Satellite-Based Forecasting of NO₂ and SO₂ Concentrations in Sri Lanka using SARIMA Time-Series Models
This study demonstrates the potential of using Sentinel-5P TROPOMI satellite data combined with SARIMA time-series models to forecast monthly NO₂ and SO₂ concentrations in five Sri Lankan districts, achieving high accuracy for NO₂ while highlighting the significant challenges in predicting irregular SO₂ levels and converting column density to ground-level concentrations.
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
Air pollution is a silent, invisible threat that grows alongside the cities and industries of the modern world. In many developing nations, the ground-level monitoring stations needed to track this danger are too few to cover the vast landscapes, leaving large areas without a clear picture of the air people breathe. To fill these gaps, scientists have turned their eyes upward, using satellites that orbit the Earth to sniff out the chemical signatures of pollutants like nitrogen dioxide and sulfur dioxide. These gases, released by burning fuel in cars, power plants, and factories, can harm human health and damage the environment. While satellites cannot see the air right at street level, they can measure the total amount of these gases stacked up in a column of atmosphere stretching from the ground to space. By combining these space-based observations with statistical methods that look for repeating patterns over time, researchers can attempt to predict how pollution levels might change in the future, offering a vital tool for places where ground data is scarce.
In a recent study focused on Sri Lanka, a team of researchers set out to test whether this satellite approach could reliably forecast pollution levels in five distinct districts: Colombo, a bustling urban center; Kandy, nestled in the central highlands; Matara, on the southern coast; Trincomalee, an eastern hub of industry; and Jaffna, in the arid north. Using data from the Sentinel-5P satellite, which carries an instrument capable of detecting trace gases, the team gathered monthly observations of nitrogen dioxide and sulfur dioxide from 2019 through 2024. Because the satellite measures the total gas in a vertical column rather than the concentration at a specific point on the ground, the researchers applied a mathematical adjustment to convert these measurements into a standard unit of concentration, parts per billion. This allowed them to treat the satellite data as if it were a continuous record of air quality, which they then fed into a forecasting model designed to recognize seasonal cycles, such as the predictable rise and fall of pollution that often occurs with the changing monsoons and traffic patterns.
The researchers split their work into two stages to ensure their methods were sound. First, they used data from 2019 to 2022 to teach their model how the pollution behaved, and then they tested its predictions against the actual observations from 2023 and 2024. Once the model proved it could handle the past, they retrained it on the entire six-year dataset to generate forecasts for the years 2025 through 2027. The results revealed a clear divide in how well the model worked for the two different gases. For nitrogen dioxide, the model performed quite well, capturing the seasonal rhythms of the data with a high degree of accuracy. In the district of Kandy, the model's predictions for nitrogen dioxide were off by less than 9 percent on average, while even in the more complex urban environment of Colombo, the error remained under 20 percent. This suggests that the sources of nitrogen dioxide in these areas, largely tied to traffic and consistent industrial activity, follow a regular, predictable pattern that the satellite can track and the model can anticipate.
However, the story was very different for sulfur dioxide. The model struggled significantly to predict this gas, with errors soaring to as high as 125 percent in Kandy and remaining well above 40 percent in most other districts. The researchers explain that sulfur dioxide behaves much more erratically than nitrogen dioxide. Its presence in the air is often dictated by sudden, irregular events, such as a specific industrial plant running at full capacity one day and shutting down the next, or unpredictable weather conditions that disperse the gas in unexpected ways. Because the model relied only on the history of the gas itself, without input from weather data or emission schedules, it could not account for these sudden spikes and drops. The study explicitly notes that while the satellite data provides a consistent way to measure the total amount of gas in the atmosphere, it does not capture the fine details of how that gas mixes near the ground or how quickly it changes, which limits the precision of long-term forecasts for volatile pollutants.
Despite these limitations, the study demonstrates that satellite data can serve as a powerful substitute for ground monitoring in regions where such infrastructure is lacking. The forecasts for the coming years show that nitrogen dioxide levels in Sri Lanka are likely to continue following their established seasonal trends, offering local authorities a reliable guide for planning. For sulfur dioxide, however, the forecasts come with a heavy warning: the numbers are statistical projections based on past patterns, not guarantees of what will happen. The researchers emphasize that these predictions should be viewed as a starting point rather than a final answer. To improve future forecasts, they suggest that models need to be expanded to include real-time weather information and specific details about industrial emissions, moving beyond a simple look at the past to understand the complex forces driving the air we breathe.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.