The Indian summer monsoon feeds a fifth of humanity, but its seasonal forecast has been considered nearly hopeless for decades: the potential predictability limit was estimated at only 45% of explained variability. The authors showed that the so-called "climate noise" (subseasonal fluctuations with periods of 3–7 days) is actually partly predictable because it is linked to slow modes such as El Niño. By improving the physics of the CFSv2 model (cloud microphysics and snow scheme) and running retrospective forecasts totaling 1920 model-years, they achieved a correlation of 0.71, while the maximum achievable correlation under the new estimate of the predictability limit reached 0.82. This means that about 70% of the monsoon's interannual variability can be predicted today, and up to 80% in the future.

Picture an octopus with its tentacles on monsoon storm statistics. Previously, those tentacles seemed to be pure chaotic noise: rapid rainfall fluctuations with periods of a few days that cannot be predicted a season ahead. But the authors found that part of this "noise" is actually driven by a slow signal: sea surface temperature in the tropical Pacific, known as El Niño.

This means that fast disturbances are not random: they are tied to slow climate modes, and therefore they can be predicted.

To test this idea, the scientists took the global atmosphere-ocean model CFSv2 and created four versions: standard, with improved snow physics, with improved cloud microphysics, and with both improvements. For each version, they ran retrospective forecasts over 30 years (1981–2010) with 16 ensemble members, totaling 1920 model-years. They then estimated the predictability limit using three methods: "perfect model" correlation (where one ensemble member is compared with the rest), variance analysis, and a new method based on exhaustive search over all possible subsets of the ensemble to find the maximum correlation with observations.

The result was unexpected. The best model version (with improved cloud microphysics) yielded a correlation of 0.71 between forecast and observations, higher than the previous predictability limit of 0.65. And the new method for estimating the limit, based on subset search, showed a maximum achievable correlation of 0.82. This means the model can explain about 70% of the interannual variability of Indian monsoon rainfall, whereas previous estimates suggested 45%.

In other words, the "fast noise" is actually governed by the slow El Niño signal.

Previously, it was thought that the Indian summer monsoon was only 45% predictable due to chaotic subseasonal variability, and that this limit was insurmountable. Now it has been shown that part of that variability (especially 3–5 day disturbances) is linked to El Niño and is therefore predictable. The actual predictability limit has been raised to a correlation of 0.82, and the model already achieves 0.71. This was accomplished by linking the statistics of fast rainfall fluctuations to slow climate modes, improving the model's physics, and then exhaustively searching all possible ensemble combinations to find the maximum achievable correlation.

Already today, the improved model can provide more reliable seasonal monsoon forecasts for agriculture and water resource management in India. If the representation of the contribution of subseasonal anomalies (especially 10–20 day oscillations) to the seasonal anomaly is improved, the predictability limit could rise to 0.9, corresponding to about 80% of explained variance. This requires continued refinement of cloud physics and ocean-atmosphere interactions. It is also important to investigate whether the minimum predictability will increase when the ensemble size grows beyond 16 members.

So the octopus of predictable climate modes, which once seemed elusive, is now taking shape. Its tentacles are not just chaotic movements but responses to slow ocean changes. And the better we understand these connections, the more reliably we can predict when the rain that feeds billions of people will arrive.