Climate network and complexity based ENSO forecast for 2026

Abstract

The El Niño Southern Oscillation (ENSO) is the dominant driver of interannual global climate variability and can lead to extreme weather events such as droughts or flooding. Recently, we have developed several statistical approaches for early ENSO forecasting, in particular, its El Niño phase. The climate network-based approach allows forecasting the onset of an El Niño event or its absence about 1 year ahead [1]. The complexity-based approach allows additionally to forecast the magnitude of an upcoming El Niño event in the calendar year before the onset [2]. Additionally, we have developed methods for forecasting the type (Eastern Pacific or Central Pacific) of an El Niño [3] and for probabilistic forecasting of La Niña and neutral events [4], also by the end of the calendar year before the event. Here we present the forecasts of these methods for 2026. The climate network and the complexity-based approach do not provide concurring signals for this year. The combined forecast indicates that a neutral event is more likely than an El Niño. If an El Niño develops in 2026, the complexity-based approach predicts a weaker event with a magnitude of $0.84\pm0.36$°C.

Figures (6)

• Figure 1: The nodes of the climate network. The network consists of 14 grid points in the central and eastern equatorial Pacific (red dots) and 193 grid points outside this area (blue dots). The green rectangle indicates the Niño3.4 area. The grid points represent the nodes of the climate network that we use here to forecast the onset or absence of an El Niño event. Each red node is linked to each blue node. The nodes are characterized by their surface air temperature (SAT), and the link strength between the nodes is determined from their cross-correlation (see main text).
• Figure 2: The network-based forecasting scheme. ( a) We compare the NCEP reanalysis 1 based average link strength $S(t)$ in the climate network (red curve) with a decision threshold $\Theta$ (horizontal line, here $\Theta = 2.82$), (left scale), and the Oceanic Niño Index (ONI), (right scale), between January 1981 and December 2025 (present). Please note that we show the 2025 version of the ONI values here. When the link strength crosses the threshold from below, and the last available ONI is below $0.5^\circ$C, we give an alarm and predict that an El Niño episode will begin in the following calendar year. The El Niño episodes (when the ONI is at or above $0.5^\circ$C for at least 5 months) are shown by the solid blue areas. Correct predictions are marked with green arrows and false alarms with dashed grey arrows. The index $n$ marks not-predicted events. The threshold was learned in a learning phase between 1950 and 1980 Ludescher2013 (not shown). The black vertical dashed line separates the hindcasting and real-time forecasting phases. Between 1981 and December 2025, there were 12 El Niño events. The algorithm generated 13 alarms, and 9 of them were correct. In the more restrictive version (ii) of the algorithm, only those alarms are considered where the ONI remains below 0.5°C for the rest of the year. In this version, the incorrect alarms in 1994, 2004, 2019 and 2023 are not activated. Between 1981 and December 2025, version (ii) of the algorithm gave 9 alarms, all of which were correct. In 2025, the average link strength $S(t)$ increased from the very low values in 2024 but remained below the critical threshold band throughout the year, thus forecasting the absence of an El Niño in 2026. ( b) Analogous to ( a), but here $S(t)$ is based on the near-surface air temperature data from ERA5 (1000hPa). We consider version (ii) of the algorithm. Both data sets yield nearly the same forecasts, except for a false alarm in 2020 (ERA5) and correct alarms in 2005 (ERA5) and 2017 (NCEP). By the end of 2025, $S(t)$ based on ERA5 has crossed the lowest threshold, but not all thresholds. This might be interpreted as a partial alarm, potentially consistent with a weak El Niño as forecasted by the SysSampEn (see below).
• Figure 3: The Niño3.4 area and the SysSampEn input data. The red circles indicate the 22 nodes covering the Niño 3.4 region at a spatial resolution of $5^\circ \times 5^\circ$. The curves are examples of the temperature anomaly time series for 3 nodes in the Niño 3.4 region for one specific year. Several examples of their subsequences are marked in black. In the calculation of the SysSampEn, both the similarity of subsequences within a time series and that of subsequences across different time series are considered. Figure from Meng2019.
• Figure 4: Forecasted and observed El Niño magnitudes. The magnitude forecast is shown as the height of rectangles in the year when the forecast is made, i.e., one year ahead of a potential El Niño. The forecast is obtained by inserting the regarded calendar year's SysSampEn value into the linear regression function between SysSampEn and El Niño magnitude. To forecast the following year's condition, we use the ERA5 daily near-surface (1000 hPa) temperatures with the set of SysSampEn parameters ($m = 30$, $p = 30$, $\gamma=8$ and $l_{eff} = 360$). The red curve shows the 2025 version of the ONI, and the red shades highlight the El Niño periods. The blue rectangles show the correct prediction of an El Niño in the following calendar year. The onset of an El Niño in the following year is predicted if the forecasted magnitude is above $0.5^\circ$C and the current year's December ONI is $<0.5^\circ$C. White rectangles show correct forecasts for the absence of an El Niño. Grey bars with a violet border show false alarms, and the only missed event is shown as a pink rectangle. The SysSampEn value for 2025 is $1.60$, which is above the threshold value of 1.50. There were 14 occurrences of high SysSampEn accompanied by a low ONI in December, as is the case in 2025 (green rectangle). In 10 out of these 14 cases, the hindcast was correct. Thus, the method predicts with 71.4% probability the onset of an El Niño in 2026.
• Figure 5: The probability of a La Niña. We regard all years that are non-El Niño years and are also followed by a non-El Niño year. Under the assumption that no El Niño starts in 2026, these events correspond to the current state. The outcome of the second year is encoded as 1 for La Niña and 0 for a neutral event (blue circles). To obtain the probability of a La Niña event in 2026, given that no El Niño starts in 2026, we apply a logistic regression and use the current OND ONI value of $-0.5$°C as a predictor. Excluding El Niño, we obtain a $29.1\%$ probability for a La Niña vs. a $70.9\%$ probability for a neutral event.