over Gaza. On 31 December 2024, in an otherwise very dry winter, IMERG had 34 mm (ERA5 9 mm). In 2025/26 the rain-driven events ranged from about 13 to 66 mm on their wettest day in IMERG. For comparison, IMERG averages 9 days of 10 mm or more and 3.3 of 20 mm or more over Gaza every winter. Byron (rain from 8 December, the storm proper from 10 December; about 150 mm in IMERG) was large, but the benchmark from before October 2023, Storm Alexa in December 2013, was larger (about 185 mm over 11–14 December in IMERG).
• Several damaging events came with little or no rain. These were the coastal flooding of 24–25 November 2024 (about 3 mm; high tides), the wind storms of 5–6 February 2025 and 14 March 2026, and the cold spells in which newborns died of hypothermia (ERA5 minimum 9 °C on 24–29 December 2024, 4 °C on 24–25 February 2025). A seasonal rainfall forecast says nothing about these.
• Reporting changed between winters, from descriptive in 2023/24 to authority counts in 2024/25 and daily cluster alerts in 2025/26. So impact counts cannot be compared across winters like for like; the dates can. Section 5 sets the counts side by side for scale only, and says what is mixed into that comparison.

How often the rain that has caused these impacts comes: over 1998–2025, IMERG puts an average of 9.2 days of ≥ 10 mm, 3.3 of ≥ 20 mm, 1.5 of ≥ 30 mm and 0.6 of ≥ 50 mm over Gaza in each October–April.

Hazard: R rain and flooding, S sea surge or high tide, W wind, C cold.

Notes and method

Weather columns cover the window from the day before the reported start to the reported end: IMERG wettest day over Gaza (north = cells at 31.4–31.6°N: North Gaza, Gaza governorate and most of Deir al Balah; south = 31.2–31.4°N: Khan Younis and Rafah); window totals from IMERG and from ERA5 (which smooths rain over 25 km cells and runs lower on heavy days; the two disagree on single days by a factor of two or more, so read them as a range); ERA5 lowest daily minimum temperature and highest hourly 10 m wind, averaged over the three cells over Gaza (partly sea, so milder and less gusty than an exposed displacement site). Figures in italics are from the Ministry of Health in Gaza, the Palestinian Civil Defense or the Government Media Office, as relayed by the UN; the rest are UN agency or cluster figures.

Before October 2023

Before October 2023, winter floods hit homes, refugee camps and the low ground around Wadi Gaza and the Sheikh Radwan stormwater lagoon, mostly because drainage was overwhelmed or pumping stations had no power. Even then the threshold was low. “Mild rainfall” (about 3 mm in IMERG) flooded houses and a sewage pump station in Gaza city in November 2017, and four children died of cold in the January 2015 storm, all from families whose homes had been destroyed in the 2014 hostilities.

Counts by winter and ENSO phase

The counts do not separate by ENSO phase. Before October 2023, the three counted El Niño winters (20, 501 and 800 people) and the three counted neutral winters (110, 560 and 21,000 people) overlap, and six counted winters are too few for any test to separate them; no La Niña winter has a count. Since October 2023 both counted winters were La Niña winters, and the El Niño winter of 2023/24 has reports but no number. An average per phase would only restate the change since October 2023, so none is given.

Winter weather does much less harm in the West Bank than in Gaza, and what it does is poorly counted. The record has two large events from before October 2023:

• the floods of 7–10 January 2013: about 1,000 homes damaged, more than 260 mm in 48 hours in parts of the north, 4 deaths;
• Storm Alexa's snow in December 2013: 7,128 Palestinian Civil Defense operations, communities cut off for days, at least 300 people displaced in Bedouin and herding communities.

Since October 2023, UN reporting on the West Bank has centred on casualties, settler attacks, demolitions and displacement, and the UN has quantified one weather event in three winters. Most other figures come from the Palestinian Civil Defense via the media. That gap shows in the timeline below. Storm Byron gave the West Bank its wettest day in IMERG's 28-year record, about 97 mm averaged over the whole territory on 11 December 2025, yet no West Bank impact from it was found in public reporting. Whether that reflects little damage or little counting cannot be told from what is published.

The impacts fall on three groups, set out by theme in section 5:

• refugee camps and towns, where drainage is blocked by the Barrier or destroyed (water, sewage and stormwater networks in the northern camps were bulldozed during Israeli forces' operations), and the hill towns when it snows;
• Bedouin and herding communities in Area C living in tents and makeshift structures, who have lost shelters, animal pens and fodder in the storms that have been reported;
• farmers, whose greenhouses in the north and the Jordan Valley take the wind, frost and flood damage, and whose rainfed crops and pasture depend on the season's rain (section 6).

Deaths come from flash floods in wadis and from heating fires. No hypothermia death was found for the West Bank in any winter.

How often the rain that has caused these impacts comes: over 1998–2025, IMERG puts an average of 11.4 days of ≥ 10 mm, 4.5 of ≥ 20 mm, 2.4 of ≥ 30 mm and 0.4 of ≥ 50 mm over the West Bank in each October–April.

Hazard: R rain and flooding, F flash flood, S snow, C cold, W wind, X structural collapse.

Notes and method

Weather columns cover the window from the day before the reported start to the reported end: IMERG wettest day averaged over the whole West Bank (a downpour on one hillside is diluted over 5,600 km²); window totals from IMERG and from ERA5; ERA5 lowest daily minimum temperature and highest hourly 10 m wind averaged over the 16 cells (the hill towns are colder and windier). Figures in italics come from Palestinian Authority bodies (the Palestinian Civil Defense, ministries) or the media; the rest are UN, cluster or Red Crescent/IFRC figures.

Before October 2023

Before October 2023: the two large events, and the recurring ones (snow on the hills, heating fires, flash floods). Several figures are counts of rescues or assistance rather than of people affected.

Counts by winter and ENSO phase

The counts do not separate by ENSO phase. The two counted El Niño winters (48 and 2,800 people), the three counted neutral winters (300, 640 and 5,200 people) and the one counted La Niña winter (300 people) overlap, and six counted winters are too few for any test to separate them. The counts are also of different kinds, so no average per phase is given.

5. What this winter could bring

In this part: How many storms a winter brings · When the first storm of 20 mm or more comes · How big a week of rain is · What one storm of each kind did last winter · Last winter in numbers · Exposure then and now · Deaths · Rain and reported impact, winter by winter · Storm for storm, before and since October 2023 · What the October forecast implies for October–December · Scenarios for winter 2026/27

The question for planners is what this winter could do. The record cannot answer that with a forecast: there is one well-documented winter, reporting that changed every year, and a population whose exposure keeps changing. What it can do is bound the answer. This section separates the weather (how many storms, how big, how early, and what El Niño does or does not change) from what each kind of storm did last winter, and then uses last winter as two reference points. It ends with the longer record: every winter since 2008/09 on one row, the same sizes of storm before and since October 2023, and three scenarios for this winter scaled from last winter's counts.

How many storms a winter brings

Over the 28 winters of IMERG (1998/99–2025/26), a Gaza winter brought on average 6 storms whose wettest day reached 10 mm (in most winters 4–8), 3 that reached 20 mm (1–4), and a storm of 50 mm or more, like Byron, in 50 per cent of winters. El Niño winters brought slightly more of the smaller storms (7 of 10 mm or more, against 5 in La Niña winters) but not more of the largest: 4 of 8 El Niño winters had a storm of 50 mm or more, against 6 of 14 La Niña winters. Last winter, with 7, 4 and 1, was among the more active.

El Niño's extra storms come early in the season. In ERA5 since 1979/80, El Niño winters have brought 1–4 storms with a total of 20 mm or more before January (1.9 on average), against 0–4 in neutral winters (1.2) and 0–3 in La Niña winters (0.8): about twice as many as in the other winters. From January to April the phases do not differ (2.9 against 2.4). Every one of the 14 El Niño winters had at least one before January; 13 of the other 33 had none.

How far to trust it. The window and the threshold were chosen after about a hundred combinations were compared, so the p < 0.01 of this one overstates the evidence. The gap is wider before 1998 (2.7 against 1.0) than since (1.4 against 0.9, within chance on its own, p = 0.10). IMERG's 28 winters, at the matched size of about 35 mm, show no gap (1.1 against 1.0). These are past ranges, not a forecast.

ERA5 averages rain over 25 km cells: 20 mm there is about 35 mm in IMERG's terms.

Notes and method

At the other totals tried from 10 to 50 mm the full-record gap has p from 0.001 to 0.031. IMERG at 20 mm, a smaller class of storm: 2.5 against 1.6 (p = 0.09). Before 1979/80 there was no gap (1.3 against 1.3 from 1950/51), like the rest of the El Niño link (section 2).

Over the whole winter the difference is proportionally smaller. El Niño winters have brought 2–7 storms of 25 mm or more (4.6 on average, IMERG), against 2–6 in neutral winters (4.5) and 1–5 in La Niña winters (3.3). On IMERG's eight El Niño winters the difference from the other 20 is within chance (p = 0.17); in ERA5 since 1979/80, with 14 El Niño winters and thresholds matched to IMERG's, the gap is 1.6 storms a winter and is unlikely to be chance (p < 0.01). The bigger storms do not follow the phase: 50 mm or more came 0–2 times in an El Niño winter and 0–3 in the others, and a storm of 100 mm or more came in 3 of 8 El Niño winters and 6 of 20 others. These are past ranges, not a forecast.

Fewest to most storms per winter, IMERG 1998/99–2025/26.

Notes and method

A storm is a run of rainy days as defined above, here sized by its total rain over Gaza. The last column tests El Niño winters against all others (Mann–Whitney): first on IMERG, then on ERA5 1979/80–2025/26 with thresholds that give the same number of storms as IMERG's over the shared years (17, 39, 61 mm); that equalizes the counts, not the storms themselves. ERA5's winters include IMERG's, so the two tests are not independent.

When the first storm of 20 mm or more comes

This is where El Niño may matter, though not by much more than chance on eight winters. In 5 of the 8 El Niño winters of the IMERG record, the first storm of 20 mm or more came before 1 December, against 12 of 28 winters overall, and the forecasts put this year's wet anomaly in October–December. Damage does not wait for a big storm, though. In each of the last three winters the first weather damage was reported in mid to late November (13–14 November 2023, 24–25 November 2024, 14–15 November 2025), and in two of them that came before the winter's first 20 mm storm. Last winter's first storm, with 22 mm, damaged about 13,000 tents.

How big a week of rain is

A 7-day rainfall forecast can be read against the year: in a typical winter the wettest week over Gaza brings about 25 per cent of an average year's rain (IMERG), and one winter in five brings a week of 37 per cent or more (11 of the 28 winters since 1998/99 had a week of 30 per cent or more). The table is a scale for reading a forecast, not a damage predictor.

More detail

IMERG and the Beer Sheva gauge differ in their totals but agree closely on the share; Beer Sheva is 46 km inland and drier than Gaza, so compare its shares, not its millimetres. Byron's week came close to half an average year.

Highest 7-day running total in each season (August–July), divided by the same source's mean annual total: IMERG 1998/99–2025/26 (area-weighted over Gaza), Beer Sheva gauge 1922/23–2014/15 (seasons with at least 330 days reported).

Notes and method

These are observed scales: a forecast system's 7-day totals carry its own biases, so a forecast compares best as a share of that system's own climatological annual total. Totals for a week ahead are much less certain than for the next two or three days, and a 7-day total says nothing about how much fell in an hour, which is what drives flash floods.

What one storm of each kind did last winter

Each row is one real event, mostly from 2025/26, with how its numbers were counted. None is an average: every row is a single case, and counts made at the sites assessed are floors.

Last winter in numbers

Last winter's observed totals, split into Byron and everything else. About half of winters bring a storm with a day of 50 mm or more, as Byron had; that held across IMERG's 28 winters, and El Niño did not change it. So the split shows what last winter's counts looked like with and without the one storm that dominated them.

All of these numbers understate last winter. They come from alerts raised by the sites that were reached. In June 2026 the Shelter Cluster's household survey found 71 per cent of households had been flooded at least once; applied to the roughly 355,000 households in displacement sites, that is about 250,000 households. So the alert-based totals probably caught a third or less of the households flooded. Exposure is also higher this winter. About 1.7 million people live in roughly 1,600 sites, against about 1.5 million in about 940 sites last December (counted differently). 65 per cent of households are in tents or makeshift shelters, and more than half of those tents are in poor or very poor condition after a year's use. Last winter was a weak La Niña, not an El Niño. None of this is a forecast.

Household-impacts count a household once for every storm that affected it; the same households were hit repeatedly.

Notes and method

The lower figure sums the per-event counts; the higher one uses the Shelter Cluster's monthly snapshots (December 65,000 families, January 17,221) with the first storm and the February–March events. “All other events” takes Byron's own counts (55,000 households, 42,000 tents) out of both; for December that subtracts a Site Management Cluster count (132 sites) from a Shelter Cluster snapshot (537 sites), two sources with different coverage, so the remainder is approximate. Tents are the sum of the UN and cluster counts by event. Deaths: 35 is the Ministry of Health's storm-related total to 14 January 2026 (relayed by OCHA) plus the child cold deaths it reported afterwards (11 in all); the higher figure uses its other count of 25 collapse deaths for the same period. Neither is UN-verified.

Exposure then and now

Site Management Cluster, Shelter Cluster and IOM figures via OCHA (SitRep 40, December 2025; HSU #353, January 2026; Humanitarian Situation Report, 25 September 2026); Shelter Cluster household survey, June 2026. The 2025 and 2026 site counts use different methods.

Deaths

No range is given for deaths. Cold deaths followed shelter more than temperature: last winter had only 2 nights at or below 8 °C in ERA5 over Gaza and 11 child deaths; the winter before had 8 such nights. Collapse deaths came with Byron, when rain soaked buildings already damaged in the hostilities. Almost every death count comes from the Ministry of Health in Gaza, and OCHA does not systematically monitor deaths unrelated to the hostilities.

Rain and reported impact, winter by winter

Every winter since 2008/09 on one row. The impact columns are not one series: before October 2023 they count people evacuated, displaced or with damaged homes, in the few storms that OCHA, UNRWA or UNICEF quantified in their reports; since, people whose tents and shelters flooded or were damaged, in counts published by OCHA, mostly from the Site Management Cluster and the Shelter Cluster. They are set side by side for scale only. Each count and its source is listed in the table of events under the next heading.

People are as the source gives them, or households multiplied by 5.6 before October 2023 and 4.8 since, shown as “about” or rounded.

Notes and method

Rain: IMERG, area-weighted over Gaza; the wettest 7 days is the highest 7-day running total of the season; storms as defined above, sized by their wettest day. Household sizes: the average household in Gaza (PCBS census 2017) and people per household in displacement sites (September 2026). A person is counted once for each storm that affected them. 2025/26: the lower figure sums the per-event counts and the higher uses the Shelter Cluster's monthly snapshots, as in “Last winter in numbers”. Tents and shelters: UN and cluster counts. Figures in italics come from the Government Media Office and Palestinian Civil Defense (tents) or the Ministry of Health in Gaza (deaths), relayed by OCHA. Deaths before October 2023 are those UN or Red Cross and Red Crescent reports attribute to the storms. “Not counted” means an impact was reported without a number; “none found” means no weather impact was found in UN reporting for that winter.

Before October 2023 the counts followed the winter's biggest storm. Across the six counted winters, the wettest day (rank correlation +0.94) and wettest week (+0.89) rank them best, ahead of the season total (+0.54); rain days run the other way (-0.52), and days of 10 mm or more, the number of storms of 20 mm or more, cold nights and windy days show no relation. Six winters is a small sample. That rests on two floods, Wadi Gaza in January 2010 and Storm Alexa in December 2013, and what El Niño adds is rain days, not bigger storms (section 3). Since October 2023 the two counted winters sit far above every earlier one on every measure, and the driest winter of the 28-winter record, 2024/25, has more people reported affected (at least 37,000) than the wettest, 2013/14 (21,000).

Storm for storm, before and since October 2023

How much more does the same weather do now? The record supports a comparison by size of storm, not a rate per millimetre: two counted winters cannot fix a slope, and events with almost no rain were counted in the thousands.

Storms over Gaza in IMERG, as defined above.

Notes and method

“With a reported impact”: any entry in the event catalogue, with or without a count, overlaps the storm (from the day before the entry starts to the day it ends). “People counted”: the range across events with a dated count, sized by the wettest day in the event's own dates, with the number of counts in brackets. Counts of assistance delivered, with no rain date, are left out. The event of 12–16 January 2026 (19,200 people) is not in the size tiers above: IMERG has 1.5 mm on its wettest day and ERA5 12 mm, so IMERG missed most of its rain. The one count of 10–20 mm before October 2023 is from January 2020; its date is approximate and its wettest day, at 20 mm, sits on the boundary.

“About”: the source gives households or families, converted at 5.6 people before October 2023 and 4.8 since.

Notes and method

Rain: IMERG and ERA5 over Gaza, from the day before the event to its last day; the two disagree on single storms. Wind and cold are not plotted: ERA5's area-mean wind and night minimum do not resolve the gusts and the cold in tents that the reports describe; the hazard column is from the reports. Events reported without a number, and counts of assistance delivered with no rain date, are in the tables of section 4 but not here.

Two things changed. More storms do harm: 7 of the 35 storms of 20–50 mm left an entry in the event catalogue before October 2023, and all 5 since. Three events with less than 10 mm of rain (high tides and wind), with no counterpart before October 2023, were counted at 3,800–34,000 people each. And each does more: the counts are tens to hundreds of times higher, about 18 times on medians (6,200 people against 350) and, as a scale check allowing for each storm's rain, about 120 times (95 per cent interval about 11 to 1,300, on 16 events). Byron (about 150 mm in IMERG; at least 235,000 people) had 11 times the count of Storm Alexa (about 190 mm; homes of 21,000 people damaged, about 10,000 evacuated), with less rain.

More detail

The scale check is a model fitted to the 16 events with a dated count: one slope on the storm's rain total, plus a step in October 2023. Without the largest storm of each period the step is about 46 times; on geometric means, ignoring rain, about 34. It treats a high-tide event with 3 mm of rain as a very small rainstorm, and takes the storm of 12–16 January 2026 at the 3 mm that IMERG recorded. Before October 2023 only the largest impacts were counted at all (seven counts in 15 winters), which pulls the ratio down; closer counting since pulls it up. Since October 2023 rain alone does not order the counts: across the nine dated counts, people affected and rain are barely related (rank correlation +0.28 with the wettest day), though among the five with a day of 10 mm or more the bigger storms were counted higher (+0.80).

What the multiplier mixes. It is a ratio of reported counts, and three changes are folded into it: how people are sheltered (tents and damaged buildings, many on flood-prone ground, where before they had houses), which is the real change in vulnerability; what is counted (people evacuated or with damaged homes then, people whose tents flooded now); and how closely (occasional situation reports then, daily site alerts last winter). The record cannot separate them, which is why section 4 warns that counts cannot be compared across winters like for like. Read “tens to hundreds of times” as the scale of the change, not a measurement of it.

What the October forecast implies for October–December

SEAS5's October 2026 forecast for October–December is wetter than 41 of its 45 past October forecasts once the trend is taken out (section 1). In the nine years with a forecast in that wettest fifth, ERA5 rain over Gaza in the quarter was 82–202 mm: four were close to the normal 82 mm and five had 119 mm or more. Last year's quarter, with Byron, had 130 mm after an October forecast in the driest fifth. ERA5 flattens the biggest storms: it has the quarters of Storm Alexa and Byron at 88 and 130 mm, where IMERG has 322 and 218 mm. Read where a winter sits, not its millimetres.

Scenarios for winter 2026/27

These are planning scenarios, not a forecast. Two counted winters cannot support a fitted model. The cases take what last winter's weather did, as it was counted, and scale it to this September's number of people in displacement sites: hazard, times what that hazard did last winter, times exposure now against then. They differ by what dominated last winter's counts, a storm with a day of 50 mm or more, which El Niño has not made more likely and no forecast sees more than days ahead.

People-impacts count a person once for each storm that affected them.

Notes and method

Each range runs from the sum of the per-event counts (people as reported where a source gives them, 4.8 per household otherwise) to the Shelter Cluster's monthly snapshots (households at 4.8 people), as in “Last winter in numbers”. Byron is taken out or added at the Shelter Cluster's 235,000 people in the lower figures and at the Site Management Cluster's 55,000 households (about 264,000 people) in the higher ones. “How often” counts IMERG winters 1998/99–2025/26. The only winter with two such storms, 2018/19, was before October 2023 and left no entry in the event catalogue. The no-storm case is last winter with Byron taken out: 3 storms of 20 mm or more remain (3 in an average El Niño winter). Exposure: about 1.7 million people in displacement sites in September 2026 against about 1.5 million in early December 2025; the two counts used different methods.

• One winter, one storm. Every figure is last winter's count rescaled, and the two larger cases rest on Byron alone.
• Alert counts are floors. The Shelter Cluster's household survey found 71 per cent of households flooded at least once last winter (June 2026; households in all types of shelter in accessible areas; 83 per cent of those in tents). At the same share, about 1.2 million of the 1.7 million people in displacement sites would be flooded at least once this winter, each counted once. Last winter the alerts caught a third or less of the households the survey found.
• The scaling is crude. It assumes impact rises in step with the number of people in sites. Older tents (54 per cent in poor or very poor condition) would raise it; site preparation, drainage and relocation would lower it. Neither is in the figures.
• A dry winter is not a safe one. 2024/25, the driest winter of the record (115 mm), still had at least 37,000 people reported affected, by high tides and wind. No El Niño winter since 1998/99 has been that dry, but sea surge, wind and cold come whatever the rain.
• Deaths are not scaled. They came from collapsing buildings and cold nights, which do not follow the number of people in sites.

In this part: How many storms a winter brings · When the first storm of 20 mm or more comes · How big a week of rain is · Snow · Who is exposed, by theme · Deaths · Plans

There is not enough counted impact data for the West Bank to say how many households a winter will affect. What the record can show is how often the weather that has caused harm comes, whether El Niño changes that, how big a week of rain is against the year, and who is exposed now.

How many storms a winter brings

Over the 28 winters of IMERG (1998/99–2025/26), a West Bank winter brought on average 6 storms whose wettest day reached 10 mm (in most winters 4–8), 3 that reached 20 mm (2–5), and a storm of 50 mm or more in 29 per cent of winters. El Niño winters brought slightly more of the smaller storms (7 of 10 mm or more, against 6 in La Niña winters) but not more of the largest: 1 of 8 El Niño winters had a storm of 50 mm or more, against 5 of 14 La Niña winters. Last winter, with 8, 7 and 2, was among the more active.

El Niño's extra storms come early in the season. In ERA5 since 1979/80, El Niño winters have brought 2–4 storms with a total of 20 mm or more before January (2.8 on average), against 0–5 in neutral winters (2.0) and 0–3 in La Niña winters (1.4): 1.7 times as many as in the other winters. From January to April the phases do not differ (3.6 against 3.2). Every one of the 14 El Niño winters had at least two before January; 15 of the other 33 had fewer.

How far to trust it. The window and the threshold were chosen after about a hundred combinations were compared, so the p < 0.01 of this one overstates the evidence. The gap is wider before 1998 (3.2 against 1.8) than since (2.5 against 1.6, within chance on its own, p = 0.06). IMERG's 28 winters, at the matched size of about 30 mm, point the same way (2.4 against 1.6, p = 0.03). These are past ranges, not a forecast.

ERA5 averages rain over 25 km cells: 20 mm there is about 30 mm in IMERG's terms.

Notes and method

At the other totals tried from 10 to 50 mm the full-record gap has p from 0.003 to 0.007. IMERG at 20 mm, a smaller class of storm: 3.2 against 2.1 (p = 0.02). Before 1979/80 there was no gap (1.7 against 1.7 from 1950/51), like the rest of the El Niño link (section 2).

Over the whole winter the difference is proportionally smaller. El Niño winters have brought 3–6 storms of 25 mm or more (5.5 on average, IMERG), against 2–6 in neutral winters (4.2) and 2–8 in La Niña winters (4.6). On IMERG's eight El Niño winters the difference from the other 20 is suggestive, with three sizes tested (p = 0.03); in ERA5 since 1979/80, with 14 El Niño winters and thresholds matched to IMERG's, the gap is 1.4 storms a winter and is suggestive, with three sizes tested (p = 0.04). The bigger storms do not follow the phase: 50 mm or more came 1–4 times in an El Niño winter and 1–6 in the others, and a storm of 100 mm or more came in 5 of 8 El Niño winters and 10 of 20 others. These are past ranges, not a forecast.

Fewest to most storms per winter, IMERG 1998/99–2025/26.

Notes and method

A storm is a run of rainy days as defined above, here sized by its total rain over the West Bank. The last column tests El Niño winters against all others (Mann–Whitney): first on IMERG, then on ERA5 1979/80–2025/26 with thresholds that give the same number of storms as IMERG's over the shared years (21, 37, 71 mm); that equalizes the counts, not the storms themselves. ERA5's winters include IMERG's, so the two tests are not independent.

When the first storm of 20 mm or more comes

In 7 of the 8 El Niño winters of the IMERG record, the first storm of 20 mm or more over the West Bank came before 1 December, against 14 of 28 winters overall. Against the other winters that is unlikely to be chance alone (p ≈ 0.03), though eight winters is a small sample. Last winter's most damaging West Bank event, on 28–30 December 2025, came in a December that gave Jerusalem more than 237 mm, its wettest December since 2013 according to the Palestinian Meteorological Department (via Al Jazeera).

How big a week of rain is

A 7-day rainfall forecast can be read against the year: in a typical winter the wettest week over the West Bank brings about 19 per cent of an average year's rain (IMERG), and one winter in five brings a week of 30 per cent or more. The table is a scale for reading a forecast, not a damage predictor.

More detail

The Jerusalem gauge, a single hill-top point whose last complete season is 2013/14, runs a few points higher on every rung than the IMERG area average: compare the IMERG column with an area forecast and the gauge column with a point. Byron's week, half an average year and the wettest week in IMERG's record, had no impact attributed to it in public reporting; the second-wettest, 7–13 January 2013 (166 mm, about 40 per cent of a year), brought the largest flood event in this page's record, with more than 260 mm in 48 hours locally in the north.

Highest 7-day running total in each season (August–July), divided by the same source's mean annual total: IMERG 1998/99–2025/26 (area-weighted over the West Bank), Jerusalem gauge 1909/10–2013/14 (seasons with at least 330 days reported).

Notes and method

These are observed scales: a forecast system's 7-day totals carry its own biases, so a forecast compares best as a share of that system's own climatological annual total. Totals for a week ahead are much less certain than for the next two or three days, and a 7-day total says nothing about how much fell in an hour, which is what drives flash floods.

Snow

Snow with impacts has come in every ENSO phase; over the last 15 years, more often in neutral and La Niña winters than in El Niño ones. The Israel Meteorological Service finds a slight, non-significant tilt towards more Jerusalem snow days in strong El Niño winters, from seven cases. With every forecast pointing to a warm winter, a snowstorm like Alexa is less likely than usual this year but not ruled out. The ERA5 cells are too smooth to see the hill-town snow, so this page does not estimate snow odds.

Who is exposed, by theme

Three groups carry most of the West Bank's winter risk, and each meets the weather in a different way: camps and towns through drainage, power and roads; herding communities through shelters, animals and fodder; farmers through their crops. Since 2023, displacement and losses in all three have been driven mainly by Israeli forces' operations, settler violence and access restrictions (OCHA); the figures below keep those causes apart from the weather.

Refugee camps and towns

Going into this winter:

• About 34,000 Palestine refugees remain forcibly displaced from the Jenin, Tulkarm and Nur Shams camps since the Israeli forces' operation in the northern West Bank that began in January 2025 (UNRWA, 10 September 2026). None have returned; Israeli military closure orders ran to 20 August 2026 for Jenin camp and 30 September 2026 for Tulkarm and Nur Shams, and their status since was not found. Over 80 per cent of those displaced now rent (UNRWA, September 2026), and only 8 per cent could pay rent regularly in mid-2025 (Shelter Cluster); a May 2025 assessment had found just over half renting, 30 per cent staying with relatives and 20 per cent in public shelters or other temporary places (OCHA). Most of the about 4,000 displaced from El Far'a refugee camp returned in February 2025.
• Satellite imagery from January–February 2026 shows 637 structures destroyed and 1,521 damaged or destroyed in total across the three camps: about half of all structures in Jenin and Nur Shams camps and 37 per cent in Tulkarm camp (UNOSAT, preliminary). Water, sewage and stormwater infrastructure in the camps was bulldozed during the operations (OCHA).
• Elsewhere, towns partly enclosed by the Barrier drain badly in heavy rain. In the WASH Cluster's 2022 community inventory, 167 of 845 communities reported flooding at least once in the previous three years (10,173 households, most of them in Jerusalem governorate).

No numbers on camp flooding, power cuts or service interruptions from weather were found for any winter since 2023.

Notes and method

Those displaced from the northern West Bank camps now mostly rent, which suggests their winter risks lie more in heating costs and the condition of rented housing than in flooding (in January 2026 about 100 Tulkarm families were living in buildings without doors or windows, Al Jazeera).

Herding communities in the Jordan Valley and Area C

Going into this winter:

• The Shelter Cluster's Area C dataset covers 315 communities (215,703 people) with 30,173 residential shelters, of which 2,242 need rehabilitation (January 2026). Bedouin and herding communities are a subset; 42 of the listed communities are marked as forcibly displaced and no longer existing. No count of people in tents was found.
• Displacement by Israeli settler attacks and related access restrictions, separate from the weather: 130 communities fully or partly displaced and more than 6,400 people since January 2023, over 2,500 of them in 2026, mostly in Bedouin and herding communities in Area C (OCHA, 4 September 2026). Ras Ein al 'Auja near Jericho, about 600 people, was fully displaced in January 2026.
• After the 2024/25 drought (33 per cent of the mean rainfall at Jericho by early March 2025), 16 per cent of livestock-keeping households in Area C reported animal deaths from malnutrition, drought or lack of feed, and 81–90 per cent of livestock keepers cited the cost of feed as a problem (FAO, survey of July–August 2025). The survey does not separate drought from pasture lost to settler violence: 65 per cent of Area C livestock keepers reported difficulty reaching pasture.

No count of livestock killed by cold, floods or snow was found for any winter since 2015.

Notes and method

Lambing falls in mid-winter; OCHA flagged the risk to lambs during Alexa, but no lamb losses were measured. No herding community is known to have been displaced by weather alone beyond a few days; long-term displacement has been driven by settler attacks, demolitions and access restrictions.

Farming: field crops, greenhouses, trees and rangeland

• Field crops as a whole (about 198,000 dunums) are 95 per cent rainfed and tree crops (about 641,000 dunums) 91 per cent rainfed; Jericho and Al Aghwar is the exception, where about 96 per cent of the tree area is irrigated (the valley grows dates, citrus and bananas). Section 6 shows how these follow the winter's rain and El Niño.
• Greenhouses: 28,743 dunums of plastic houses, 77 per cent of them in Jenin, Tulkarm and Tubas and another 8 per cent in Jericho; vegetables as a whole are only 13 per cent rainfed. Their winter hazards are wind, snow load, frost and flooding, not the season's rainfall.
• Rangeland: about 2 million dunums, of which Palestinians had access to about 621,000 in 2014 (PA figures); settler violence and displacement have restricted it further since (no updated figure).

PCBS and MoA Agricultural Census 2021 (areas, irrigation, greenhouses); PA data in the Alexa damage assessment (rangeland). No storm damage to greenhouses or orchards was found reported for 2025/26.

In the three event tables, italics mark impact figures from Palestinian Authority bodies (ministries, the Palestinian Civil Defense) or the media, including when the UN relays them.

Deaths

Attribution is mixed (UN, IFRC, Civil Defense via the media), so the rows should not be summed.

Plans

No 2026/27 winterization plan written for the West Bank was found by 2 October 2026. The West Bank's winter lines sit inside the calendar-2026 appeals:

• the Flash Appeal's West Bank shelter response: 62,900 people, US$28.4 million, including weatherproofing, insulation, heaters and blankets;
• UNRWA's West Bank shelter and household items: 11,500 households, US$21.9 million.

UNICEF's Winterization Plan 2026–2027 (17 September 2026) seeks US$97.5 million for 540,000 children in Gaza and the West Bank; its costed lines are almost all for Gaza. A US$1.0 million oPt Humanitarian Fund allocation for West Bank shelter gaps, in Area C and for displaced people, was approved in January 2026. Funding is reported only for the Occupied Palestinian Territory as a whole.

6. Farming in the West Bank: crop years after El Niño winters

In this section: Spring greenness, zone by zone · Field crops, trees and rangeland · Harvests after El Niño winters · Rainfed cereals and pulses · Olives · Spring vegetation since 1982 · Every strong El Niño crop year since 1983 · What the record suggests for 2026/27

Most West Bank farming depends on the winter rain. The West Bank is commonly divided into four agro-ecological zones (Dudeen 2001): the Central Highlands from Jenin to Hebron, almost all rainfed and mostly under olives, grapes and other fruit trees, with winter cereals and pulses on the rest; the semi-coastal plain in the northwest, the wettest zone, with rainfed field crops and irrigated vegetables; the Eastern Slopes, a dry transition zone of marginal rainfed barley, wheat and olives and spring grazing; and the Jordan Valley, which gets 100–200 mm and is farmed on springs and groundwater. Crop statistics are published by governorate, and the Eastern Slopes cut across the highland governorates, so the yield figures below use three zones and the Eastern Slopes appear through satellite vegetation data.

One convention throughout this section: a crop year is named after its harvest and follows the winter before it. This winter, 2026/27, will shape the cereal and pulse harvest of May–June 2027, spring grazing in 2027 and the olive harvest of October–November 2027. The olive harvest that starts this month was set by last winter and by the size of the previous crop.

Wheat and barley area: PCBS Agricultural Census 2021, by governorate.

Notes and method

The Jordan Valley's share is almost all Tubas governorate (which PCBS reports with the northern valleys), much of it on the governorate's hills and plains above the valley floor. Olives pressed: PCBS Olive Presses Survey, share of all olives pressed in the Occupied Palestinian Territory, 2003–2019 mean (the West Bank as a whole pressed 89%). PCBS merges Jenin and Tubas, so Tubas's olives sit in the semi-coastal row; Jericho has no presses; Jerusalem is reported only from 2008 and left out (about 3% of the highland pressings since). Rain: October–April ERA5 over each zone's governorates, 1979/80–2025/26, against December–February Niño3.4 (pinned series). Spring greenness: MODIS NDVI, February–April, 2001–2026, against the 2001–2020 normal (see below). The zones share ERA5 cells and their winters move together (r ≥ 0.89 between any two), so ERA5 cannot tell their El Niño responses apart: on it they are the same. Its 25 km cells also smooth the steep rain gradient (its Jordan Valley mean is about 300 mm, against 100–200 mm on the valley floor), so only anomalies are used.

Spring greenness, zone by zone

The one yearly measure available for every zone is satellite greenness. In the spring after an El Niño winter, it was above normal in 6 of 8 years in the semi-coastal plain, 8 of 8 years in the Central Highlands, 6 of 8 years in the Jordan Valley; after La Niña winters, in 5 of 12, 4 of 12, 3 of 12. Several of those El Niño springs were only slightly above normal.

More detail

Most El Niño winters of the MODIS years brought close to normal rain (percentage above or below the 1991–2020 mean across the West Bank: 2002/03 +38, 2004/05 −2, 2006/07 −5, 2009/10 −10, 2014/15 +5, 2015/16 +4, 2018/19 +44, 2023/24 +2). The drier the zone, the larger the swing: the median El Niño spring was 3 per cent above normal in the semi-coastal plain, 8 per cent above normal in the Central Highlands, 12 per cent above normal in the Jordan Valley, though the drier zones also vary more in every year (standard deviation 5 per cent, 9 per cent, 13 per cent in the same order). The semi-coastal plain, the wettest zone and partly irrigated, has the weakest link to El Niño: rain there is less often what limits growth.

Field crops, trees and rangeland

Split by land cover instead of by zone, the same satellite record separates the crops that ride out a dry year from those that do not. Tree cover (denser orchards, groves and woodland) has the weakest link to the winter's rain and, in the hills and on the plain, the smallest swings; its El Niño springs were above normal in 8 of 8 years, but only slightly. Cropland and the shrubland and grassland that cover most of the West Bank (rangeland and sparse olive groves) swing more and follow El Niño about equally.

More detail

On the semi-coastal plain no group has a significant link to El Niño, and its cropland, much of it irrigated, has none at all. In the Jordan Valley, every group swings widely, and the valley's cropland and rangeland had the largest El Niño and La Niña departures.

Spring (February–April) MODIS NDVI as above, averaged over the 250 m pixels that are at least 60% one land-cover group in ESA WorldCover 2021 (10 m): cropland (class 40), tree cover (10), shrubland and grassland (20, 30).

Notes and method

Built-up and bare land are left out. Shares of the West Bank are of all land, including built-up and bare. r with rain: against the October–April ERA5 total over the whole West Bank. “Year-to-year spread” is the standard deviation of the spring anomaly, 2001–2026. The 2021 map is applied to every year. WorldCover has no orchard class, and how an olive grove is classed depends on how dense its canopy is, so the shrubland and grassland group mixes rangeland with sparse groves, and tree cover is the denser orchards, groves and woodland. Requiring 60% of a pixel favours large blocks: the cropland rows describe the plains and valley fields more than the fragmented terraces. NDVI is used without quality or snow masking, so February composites in snowy springs (2015, 2021, 2022) may read low; 2023 and 2026 miss one and two composites. Greenhouses cannot be told apart in these maps.

Harvests after El Niño winters

Cereal yields are published only for the Occupied Palestinian Territory as a whole (the West Bank and Gaza together), but the West Bank holds 90 per cent of its wheat and barley area and presses 89 per cent of its olives. Rainfed wheat and barley yields were above trend in 7 of 9 crop years after El Niño winters, against 3 of 11 after La Niña winters. The four worst harvests since 1994, 1999 (60 per cent below trend), 2008 (35 per cent), 2011 (20 per cent), 2001 (17 per cent), all followed La Niña winters. El Niño years were not bumper years: their median, 7 per cent above trend, is close to that of neutral years (5 per cent above trend). What an El Niño winter has done is make a bad cereal year less likely.

More detail

The longer vegetation record, which also covers pasture, tells the same story: spring vegetation was above its median in 10 of 14 El Niño years on cropland and 12 of 14 on all land, against 9 of 20 and 4 of 20 after La Niña winters. Olives show no clear El Niño signal (3 of 6 El Niño years above what the previous crop predicts, 5 of 10 La Niña years; r with Niño3.4 +0.14).

These are small samples. With 9 and 11 years, the true share of good cereal years could plausibly lie anywhere from 45 per cent to 94 per cent after El Niño winters and from 10 per cent to 57 per cent after La Niña winters (95 per cent intervals), and the El Niño–La Niña contrast is only borderline significant (Fisher exact test p = 0.07; Mann–Whitney p = 0.08). The clearest part is the bad tail: no El Niño year fell more than 15 per cent below trend, against 4 of 11 La Niña years. The correlation with the season's rain also leans on one year: without 1999 it drops from +0.41 to +0.19.

“r with rain”: against the October–April ERA5 total over the West Bank before the harvest; “r with Niño3.4”: against December–February Niño3.4 of that winter.

Notes and method

Medians are percentages above or below trend or normal, or VCI points above or below the record's median. All of these years fall after 1979, inside the period in which the El Niño link to West Bank rain exists.

Rainfed cereals and pulses

Wheat and barley are sown in November–December and harvested in May–June. In the West Bank, 98 per cent of their area is rainfed (2021 census); about a third of it is in Hebron governorate, on the southern hills and the edge of the eastern slopes, and more than a quarter in Jenin. Yields follow the winter's rain, loosely, and a wet winter is no guarantee.

More detail

The two worst harvests followed the two driest winters with official figures (1998/99 and 2007/08), but two other dry winters (2013/14 and 2016/17) have no official yield, and two poor harvests (2001, 2011) followed winters that were only somewhat dry or near normal. The total is not everything: when the rain comes, frost, and heat during grain filling also matter, so a wet winter is no guarantee (2019, after a very wet El Niño winter, was slightly below trend) and a dry one can still yield well (2000).

Lentils, the pulse with the most official figures (19 years), follow the winter's rain more closely than the cereals (r = +0.80, with Niño3.4 +0.53; 5 of 6 El Niño years above trend, 3 of 9 La Niña years). Chickpeas show no clear link (r = +0.28 with rain).

Which part of the winter matters (four windows tried, so read loosely): yields correlate with rain in Oct–Dec r = +0.44, Nov–Jan r = +0.65, Jan–Mar r = +0.18, Feb–Apr r = +0.09.

Notes and method

The early part of the season, around sowing and establishment, carries the link, and that is the part the forecasts favour as wet.

Olives

Olives are the West Bank's main tree crop and almost all rainfed. The crop swings by a factor of two to five from one year to the next because trees alternate heavy and light crops (alternate bearing), and that cycle, not the winter, sets most of the change. With it taken out, national production shows no clear link to El Niño (official years only). By zone (olives pressed, PCBS, 2003–2019), bigger crops lean weakly towards wetter winters, more in the semi-coastal plain than in the highlands, but the lean rests largely on 2019, a very wet winter followed by the largest crop of the period, and it does not carry through to El Niño: in these years El Niño winters were not reliably wet (2009/10 was below average).

Since October 2023, access has shaped the harvest as much as the weather. Between 1 and 27 October 2025 OCHA documented 126 olive-harvest-related attacks by Israeli settlers, against 110 in the same period of 2024 and 30–46 in the same window of 2020–2023, with more than 4,000 trees and saplings vandalized. The 2025 harvest, after a winter (2024/25) tied with 1998/99 as the driest in ERA5's record, was among the smallest in years: early estimates of 7,000–8,000 tonnes of oil, against 27,300 tonnes in 2024 and an average of about 22,500 (MAS, as reported by Al Jazeera). 2024 was a large crop, so alternate bearing alone would have made 2025 smaller; bearing, drought and access cannot be separated in these figures.

Spring vegetation since 1982

The NOAA record starts in 1982 and so covers all six strong El Niño winters since then. The springs after 1982/83 and 1997/98 were among the greenest on record, the one after 2015/16 close to the middle, the one after 2023/24 somewhat above it, and the spring after 2009/10, a below-average winter, well below it. The wettest winter in ERA5's record, 1991/92, left cropland below the middle: its December–February was also one of the two coldest in that record. All land, which includes the rangeland of the eastern slopes, responds more to El Niño than cropland does, as the zone figures above show for the drier zones.

Every strong El Niño crop year since 1983

The six winters with December–February Niño3.4 of +1.5 °C or more since 1982, the closest analogues for 2026/27.

Niño3.4 ≥ +1.5 °C in December–February.

Notes and method

Rain as a percentage above or below the 1991–2020 mean (ERA5, West Bank); yields and olives as in the table above (— = no official figure); VCI 50 is the middle of the record's range; spring NDVI for the whole West Bank against the 2001–2020 normal (from 2001).

What the record suggests for 2026/27

If the forecasts' wet October–December holds, the record leans towards a reasonable spring for rainfed cereals, pulses and pasture in 2027, in every zone and most visibly in the drier highlands, eastern slopes and Jordan Valley margins. What it shows is mainly a lower chance of a bad year, not a bumper one, and it rests on 9–14 El Niño years per record: the intervals around every count above are wide, and this is not a crop forecast. Olives will be set more by alternate bearing and by access than by the winter: a large 2026 crop would make a smaller 2027 crop likely whatever the rain.

Three limits apply. Every agricultural record here starts after 1979, inside the period in which the El Niño link to West Bank rain exists, and the coming El Niño is likely to be stronger than any in them. The timing of the rain, frost and spring heat also matter, and the season total does not capture them. And much of what will decide the 2027 harvest is not weather: access to land near Israeli settlements and behind the Barrier, settler violence, and the cost and availability of water and inputs.

Caveats

• A relationship that switched on can switch off. The El Niño link appears only after the late 1970s in every long record for both areas. The page's numbers describe 1979–2025, not a physical law, and the coming event is likely to be stronger than any in that period.
• Small samples and index choice. 14 El Niño winters since 1979 on the pinned index (16 on HadISST), six of them strong and two of those following major eruptions. Every hit rate here carries wide uncertainty, and the break at 1979 was identified from the data and the literature together.
• The records, Gaza. No rain gauge inside Gaza reports internationally. GPCC is a 1° analysis that takes in the southern coastal plain and north-western Negev; after 2019 it is the Monitoring product, built on fewer gauges. Beer Sheva is 46 km inland, drier, and one of GPCC's inputs, so the two are not independent. ERA5 is three 25 km cells, partly over the sea, and assimilates no rain gauges. IMERG is the satellite-only late run, which tracks the gauges less well year to year. Their agreement on the sign and the timing is the evidence; no single number is Gaza's.
• The records, West Bank. GPCC's 1° cells take in the Jordan Valley and the hills on both sides. The Jerusalem gauge is one hill-town site, is one of GPCC's inputs, and its record ends in 2015. ERA5's cells are smoother and lower than the hills, so it understates heavy days and cannot see snow. A storm that soaks one hillside is diluted in any West Bank average, so the storm counts here understate local downpours.
• Forecast calibration. The September 2026 SEAS5 issuance sat in the tails of its hindcast unusually often across the globe (see the Zimbabwe page); a very strong El Niño can do that, but the normalization check is not complete, and it has not been repeated for the October issuance. Read its return periods as a ranking. The last time El Niño and a positive Indian Ocean Dipole lined up, a “forecast of opportunity” for October–December 2023 gave better than 60 per cent odds of above-normal rain over Southwest Asia, and the season came in at the 15th percentile (Hoell et al. 2025).
• Rainfall is one of several winter hazards. Sea surge, wind and cold caused a large share of the damage and deaths in Gaza in the last three winters, and snow and flash floods have caused much of the West Bank's; none of them has a clear El Niño signal in this analysis or in the literature (windy days may be somewhat more frequent in El Niño winters; cold nights and snow are not).
• Impact figures. Counts from the Ministry of Health in Gaza, the Palestinian Civil Defense and the Government Media Office are relayed, not verified, by the UN; OCHA does not systematically monitor deaths that are not caused by the hostilities. Most West Bank weather impacts are reported by the Palestinian Civil Defense through the media, often as counts of incidents or rescues, and the UN quantified one West Bank weather event in three winters. Reporting practice changed between winters, several figures are cumulative or count exposure rather than impact, and the absence of counts is not an absence of impact.

References

• Price, C., Stone, L., Huppert, A., Rajagopalan, B. & Alpert, P. (1998). A possible link between El Niño and precipitation in Israel. Geophysical Research Letters 25, 3963–3966. doi:10.1029/1998GL900098
• Alpert, P. et al. (2005). Tropical tele-connections to the Mediterranean climate and weather. Advances in Geosciences 2, 157–160. doi:10.5194/adgeo-2-157-2005
• Mariotti, A., Zeng, N. & Lau, K.-M. (2002). Euro-Mediterranean rainfall and ENSO — a seasonally varying relationship. Geophysical Research Letters 29, 1621. doi:10.1029/2001GL014248; Mariotti, A., Ballabrera-Poy, J. & Zeng, N. (2005). Tropical influence on Euro-Asian autumn rainfall variability. Climate Dynamics 24, 511–521. doi:10.1007/s00382-004-0498-6
• Halfon, N., Ilotoviz, E., Carmona, Y., Yosef, Y. & Givati, A. (2026). The link between extreme El Niño values and precipitation in Israel (Hebrew; author names transliterated). Israel Meteorological Service technical report, July 2026. ims.gov.il (not peer-reviewed)
• Hochman, A. & Gildor, H. (2025). Synergistic effects of El Niño–Southern Oscillation and the Indian Ocean Dipole on Middle Eastern subseasonal precipitation variability and predictability. QJRMS 151, e4903. doi:10.1002/qj.4903
• Kang, I.-S. et al. (2015). Multidecadal changes in the relationship between ENSO and wet-season precipitation in the Arabian Peninsula. Journal of Climate 28, 4743–4752. doi:10.1175/JCLI-D-14-00388.1; Kamil, S. et al. (2017). Multidecadal changes in the relationship of storm frequency over Euro-Mediterranean region and ENSO during boreal winter. Earth Systems and Environment 1. doi:10.1007/s41748-017-0011-0
• Dayan, U., Nissen, K. M. & Ulbrich, U. (2015). Atmospheric conditions inducing extreme precipitation over the eastern and western Mediterranean. NHESS 15, 2525–2544. doi:10.5194/nhess-15-2525-2015
• Nissen, K. M. et al. (2010). Cyclones causing wind storms in the Mediterranean. NHESS 10, 1379–1391. doi:10.5194/nhess-10-1379-2010; Sandler, D. et al. (2024). The connection between North Atlantic storm track regimes and eastern Mediterranean cyclonic activity. Weather and Climate Dynamics 5, 1103–1116. doi:10.5194/wcd-5-1103-2024
• Dogar, M. M. et al. (2017). Sensitivity of the regional climate in the Middle East and North Africa to volcanic perturbations. JGR Atmospheres 122, 7922–7948. doi:10.1002/2017JD026783
• Hoell, A. et al. (2025). An unexpected outcome followed an apparent seasonal forecast of opportunity and prolonged drought in Southwest Asia. International Journal of Climatology 45. doi:10.1002/joc.8851
• NOAA CPC ENSO diagnostic discussion, 10 September 2026 link; IRI ENSO forecast, 21 September 2026 link.
• IOM, Site Management Cluster & Shelter Cluster (2026). Sites exposed to flood risk: Phase one outlook for rainy season 2026–27 link; Shelter Cluster (2026). Exposed on all sides: shelter severity and household needs across accessible areas in Gaza link; OCHA Humanitarian Situation Report, 25 September 2026 link.
• Rainfall: GPCC Full Data v2020 to 2019 + Monitoring v2020 after (DWD), combined 1° product via NOAA PSL; ERA5 hourly time series (Copernicus Climate Change Service); IMERG late run v7 (NASA), from the team blob; GHCN-Daily Beer Sheva IS000051690 (NOAA NCEI). Niño3.4: NOAA PSL (ERSST v5 pinned; HadISST before 1950). SEAS5 skill from the seas5-skill app's detrended skill cube; the October 2026 return periods from the team's admin-1 raster statistics with the same method. Boundaries: COD-AB (OCHA) via FieldMaps.
• Dudeen, B. (2001). The soils of Palestine (the West Bank and Gaza Strip): current status and future perspectives. In Zdruli et al. (eds), Soil resources of Southern and Eastern Mediterranean countries, Options Méditerranéennes B 34, CIHEAM, 203–225.
• FAO (2026). FAOSTAT: Crops and livestock products (listed by FAO as “Palestine”), 1994–2022.
• PCBS. Olive Presses Survey, main results, 2003–2019 (e.g. 2019); Agricultural Census 2021, final results. Rows and sources in deep_dives/data/pcbs_*.csv.
• NOAA STAR. Blended Vegetation Health Products, province-averaged (West Bank), 1982–2026.
• Didan, K. (2021). MODIS/Terra Vegetation Indices 16-Day L3 Global 250m (MOD13Q1 v061), NASA LP DAAC; read from the Microsoft Planetary Computer.
• OCHA (2025). Humanitarian Situation Update #335, West Bank, 30 October 2025.
• Al Jazeera (2025). Oil presses stand silent as West Bank has its worst olive harvest in years, 3 December 2025 (estimates from MAS).

Generated by levant_deep_dive.py from deep_dives/gaza-west-bank.toml and deep_dives/parts/gaza.toml, deep_dives/parts/west-bank.toml. Grid method and Niño3.4 series as in the global survey.