This week Elon Musk and Company are calling for an “AI-First” strategy in government. I am not sure I am onboard with a totally autonomous system nor the removal of millions of workers in the federal workforce, but what I am certain of is that AI is useful, complicated, and evolving. A good definition of an AI system is that it is an ecosystem of infrastructure for models, data pipelines, processes, and teams. The ecosystem promotes end-to-end computer solutions of real-world problems that are sustainable and align with societal values. An AI system is the practical deployment of models delivering value at scale.

Following 20+ years as a geospatial professional in the USDA Forest Service (35+ years in federal service), I had been tasked with the coordination and integration of geospatial technology into the federal agency on a local unit (i.e., on the Monongahela National Forest), at a regional level (i.e., Region 1 based in Missoula, Montana), and at a national level at the Geospatial Technology and Applications Center (GTAC) in Salt Lake City. This experience taught me that the evolution of this new technology and applied use in an ecological land management context was fraught with complexity and required an interdisciplinary framework for capturing programmatic content – lands, engineering, recreation, wildlife, soils, geology, law enforcement are representative programs with professional staff and a wealth of data that required analysis, management, and coordination.

AI models and systems hold great promise for augmenting geospatial professionals and resource specialists in their work as well as providing foundational insights for Agency leadership in making decisions. Current techniques and tools exist for automating much of the processes right now, but the integration of AI systems and task-based models would help to not only solve real-world problems but potentially reveal emergent and innovative approaches to natural resource management.

Current AI applications in the natural resource and land management realm are primarily task-based and concentrated in the remote sensing arena.  Several remote sensing examples include feature extraction and change detection, predictive analytics, and natural language processing. Specific real-world examples that have been developed are:

§  Future of AI in natural resource management: Self-Learning Forest Growth Model (Liang, 2023)

Mr. Liang is co-director of the Forest Advanced Computing and Artificial Intelligence Lab and is leading a project to use AI to map global forest carbon accumulation rates (Liang, 2023).

§  Exploring artificial intelligence for applications of drones in forest ecology and management (Buchelt et al, 2024)

This article is an in-depth look at utilizing drone technology for real-time tracking of forested landscapes for wildfires, pests, and inventory needs to name a few.

Another great source of information on applied AI in the geospatial world is Esri. Esri is the leading commercial vendor for geospatial software with their ArcGIS suite of applications. They are leading the GeoAI charge which is what they define as:

 “…the application of artificial intelligence (AI) fused with geospatial data, science, and technology to accelerate real-world understanding of business opportunities, environmental impacts, and operational risks. Organizations are modernizing operations to run at scale through automated data generation and approachable spatial tools and algorithms” (Esri, 2024). 

Esri also has a free ebook available for download which describes “AI + Location Intelligence” here.

A major challenge is that data integrity has been and is a major concern for grappling with Big Data – those big data pipelines are integral to not only the training and testing of data for AI, but also for the long-term maintenance of an AI system. At the same time Agency leadership has struggled with data governance and staffing – there will be similar struggles for implementing AI across an organization.

Key tools for fighting wildfires (and post-fire mitigation) will be the expanded use of drones and the real potential of integrating AI into the battle. Fundamentally though, it will be a mistake to dismiss the need for a highly structured methodology in not only integrating the technologies (with robotics), but for prioritizing training needs for staff use and incident coordination.

As AI technology matures and begins to deploy autonomously – with model methodology making a transition to AI systems from task-based models – there will a greater emphasis and success for working across disciplines and engaging interdisciplinary teams in land and natural resource management.