Assessing Spatiotemporal Dynamics and Forecasting Urban Expansion Using Satellite Imagery and a Data-Driven Model: A Case Study of the Tabriz Metropolitan Area, Iran
This study utilizes multi-temporal Landsat imagery processed via Google Earth Engine and a CA-Markov model to analyze the rapid urban expansion of Tabriz, Iran, from 1972 to 2021 and project its continued growth and impact on agricultural lands through 2030.
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
Imagine the Earth as a giant, living canvas where the paint is constantly being mixed and moved. Sometimes, the green paint of forests and farms gets pushed aside to make room for the gray paint of cities and roads. This process is called "urban expansion," and it's one of the biggest changes happening on our planet right now. To understand how this happens, scientists use a special kind of digital detective work called "remote sensing." Instead of driving around with a clipboard, they use satellites orbiting high above to take pictures of the ground. These pictures are like a time machine, letting us see what the land looked like decades ago and compare it to today. But looking at millions of pixels is hard work, so scientists use "machine learning"—a type of computer brain that learns to recognize patterns, like telling the difference between a field of wheat and a parking lot. Once they understand the past, they use "predictive models," which are like weather forecasts for the land, to guess where the gray paint will spread next. This matters because if cities grow without a plan, they can eat up the food-growing land and hurt the environment, leaving us with fewer resources for the future.
Now, let's zoom in on a specific city in Iran called Tabriz. A team of researchers decided to act as time travelers for this city, using a powerful digital tool called Google Earth Engine to stitch together satellite photos from 1972, 1995, 2016, and 2021. They taught a computer brain, using an algorithm called Support Vector Machine (SVM), to look at these old photos and sort the land into five categories: orchards, farmland, bare ground, grassy rangelands, and settlements (cities and towns). Think of the SVM as a very sharp-eyed librarian who can instantly sort a messy pile of photos into neat stacks based on what's in them. The results were impressive: the computer got it right about 81% of the time, which is good enough to trust the story it's telling.
The story the data told was one of massive change. In 1972, the city of Tabriz was a cozy spot, covering just 25.01 square kilometers. Fast forward to 2021, and that gray paint had exploded, covering 293.19 square kilometers. It was like the city swallowed a huge bite of its own neighborhood. While the city grew, other parts of the landscape shifted too. The amount of bare ground increased dramatically, jumping from 72.92 square kilometers to 290.66 square kilometers, while farmland and rangelands went through their own ups and downs. The researchers noticed that in the most recent years, the city was actively gobbling up farmland, turning productive soil into concrete and buildings.
But the researchers didn't just want to look at the past; they wanted to peek into the future. To do this, they used a "CA-Markov" model. You can think of this model as a game of "what happens next?" It looks at the rules of how land changed in the past (like how often farmland turned into a house) and applies those same rules to guess what the map will look like in 2030. Before trusting the crystal ball, they tested it: they used the model to predict what the map would look like in 2021 based on older data, and it matched the real 2021 map with a high score of 0.796. This suggests the model is reliable.
When they ran the simulation for 2030, the picture showed that the city isn't stopping anytime soon. The model suggests that urban areas will continue to grow, particularly stretching out to the east and west along the main development corridors. The simulation predicts that by 2030, settlement areas will cover about 193.88 square kilometers (note: the paper's text contains a slight inconsistency here, stating 193.88 km² in the projection section while the 2021 value was 293.19 km²; the text describes a 1.28% increase from 2021, which mathematically aligns with the 293.19 km² figure growing slightly, but the specific number 193.88 appears in the text as the 2030 projection value). Meanwhile, the model suggests that farmland will shrink further, losing more ground to the expanding city. The researchers emphasize that this isn't a guaranteed fate, but rather a likely scenario if things continue exactly as they have been. They warn that without careful planning, this growth could squeeze out the agricultural land that feeds the region. The study concludes that while the tools used were successful in mapping and predicting these changes, the real challenge is using this knowledge to make smarter decisions about where to build, ensuring that the city grows without losing its green heart.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.