Spatiotemporal Dynamics and Projection of Forest Cover Using Remote Sensing and Machine Learning in Baishari Bangdopa, Bangladesh
This study utilizes remote sensing and machine learning to reveal a 74% decline in forest cover in Baishari Bangdopa, Bangladesh, between 1988 and 2025, and projects that future forest stability depends critically on management interventions, with strict protection potentially tripling current cover by 2050 compared to a near-total collapse under high deforestation scenarios.
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
Forests are more than just collections of trees; they are living systems that regulate the climate, hold soil in place, and provide homes for countless species. When these systems are broken apart into smaller, isolated pieces, they lose their ability to function properly, a process known as fragmentation. This happens when human activities like farming, logging, or building roads cut through continuous woodlands, leaving behind scattered patches that are vulnerable to wind, fire, and invasive species. Understanding how forests change over time and predicting what might happen to them in the future is essential for protecting biodiversity and ensuring that these ecosystems can continue to support life. Scientists have long used satellite images to watch these changes from above, but predicting the future requires more than just looking at the past. It demands tools that can recognize complex patterns in how land use shifts from year to year, allowing researchers to test different management strategies before they are implemented on the ground.
In the rugged, hilly terrain of southeastern Bangladesh, a specific forest landscape known as Baishari Bangdopa has become a critical focus for such study. This area, which straddles the borders of several local districts, is a biodiversity hotspot and a vital corridor for wildlife, including elephants and leopards. However, it has faced intense pressure from shifting cultivation, illegal logging, and the expansion of settlements. To understand the true scale of this degradation and to explore possible futures, a researcher named Niamjit Das conducted a detailed analysis of the region. By combining decades of satellite imagery with advanced computer models, the study paints a stark picture of what has been lost and offers a clear view of the choices that lie ahead for this fragile landscape.
The investigation began by looking at the history of the land. The researcher gathered satellite images taken from 1988 up to a recent point in 2025. These images allowed for a direct comparison of the forest cover over nearly four decades. The results revealed a dramatic transformation. In 1988, the forest covered 4,867 hectares. By 2025, that number had plummeted to just 1,270 hectares, representing a loss of roughly 74 percent of the original forest. As the trees disappeared, the landscape did not simply become empty; it was replaced by shrubland and bare earth. The area covered by shrubs nearly doubled, and bare land also expanded significantly. This shift indicates that the forest is not just shrinking but is being actively converted into other types of land use, primarily driven by agricultural expansion and the shortening of fallow periods in traditional farming practices.
To ensure these observations were accurate, the study employed sophisticated computer algorithms designed to distinguish between different types of land cover. The researcher tested three different methods: one that builds many decision trees to reach a conclusion, another that finds the best boundary line between different categories, and a third that learns from sequences of data over time. The method that learned from sequences proved to be the most effective, correctly identifying the land cover with an accuracy exceeding 93 percent. This high level of precision gave the researcher confidence that the maps created were a true reflection of reality, allowing for a reliable assessment of how the forest has changed.
Beyond just measuring the total area lost, the study examined how the remaining forest is arranged. A healthy forest is often a large, continuous block, but the analysis showed that the Baishari Bangdopa landscape has become increasingly fragmented. The number of separate forest patches has more than doubled, while the average size of each patch has shrunk by over 60 percent. Perhaps most concerning is the loss of "core" forest—the deep interior areas far from the edges where sensitive species can thrive. In 1988, 72 percent of the forest was this protected interior type. By 2025, that figure had dropped to just 22 percent. This means that almost all of the remaining forest is now exposed to the harsh conditions found at the edges, such as higher temperatures and invasive plants, making it much harder for the ecosystem to recover or support wildlife.
With a clear understanding of the past, the researcher used the most accurate computer model to simulate what might happen in the future, looking ahead to the year 2050. Three different scenarios were tested to see how different management choices could alter the outcome. In the first scenario, which assumes that current trends continue without significant change, the forest cover is projected to stabilize at a mere 8 percent of the landscape, leaving behind only scattered, isolated fragments dominated by shrubs and agriculture. In a second scenario, where strict protection measures are enforced and local communities are empowered to manage the forest, the forest cover could stabilize at 33 percent. This would nearly triple the amount of forest compared to the first scenario, preserving vital corridors for wildlife. The third scenario, representing a worst-case situation with unchecked deforestation, projects a near-total collapse of the forest, reducing it to less than 1 percent of the landscape by 2050.
These simulations highlight a crucial truth: the future of the Baishari Bangdopa forest is not predetermined. The data suggests that the landscape is at a tipping point where human decisions will determine whether the forest survives or disappears entirely. The study concludes that without intervention, the region faces continued degradation and the potential loss of its ecological functions. However, the possibility of stabilizing and even regenerating the forest exists if strong conservation policies are implemented and local communities are involved in stewardship. The research provides a replicable framework for other regions facing similar pressures, demonstrating that by combining satellite monitoring with predictive modeling, it is possible to see the consequences of our choices before they happen and to act accordingly to protect the natural world.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.