Seasonal satellite observations improve shrub-encroachment mapping
Researchers at the Aerospace Information Research Institute of the Chinese Academy of Sciences have developed a method to map shrub cover in the grasslands of Inner Mongolia by using satellite imagery collected throughout the growing season. The approach is intended to distinguish spreading shrubs from dense grass, a problem that can be difficult to solve from a single satellite image.
Shrub encroachment is a consequential ecological change in semi-arid grasslands. As woody plants expand into areas dominated by grasses, they can alter plant communities and affect the forage available to livestock. Detecting the change early is especially challenging because young shrubs may be sparse and small, leaving individual satellite pixels to capture a mix of shrubs, grass and bare soil.
Combining fine detail with seasonal change
The team tested the approach near Xilinhot in northern China’s Inner Mongolia Autonomous Region. It paired PlanetScope imagery, which has a spatial resolution of 3 meters, with a U-Net deep-learning model. The researchers checked their shrub-cover estimates against drone surveys and very-high-resolution imagery.
Fine spatial detail was important because it enabled the model to identify the shapes of shrub patches. But the study’s central advantage came from adding time. The researchers incorporated observations at 15-day intervals over the growing season, allowing the model to use differences in the timing of growth between shrubs and grasses.
Compared with estimates based on imagery from a single period, the seasonal observations increased the proportion of variation in shrub cover explained by the model, measured as R², by about 17%. The method also retained fragmented shrub patches more effectively than approaches that estimate cover one pixel at a time.
Why early detection remains difficult
The results do not eliminate the hardest part of the problem. Estimates were less reliable for plots excluded from model training, and sparse shrubs remained difficult to identify. That limitation matters because scattered shrubs are often the earliest sign of encroachment and the point at which land managers may have the greatest opportunity to monitor or respond to changing vegetation.
Still, the work shows how repeated observations can add useful ecological information beyond a sharper single image. In landscapes where grass and shrubs can appear similar at one moment in time, their seasonal growth patterns can provide a clearer signal. The findings were published in the International Journal of Digital Earth.
A monitoring tool for grassland management
For land managers, more reliable shrub-cover maps could support efforts to track changes in semi-arid rangelands, including areas where forage availability is a concern. The study also illustrates a broader use of high-frequency commercial satellite imagery: combining spatial resolution, time-series data and machine learning to monitor fragmented vegetation patterns that are difficult to identify with conventional approaches.
The next challenge will be improving performance where shrub cover is low. The researchers note that stronger estimates in those early, sparse stages would make the system more useful for identifying encroachment before shrub patches become extensive.
This article is based on reporting by Phys.org. Read the original article.
Originally published on phys.org
