Measuring Azuero

Testing the Biome App in the Field

January 18, 2023

Last September, we traveled to the Azuero peninsula in Panama. We went to measure trees. We went to deepen our sense of place.

I am a forest ecologist whose research skews toward developing high tech methods to detect forest change. This year at Earthshot Labs I joined a team of technologists – a software engineer, a user interface designer, and two computer vision experts – to change the way we measure forests.

Our work concerns locales that to us are exotic, but much of it is mediated through screens, be it writing code, creating models, examining maps, analyzing satellite imagery, or sitting in zoom calls with collaborators. Most of our interactions, with each other and with the places we try to understand, come through text, images, and graphs. But to know a place, to gain a true sense of its flow and process, requires being present within it. In September, we traveled to the Azuero peninsula in Panama. We went to measure trees. We went to deepen our sense of place.

Improving Carbon Sequestration Estimations

Earthshot is developing projects to regrow forests throughout the tropics. Every project has an associated carbon projection–an estimate of the amount of carbon that will be sequestered in each year through the lifespan of the project. Young forests have small trees that grow slowly. As individual trees grow, they have more machinery to fix carbon from the atmosphere and thereby grow faster. But eventually neighboring trees begin to compete for limited resources like light and water and the forest stand is constrained to a maximum size. This maximum is dependent both on the natural environment and the species of trees that make up the forest.
Since the growth rate of a forest changes as it matures, accurate representations of carbon sequestration require measurements of forest biomass at a variety of stand ages. Some forest types, like deciduous forests in the eastern United States, are extremely well studied and this data is easily acquired. Others, such as humid tropical forests in Central America, have little data available in the scientific literature. This lack of data translates directly into uncertainty in our carbon projections, which in turn translates to investment risk in the reforestation project. Our project on the Azuero peninsula in Panama spans at least three different forest types, and the available literature was ambiguous about precisely what kind of forests were measured.
Traditional tools for forest inventory:
waterproof notebook and plastic tape.
Biome app
A visit to the Azuero forests was needed, with the goal of measuring stands like those in our project areas ourselves.

Another goal was to test the Biome app, a new tool being developed at Earthshot for conducting forest inventories. The app uses a cellphone’s camera, lidar sensor, and gyroscope within an Augmented Reality (AR) framework to allow a user to establish a forest inventory plot and measure the trees within it.  It calculates the tree diameter directly from a photo taken within the app, measures heights using the gyroscope and a few screen taps, tracks all of the tree locations within the plot, and then exports the data in a useful format eliminating transcription time and error.

We wanted to make certain that the underlying estimation models were accurate, which meant measuring the same plots using the app as well as traditional forestry methods. We also wanted to evaluate Biome under actual use conditions to inform the trajectory of development and make certain that it would meet the needs of the scientists who will use it.
But more, the embodied experience of the forest and landscape we are all working to restore creates a connection and sense of place that is otherwise impossible to replicate.

This translates into better intuitions about the system and the problems it faces.
Red highlights the Azuero Corridor, a region Pro Eco Azuero has identified as being crucial to restoring the wild range of the critically endangered Azuero Spider Monkey.


We first met up with Andrew Coates, our colleague and host, in Gamboa. The town of Gamboa is famous amongst ecologists for Barro Colorado Island, the most intensively studied tropical research site in the world. After a brief stay to introduce us to both the tropical forest and scientists with the Smithsonian Tropical Research Institute, we travel south to the Azuero peninsula.
Azuero is remote and rural. The landscape is hilly and steep, the result of a multitude of small volcanic uprisings. The soil is clay, and from May through November during the wet season, certain areas are impassable by car. Transportation is by horse or going the long way around by boat. The majority of the primordial forests have been cut to make way for cattle.
Decades of cattle grazing have created a stepped terrace effect on the hillsides.
Native forests persist in narrow bands along drainages and streams, supported by the water and providing natural posts for boundary fences. Capuchins, howler monkeys, and critically endangered spider monkeys live in these fragmented forests. Individual ranchers own upward of a thousand hectares of pastureland, maintained as a grassland through grazing and fire.  But the ranchero way of life does not appeal to many in the younger generations, and the possibility of carbon credits provides an alternative.

Our project areas are designed to create a biological corridor linking a diverse habitat. The corridor stretches across the tropical rainforest in Cerro Hoya national park in the southwest to the dry tropical forests and mangroves in the southeast coast, running through the humid tropical forests in the central highlands.
Within the next few years Earthshot hopes to plant 10,000 hectares of forest across the peninsula. These millions of trees will come from a network of nurseries. By the time of our visit, Roberto Machezek had set up three main nurseries run by Earthshot Labs, Panama. He’s coordinating growing 86 different species of native trees to be planted in Earthshot projects.  

But these nurseries will never be able to meet the intense demand for saplings reforesting the Azuero peninsula will require. And so the bulk will come from a multitude of micro-nurseries, saplings grown, predominantly by women, with tools and techniques provided by nursery staff in hundreds of local backyards and then purchased by Earthshot for use in reforestation.
One of Earthshot Panama's Hub nurseries, in Venao

Forest Inventory

To properly calibrate our carbon projections, we need to inventory forests with a known age. We could identify areas using the historical Landsat satellite record, but a history of mismanagement during privatization in the 1990s and perennial cloudiness during the wet season makes these estimates less reliable. Instead, we rely on conservationist landowners who started to regrow forests in the past few decades and who are excited about the prospect of reforesting the peninsula.
Vern Scholey has been directing tuna research at the Achotines Laboratory since 1985 and has been conserving and regrowing forests in the southwest Azuero peninsula since the mid 2000s. He met us at his barn to show us the regeneration work he is doing at his property. We followed him past different planting regimes, including a commercially valuable rosewood plantation, to the top of a central hill. From this vantage, he pointed out different sections of the surrounding forest mosaic as I drew a map and annotated it with ages and land use. Many areas are mixed-use silvopasture, where cattle are allowed to graze between trees after they become established. Others are fenced forests. Some sections are planted with a mixture of timber species and others are allowed to regenerate naturally.
A plantation of rosewood (Cocobolo) trees on Verne's property
An area of naturally regenerated forest near a creek
In Eco Venao, an eco lodge known for surfing and spider monkeys, Nicolas Nickson-Casal guided us through old cattle lands that have been regenerating into forests over the past two decades. A large area of mature secondary forest provided a baseline for what the future could hold for the rest of the land. Other parts of the property reflect a variety of informal planting experiments, fruit trees within forests, firebreaks, fences to prevent grazing, or some combination.

During early restoration, they planted several hectares with a mixture of teak, expecting to thin the valuable hardwood as the forest regrows. But teak exhibits allelopathy; chemicals in their fallen leaves remain in soil and inhibit the growth of other plants around them. Instead of a tropical forest with a few valuable trees that could be extracted, this section of property became a sparse teak savanna. In the past year, Earthshot and Pro Eco Azuero have planted more saplings, starting the process of converting fields to forests.
The team compares manual measurements to the Biome app.
At each of the sites we traveled through the mosaic of forest stands to measure as much variability in forest age, planting regime, and land use history as possible.  Within each different forest type we established multiple plots, each 10 meters in diameter, and recorded the diameter, height, and species of every adult tree inside the plot.  In each locale, we used two methods for measurement: both the traditional forestry method, and Earthshot's Biome app, which uses computer vision, augmented reality, and machine learning to automate much of the process.

Measuring both ways, we found, were essential for confirming Biome’s accuracy. We also recorded the number of saplings within the plot, and took photos to record ground cover, canopy closure, and vegetation density. Knowing the size and species of trees allows us to estimate their biomass, and consequently, the amount of carbon stored within them. Having measurements from a variety of forest stands of a known age and history enables us to make predictions about how much carbon will be sequestered in the future if we planted a similar stand today.
Measuring individual trees obviously cannot be done from a desk a continent away, though we could have asked local people to take these measurements, which we will do as we expand and scale. But physically doing a task is essential when faced with uncertainty. Our inventory methodology specifies 10 meter plots, but because much of the regeneration plots are very sparse and savanna-like, we discovered they may not be large enough. At the same time, the mature forest sections are quite dense and so the two use cases may require different protocols. In the Biome app, we tested several methods of measuring height, one that required the phone to be kept at a stable height and position and another where the user touches the tree and then moves back to measure the distance. Either works fine in a park, but neither are suitable in a treacherous forest floor full of down limbs, spiny trees, and loose rock. Having technologists do the measurements enabled us to viscerally understand the inherent difficulties of the work, have empathy with those who will use our methods and technologies, and create designs and methods that are tolerant of error.
But more, the embodied experience of the forest and landscape we are all working to restore creates a connection and sense of place that is otherwise impossible to replicate. This translates into better intuitions about the system and the problems it faces. One morning, as we walked to measure trees, we watched as a troop of spider monkeys crossed the trail ahead of us. Seeing these beings watching us, waiting for us to pass, and believing that our actions there can and will help them deepens our sense of purpose in a way nothing else can.
Enjoying a traditional Panamian breakfast at a local farm before a day of field work
Marking plot boundaries on a steep hillside. The task was made more difficult due to the thorny trees covered with poisonous ants.
In March, the entire Earthshot team will visit Panama and, I hope, reap the benefits of being grounded in a sense of place. Afterward, some of us will stay for several more weeks to collect additional data from the forests. We expect to measure along the gradient of forest types that occur on the peninsula, from rainforest to dry tropical forests. We also hope to measure more sites of known age and also image them using drone technology to investigate the accuracy of biomass and forest structure estimates from that imagery. Additionally, we will perform a rigorous accuracy assessment of the Biome app to ensure we can rely on its measurements. Finally, we will expand our measurements of biodiversity throughout our study sites as we develop the methods to assess how well our projects meet our high standards for regenerating thriving forests.

Read More