A 2026 study examined 135 bird species over 16 years in Peru’s Amazon and found that satellite measurements of forest conditions predicted bird occurrence better than traditional habitat categories, suggesting remote sensing could improve biodiversity monitoring across vast tropical forests

Satellite observations of forests may offer a more effective way to predict where bird species occur in the Amazon than traditional habitat categories, according to new research. But the study also found a major obstacle to monitoring biodiversity: many tropical birds are simply too difficult to detect during field surveys.

Researchers analyzed 3,129 bird surveys conducted between 2004 and 2020 in the Tambopata region of southeastern Peru. They combined field observations with data from NASA and USGS Landsat satellites to examine whether vegetation information could help predict the distribution of bird communities across the Peruvian Amazon. The research led by Environmental Scientist Andrew Christopher Slater and colleagues was published in the peer-reviewed journal PLOS ONE on June 3, 2026.

The researchers found that models based on satellite-derived measurements of vegetation and forest conditions performed better than models based on traditional habitat labels, such as primary forest, secondary forest and agricultural land.


The findings suggest that satellites could become an increasingly useful tool for tracking biodiversity across vast and difficult-to-access tropical forests.

Satellite data performed better

The team examined 135 frequently recorded bird species and compared different ways of predicting whether they occurred at particular locations.

One approach relied on traditional habitat descriptions. The other used satellite measurements, including how vegetation reflected different wavelengths of light and indicators related to greenness, moisture and forest structure.

The satellite-based model achieved an average predictive score of 0.68, compared with 0.58 for the habitat-category model, according to the study. It also successfully achieved a high level of predictive accuracy for 49 species, compared with just 20 species using traditional habitat categories.

In simple terms, the researchers found that the detailed environmental information captured by satellites was generally more useful for predicting where individual bird species were likely to occur than broad habitat labels.

Traditional categories can oversimplify forests, the researchers noted. Two areas classified as the same type of forest may still differ in important ways, including vegetation structure, moisture and canopy condition.

Satellite data, by contrast, can capture these environmental differences as gradual changes rather than placing landscapes into a limited number of categories.

More than 3,000 surveys over 16 years

The research drew on bird surveys conducted at 637 stations in the Madre de Dios region near the Tambopata River and Puerto Maldonado. Researchers and volunteers used two main methods: point counts, in which birds were recorded visually or by their calls, and mist nets, which capture birds for identification.

The surveys identified 358 bird species, although the researchers included 138 species with enough observations for their models. The final analysis focused on 135 species.

To match the field observations with environmental conditions, the researchers used Landsat satellite data collected throughout the study period. They examined several vegetation measures, including the widely used Normalized Difference Vegetation Index (NDVI), as well as indicators of moisture, vegetation health and forest structure.

The biggest challenge: Birds are hard to find

However, the study also revealed an important limitation. Even with thousands of surveys, many bird species had very low probabilities of detection in a single survey.

Morning surveys generally performed better than afternoon surveys, while point counts produced higher overall detection rates than mist-net surveys. Still, detecting the full bird community at any single location remained extremely difficult.

This matters because scientists cannot reliably monitor changes in biodiversity if they fail to observe a large proportion of the species that are actually present.

The researchers found that the model's predicted number of species at survey locations was broadly similar to the observed number. But repeated surveys suggested that field observations were still missing many species present in those areas.

That limitation became especially clear when researchers examined whether their survey system could detect population changes over time.

On average, there was only a 20% probability of detecting a 50% reduction in occupancy across all the species studied. Only five of the 135 species had at least a 70% chance of revealing such a decline under the existing survey design.

Satellites could help monitor vast forests, but cannot replace fieldwork

The findings do not mean satellites can directly see birds from space. Instead, satellites measure characteristics of forests and vegetation that may be connected to the environments different bird species prefer. Scientists can then combine those environmental measurements with field surveys to predict where species are likely to occur across much larger areas.

That could be particularly valuable in tropical forests, where biodiversity surveys are expensive, time-consuming and difficult to conduct across enormous landscapes.

But the researchers emphasized that satellite technology alone is not enough. The quality of predictions still depends heavily on the quality of field data used to train the models. If surveys repeatedly miss species, scientists may struggle to accurately understand how bird communities are changing.

The study concluded that combining remote sensing with well-designed, repeated field surveys offers a potentially scalable approach to monitoring biodiversity in tropical forests. However, improving the way birds are surveyed will be essential if researchers want to reliably detect changes over time.