Influence of Geographical Location on the Temporal Dynamics of Reservoir Storage: A Case Study of Bogotá, Colombia

Fresh water is the lifeblood of modern cities, yet the systems that capture and store it are increasingly fragile. As urban populations swell and weather patterns grow more erratic, the old habit of managing water based on past averages is no longer enough. The challenge lies in predicting how reservoirs will behave when the climate shifts or when human needs change. To do this, scientists often turn to time-series analysis, a method that looks at how a single variable, like water volume, changes over time to find hidden patterns. These patterns can reveal whether a system follows a steady, predictable rhythm or if it is being pushed by irregular forces, such as sudden droughts or deliberate human decisions to release water for flood control. Understanding these rhythms is critical for cities that rely on these reservoirs, because a failure to anticipate a drop in water levels can lead to severe shortages for millions of people.

In 2024, the city of Bogotá, Colombia, faced a stark reminder of this vulnerability. A severe drought linked to a global climate phenomenon known as El Niño drove the city's reservoirs to critically low levels, forcing authorities to ration water for nearly ten million residents. This crisis highlighted a gap in the tools available to water managers. While some models try to simulate the complex physics of rain and evaporation, others rely on statistical patterns found in historical data. The researchers behind this study, a team from several universities in Bogotá, asked a specific question: does the geographical location of a reservoir determine which statistical tool is best for predicting its future? They focused on four key reservoirs that supply the city: Chuza, Neusa, Sisga, and Tomine. These bodies of water are not identical; they sit in different landscapes, receive rain in different patterns, and are managed for different purposes. The team hypothesized that a single modeling approach would not work for all of them. Instead, the best prediction method would depend on the unique "signature" of each reservoir's location and its interaction with human management.

The researchers began by examining the history of water storage in these four reservoirs. They found that the reservoirs behaved in two distinct ways, dictated largely by where they are located. The Chuza reservoir, situated in a national park on the eastern slope of the Andes, follows a simple, single-peak rainfall pattern typical of that region. It fills up during one main rainy season and empties during the dry season. Because its behavior is driven almost entirely by this natural cycle, the researchers found that a standard statistical model, which looks for repeating seasonal patterns, worked well for it. However, the other three reservoirs—Neusa, Sisga, and Tomine—are part of a northern system closer to the city's urban core. These reservoirs experience two rainy seasons a year, creating a more complex, double-peak rhythm. Furthermore, because they are used for flood control and to supply water to the city, their levels are frequently adjusted by human operators who open and close gates. This human intervention disrupts the natural rhythm, adding a layer of complexity that a simple seasonal model cannot easily capture.

To handle this complexity, the team developed a hybrid approach for the northern reservoirs. They combined a method that breaks down the data into its repeating seasonal parts with another method that analyzes the remaining irregular fluctuations. Think of it as separating a song into its melody and its background noise; the first part captures the predictable, repeating climate cycle, while the second part accounts for the unpredictable, human-driven changes. When they tested this hybrid model against the traditional one, the results were striking. For the three northern reservoirs, the hybrid approach produced very accurate forecasts, with errors ranging from roughly 5 to 9 percent. In contrast, the traditional model for the Chuza reservoir, which relies on a single seasonal pattern, produced a much higher error rate of nearly 29 percent. However, the researchers were careful to explain that this high error was not a failure of the model itself. The model had correctly identified the normal seasonal pattern, but the 2024 drought was an extreme event that pushed the reservoir far outside its usual historical range. The model was right about the rules, but the weather broke the rules.

The study also looked closely at how well the models performed during the actual 2024 crisis. For the northern reservoirs, the hybrid models kept their predictions within a realistic range of uncertainty, even as the drought intensified. This suggests that by separating the natural climate signal from human operational decisions, the model could better absorb the shock of the drought. The Chuza reservoir, however, saw its water levels drop so low that they fell below the model's predicted safety margins. This happened because Chuza depends on a single rainy season; when that season failed, there was no second chance to recover, unlike the northern reservoirs which benefit from two rainy seasons and a more flexible management system. The researchers concluded that the high error for Chuza was not because the model was wrong, but because the reservoir was subjected to an extraordinary event that no historical model could fully anticipate.

Ultimately, the paper suggests that there is no single "best" way to predict water storage for an entire city. The geography of a basin defines its rainfall regime, and the operational role of the reservoir defines how humans interact with it. For a reservoir like Chuza, which follows a simple natural rhythm, a standard seasonal model is sufficient. For reservoirs like Sisga and Tomine, which are subject to both complex climate patterns and active human management, a hybrid approach that separates these two influences is far more effective. The study offers a practical tool for water managers, showing that by tailoring the prediction method to the specific physical and operational reality of each reservoir, cities can better anticipate crises. The work does not claim to solve the problem of climate change or to predict extreme droughts with perfect certainty, but it does provide a clearer, more reliable way to understand the dynamics of water storage in a changing world. By recognizing that location and human intervention shape the data, managers can plan with greater confidence, distinguishing between a system that is behaving normally and one that is facing an unprecedented challenge.