Satellite-Based Estimation of Monthly Rainfall Across the Indonesian Archipelago: A Month-Specific Multiple Linear Regression Framework Using Bias-Corrected NASA POWER Data over Java, Bali, Nusa Tenggara, and Maluku

Rainfall is the heartbeat of the tropical world, a variable force that dictates when crops grow, when rivers swell, and when the land turns to mud. In Indonesia, an archipelago of thousands of islands stretching across the equator, this rhythm is governed by massive seasonal winds that shift moisture from the Indian Ocean to the Australian continent. For farmers, city planners, and disaster responders, knowing exactly how much rain will fall in a given month is not just a matter of curiosity; it is a necessity for survival and economic stability. However, a persistent problem has long hindered accurate planning: the ground-based weather stations that measure this rain are often sparse, and the historical records they keep are frequently incomplete. In many parts of the country, the freely available data stops after just a couple of years, leaving scientists without the long-term view needed to understand climate trends or to calibrate reliable prediction models. Without these deep archives, the ability to forecast future water availability or prepare for floods remains limited.

To bridge this gap, a team of researchers from several Indonesian universities turned their attention to the sky, specifically to a satellite-based dataset that offers a continuous, unbroken record of weather conditions. Their goal was to build a new way to estimate monthly rainfall across four distinct regions of the Indonesian archipelago: the densely populated island of Java, the small island of Bali, the dry and seasonal Nusa Tenggara islands, and the scattered, convective islands of Maluku. Instead of relying on the short, fragmented records from ground stations, they utilized a global dataset called NASA POWER, which provides a consistent stream of meteorological information derived from atmospheric models and satellite observations. This dataset includes variables like air temperature, humidity, and wind speed, all of which are physically linked to how clouds form and how rain falls. The researchers hypothesized that by combining these long-term satellite readings with the limited ground data available, they could create a mathematical model capable of estimating rainfall with high accuracy, even in areas where weather stations are few and far between.

The team focused on a method that treats each month of the year as a unique event, recognizing that the forces driving rain in January are fundamentally different from those in July. They gathered rainfall observations from 66 official weather stations across the four regions and matched them against the satellite data for the period between 2011 and 2026. Before building their prediction tool, they first corrected the satellite data to ensure it aligned with the reality measured on the ground, a process akin to tuning a radio to remove static so the signal comes through clearly. Once the satellite data was adjusted, they used a statistical approach to find the relationship between the satellite's measurements of temperature, humidity, and wind, and the actual rainfall recorded by the stations. Because the influence of these weather factors changes with the seasons, they did not create a single formula for the whole year. Instead, they developed twelve separate sets of rules, one for each month, allowing the model to adapt to the shifting monsoons.

The results of this effort were encouraging. The model proved capable of estimating monthly rainfall across all four regions with a level of accuracy that the researchers described as good. In the region of Bali, the model's predictions matched the observed rainfall patterns with a correlation of 0.73, meaning the estimated values rose and fell in close step with the actual measurements. The Nusa Tenggara region showed similar strength, with a correlation of 0.70. Even in Java, which is home to complex urban environments and varied terrain that can make rainfall difficult to predict, the model achieved a correlation of 0.65. Perhaps most significantly, the model performed well in the data-sparse eastern regions of Maluku and Nusa Tenggara, where ground observations are scarce. In these areas, the model successfully captured the general trends of rainfall, suggesting that satellite data, once corrected, can serve as a reliable substitute for long-term ground records.

A closer look at the findings revealed how the different weather factors influenced the predictions. Across all regions, humidity emerged as the most consistent driver of rainfall; when the air was more humid, the model correctly predicted more rain. Air temperature also played a major role, though its effect varied by season and location. Wind speed showed a distinct pattern, often acting as a suppressor of rain during the dry monsoon season, particularly in the eastern islands where dry air from Australia sweeps across the archipelago. The researchers noted that while the model worked well for the broad monthly totals, it faced challenges in capturing the extreme, localized bursts of rain that occur in Java's urban centers or the highly convective storms of the Maluku islands. These areas involve complex interactions between the land, the sea, and the atmosphere that a simple linear model cannot fully resolve. Nevertheless, the overall performance confirmed that the approach is viable for regional planning.

The study concludes that this method offers a practical, low-cost solution for a country struggling with data gaps. By using the long, continuous record from NASA POWER and correcting it with the best available ground data, scientists and policymakers can now generate reliable monthly rainfall estimates for regions that were previously difficult to analyze. This capability is crucial for managing water resources, designing flood defenses, and planning agricultural cycles in a changing climate. The researchers emphasize that while their model is not a perfect crystal ball, it provides a solid foundation for evidence-based decision-making. They suggest that future work could refine these estimates by incorporating more complex climate drivers and testing the approach against advanced machine learning techniques, but for now, the study demonstrates that a transparent, month-by-month statistical framework can effectively illuminate the rainfall patterns of the Indonesian archipelago.