Artificial intelligence (AI) technology is changing the management and understanding of natural resources in a significant manner. The technology offers a powerful toolkit to tackle immense challenges associated with natural resource management, ranging from optimizing energy production to safeguarding forests. Specifically, the recent data science advancements, coupled with the revolution in satellite and digital technology, have further improved the potential for AI applications in the forestry and wildlife sectors. This article deliberates on the application of AI in forest management.
Image credit: Summit Art Creations/Shutterstock
Importance of AI for Forest Management
Forests are vast and vibrant terrestrial ecosystems covering roughly one-third of the global land area, cradling 90% of terrestrial biodiversity. This vital natural resource is critical for sustaining life. For instance, forests absorb approximately 30% of global annual atmospheric carbon dioxide emissions and provide breathable, clean air by stabilizing greenhouse gasses in the atmosphere.
Forests support the hydrological cycle, natural predators of agricultural pests, and pollinators, ensuring global food security. Additionally, they act as a key medicinal plant source, supply a significant amount of global renewable energy, and are crucial to ensuring the hydrological integrity of different ecosystems.
However, forests are witnessing rapid degradation due to growing urbanization, agriculture expansion, and exploitation of timber. Effects of climate change, like wildfires, further contribute to forest decline. Despite these threats to the forests, the sector relies on conventional methods to manage them. These methods have many drawbacks, such as lack of scalability, introduction of personal bias, and slow data analysis and collection. The forest sector remains one of the few sectors where new technology adoption is slow.
To overcome these drawbacks, the forest sector can implement the AI technology. This technology can assist in effective forest conservation, monitoring, and management. The sector will immensely benefit from the technology’s inherent ability to support innovation rapidly at various scales and in different geographies.
Addressing Deforestation using AI
AI and machine learning (ML) algorithms, coupled with spatial analysis, have been employed to monitor and predict the rate of deforestation around the world. For instance, artificial neural network and geographic information system (GIS) were used for deforestation prediction.
Similarly, Rainforest Connection is focusing on developing solutions to mitigate the challenge of deforestation. The company is utilizing discarded, old cellphones, powering them using solar power and deploying them on treetops for recording chainsaw sounds from the forest.
Then, these recordings/data are sent to cellphone towers, which transmit the data to the base station. In the base station, TensorFlow/Google’s ML and AI library is utilized to detect the chainsaw noise from the recordings. After identifying the noise, this information and the sensor’s location information are shared with forest managers to enable them to carry out further due diligence for identifying and preventing illegal tree felling.
Non-profit organizations and startup companies, such as Future Forest Map project, Global Forest Watch, Terramonitor, and Outland Analytics, also use AI technology and open-source satellite data to monitor and map deforestation in real-time. For instance, Outland Analytics utilizes audio recognition AI algorithms to detect unauthorized vehicles/chainsaw sounds and sends alerts in real time to officials through email to manage environmental crime efficiently.
Similarly, Satelligence and Terramonitor employ a satellite image database collected daily by multiple satellites and AI to generate low-cost satellite data for natural area management and to monitor forest health and deforestation in real-time. A collaboration of the World Resources Institute and Central Africa Regional Program for the Environment adopted AI and spatial modeling to identify the factors influencing deforestation in the Democratic Republic of Congo (DRC) and map the regions where forest loss is expected in the future.
Their analysis indicated that factors related to human presence, like the presence of roads and shifting cultivation, had the maximum influence on forest loss. Climatic variable precipitation also adversely impacted the forest cover. The analysis also indicated that forests near farmland are the most susceptible to deforestation. This study can assist the DRC authorities in making land-use decisions proactively to shift the development pressure from high-value forests.
Forest Inventory, Mapping, and Biomass Estimation
Many startups are using both proprietary and openly available high-resolution satellite imageries and merging them with other datasets to create highly detailed maps of forests. For instance, SilviaTerra combines field survey data from the United States (US) forest department with openly available high-resolution satellite imagery to produce a predictive model that estimates forest conditions at a 15 m × 15 m resolution.
The data consists of information about tree type, height, and diameter, which is utilized by different timber and conservation organizations to guide their plans. The Chesapeake Bay Conservancy, Esri, and the Microsoft Azure team joined forces to create a highly detailed landcover map of the Chesapeake Bay using ML libraries.
Similarly, Portugal-based startup 20tree.AI used AI, cloud computing, big data, and remote sensing for forest inventory and monitoring in real-time. The Finnish forest center leverages AI, climate and weather data, imagery sources, and GIS data for precise forest stand measurements and better prediction of forest inventory.
Germany and Finland-based for-profit company CollectiveCrunch has built an AI platform “Linda Forest” that predicts wood quality, mass, and species of target areas more accurately compared to existing conventional methods. Linda Forest leverages various data sources, like Copernicus Climate Change Reanalysis data for microclimate modeling and growth predictions, Sentinel-2 images for growth modeling, and the VHR2 image of Europe from the Copernicus Land Monitoring Service, to precisely estimate the wood quality and mass in standing forest of the target area.
Companies can estimate resource-efficient consumption and production of wood products based on this information. AI and ML techniques have displayed a high potential in monitoring and mapping carbon dioxide stock and other ecosystem services in forests.
For instance, a startup GainForest has utilized a video prediction model, game theory, huge amounts of unlabeled satellite imagery, and ML-based measurement, reporting, and verification (MRV) processes to forecast and monitor deforestation and design carbon payment schemes.
Similarly, another startup Panchama applied ML on a combination of LiDAR, drone, and satellite images to accurately estimate the individual tree volume, size, and carbon density. Additionally, the non-profit collaboration between the Center for Global Discovery, and Conservation Science (GDCS) at ASU, Erol Foundation, and Planet.Inc. utilizes satellite imagery at 3–5 m resolution, LiDAR, and computer vision models for cost-effective and automatic direct mapping and measurement of carbon emission and stock at high frequency and resolution in the Peruvian forest.
Hazard Prediction and Assessment
Technological advances have facilitated the development of novel methods for preparing and collecting highly accurate, high-resolution data that will substantially improve forest management and conservation activities.
For instance, technologies like natural language processing and optical character recognition can digitize vast amounts of data from past forest monitoring that exist in paper format. Data in digital format can be used as input in many analytical algorithms to perform analyses.
Similarly, near-real-time information can be obtained about forest conditions and activities by installing Internet-enabled sensors that measure moisture and temperature in forests. Such data facilitates the development of predictive models to identify and obtain insights into forest health and threats like storm damage, disease outbreaks, pests, wildfires, drought, and deforestation. For instance, a Canada-based startup Terrafuse utilizes physics-enabled AI models to comprehend climate-related risk at the hyperlocal level.
The company leverages numerical simulations, historical wildfire data, and satellite imagery on Microsoft Azure for wildfire risk modeling for all locations. Terrafuse also measures the temporal change in carbon density due to calamities such as deforestation and fire.
Illegal Wood Trafficking Tracking
An Estonia-based startup Timbeter employs AI and the world’s biggest database of photometric measurements of roundwood to track roundwood to individual piles and shipments in the online mode to combat timber trafficking and illegal logging.
Similarly, a Germany-based startup Xylene utilizes a fusion of blockchain and supply chain mapping, space technology, automatic data collected by IoT devices, and AI technology for real-time tracking of the wood supply chain.
To summarize, AI technology has been effective in forest management and has a huge application potential in this field. However, the efficacy and reliability of AI/ML systems must be evaluated rigorously before their practical application, as the predictions made by these systems are not always reliable due to uncertainty associated with expert knowledge and data.
References and Further Reading
Ahmadi, V. (2018) Using GIS and Artificial Neural Network for Deforestation Prediction. https://www.researchgate.net/publication/345673436_Using_GIS_and_Artificial_Neural_Network_for_Deforestation_Prediction
Shivaprakash, K. N., Swami, N., Mysorekar, S., Arora, R., Gangadharan, A., Vohra, K., Jadeyegowda, M., Kiesecker, J. M. (2021). Potential for Artificial Intelligence (AI) and Machine Learning (ML) Applications in Biodiversity Conservation, Managing Forests, and Related Services in India. Sustainability, 14(12), 7154. https://doi.org/10.3390/su14127154
Mayfield, H., Smith, C., Gallagher, M., Hockings, M. (2016). Use of freely available datasets and machine learning methods in predicting deforestation. Environmental Modelling & Software, 87, 17-28. https://doi.org/10.1016/j.envsoft.2016.10.006