Calculating Mill Level Deforestation & Carbon Risk Scores Across the Palm Supply Chain

Illustrating the power of the Descartes Labs Platform to explore agricultural traceability challenges

Zoomed out view of mill-level carbon scores. Lower scores mean less nearby deforested area and estimated forest carbon loss.

According to the World Wildlife Federation, palm oil can be found in almost 50% of all packaged supermarket goods. As demand continues to increase, global agriculture and manufacturing companies are searching for ways to determine risk factors for deforestation and carbon emissions in the palm supply chain. Adding risk factors and associated traceability can help verify sustainable sourcing commitments and provide tools to help non-compliant participants to adopt more sustainable practices.

While there is some sourcing transparency from mills to manufacturers, a high number of third-party growers supply to these mills, comprising a challenge for NDPE compliance. Therefore, companies need to understand probable source locations, land-use histories, and aggregate sourcing practices for each mill in their supply chain. They also need to attribute ongoing deforestation activity to a given mill and supplier.

However, there is currently no widely available information that traces palm throughout the supply chain. And there is no direct way of attributing palm harvested within a plantation to the mill that eventually processes it. Further, concession ownership and permits have little transparency, making it challenging to attribute deforestation to specific growers. Still, there are a few factors that bound the sourcing region for a given palm mill (e.g. distance, transportation network, mill capacity, nearby mill capacity, etc.), that could allow us to estimate the probability and risk of a given mill processing palm from a given plantation.

Palm concessions and mills over a subset of the island of Sumatra on top of a Sentinel-1 radar composite image

Application of Deforestation & Forest Carbon Loss Risk Scores

This post explores a relatively simple method to attribute risk from deforestation and forest carbon loss to individual mills throughout the palm supply chain in Southeast Asia. It applies deforestation and carbon loss risk scores to the entire Universal Mill List maintained by the World Resources Institute (WRI), Rainforest Alliance (RA), Proforest, and Daemeter Consulting.

Our analysis leverages some key datasets and the compute power of the Descartes Labs Platform. The creation of these datasets in a singular development environment natively allows for the prediction of carbon equivalent of deforestation events. Each has been built with the other in mind.

  • Sentinel-2 and GEDI derived forest carbon composite
  • Sentinel-2 derived palm area growing mask
  • Sentinel-1 InSAR derived deforestation detections
  • Forest carbon predicted using an allometric equation from GEDI TCH at RH 95

Methodology and Replication

First, we show how the methodology works for a given mill. Then we replicate the methodology over every mill in the Universal Mill List. We start by drawing a radius around each mill, quantifying the deforested area and forest carbon lost within the region over a set period of time, and then normalize by the total forest area and forest carbon.

Creating a radius around each mill and displaying deforestation and forest carbon
  • Total area deforested from July 2020 — December 2020: 3,355 ha
  • Total forest carbon from July 2020 — December 2020: 575,478 Mg C

Weighted Average Using Inverse Distance

We also create a weighted average over the mill area using inverse distance such that deforested pixels closer to the mill are weighed more than those found farther from the mill.

Creating a simple distance-weighted risk score for deforestation and forest carbon loss.

Calculating Deforested Area Score & Carbon Loss Score

Once we have our totals and our distance weights, we can calculate a deforested area score and carbon loss score for the mill in question. Scores generally range between 0 and 4, with higher scores indicating higher risk

Deforested area score = 2.91

  • Distance-weighted average of deforested area ÷ maximum weighted average if the entire area was deforested

Carbon loss score = 3.27

  • Distance-weighted average of carbon lost ÷ maximum weighted average if the entire area was deforested

Applying the Descartes Labs Platform to the Full Universal Mill List

Next, we use the Descartes Labs Platform to apply the same methodology to the Universal Mill List within Indonesia and Malaysia.

Visual Results Using Descartes Labs Workflows API

Finally, we can visualize it all together using Descartes Labs Workflows API.

Zoomed out view of mill-level carbon scores. Lower scores mean less nearby deforested area and estimated forest carbon loss.
Histograms showing the frequency distribution of total deforested area, total carbon lost, and the overall area & carbon scores for each mill

When we zoom in on a particular location, we see the layers for each dataset along with the deforested area and carbon scores for nearby mills. Note that the size of the buffer around each mill has been reduced to prevent overlap during visualization.

Using the Dataset to Inform Supply Chain Risk Assessment

Now that the dataset has been created, it can be used to inform supply chain risk assessment for any company that sources from the Universal Mill List. To summarize, we built a mill-level deforestation & carbon emissions risk score across Malaysia and Indonesia. To do this, we leveraged proprietary Descartes Labs datasets to determine historical deforestation activity and carbon loss potential based on a distance-weighted formula to each mill. This produces a dataset that allows mills to be ranked by deforested area score and carbon loss score to determine focus areas for ongoing sustainability decisions.

Click here to download a datasheet with more details about our Tropical Deforestation Monitoring Package.

--

--

--

Explore posts from the Descartes Labs team

Recommended from Medium

Save The Turtles! Oh Wait, You Were Serious?

Let’s Hope New Texas Law Turns Out to Be ‘All Hat, No Cattle’

Achieving net zero will be the new Space Race, here’s why

What To Do When The World Makes You Sick And Tired

Ban All-Inclusive Holidays

Investing for Purpose and Performance

Climate Wave Coffee Company

Goats in The Park

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store
Descartes Labs

Descartes Labs

Descartes Labs is building a data refinery for geospatial data.

More from Medium

Mapping the world with unmatched frequency

Area Monitoring — Going Large and Towards Finer Detail — Partnership with Planet and Batch Stat API

Protests spot report 24–30 November 2021

Introduction to the openEO Platform