Artificial Intelligence and Forest Management

How Can We Make the World More Green and Lush With the Help of AI?

Alex Moltzau
Sep 3 · 13 min read

This article is co-written together with Syed Nazmus Sadat who Studies Forestry and Environmental Science at Shahjalal University of Science & Technology, Sylhet in Bangladesh.

How can artificial intelligence help in efforts to prevent deforestation? Deforestation has an incredibly adverse impact on planet earth. The forests cover close to a third of the land area on our planet and provide us with purer air and fresher water. Eighty percent of the world’s land based wildlife live in forests [1]. We should of course not need to argue for why forests are important yet it seems we have to. Between 1990 and 2016, the world lost 502,000 square miles (1.3 million square kilometers) of forest, according to the World Bank [2]. With the recent events of increasing forest fires in the Amazon protecting our forests across the world has again grown in relevance [3]. So let us touch upon the current issue of deforestation and then discuss how artificial intelligence can contribute to solutions that assist in solving this wicked problem. Artificial Intelligence can of course only be part of a coordinated effort, yet it has to be duly considered in addressing this issue.

Let’s Talk Deforestation

Numbers are not looking great for our forests. The Global Forest Resources Assessment (FRA), coordinated by FAO, found that the world’s forest area decreased from 31.6 percent of the global land area to 30.6 percent between 1990 and 2015, but that the pace of loss has slowed in recent years [4]. So amongst this devestating message there is the respite of possibility to prevent further damage or even reverse the trend completely. Another source however estimates that up to 420 million acres of forest could be lost between 2010 and 2030 in these “deforestation fronts” if current trends continue [5].

If we go further back It said that the number of trees worldwide has fallen 46 percent since the dawn of agriculture 12,000 years ago and more than 15 billion trees are felled every year. That loss has significant implications for the planet in terms of climate change, biodiversity, and human well-being, according to Crowther [6]. Returning from ancient to recent times research indicates that anywhere from 3 trees to an area the size of a football field is cleared per second. Nearly 100,000 acres of forest is cleared per day. About 15,000,000,000 trees are cut down every year, according to Nature [7]. Numbers in a representation by Statista shows as well that we should not only be worried about Brazil. The forests in the north, or Russia in particular, is rapidly being lost.

How do we manage forests better?

Managing forests is not only a technological exercise and technology can not solve everything. We already understand that forests are being lost and this fact must be addressed. As such mapping forest reserves seems only part of the solution. Political will and understanding of the severe consequences by a larger population is necessary. If change does not come from one place or one person — and emerges from a collective of action then we can assume that every action you do towards improving conditions for our forests will matter.

As such technologists or technology enthusiasts can of course play a part, and why should not technological advances benefit humanity? You may silently think about what better way to use artificial intelligence than to ensure cleaner air and ensure increased natural carbon capture by managing our forests better. If you manage to think of a better way for humanity to use machine learning techniques or other advances within the field of artificial intelligence then I am very happy for you and the world (please let us know).

Let us now however now consider how artificial intelligence can contribute to a better environment for humans and animals alike. In writing this article we collaborated from Bangladesh and Norway as such we asked locally how we could approach this topic. One valuable suggestion came from Swapan Kumar Sarker who is an Associate Professor of Forestry and Environmental Science Department of Shahjalal University of Science and Technology. He had the following to say:

“AI can play a tremendous role in the control of forest fire. We can train the software by providing it the information about the places which are most vulnerable to catch fire. So that it can detect the vulnerable places of a forest and send notifications to the forest officers so that they can rapidly go to those places and put out the fire. Thus they will require less time and can save more natural resources by putting out the fire easily. It could save millions of dollars.”

Immediately as such detection of fires seem to one possible use case for artificial intelligence. If firefighters can respond or appropriate measures can be taken to conserve forests that would be beneficial if implemented with considerations to the consumption of energy by the algorithms used.

How can artificial intelligence techniques be used?

In the article from 2008 Artificial Intelligence techniques: An introduction to their use for modelling environmental systems [8] a few approaches to forest management is suggested. Two techniques were described as relevant for use in regards to aspects of forest management, although more techniques can likely be used:

  1. Fuzzy systems (FS) use fuzzy sets to deal with imprecise and incomplete data. In conventional set theory an object is a member of a set or not, but fuzzy set membership takes any value between 0 and 1. Fuzzy models can describe vague statements as in natural language. This can be used for evaluating habitat suitability for riverine forests.
  2. A multi-agent system (MAS) comprises a network of agents interacting to achieve goals. An agent is a software component containing code and data. MASs are well suited to this field because of their ability to represent complex systems with several stakeholders and allow exploration of alternative management approaches. They do, however, seem more suited to social learning among interest groups than prediction of system behaviour. Examples could be in natural resource management within forestry industries and forest ecosystems undergoing land-use change.

Another interesting mapping of possibilities can be seen in a more recent paper from June 2019 by leading AI researchers called Tackling Climate Change with Machine Learning [9]. Researchers from Google AI, Microsoft Research, DeepMind, MIT, Stanford etc. united to consider how different challenges relating to climate change could be handled with the help of insights and applications within the field of artificial intelligence.

Thus in this manner, the technologies that were suggested as relevant for farms & forests were computer vision, RL (Reinforcement learning) & Control and transfer learning. The titles correspond to 5.1 remote sensing of emissions; 5.2 precision agriculture; 5.3 protecting peatlands; and 5.4 forests. While we could jump straight to the last section with the appropriate name (‘forests’) it is advantageous to have a quick look at all the other sections mentioned from the paper, let us run through it quickly.

Remote sensing of emissions suggest real-time maps for greenhouse gas emissions. In this way we could instantly know which areas of farmland on the planet that emits the most greenhouse gases and do our best to adjust with regulations, incentives or collaboration. “While greenhouse gases are invisible to our eyes, they must by definition interact with sunlight. This means that we can observe these compounds with hyperspectral cameras.” (p.28) This is described as an open problem with a high potential impact.

Precision agriculture is another opportunity. Deforestation often happens to give way to food for humans. Beef is not the most efficient, yet this seems to be a large reason for the clearing of large areas of the Amazon forest. If we switch to more plant-based diet and still keep animals in areas suitable for animal husbandry then we can consider what else that could be done. Overall, agriculture is responsible for 14% of greenhouse gas emissions. Precision agriculture could lead to less pesticides and more efficient use of water with a combination of robotics, hardware and clever use of algorithms. Machine learning tools for policy makers and agronomists could additionally help to encourage climate positive action.

Protecting peatlands. Peatlands cover only 3% of the Earth’s land area, yet hold twice the total carbon in all the world’s forests, making peat the largest source of sequestered carbon on Earth (p.29). When peat dries it releases a lot of carbon and can catch fire. A single peat fire in Indonesia in 1997 is reported to have released emissions comparable to 20–50% of global fossil fuel emissions during the same year. Advanced machine learning techniques could help develop advanced monitoring tools at low cost and predict the risk of fire.

In terms of direct forest management we can mention:

  • Carbon stock estimation. An estimator can perform predictions on the scale of the planet. Transfer learning techniques might help in this regard, we will come back to what transfer learning is.
  • Automated afforestation. there is a capacity for 1.2 trillion more trees on the planet. This has the potential to cancel out a decade of carbon emissions. In this sense automation can be useful. Two startups BioCarbonEngineering and Droneseed is mentioned in this regard. Machine learning can be used to locate appropriate planting sites, monitor plant health, assess weeds, and analyze trends.
  • Forest fire management. As mentioned previously we can prevent large forest fires. It is important to mention that small forest fires can be good. Reinforcement learning can be used to predict the spatial progression of fire (how fire spreads). With good tools to evaluate regions that are more at risk, firefighters can perform controlled burns and cut select areas to prevent progression of fires.
  • Forestry. While some deforestation is the result of expanding agriculture or urban developments, most of it comes from the logging industry. Clearcutting has a particularly ruinous effect and remains a widespread practice across the world. Tracking deforestation can inform policy-makers. Rainforest Connection has installed old smart-phones powered by solar panels in the forest. Then, an ML algorithm can detect chainsaw sounds within a radius of a kilometer and report them to a nearby cell phone antenna. Machine learning can also be applied to logistics and transport, although this can have a negative effect and has to be combined with good policies.

What companies are using AI to help manage forests?

#Forest intelligence — 20 trees AI

For near real-time forest inventory & sustainable forest management

Forests produce the air we breathe and the products we use. They are massive and complex, that’s why we need AI and a view from the sky to understand them. Thanks to very high-resolution satellite imagery and radar data, 20tree.ai is able to extract insights on tree level on a global scale.

This includes insights into forest composition like tree species, tree height and diameter (DBH), tree growth & productivity, and harvesting insights. Providing near real-time intelligence into forest and wood inventory.

Combining satellite imagery and AI also unlocks insights that are not directly visible for the human eye. Insights into forest health & threats, like deforestation, drought, insect plagues, soil health, storm damage, and other forest disturbances. Resulting in more efficient use of resources and limited negative impact.

Every day, the company uses NVIDIA GPUs to process almost 100TB of new satellite data — obtained from partners such as Airbus Defence and Space and the European Copernicus program — which is used to train a series of deep neural networks. GPUs, running in-house and in the cloud via AWS and Google, provide the muscle power for the training, enabling it to be completed in just a few hours. The deep neural networks can then draw insights into forest health that are otherwise invisible to the human eye.

“Thanks to the power of artificial intelligence and NVIDIA’s GPUs, we are enabling faster, better decision making for our planet,” said Indra den Bakker, co-founder and deep learning engineer at 20tree.ai.”

For more information: https://20tree.ai/

#aiTree:

aiTree Ltd has been focusing on systematic technologies to solve Demand & Supply problems with Artificial Intelligence algorithms for over 25 years. The typical application Forest Simulation Optimization System (FSOS) has been applied and improved for over 20 years in British Columbia, Canada. The demands from a forest include wildlife habitat, biodiversity, water quality, visual quality, carbon storage, timber production and economic contributions. FSOS focuses on both “what we can take from the forest” and “what we can create in the forest”. Forest design is the most complicated problem because the trees are growing and dying, and all the values have to be considered every year for over 400 years. FSOS is a good example that uses Artificial Intelligence, Big Data and Cloud computing technologies to solve the complicated Demand & Supply problems. You can find more information in the Canadian forest cloud: http://forestcloud.ca and in the Chinese forest cloud: http://forestcloud.cn

Their products:

  1. Forest Simulation Optimization System (FSOS): FSOS is developed for multiple-objective forest analysis and planning, it integrates long-term strategic planning and short-term operation planning in one model. It is a great tool for forest management simulations, animations and optimizations. You can compare management scenarios and see the future forests with different management scenarios.
  2. Forest Carbon Analysis and Accounting System (FSOS-C): We have developed a forest Carbon Analysis and Accounting System FSOS-C based on Forest Simulation Optimization System (FSOS).Unlike other forest carbon accounting system, FSOS-C can not only calculate carbon amounts stored in different pools such as stands, wood products, floor, and soil with different stand dynamics and management scenarios, but also optimize and balance with other management objectives such as timber production, profit, wildlife habitat, biodiversity, water quality, etc.
  3. Integrate Field Data Collection, Cloud Data Storage and Computing Technologies (FC): You can design your own field data collector and cloud data storage system. In the cloud, you can define a field data collection task including tables, points, lines, polygons, photos, videos and assign the task to a number of field people. The field people can start the applications, login and start the field data collection work. Many field people can work with the same task and all data can be stored in your mobile devices and the cloud database when the internet connection is available.

For more information: http://aitree.ltd/

#Microsoft AI for earth project:

AI for Earth awards grants to projects that use artificial intelligence to address four critical areas that are vital for building a sustainable future. These four critical areas are:

Climate: The changing climate threatens human health, infrastructure, and natural systems. AI can give people more accurate climate predictions to help reduce the potential impacts.

Agriculture: By 2050, farmers must produce more food, on less arable land, and with less environmental impact to feed the world’s increasing population. AI can help people monitor the health of farms in real time.

Biodiversity: Species are going extinct at an alarming rate. AI can help people accelerate the discovery, monitoring, and protection of biodiversity across our planet.

Water: In the next two decades, demand for fresh water is predicted to dramatically outpace supply. AI can help people model Earth’s water supply to help us conserve and protect fresh water.

Featured AI for Earth projects:

Terrafuse: Physics-enabled AI models help everyone understand climate-related risk at the hyperlocal level. Terrafuse uses machine learning to forecast climate-related risks. Terrafuse leverages historical wildfire data, numerical simulations, and satellite imagery on Microsoft Azure to model wildfire risk for any location. Anyone can access wildfire forecast information via APIs and graphical tools.

Ag-Analytics: Collecting farmland data in the cloud and making it available to farmers to enable precision agriculture. Ag-Analytics uses sensors to collect soil, tillage, and yield data for specific plots of farmland. The data is stored in Microsoft Azure and made available to farmers via user-friendly APIs to help them lower costs, improve yields, and minimize the environmental cost of agriculture.

OceanMind: Using satellites and AI to preserve biodiversity, protect livelihoods, and prevent slavery in the seafood industry. OceanMind works with government authorities to prevent illegal, unreported, and unregulated fishing by analyzing vessel movements in real time. AI algorithms identify suspicious behaviour, which OceanMind shares with agencies to direct patrol boats more effectively.

SilviaTerra: Transforming how conservationists and landowners measure and monitor forests. It’s essential for conservationists, governments, and landowners to inventory forests for ecological, social, and economic health. By utilizing AI, cloud software, and machine learning these groups can work together to study the effects of climate change and improve habitats. SilviaTerra is using Microsoft Azure, high-resolution satellite imagery, and US Forest Service inventory and analysis field data to train machine-learning models to measure forests. (For more information: https://silviaterra.com/bark/index.html)

Wild Me: AI and citizen scientists work together to fight extinction. AI for Earth is using AI technology and advanced cloud software to identify animal species that are on the verge of extinction. Wild Me uses computer vision and deep learning algorithms to power Wildbook, a platform that utilizes technology to scan and identify individual animals and species. (For more information: http://wildbook.org/doku.php)

Land Cover Mapping: Scientists are using real time data mapping to enable precision conservation. In an ongoing effort to protect complex eco-systems, we partnered with Chesapeake Conservancy to build a dynamic system for generating one-meter resolution land cover data anywhere in the United States. Our advanced mapping utilizes AI and key data sets to revolutionize precision conservation.

PAWS: Protection Assistant for Wildlife Security (PAWS) uses AI to aid conservationists in the fight against poaching by utilizing machine learning, AI planning, and behavior modeling. PAWS collects information from previous poaching activities, then uses machine learning and behavioral modeling to generate predictions about poaching locations and optimal patrol routes. The outcome is more effective patrols and better use of resources in the fight against poaching vulnerable species. (For more information: https://www.cais.usc.edu/projects/wildlife-security/)

Lastly some good news!

Large areas of the continent have seen a forest boom that means today more than two-fifths of Europe is tree-covered. Between 1990 and 2015, the area covered by forests and woodlands increased by 90,000 square kilometres — an area roughly the size of Portugal. As such it seems it is possible that we can increase afforestation and make the world more lush and green. So let us do everything we can to ensure thriving forests around the planet.

Sources from the first sections:

1. https://www.worldwildlife.org/threats/deforestation-and-forest-degradation

2. https://blogs.worldbank.org/opendata/five-forest-figures-international-day-forests

3.https://www.theguardian.com/environment/2019/aug/23/amazon-fires-global-leaders-urged-divert-brazil-suicide-path

4. http://www.fao.org/state-of-forests/en/

5. https://www.worldwildlife.org/threats/deforestation-and-forest-degradation

6. https://news.yale.edu/2015/09/02/seeing-forest-and-trees-all-3-trillion-them

7. https://www.nature.com/articles/nature14967

8.https://www.researchgate.net/publication/257219838_Artificial_Intelligence_techniques_An_introduction_to_their_use_for_modelling_environmental_systems

9.https://arxiv.org/pdf/1906.05433.pdf


This is day 92 of #500daysofAI. My current focus for day 50–100 is on AI Safety. If you enjoy this please give me a response as I do want to improve my writing or discover new research, companies and projects.

ODSCJournal

Collecting all of the best open data science articles, tutorials, advice, and code to share with the greater open data science community!

Alex Moltzau

Written by

Co-founder of AI Social Research. Student at the University of Oslo, major anthropology + minor computer science. All views are my own. twitter.com/AlexMoltzau

ODSCJournal

Collecting all of the best open data science articles, tutorials, advice, and code to share with the greater open data science community!

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