For eighteen years, I drove past the Morgantown Generating Station on my way to work. Its stacks were part of the background geography of my daily life, sitting along the Potomac River in Newburg, Maryland. Like many pieces of industrial infrastructure, it was both conspicuous and easy to stop seeing. It was just there.
Then, in June 2022, Morgantown’s two coal-fired units went offline, five years ahead of the retirement date that had been announced just eighteen months earlier (GenOn Holdings, Inc., 2020; GenOn Holdings, LLC, 2021; Global Energy Monitor, 2026). Satellites had been tracking the plant’s emissions for years. Algorithms could model its flood exposure on the Potomac River, estimate its remaining generating capacity, and flag it as a candidate stranded asset. This is spatial finance at its most compelling: the integration of geospatial data and analysis into financial theory and practice (Caldecott et al., 2022), made possible by advances in earth observation, machine learning, and cloud computing. The field has genuine transformative potential, and the institutions building it are serious and sophisticated (UK Centre for Greening Finance and Investment [CGFI], n.d.-b).
But there is a problem the satellite could not see. Morgantown’s financial trajectory was not shaped only by emissions, elevation, generating capacity, or flood exposure. It was shaped by regulation, market structure, local politics, redevelopment pressure, legacy contamination, and the community around it. In other words, it was shaped by human geography. That is the challenge at the center of this post.
What Tools See Well
The Spatial Finance Initiative (SFI), established by the University of Oxford, the Alan Turing Institute, and the Satellite Applications Catapult, has built something remarkable. Its GeoAsset project is creating open, global databases of physical assets across high-emission industries such as cement, steel, and power generation. Those databases link individual facilities to ownership structures and financial instruments (CGFI, n.d.-a).
Overlaid with climate models, elevation data, and remote sensing, they allow investors and regulators to ask which assets are exposed to which physical hazards, with what probability, and on whose balance sheet. The SFI’s 2021 state-of-the-field report found growing applications in insurance pricing, supply chain monitoring, climate stress testing, and portfolio management (Spatial Finance Initiative et al., 2021). For a field barely a decade old, this is a serious intellectual and practical achievement.
This is the appeal of spatial finance. It can make a place like Morgantown legible to capital markets without requiring anyone to stand beside the Potomac and look at the plant. The power is obvious. So is the caution.
What the Tools Still Cannot See
Spatial finance’s reliance on earth observation and physical asset data reflects the tools available to it. But tools shape attention, and attention shapes what gets counted as risk.
The current state of spatial finance is strongest in physical geography, such as geolocated assets, environmental exposure, remote sensing, climate hazards, emissions, land cover, and infrastructure. It is beginning to reach into social, institutional, and governance questions, but those human-geography dimensions remain less mature, less standardized, and less systematically integrated.
This is where the abstraction starts to bother me. I understand why the model wants clean inputs. I also know that places like Morgantown do not become known just because their coordinates are known.
Human geography, broadly defined as the study of how people, societies, and institutions are distributed across and shaped by place, covers factors that are largely invisible to satellites. They still matter to asset value, liability, and investment outcomes.
Social vulnerability. A flood model can place a warehouse in a one-in-fifty-year inundation zone. It cannot show whether the workforce can absorb a two-week shutdown, whether the municipality has the fiscal capacity to rebuild access roads, or whether the community has enough insurance coverage to support recovery. Those questions help determine whether a physical hazard becomes a financial loss. Fox et al. (2024) demonstrated this directly, constructing a vulnerability-adjusted flood risk index that integrates high-resolution fluvial flood modeling with population and poverty data. Their findings reordered the global geography of flood risk by identifying hotspots that conventional physical risk mapping had systematically underestimated. The implication for spatial finance is pointed: a physical risk layer without social vulnerability context can send capital looking in the wrong direction.
Migration and demographic shift. Population movement changes where assets, infrastructure, and labor markets will make sense over the time horizons that matter to long-term investors. A logistics facility that works today may face labor shortages within a decade if regional demographic trends turn against it. A retail asset in a climate-threatened coastal zone may lose demand years before the physical hazard materializes. These dynamics sit squarely within the human layer of geography, and they require demographic, administrative, and migration data that spatial finance frameworks have not yet incorporated consistently.
Land tenure and governance. In emerging market investment, who controls land, under what legal arrangements, and with what institutional stability can determine whether a project moves forward or stalls. Tenure insecurity, customary land rights, and contested ownership create risks that no satellite can observe and no climate model can price clearly. For infrastructure investment in particular, ignoring the social and institutional geography of land leaves a major exposure outside the model.
Community resilience and political economy. Whether a workforce or community will accommodate or resist economic transition is a place-based issue that can change the economics of an asset. Morgantown makes the point concrete. The plant was not just a point on a map or a polygon in an asset database. It was a workplace, a tax base, an environmental legacy, a power market participant, and a neighbor to the communities around it.
Following the coal units’ early retirement, TeraWulf moved in February 2026 to acquire the site, subject to FERC review. The plan involved gas-fired generation, battery storage, and data center load. Google held a 14 percent stake in TeraWulf, which became part of the disclosure dispute around the transaction (Howland, 2026). The transaction immediately drew opposition from the PJM market monitor, the Maryland Office of People’s Counsel, the Sierra Club, local citizens, and public interest groups.
The objections were not prompted by anything a satellite could detect, but by contested questions about community impact, coal ash cleanup liability, water quality, and regulatory disclosure (Howland, 2026). This is the part of the asset’s geography that is hardest to model: not where the plant is, but what claims, obligations, and conflicts have accumulated around it. The varying political responses to coal plant closures across Appalachia, the Ruhr Valley, and the Hunter Valley in Australia reflect the same pattern. Not of physical geography but the social, historical, and institutional fabric of place (Thomas et al., 2019). Rosenman (2019) argued that finance practitioners’ framings of geography have consistently been too thin to capture these dynamics. That critique applies with particular force to spatial finance. The better the physical data layer gets, the more conspicuous the missing human layer can become.
Why the Blind Spot Exists
The absence of human geography from spatial finance is not an oversight born of indifference. It reflects a real methodological problem.
Physical data derived from earth observation is observable, comparable, consistently structured, and increasingly available in near-real time. Social data has none of these qualities. Migration patterns, land tenure arrangements, community resilience, and institutional quality vary widely across geographies. They resist standardization and are often politically sensitive to collect and publish. Turning them into inputs that can be used across an investment portfolio remains difficult.
ESG represents the financial sector’s most sustained attempt to incorporate social and governance factors alongside environmental ones. The “S” in ESG points toward human geography concerns: labor standards, community relations, and human rights due diligence. But as Control Risks (2023) observed, the industry often treats social risk as fundamentally harder to quantify than environmental risk, which tends to follow science-led measurements and targets. As a result, social factors in ESG are typically assessed at the company level. They are abstracted away from the specific places and communities where financial consequences materialize.
That abstraction from place is exactly the sort of problem spatial finance, with its asset-level granularity, should be able to help correct and it makes the gap more striking.
Why It Matters Now
The methodological difficulty of incorporating human geography into spatial finance explains the gap. It does not justify it, particularly as the financial consequences of its absence become more visible.
This is not just a matter of adding nuance. The map changes when social vulnerability is added. Fox et al. (2024) showed that integrating social vulnerability into flood risk modeling does not simply add nuance to existing assessments. It changes them materially, identifying high-risk hotspots that physical hazard mapping alone had missed and downgrading others it had overstated. If the same dynamic holds across other hazard classes, then spatial finance tools built exclusively on physical data are not just incomplete; they can point investors and regulators toward the wrong conclusions.
The money case is just as pressing. Morgantown’s coal units were retired early not because a flood destroyed them, but because unfavorable economics, environmental compliance costs, competition from other generation types, and evolving market rules made continued coal generation uneconomical. That was a regulatory and political outcome, not a physical one (GenOn Holdings, LLC, 2021).
The site now carries a substantial coal ash cleanup issue under the 2024 Legacy Coal Ash Rule, a contested ownership transfer under FERC review, and unresolved community concerns about water quality and redevelopment terms (Earthjustice, 2025; Howland, 2026). None of those risks are captured by the satellite record.
More broadly, whether the transition unfolds smoothly or generates stranded asset write-downs, stranded worker liabilities, and political backlash is almost entirely a function of place-based social context. That transition, meaning the managed wind-down of fossil fuel assets and the communities economically dependent on them, depends on local labor markets, regional fiscal capacity, the history between communities and extractive industry, and the political power of affected populations (Thomas et al., 2019). Physical geography locates the coal plant. Human geography helps explain whether closing it will lead to ten years of litigation, regulatory reversal, or reputational damage.
In Morgantown’s case, the interesting question was never only where the plant sat. It was what the plant meant to the market, the regulators, the owner, the surrounding communities, and the river beside it. Spatial finance currently has sophisticated tools for the former and no comparably mature toolkit for the latter.
Migration-driven demand shifts are another way that human geography can show up on the balance sheet, and they are only beginning to receive serious attention. As climate change accelerates internal and cross-border population movement, the human geography of asset demand becomes as important to long-term valuation as physical exposure to hazard.
Investors will need to understand who will live near, work in, or use a given asset in twenty years. A spatial finance toolkit without that human layer is poorly equipped to answer that question.
Toward a Fuller Geography of Financial Risk
None of this requires the field to abandon its strengths. The field’s asset-level granularity and earth observation infrastructure are what make the human layer worth adding. The physical layer provides the scaffold onto which social data can be meaningfully anchored.
The building blocks are already available. Social vulnerability indices, developed extensively in the disaster risk and public health literature, are increasingly available at fine spatial resolution. They could sit alongside elevation and flood layers in climate risk workflows (Fox et al., 2024). Census, administrative, and migration datasets offer tractable proxies for demographic shift and labor market exposure. Meanwhile, the financial geography literature has spent decades studying how money, institutions, and power move through places in ways that spatial finance still underweights (Wójcik, 2025).
The immediate need is not only new data. It is also stronger interdisciplinary connections between the satellite and data science community that built these tools and the human geographers, demographers, and political economists who understand what satellites cannot see. Regulatory momentum may help. The social provisions of the EU’s Sustainable Finance Disclosure Regulation and the proposed Corporate Sustainability Due Diligence Directive are beginning to ask questions about people and places that purely physical frameworks cannot answer (Control Risks, 2023).
Conclusion
Morgantown’s satellite record is comprehensive. Its dimensions, emissions history, and flood exposure on the Potomac are all, in principle, knowable from orbit. That matters. Those measurements are useful, and spatial finance is right to take them seriously.
But they do not explain why the plant closed when it did, who bears the cost of what it left behind, or whether the community that lived alongside it for fifty years will accept what comes next. Those regulatory, political, social, and historical questions determined Morgantown’s financial trajectory, and they are the ones spatial finance has no systematic way to answer yet.
That does not make spatial finance wrong. It makes it incomplete in a familiar way. The map is getting better, but the territory still includes institutions, memory, power, liability, and trust. Until those are part of the model, some of the most important risks will remain outside the margins.
References
Caldecott, B., Braemer-Evans, S., Hickey, C., Jahn, M., Kruitwagen, L., McCarten, M., & Tomlinson, S. (2022). Spatial finance: Practical and theoretical contributions to financial analysis. Journal of Sustainable Finance & Investment. https://doi.org/10.1080/20430795.2022.2153007
Control Risks. (2023). Managing social factors in investments: Not losing sight of the “S” in ESG. https://www.controlrisks.com/our-thinking/insights/managing-social-factors-in-investments
Earthjustice. (2025). Toxic coal ash in Maryland: Addressing coal plants’ hazardous legacy. https://earthjustice.org/feature/coal-ash-states/maryland
Fox, S., Agyemang, F. S. K., Hawker, L., & Neal, J. (2024). Integrating social vulnerability into high-resolution global flood risk mapping. Nature Communications, 15, Article 3155. https://doi.org/10.1038/s41467-024-47394-2
GenOn Holdings, Inc. (2020, December 18). GenOn Holdings, Inc. announces retirement of Morgantown coal units. https://www.genon.com/genon-news/genon-holdings-inc-announces-retirement-of-morgantown-coal-units
GenOn Holdings, LLC. (2021, June 9). GenOn Holdings, LLC announces retirement of three coal-fired power plants. https://www.genon.com/genon-news/genon-holdings-llc-announces-retirement-of-three-coal-fired-power-plants
Global Energy Monitor. (2026, February 12). Morgantown Generating Station. https://www.gem.wiki/Morgantown_Generating_Station
Howland, E. (2026, April 3). FERC urged to reject TeraWulf’s power plant purchase due to undisclosed Google ownership stake. Utility Dive. https://www.utilitydive.com/news/ferc-terawulf-morgantown-power-plant-google/816582/
Rosenman, E. (2019). The geographies of social finance: Poverty regulation through the ‘invisible heart’ of markets. Progress in Human Geography, 43(1), 141–162. https://doi.org/10.1177/0309132517739142
Spatial Finance Initiative, Satellite Applications Catapult, & ConsultingWhere. (2021). State and trends of spatial finance 2021: Next generation climate and environmental analytics for resilient finance. Spatial Finance Initiative. https://www.cgfi.ac.uk/wp-content/uploads/2021/07/SpatialFinance_Report.pdf
Thomas, K., Hardy, R. D., Lazrus, H., Mendez, M., Orlove, B., Rivera-Collazo, I., Roberts, J. T., Rockman, M., Warner, B. P., & Winthrop, R. (2019). Explaining differential vulnerability to climate change: A social science review. WIREs Climate Change, 10(2), e565. https://doi.org/10.1002/wcc.565
UK Centre for Greening Finance and Investment. (n.d.-a). GeoAsset Project. https://cgfi.ac.uk/spatial-finance-initiative/geoasset-project/
UK Centre for Greening Finance and Investment. (n.d.-b). Spatial Finance Initiative. https://cgfi.ac.uk/spatial-finance-initiative/
Wójcik, D. (2025, November 15). Finance and economic geography: Where does money come from and where is it going? Journal of Economic Geography, lbaf051. https://doi.org/10.1093/jeg/lbaf051
Header image: Ravens326 at English Wikipedia, Public domain, via Wikimedia Commons
