📄 earth_science

Spatiotemporal Heterogeneity of Daily Rainfall Persistence across the Maluku Archipelago: A First-Order Markov Chain Analysis of ERA5 Reanalysis Data (2000–2025)

This study utilizes a first-order Markov chain analysis of ERA5 reanalysis data (2000–2025) to characterize the pronounced spatiotemporal heterogeneity of daily rainfall persistence across the Maluku Archipelago, revealing how the interplay between monsoonal circulation and complex island topography governs dry–wet spell dynamics and extreme rainfall thresholds in this data-sparse tropical region.

Original authors: Imel Putri Apriliyanti, Adi Mulsandi

Published 2026-08-04

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Imel Putri Apriliyanti, Adi Mulsandi

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 Weather's Memory: Why Some Islands Get Stuck in Rain or Drought

Imagine the weather not just as a series of random events, but as a creature with a memory. Sometimes, once it starts raining, it just can't seem to stop. Other times, once the sun comes out, it refuses to let a single cloud cross the sky. In the world of science, this "memory" is called persistence. While we often talk about how much it rains, scientists are increasingly interested in how long the rain (or the lack of it) sticks around. This is crucial for places like small islands, where a few extra days of rain can cause flash floods, and a few extra days of dryness can drain the water supply. To study this, researchers use a mathematical tool called a Markov chain. Think of this as a simple rulebook for the weather: it assumes that what happens tomorrow depends mostly on what happened today. If today is wet, the rulebook calculates the odds of tomorrow being wet too. By applying this logic to vast amounts of data, scientists can map out which parts of the world are prone to long, sticky wet spells and which are prone to stubborn droughts.

The Story of the Maluku Islands: A Tale of Two Weather Personalities

This study takes a deep dive into the Maluku Archipelago in Indonesia, a stunning collection of hundreds of islands scattered across warm seas. The researchers, Imel Putri Apriliyanti and Adi Mulsandi, wanted to solve a mystery: Why does the weather behave so differently from one island to the next, even when they are close together? They knew that the islands have tricky shapes—some are flat and low, while others have towering mountains—and that huge wind patterns (monsoons) sweep across the region. But they needed to figure out exactly how these factors combine to make the weather "stick" in one state or another.

To crack the code, they didn't rely on a few scattered rain gauges, which are often missing or broken in remote island areas. Instead, they used a super-powerful digital weather map called ERA5, which acts like a time machine, reconstructing daily weather details from the year 2000 to 2025. They picked eight distinct "zones" to represent the different faces of the archipelago: from the deep, mountainous interiors of islands like Seram, to the flat coastal plains, and all the way to the dry, southern islands near Australia. They treated every single day as either "wet" (if it rained at least 1 mm) or "dry," and then ran their Markov chain math to see how likely the weather was to stay the same the next day.

The Big Discovery: Mountains Get Stuck in Rain, Plains Get Stuck in Drought

The results revealed a dramatic split in the weather's personality across the archipelago. The study found that rainfall persistence is not random; it follows a clear pattern based on the landscape.

The mountainous interior of Seram Island (Zone 3) turned out to be the wettest and most persistent place. Here, the mountains act like a giant wall that forces moist air to rise, cool, and dump water. The data showed this area had the highest average rainfall at 11.85 mm per day and the highest extreme rain threshold, where the top 5% of rainy days hit 72.8 mm. Because of this, the rain here tends to stick around. In June, the "wet spell" (a streak of rainy days) lasted an average of 8.87 days, and in extreme cases, it could stretch to 21.3 days (the 90th percentile). This suggests that if it starts raining in the mountains, it is very likely to keep raining for over a week, creating a high risk for flash floods in areas that can't hold much water.

On the other end of the spectrum, the southern Tanimbar region (Zone 8, near Saumlaki) told a very different story. This area is low and flat, and during the Southeast Monsoon, it gets hit by dry air blowing in from the Australian continent. The study found this zone had the lowest average rainfall at just 4.45 mm per day, with many days having zero rain at all. Here, the "dry spell" (a streak of sunny days) was the dominant feature. In July, the average dry spell lasted 6.72 days, and in extreme cases, it could stretch to 14.8 days. The weather here gets "stuck" in a dry mode, making these islands particularly vulnerable to drought.

The Seasonal Dance of Wet and Dry

The researchers also discovered that the weather's memory changes with the seasons. They found that the transition probabilities (the odds of the weather changing) shift systematically throughout the year.

June was the peak for wet persistence, with the longest expected wet spells.
July was the peak for dry persistence, with the longest expected dry spells.

This seasonal shift is driven by the massive Asian-Australian monsoon system. When the monsoon shifts, it brings moist air to the central and northern islands, making the rain stickier, while simultaneously pushing dry air over the southern islands, making the sun stickier. The study showed that you cannot explain the weather just by looking at the wind or just by looking at the mountains; it is the combination of the two that creates these distinct patterns. The mountains amplify the wind's effects, turning a wet wind into a long, soaking deluge, or a dry wind into a prolonged drought.

Why This Matters for the Future

The authors suggest that this "one-size-fits-all" approach to weather planning doesn't work for the Maluku Archipelago. Because the mountains and the lowlands have such different "weather memories," a single strategy for the whole region would fail. The mountainous areas need to prepare for long, consecutive wet days that could trigger floods, while the southern islands need to prepare for long, consecutive dry days that could run out their water supplies.

By using this simple but powerful mathematical model on high-quality digital data, the study provides a new way to understand weather in places where we don't have enough physical rain gauges. It shows that even in a complex, data-sparse world, we can map out the "personality" of the weather to help communities prepare for the specific risks they face—whether that's a flood that won't stop or a drought that won't break.

Technical Summary: Spatiotemporal Heterogeneity of Daily Rainfall Persistence across the Maluku Archipelago

Problem Statement
Tropical archipelagos, particularly the Maluku Archipelago within the Indonesian Maritime Continent, exhibit extreme hydroclimatic heterogeneity driven by the interaction of complex island topography, surrounding warm seas, and multi-scale atmospheric circulation. While mean rainfall climatology and extreme precipitation indices have been extensively studied in this region, the temporal persistence of daily rainfall—the tendency for wet or dry conditions to continue from one day to the next—remains poorly characterized. This gap is critical for hydrological risk management in small, water-limited island systems where the duration of dry or wet spells directly influences drought susceptibility and flash-flood potential. Furthermore, the scarcity of long-term, spatially continuous ground-based observations in the region has hindered systematic characterization of these dynamics.

Methodology
This study addresses the data gap by coupling high-resolution ERA5 reanalysis data (2000–2025) with a first-order, two-state Markov chain framework. The analysis focuses on eight representative climate zones spanning diverse physiographic settings: mountainous interiors (e.g., Seram Island), coastal plains, open-ocean islands, and the dry southern frontier (Tanimbar Islands).

Key methodological steps included:

Data Processing: Daily precipitation totals were extracted from ERA5 grid cells nearest to reference locations. A fixed threshold of 1 mm day⁻¹ was used to classify days into discrete "dry" (state 0) and "wet" (state 1) states.
Markov Modeling: A first-order Markov chain was applied to calculate monthly transition probability matrices ( $P_{DD}$ , $P_{DW}$ , $P_{WD}$ , $P_{WW}$ ). These matrices quantify the probability of transitioning between dry and wet states.
Derived Metrics: From the transition probabilities, the study derived:
- Stationary probabilities ( $\pi_0$ , $\pi_1$ ) representing the long-term likelihood of dry or wet states.
- Expected spell durations ( $E(L_d)$ , $E(L_w)$ ) calculated as the inverse of transition probabilities ( $1/P_{DW}$ and $1/P_{WD}$ ).
- Extreme spell durations defined by the 90th percentile ( $P_{90}$ ) of continuous spell lengths.
- Extreme rainfall thresholds ( $P_{90}$ , $P_{95}$ ) based on the distribution of wet-day precipitation.
Validation: The model's validity was assessed by comparing the reconstructed annual cycle of precipitation (via stochastic simulation) against the observed ERA5 annual cycle, confirming the framework adequately captures seasonal evolution.

Key Results
The analysis reveals pronounced spatial heterogeneity in rainfall persistence across the archipelago, governed by the interplay of orography and monsoonal circulation:

Spatial Gradients:
- Mountainous Interiors (Zone 3, Seram): Recorded the highest mean daily precipitation (11.85 mm day⁻¹) and the highest extreme rainfall threshold ( $P_{95} = 72.8$ mm). This zone exhibited the strongest wet-to-wet persistence, consistent with orographic enhancement of convective rainfall.
- Southern Islands (Zone 8, Saumlaki): Recorded the lowest mean daily precipitation (4.45 mm day⁻¹) and a median near zero. This zone showed the highest dry-to-dry persistence, driven by the seasonal intrusion of dry continental air from northern Australia during the Southeast Monsoon.
Seasonal Dynamics:
- Wet Spells: Expected wet-spell durations peaked in June (mean 8.87 days; $P_{90} = 21.3$ days), coinciding with active monsoonal moisture transport.
- Dry Spells: Expected dry-spell durations peaked in July (mean 6.72 days; $P_{90} = 14.8$ days), corresponding to the peak of the dry season.
- Variability: Monthly variability in precipitation was lowest in January and April but increased markedly during the middle and latter parts of the year, reaching a primary maximum in July and a secondary maximum in December. These elevated variability months indicate greater interannual fluctuation and reduced predictability during the peak dry season and the onset of the wet season, consistent with heightened instability in the regional monsoon transition.

Significance and Claims
The paper claims that rainfall persistence in the Maluku Archipelago is not a random or island-specific phenomenon but is organized along a coherent, physically explicable spatial gradient. The study demonstrates that this persistence is jointly governed by monsoonal circulation and island-scale physiographic complexity, rather than either factor acting alone.

The primary significance of the work lies in its methodological contribution: it establishes a transferable, reproducible approach for quantifying atmospheric memory in data-sparse tropical island regions. By utilizing a parsimonious first-order Markov framework applied to freely available reanalysis data, the study provides a viable alternative to conventional climatological descriptions where ground-based networks are insufficient.

Practical Implications
The findings offer direct relevance to:

Drought Monitoring: Identifying southern islands as high-risk zones for prolonged dry spells and water scarcity.
Flash-Flood Early Warning: Highlighting mountainous catchments as areas prone to extended wet spells and cumulative rainfall hazards.
Climate Adaptation: Arguing that a single archipelago-wide adaptation strategy is inadequate due to the asymmetric nature of hazards (wet-spell persistence in mountains vs. dry-spell persistence in the south).

The authors maintain a modest stance, acknowledging that the 0.25° resolution of ERA5 may not resolve fine-scale local circulations (e.g., sea breezes) and that the first-order Markov assumption simplifies atmospheric memory. However, they assert that the framework successfully recovers physically coherent, spatially resolved persistence signals that are otherwise inaccessible in this region.