📄 earth_science

Seasonal rainfall predictability governed by interacting oceanic and terrestrial memory reservoirs

This study reveals that seasonal rainfall predictability, particularly in the South Asian monsoon, is governed not solely by remote oceanic forcing like ENSO but significantly by the persistence and spatial organization of terrestrial memory reservoirs, where soil moisture acts as a primary driver that often mediates oceanic influences and enhances forecast skill when initialized.

Original authors: Swadhin Behera, Mukunda Dev Behera

Published 2026-07-08

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Swadhin Behera, Mukunda Dev Behera

Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

The Big Idea: Who is the Real Boss of the Rain?

For a long time, scientists and weather forecasters have believed that the ocean is the main boss of seasonal rainfall. They think that big ocean events, like El Niño (a warming of the Pacific Ocean), act like a giant remote control that dictates whether a region will have too much rain or too little.

However, this new study suggests that while the ocean starts the conversation, the land is often the one who actually remembers the details and decides the outcome. The researchers found that seasonal rainfall predictability isn't just about how strong the ocean signal is; it's about where the water "memory" is stored in the climate system.

The Analogy: The Orchestra and the Echo Chamber

Imagine the climate system as a giant orchestra playing a song (the rainfall).

The Ocean (El Niño) is the Conductor. The conductor waves their baton to start the music. Everyone knows the conductor is important.
The Land (Soil Moisture) is the Echo Chamber. Once the music starts, it bounces off the walls of the room (the land). If the room is made of hard stone (dry soil), the sound dies out quickly. If the room is filled with soft, damp cushions (wet soil), the sound lingers, echoes, and changes the music.

The Paper's Discovery:
In many parts of South Asia (specifically the dry, interior regions), the "Echo Chamber" (the soil) is so powerful that it matters more than the Conductor's initial wave. Even if the Conductor (El Niño) tries to change the song, the Echo Chamber (the soil) holds onto the previous notes so tightly that it overrides the new instructions.

How They Tested This

The researchers looked at 44 years of data (1982–2025) over South Asia. They used a statistical "sieve" to separate the rain into three buckets:

Bucket A: Rain caused directly by the Ocean (El Niño).
Bucket B: Rain caused by the Land "breathing" (Evapotranspiration—plants and soil releasing water vapor).
Bucket C: Rain caused by the Land "remembering" (Soil Moisture—water stored deep in the ground from previous months).

They found that in the Deccan Plateau (a large, dry plateau in southern India) and parts of Northwest India, Bucket C (Soil Memory) was the biggest contributor to predicting the rain. In fact, the ocean's direct influence was surprisingly small in these areas.

The "Memory" Metaphor

Think of Soil Moisture like a battery or a savings account for water.

Before the monsoon season starts (in March, April, or May), the soil stores water like a battery charges up.
When the monsoon arrives, this stored water doesn't just sit there; it releases energy and moisture into the air, helping to create more rain.
The study found that if you know how "charged" the soil battery is before the season starts, you can predict the rain much better than if you only look at the ocean temperature.

Key Findings in Plain English

The Ocean isn't the only boss: While El Niño is famous, its direct power to predict rain is actually quite limited in many regions. It often only works by changing the soil first.
Land is the real memory keeper: In dry, semi-arid regions, the soil holds onto water anomalies (too wet or too dry) for months. This "memory" is what actually drives the rainfall patterns.
The "Mediation" Effect: The study showed that El Niño often affects the rain only because it changes the soil moisture first. If you account for the soil, the direct link between El Niño and rain almost disappears. It's like a relay race: The Ocean passes the baton to the Land, and the Land runs the rest of the race.
Better Forecasts: When the researchers added "soil moisture" data to their computer models, the rain predictions got significantly better, especially in the Deccan Plateau. It's like trying to guess the weather: knowing the ground is already wet helps you predict a storm better than just knowing the ocean is warm.

Why This Matters (According to the Paper)

The paper concludes that we need to stop looking at the ocean as the only source of predictability.

Human Impact: Because soil moisture is a "memory reservoir," things humans do—like irrigation (watering crops), digging up groundwater, or changing how we use the land—can change how this memory works.
Shifting the Rules: If humans change the land (e.g., by drying out the soil or making it wetter), they might accidentally change where and how we can predict the rain, even if the ocean stays exactly the same.

In short: The ocean starts the fire, but the land (specifically the soil) decides how long the fire burns and how big it gets. To predict the rain, we need to watch the soil just as closely as we watch the ocean.

Technical Summary: Seasonal Rainfall Predictability Governed by Interacting Oceanic and Terrestrial Memory Reservoirs

Problem Statement
Seasonal rainfall predictability has traditionally been attributed to persistent oceanic anomalies, particularly the El Niño-Southern Oscillation (ENSO) and the Indian Ocean Dipole (IOD). The prevailing paradigm interprets rainfall anomalies primarily as responses to remote oceanic forcing. However, this view often conflates sensitivity to forcing with predictive dominance. The authors argue that predictability depends fundamentally on where hydroclimatic anomalies are stored and reintroduced within the climate system. While oceans provide large-scale memory, land surfaces (via soil moisture and evapotranspiration) also store anomalies on weekly-to-seasonal timescales. The central question addressed is whether predictive skill resides in the remote oceanic forcing or in the terrestrial reservoirs that sustain and reintroduce anomalies, and how these contributions are partitioned across different hydroclimatic regimes.

Methodology
The study utilizes a diagnostic framework applied to the South Asian monsoon region (60–110°E, 5–40°N) using ERA5 reanalysis data from 1982 to 2025.

Diagnostic Framework: Seasonal rainfall anomalies ( $P'$ ) are modeled as a linear combination of oceanic forcing (represented by the Niño3 index), evapotranspiration anomalies ($ET'$), and root-zone soil moisture anomalies ($SM'$):
$P' = \alpha \cdot \text{ocean}' + \beta \cdot ET' + \gamma \cdot SM' + \varepsilon$
This formulation is statistical and diagnostic, intended to partition predictive influence rather than establish mechanistic causality.
Variance Decomposition: Using ordinary least squares regression, the total explained variance ( $R^2_{full}$ ) is partitioned into:
- Unique contributions of oceanic forcing, evapotranspiration, and soil moisture.
- Shared covariance among these predictors.
  Unique contributions are isolated via partial regression (e.g., $R^2_{ocean, unique} = R^2_{full} - R^2(P' \sim ET', SM')$ ).
Land-Dominance Index: An internal index ( $D_{land}$ ) is defined to distinguish between flux-driven variability (evapotranspiration) and storage-mediated variability (soil moisture) within the terrestrial block. Negative values indicate soil moisture dominance.
Validation and Hindcasting:
- Cross-Validation: A leave-one-year-out cross-validation (1982–2025) is used to assess out-of-sample skill, avoiding in-sample overfitting.
- Hindcast Experiments: The study compares an ocean-only baseline (using antecedent March–May Niño3) against a model incorporating pre-monsoon soil moisture initialization.
- Robustness Checks: Sensitivity to vertical soil moisture weighting schemes and temporal stability (via 25-year sliding windows) are tested.

Key Results

Spatial Redistribution of Predictability: While total seasonal rainfall predictability is substantial across semi-arid and transitional regions, the unique contribution of oceanic forcing is spatially confined and generally modest ( $<0.1$ ). In contrast, soil moisture accounts for the majority of independent explained variance in interior regions.
Regional Hierarchy: Three distinct hydroclimatic regimes were analyzed:
- Deccan Plateau (DP): Exhibits the strongest storage-mediated control. Unique soil moisture contribution is highest ( $R^2 \approx 0.391$ ), while unique oceanic contribution is minimal ( $R^2 \approx 0.010$ ). The land-dominance index is strongly negative ( $D_{land} = -0.652$ ).
- Indo-Gangetic Plain (IGP): Shows a mixed but storage-weighted structure. Soil moisture remains the dominant unique predictor ( $R^2 \approx 0.315$ ), with oceanic influence largely mediated through shared covariance with terrestrial states.
- Northwest India (NWI): Represents a strongly land-regulated regime where storage-mediated variability outweighs oceanic forcing, though the oceanic contribution is slightly more evident than in the DP.
Mediation of Oceanic Signals: Mediation analysis reveals that ENSO-rainfall associations weaken substantially after conditioning on terrestrial storage. This suggests oceanic variability frequently projects onto rainfall through shared covariance with land-state memory rather than acting as an independent predictor.
Forecast Skill Gains: Out-of-sample hindcast experiments demonstrate that initializing models with pre-monsoon soil moisture significantly enhances seasonal forecast skill in storage-dominated regimes (specifically the Deccan Plateau, with a correlation skill gain $\Delta r \approx 0.39$ ). In contrast, ocean-only models exhibit weak or negative out-of-sample skill in these regions.
Temporal Stability: The dominance hierarchy (DP > IGP > NWI) and the spatial partitioning of predictive control remain stable over the 1982–2025 period, indicating these are structural features of the monsoon system rather than transient anomalies.

Significance and Claims
The paper claims to establish a unifying principle: seasonal rainfall predictability is governed not solely by the magnitude of remote forcing but by the persistence and spatial organization of interacting climate memory reservoirs.

Distinction between Sensitivity and Dominance: The study clarifies that strong correlations between rainfall and oceanic variability do not imply that predictive skill resides in the ocean forcing once interacting components are considered. Predictive leverage aligns with the reservoir that stores and reintroduces anomalies (terrestrial memory in semi-arid regions), not necessarily the forcing that initiates variability.
Implications for Forecasting: The results suggest that operational forecast systems, which often prioritize ocean-state initialization, may benefit from regionally adaptive strategies. In storage-dominated regimes, accurate representation and initialization of land-state memory (soil moisture) may be as critical as ocean-state initialization.
Anthropogenic Context: Because terrestrial memory is sensitive to land-use change, irrigation, and hydrological management, human activities may systematically redistribute regional predictability under a changing climate, even without changes in ocean forcing amplitude.

The authors conclude that identifying where predictive leverage resides requires separating rainfall sensitivity to forcing from the persistence of the reservoirs that sustain anomalies, a distinction not captured in conventional land-atmosphere coupling analyses.