In the late 1990s and early 2000s I worked in critical infrastructure protection, building geospatial applications and data sets to model how infrastructure networks behave. I worked on systems like switched telecommunications, commercial rail, and refined petroleum products, along with the ways those systems depend on one another. Working on those teams, I learned a lot about infrastructure analysis and assessment that I continue to carry forward even now.
Since then, “critical infrastructure” has become a colloquial synonym for “infrastructure,” but “critical” originally had a material distinction. The 9/11 attacks made the idea of protecting domestic assets demonstrably real. It was not possible to equally protect every asset everywhere all the time, so “critical” could not simply mean “infrastructure.” It needed a framework, and for us, that framework was called “mission assurance.”
Mission assurance reduces to four basic questions. What are you trying to do? What assets do you need to do it? When do you need those assets, and for how long? Can the mission continue without it?
The actual methodology was more structured than that, and most of its detail has receded from my memory, but its primary assertion did not. Criticality is not a property of an asset. It is a conditional state, set by use case, dependency, and time window. The same asset is critical in one context and irrelevant in another, and nothing about the asset itself tells you which.
The basis of criticality can change across geography. It is always conditional, but the conditions that make something matter can be easier to identify in some domains than in others. As you move from physical assets, through the networks that connect them, toward the human terrain that gives them a reason to matter, more of that determination has to come from outside the data that describes the asset itself. That gradient, more than any particular technique, is what has stayed with me.
At the physical geography end of the spectrum, criticality is conditional but understandable. Define the mission and its dependencies tend to fall out of it. Determining minimal cut sets in highly distributed networks is not trivial, but it is still a constrained problem. A fiber OLT that disables downstream ONTs when it goes offline is critical within the physical confines of its fiber network. It has a defined answer, even if reaching it is difficult.
Power, transport, communications, water, and logistics are multivariate and deeply interdependent, but comparatively deterministic. Disturbances propagate in complicated ways that are still structured enough to model, simulate, and trace. The work leans on field verification, complex data acquisition, and incomplete observability, but its difficulty has structure. You can see it, test it, and work around it. Set the mission, and the map of what matters mostly computes.
Such analysis transitions into human geography as higher level impacts and mitigations in terms of policy, economics, and operational behavior come into play. Disruptions to physical networks affect the humans and the organizations that use and depend on those networks.
A rail line is not critical in the abstract. It becomes critical when it is moving a specific feedstock to a refinery, linking a distribution center to a port, or carrying the components that keep a production line running. A plant that upholsters bucket seats for an auto line is not sufficient because it is intact. It has to be staffed, supplied, reachable by road or rail, connected to power and communications, and able to get its output to the next plant in time to keep the line moving.
Disruptions to networks and supply chains often involve nuance and negotiation. Inventory may be on hand, a second supplier may be available, or production may be re-sequenced around the gap and a plant may keep operating. Conversely, it could idle within hours because the line is tightly synchronized, substitutions are not approved, trucks cannot be rerouted in time, or management decides partial production is not worth the disruption. The physical event is identical in both cases. Criticality is determined through inventory policy, supplier contracts, approval chains, and someone’s judgment about acceptable loss.
Even at the most physical and computable end of the spectrum, the thing that finally sets criticality is not in the physical data. It lives in an institutional and human layer above it. Mission assurance is useful here not because it replaces physical analysis, but because it shows how human geography overlays onto it. Physical networks can be modeled. The impacts of disruptions to those networks in the context of policy, economics, and operational behavior are harder to model deterministically.
Once the analysis enters into that context, human geography becomes the main analytical problem. In physical geography the categories tend to hold steady. A substation is a substation, and a rail segment is a rail segment. In human geography they do not. Ethnic, cultural, informal, and behavioral factors cross borders, overlap one another, and shape what happens without matching any physical feature or de jure boundary. More importantly, some of the factors that finally determine criticality, especially policy, use case, and operational behavior, are among the least likely to be fully persisted in data and among the hardest to model cleanly in traditional data architectures.
This is not to suggest that human geography resists automation. Some of it reduces to constrained estimation and automates well. Automated redistricting and land-use classification have existed for decades. Recent Stanford work on AI-assisted poverty and development mapping is a clean example, and geospatial foundation models can pull real patterns out of spatial data that are relevant to human terrain. Human geography can be less deterministically automatable than physical geography, not uniformly so. The greater challenge is deciding how policy, use case, and operational behavior change the meaning of what the data shows.
That is where mission assurance becomes more than a framework from one corner of government work. It is a concrete example of what human geography does in geospatial analysis. Physical geography tells you what is there, how it is connected, and what has changed. Human geography tells you why those things matter in context. Mission assurance was our way of applying that question to infrastructure. It helped determine what was important for a specific purpose, in a specific time window, under specific constraints. In that sense, it was human geography laid over physical geography in operational form.
I’ve been thinking about this more as geospatial AI moves beyond physics-based EO. Pixels can depict floods, burned areas, damaged infrastructure, traffic, construction, and land use change and those features can be extracted from the pixels for further analysis. On their own, they don’t answer the question “so what?” That is the role of human geography. It supplies the relevance, purpose, and consequence. It helps us decide what is important and where to allocate resources. It is more difficult, though not impossible, to capture with satellites. It requires more communication, collaboration, and agreement, but that is the part that makes it human.
Header image: Carl Young, CC BY-SA 4.0 https://creativecommons.org/licenses/by-sa/4.0, via Wikimedia Commons
