History

Satellite imagery has been used to estimate forest area change for decades.

In the past, because available satellite imagery over large areas was low resolution and infrequently updated, we often missed details on the ground like forest degradation or fast-moving land conversion.

Moreover, although we could map forest cover, we could only indirectly estimate the corresponding forest carbon stocks (the amount of carbon taken from the atmosphere and stored within the forest ecosystem) and emissions (carbon dioxide released into the atmosphere).

In the past, scientists have used Airborne Light Detection and Ranging (or LiDAR) to estimate the amount of carbon that tropical forests were storing, but it has been difficult to implement this process at scale or high-frequency due to cost.

New Discoveries

However, new breakthrough research from Arizona State University (ASU) helps break down these technological and cost barriers, using machine learning to combine the insights of ASU’s airborne LiDAR with Planet’s high-spatial and temporal resolution satellite imagery.

Focusing their efforts on Peru, ASU trained and tested their machine learning models using extensive sampling of forest carbon estimates from their Global Airborne Observatory’s in-country flight campaigns. The models then ingested Planet’s analysis-ready basemaps and other auxiliary data to comprehensively measure aboveground carbon of every hectare in Peru.

When compared to traditional approaches of using lower resolution satellite imagery to measure carbon emissions from deforestation and forest degradation, the results were staggering.

The model preliminarily found that carbon emissions caused by deforestation in Peru were 24.3 percent higher than those reported by previous annual assessments.

Moreover, when the researchers totaled quarterly carbon emissions over the same one-year period, preliminary results indicated an 83 percent increase in previously reported annual emissions.

These initial results indicate that we are vastly underestimating emissions from deforestation and forest degradation, simply because we have not been able to measure them accurately—until now.

The implications of these results can be immense. If we can more practically and cost-effectively understand forest carbon stocks and emissions, then we can more readily prioritize conservation efforts and inform climate policies.

For example, it becomes possible to detect mining activity in areas such as the buffer zone of Tambopata National Reserve more quickly than ever before.

And while Peru’s forests are some of the world’s most carbon-rich and biodiverse, they are not the only ones threatened by mining and other land use conversion, so time is of the essence to take action.