Cape AI
Published in

Cape AI

The Problem of Charcoal Poaching and How Data Science Can Help to Stop It

This research is done in collaboration with Sensing Clues.

Early January I moved from Amsterdam to Cape Town to start my thesis internship at Cape AI. Since then, I learned a lot..

My interest in charcoal was first sparked when I stumbled across its miraculous benefits. Specifically, charcoal had been known for purifying skin and whitening teeth. I soon discovered, however, that charcoal had a dark side.. a very dark side.

Little did I know about that.

The Charcoal Poaching Problem

The world relies heavily on charcoal. In Africa, specifically, big cities are reliant on charcoal. For example, in Kinshasa, Congo’s capital, 90% of its inhabitants use charcoal as their primary source of fuel.

Charcoal can be produced through the cutting and burning of trees in a kiln, a type of oven. A kiln has a thermally insulated chamber that can reach temperatures of up to 900 degrees Celsius. At this temperature, pyrolysis is reached and wood is transformed into charcoal.

Typically, poachers target public land, seemingly free of any other use, to set up sites to build kilns. The lack of supervision makes it easy for them to use these lands with relative impunity. The charcoal production usually takes around 7 days. During this period, poachers head off hunting for bushmeat. Unfortunately, the most devastating consequence of the illegal production of charcoal is not always the poaching but actually desertification due to deforestation.

Deforestation is a major problem in many African countries. In Uganda, forest coverage has decreased from 24% in 1990 to 9% in 2015. Similarly, in Ethiopia, forest coverage fell from 35% in the 1900s to 4% in the 2000s.

The problem doesn’t stop there. The vast majority of charcoal trade happens between the Horn of Africa and countries in the Middle East. According to the UN, an estimated 8.2 million trees were lost to illegal charcoal production between 2011 and 2016. This amounts to approximately $289 million per year in illegal business. This trade is largely dominated by Al-Shabaab, a Jihadist fundamentalist terrorist group based in Somalia and other Eastern African countries. The money from this illegal trade serves as important funding in their wars.

A charcoal kiln made from clay in the northeast region of Thailand

The Project Area

In East Africa a conservation company tackles illegal charcoal production.

Their project area is a migration corridor for wildlife and is covered with Acacia-Commiphora dryland forest. It serves as a home for endangered mammals such as lions, leopards and elephants.

The company protects the project area by selling its carbon rights to companies that want to reduce their net carbon footprint. Part of the proceeds from their initiative are invested in the local community. Another part of the proceeds is used to hire patrol rangers to protect the land.

The number of “boots on the ground” are always scarce in comparison to the extend of the problem. Despite best efforts, some poachers go unseen. Of course hiring additional rangers may help increasing detection. However, there is a financial limit to the number of rangers that can be hired. Rather than more patrols, intelligence-drive patrolling is called for.

Time to Talk Data

In 2010, rangers in the project area started to collect data, at that time on paper, since March 2020 with the Cluey Data Collector and Tracking app of Sensing Clues. The rangers log every observation, from grazing groups of cattle to signs of poaching activities like snares and log heaps. This results in a rich geo-temporal dataset. This data holds enormous potential for combating the problems at hand!

One of the methods that can be applied to optimize patrol strategies, using the data gathered in Cluey, is Game Theory. Game Theory can be defined as the science of logical decision making: It uses mathematical models to mimic interactions between rational decision makers. Its countless applications have led to ground-breaking insights in human behavior.

In the security domain, where there is an attacker and a defender, the application of game-theoretic frameworks is very popular.

In particular, Stackelberg Security Games. In this sequentially played game, first the defender commits to a mixed strategy which determines how his available defense resources are deployed. Second, based on this information, the attacker chooses how to respond by attacking a target.

The use of Stackelberg Security Games led to promising results in a variety of security areas. However, these applications are all based on the Stackelberg assumption that the attacker is always fully aware of the defender’s strategy and reacts accordingly. In conservation, this is often far from reality. Most of the time, the attackers simply lack the resources to continuously observe the defenders strategy.

This led to the development of Green Security Games, a modification of Stackelberg Security Game that deals with the shortcomings like the one just mentioned. Green security games have successfully been applied to wildlife poaching and fishery. But can, because of its characteristics, perfectly be applied to the charcoal problem, too.

A ranger working in northern Tanzania

To obtain a pay-off matrix (containing the pay-off for every possible combination of strategies for both players), essential to play the security game, the first step is to estimate the relative risk of poaching activities throughout the project area. This is done by using Machine Learning techniques on features created using the data collected by the rangers using Cluey as well as external data.

Features that have been proven to be significant include the slope of the area, the distance to roads, settlements and ranger posts, animal density and dry or wet season.

How It Works

After the payoff matrix has been defined, the security game is played as follows:

  1. Not knowing whether or not the attackers (poachers) will be able to observe him, the defenders (rangers) use the model to plan a patrol strategy
  2. The guards execute the subscribed patrol routes, following instructions in the app
  3. Poachers attack targets within the area, with a probability of being caught by the defender
  4. New data is collected during these patrols, the model is updated and the game is played again.

The patrol routes mentioned in Step 2 will be displayed in the same app which is being used to collect the data. This enables them to always be aware of where poachers are most likely to attack in the area. This will help greatly in optimizing the rangers’ patrol strategies!

With optimized patrol routes, rangers will be able to do their work more effectively and stop more poachers in their tracks. Hopefully this project will have a positive impact on the number of trees and animals saved in the project area. If we succeed, which we will, it will be made available to the NGO’s managing other areas that suffer from poaching.

This is an ongoing research project and if you like to stay updated on its progress, please follow.

I would like to thank Sensing Clues for making their expertise and data available and providing me guidance. Organizations like theirs are massively inspiring and essential to the conservation of our beautiful planet.

— Ea Werner, Data Science Intern at Cape AI

‘Ea’s research is a fantastic use case of the effective application of Data Science in tackling environmental problems. This aligns directly with the Cape AI Mission to use AI to benefit the triple bottom line : People, Planet, and Profit!’ — Adit Mehta, Lead Data Science at Cape AI

Please leave claps to spread awareness!



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