How can machine learning help to feed 9.8 billion people in 2050?

Anniek Schouten
Mar 22, 2018 · 7 min read

Forest are under pressure of agricultural expansion. In order to meet the rising demand for food forests are converted to farms. How can technology help us protect the earth’s ecosystems and help us feed the growing population? Machine learning is leveraged to detect early-stage deforestation and forest degradation. At the same time, machine learning helps to increase productivity of agriculture. Can a combination of these two add to a more sustainable future?

Growing world population (source: United Nations Department of Economic and Social Affairs)

More people, more food, less forests
The world population is increasing at rapid pace. According to research of the United Nations, there will be 9.8 billion people on earth in 2050. This is an increase of 29% compared to the 7.6 billion people today. This rapid growth causes problems for all stakeholder on our planet: Will there be enough clean water? Will there be enough food? And: Can we feed all people without destroying our earth’s crucial ecosystems?

With growing population comes growing demand for food, wood and minerals. This is one of the key causes of deforestation and forest degradation. Take the production of meat for example. Forests are cleared for cattle ranching and for soy plantations. Moreover, about 75% of all soy produced is used as animal feed. It is clear that the production of beef is the number 1 contributor to forest loss. Furthermore the production of wood-based products is an important cause. Timber is used for buildings and furniture and paper is used for the cardboard box on your doorstep when your package has arrived. All these products are taking its toll on forests.

Biggest drivers of deforestation and forest degradation

Deforestation fronts
The pressure on our forests is happening across the world. Well known areas of deforestation are the Amazon in Brazil and Borneo and Sumatra in Indonesia. However also in other regions in South America, Africa, Asia and Australia significant amounts of forest have been converted in recent years and are projected to be deforested even more in the years to come. The drivers can differ per region. For example, in Indonesia the main driver is the well known palm plantations. In the Cerrado, Brazil, it’s cattle ranching and soy plantations.

Global deforestation map (source: WWF, “Over 80% of future deforestation confined to just 11 places”)

Technological innovations
To curb this development of deforestation, changes are needed on both demand side as supply side. On the demand side, think about recycling and deducing food waste. On the supply-side innovative ideas become reality: plant-based meat and lab-grown meat are developed to replace factory farming. However, to fight deforestation in short term and on a large scale additional actions are needed.

Increasing productivity using machine learning
Leveraging artificial intelligence, and more specifically advancements in machine learning — a subfield of artificial intelligence, can contribute to increasing productivity of converted land. When satellite imagery and machine learning are combined, automatic insights can be generated to help a farmer improve operations. It helps the farmer to make data-driven decisions about location, timing and quantity of irrigation needed and pesticides to be applied. Furthermore, it helps optimising timing of harvest. With as a result: higher yield and more sustainable use of resources.

In a similar way, machine learning and satellite imagery are applied to increase productivity of production forests. Satellite imagery is especially useful for creating forest intelligence, since it provides a view from the sky for massive forests. Forest intelligence about drought, insect plagues, tree species composition and productivity can help forest managers optimise their use of land and inventory. When done in a responsible way, efficiency improvements can result in higher productivity on the same amount of land and plantations do not have to expand into forests to increase their productivity.

An example of selective logging (source: DigitalGlobe,

Deforestation patterns from the sky
In the battle against deforestation, machine learning is also used to detect forest changes from the sky and identify high risk areas. Deforestation is associated with specific patterns that appear in forests. When these patterns appear in a forest, it tells us that the chance is high that large scale deforestation is about to happen. For example little roads appear in a previously untouched forest. We see narrow roads and winding trails reaching into a dense forest area, leading towards a river or a road. These roads are often a sign of unsustainable selective logging. Close to these logging roads, ancient and highly desired trees are cut down leaving holes in the forest, also known as selective logging. The valuable timber is transported via the river or road and is sold.

An example of a fish bone pattern (source: DigitalGlobe,

A different pattern of deforestation that is clearly visible from the sky is the so called “fish bone pattern”: alongside a road several small roads reach into the forest. It is only a matter of time until farmers settle at these roads. Forest is removed and space for crops and cattle is created. As a result, the fish bone shaped structure starts to appear. This is an indication that more deforestation is about to happen.

Another visual pattern that can be seen clearly from the sky is the “slash and burn” method. As the name suggest: it is not subtle. Large areas of natural vegetation is cut down and set on fire, burned down to the ground to create open space for agriculture. The ash is left on the ground as short-term fertiliser. When the plot is exhausted, a new part of the forest is slashed and burned. Destroyed ecosystems and depleted soil as a result.

Patterns of deforestation

So, how does machine learning work?
Machine learning is a type of artificial intelligence and is about creating algorithms that can “learn” from experience. Let’s take the example of a tree. When you see a tree in the park, you recognise immediately it is a tree. For humans this is simple, but how can we teach a machine to recognise a tree? By showing thousands of examples of a tree, the machine learns from that experience. After this training process is completed, we present the machine with a new image of a tree, an image it has never seen before. The machine can now automatically classify this image as a tree. As children do, the machine learns from many examples. Deep learning is a specific technique within the field of machine learning that has been successful for image classification. In deep learning, the algorithms are capable of learning to focus on the right features by themselves. This means you do not need to tell the algorithm that a tree is green and it looks round. It learns by itself which characteristics are important in deciding if an image contains indeed a picture of a tree or not.

In the case of deforestation it works roughly like this too. By feeding the algorithm with many examples of deforested areas, it learns what deforestation looks like. We feed it with examples from various points in time and from different geographical areas. This way we can train the machine learning algorithm to detect deforestation patterns and predict high risk areas. It even makes it possible to detect patterns that cannot be spotted by humans yet. With the latest advancements in satellite imagery, it is possible get daily snapshots of the whole planet. To analyse these amounts of images and to detect minor changes machine learning is the perfect solution.

Deep Learning — a subfield of machine learning and artificial intelligence

With these insights, local and global stake holders can put deforestation activity to a stop. Automatically detecting patterns of deforestation requires a combination of high spatial and temporal resolution satellite imagery and strong algorithms to recognise what is happening on the ground.

Reduce the pressure on forests
How can machine learning help to create a more sustainable future? Machine learning in combination with satellite imagery is applied to help increase productivity of food and wood production on farmland. On the other hand, it is used to detect patterns of deforestation and forest degradation. By focusing on conservation of forest on the one hand and on increasing productivity of existing farms on the other hand, this technology can help to reduce the pressure on forests for a more sustainable future for our planet and people. provides forest intelligence by using artificial intelligence in combination with high-resolution satellite imagery. Forest insights help corporates, governments and NGOs to combat deforestation and increase productivity and sustainability of forest management.

Exploring forests from the sky.

Thanks to Indra den Bakker

Anniek Schouten

Written by

Exploring forests from the sky.

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade