The role of Carbon in the sustainability equation
The reduction of carbon dioxide in the air down to levels that were common before industrialization and urban societies is a key axis of Sustainable Development. Forests are one of the most common and efficient way to store carbon through photosynthesis that captures carbon dioxide, releases oxygen while storing the carbon in the tree at all levels: there is carbon on the wood (the most dense part of the tree), on the roots but also on the branches and leaves.
The longer the tree stands, the longer the carbon stays captured. Notwithstanding, using forestry products is also useful for society and the environment:
- some of the carbon stays captured for decades (e.g. : to make furniture or housing materials);
- some of the carbon will be used, then partially recycled and partially incinerated or handled in a waste processing facility (e.g.: to make paper and cardboard). In this process some of it will be released back to the atmosphere;
- some of the carbon will return swiftly to the atmosphere through burning (e.g.: to cook and heat) saving less renewable sources of carbon.
In any case, knowing and monitoring the terrestrial carbon cycle is of major importance to know how efficient the carbon capture process is. A quantitative, auditable estimate of carbon stored in forests, split across multiple forms and layers and how much value it represents is the foundation of any proposed “environmental services” that will be paid by stakeholders.
Albatroz Engenharia introduces its own evidence-based method to estimate the carbon content of forests.
The need for sampling and sampling strategies
Forests are large, sparsely populated, often remote areas where it is difficult and expensive to send labourers. Moreover, revenues from most forests come in lumps with lapses of decades between successive revenue coming from the same plot. Therefore, all operational expenditure must be decided with utmost parsimony.
A mix of old established techniques and state-of-the art technology prove its value to address such a large fan of social, technical, environmental and economic challenges both on the field and in the computer  and across many types of forests.
That’s how sampling enters the scene as a key approach to efficiently acquire data from the field. While there are well-established strategies to determine the sample-to-population ratio and the homogeneity of forest stands to be represented by an individual sample, Albatroz proposes to take advantage of modern methods to
a) combine ground surveys with airborne surveys to better estimate horizontal and vertical features, respectively;
b) enlarge the samples, reducing sensitivity to outliers and approximations;
c) use nested sampling to corroborate, at least partially, different survey and sampling methods;
d) make cross-section surveys to highlight heterogeneity, hence checking if the partition of the stands fits the reality of the forest.
The results could be validated by expert users using different methods to compute the same variables to highlight constraints and estimate sampling approximations. In the end, results coming from multiple methods are likely closer to reality than from a single method. Counting the number of trees offers a simple example of the benefits of the approach. In dense pine forests, one counts tree trunks to estimate the number of trees from the ground and one counts canopy cones to estimate the number of trees from the air. It is found that the ground estimate is usually higher than the airborne estimate due to dominant trees hiding shorter ones and turning them invisible from the air. One could say that this might be corrected by a given offset – say 8%. However, a custom sampling for ground and airborne surveys would modulate this offset, supported on actual field data, and adjust it from 5% to 15% depending on the stands, yielding a much better estimate while offering a better knowledge of where these subdued trees exist as an added value product.
Currently, most ground and airborne surveys both require putting people on the ground near or inside the forest stands. Therefore, the challenge is to make the most efficient, simultaneous use of the brain and brawn available on site.
Traditional sampling requires manual measurements using tree callipers on typically 400 m2 to 1000 m2, very rarely more than an acre (approx. 4047 m2). Depending on the tree species this represents a few dozens, rarely more than one hundred trees. If one takes at least 2 minutes to move to a tree and collect data, this means that any traditional survey takes a minimum of 2 hours per sampled plot.
Handheld mobile laser scanner, which is an application of LiDAR technology, can survey such an area in a period shorter than half an hour, provided shrubs do not clutter the visibility between trees. Therefore, it makes sense to a) make the LiDAR scan for a larger sample and b) make the calliper method to a subsample of said sample to corroborate diameter at breast height (DBH) estimates and, in some cases, tree height estimation.
Differences in methods could be propagated as systematic corrections, applicable only to this sample. Moreover, pre-existing surveys made with calliper may be compared with current ones made with LiDAR provided the estimated correction coefficient is applied.
The next step of inference is to compare the ground with unmanned airborne surveys using UAS (unmanned aerial systems, also known as drones) to estimate the number and height of the trees in the previous sample, as described earlier. Since UAS flight is usually more uncluttered than walking in the woods, it makes sense to enlarge the sampled plot while keeping the ground samples (both LiDAR and measurement) in the survey to act as an approximation to a “ground truth”.
Using other airborne technologies, either with UAS, helicopter or aeroplane, it is possible to cover tens or hundreds of hectares in one single day, thus offering so much more field data – capturing the inevitable modulations and variations of reality – than what would be coming out of a few isolated ground samples performed with the most skilled staff.
Albatroz has considered the use of all terrain ground vehicles to perform ground inspections but its efficient use is yet to be demonstrated.
The first image shows the ground sample for ground LiDAR (red) with the sub-sample for calliper (gold)…
…while the second image shows the two samples as a part of the wider sample for airborne survey: Calliper to ground LiDAR ratio could be 5 to 10 while ground to airborne LiDAR ratio usually exceeds 10-fold and if airborne LiDAR is replaced with image technology that allows flying farther away from the canopy, that ratio could exceed 100.
Albatroz has also compared airborne acquisition with satellite data in the special case of post-wildfire assessment: data from Sentinel-II satellite within the EU Copernicus Program was used in the detection of surviving trees at Leiria, Portugal . However, UAS offers better value/cost ratio for the typical size of forests to be surveyed, which lie in the range of tens to hundreds of hectares.
Procedures according to technology
The Albatroz experts piloting the UAS are also competent using the ground systems (LiDAR, cameras, calliper or other). Moreover, the fact that they are observing the forests from different perspectives helps them decide which parameters are better visualized from each point of view.
The main deposit of carbon, and of economic value in case of commercial forestry, it is the trunk of a tree. To estimate it, one uses the diameter at different heights along with the height of the tree. With the calliper only one diameter is taken (DBH), while ground LiDAR allows capturing diameters at regular intervals from ground up to 5 to 10 metre height, depending on systems and methods. The height of the tree is used to complete the wood estimate. When local growth models and plot age are available, these are combined to estimate the “default” growth and compare it to the observed.
The following image shows an example of tree trunk segmentation using LiDAR: the original point cloud, the trunk segments and both combined; the enclosing box has about 0,42m2 ground area and is about 2.9m high.
Branches are a subsidiary contributor to wood and carbon, especially in non-conical trees, such as Mediterranean trees (cork and holm oak, olive trees, …) and they can be estimated indirectly from weighing the actual content from fallen trees and using proportional models from known parameters: canopy diameter, number of primary branches, height from canopy base to apex.
Finally, roots and leaves also contribute to the total carbon content but their value is usually obtained from theoretical models applied to the tree physical parameters as it is difficult or too costly to estimate them experimentally.
As explained, the combined analysis of airborne LiDAR data and ground LiDAR yields the number of trees in a plot and the height of individual trees that dominate over shorter trees. The following image shows such a combination applied to a eucalyptus plot. On the left, the ground LiDAR is shown; the airborne LiDAR cloud is on the middle; the two clouds are combined on the right.
The colours depict intensity; the scale range is the same but different colour codes were used to improve legibility. Nevertheless, it can be noticed that ground LiDAR intensity is more homogeneous due to shorter ranges and range variations and more perpendicular scanning while airborne LiDAR shows a wider range of intensities due to the difficulty in penetrating the canopies and wider range variations.
The fifth and last image is another step into sensor integration: it shows a section of a different eucalyptus plot where field data combines airborne photogrammetry (white) with ground LiDAR (red) to the same purpose. It also depicts ground (yellow) and shrubs (green).
Shrubs are more difficult to model and they are measured as volumes which are later converted to mass and carbon according to pre-established algorithms. Nevertheless, the image above shows how shrubs are automatically estimated from ground LiDAR. Comparing points from photogrammetry with airborne LiDAR, knowing that the plots are different but the species is the same (Eucalyptus globulus sp.) it is apparent how much more penetrating airborne LiDAR is compared to image.
The final step is the validation of the field data and analytics.
The most precise computation takes place when the commercial forests are taken down, weighted and sold. However, since roots, shrubs, branches and canopies traditionally have little to no value, it is important to find a metric that rewards computing this fraction of carbon capture, at least on a set of samples of the harvested plot.
Some approximations may be obtained from removing dead biomass and weighing or measuring (in terms of truckloads, for instance) from the ground at regular intervals and some tree species, such as the cork oak, also produce periodically a by-product – the bark – that can be used as an indicator of the overall carbon captured in the whole tree and stand.
In the last two decades, sustainability has been more about words than figures. And when it comes to figures, it is most often about forecasts: long term scenarios that are always open to debate and divergence. Not so about the past. To figure out Sustainability, it should be the opposite: less talk and more efficient actions to measure what’s already happening on the ground. That should provide the best possible roots for the debates about the Earth’s and our own species’ future.
To learn more, please check News, Publications & Events at Albatroz Engenharia website.
 Zack Parisa, Max Nova, “How AI will revolutionize Forestry”, IEEE Spectrum, August 2020
 Andreas Bayer, “Biomass forest modelling using UAV LiDAR data under fire effect”, Instituto Superior de Agronomia, Universidade de Lisboa, Portugal, December 2019