Forming Weather Weathering Form

kan li

photography info index

title

type

year

format
credits

Forming Weather, Weathering Form.

visual essay

2026

two-channel video · 10’29’
Haoge Gan (co-author WF)
Meiling Zhong (sound design)

Forming Weather Weathering Form (氣 · 象) is a visual investigation of weather forecasting as a process of mediation. The project contends that forecasting is not a neutral unfolding of facts, but a field in which visibility and its limits generate one another. Presented as a dual-structure video installation, the work synthesises archival footage, scientific documents, and meteorological iconography into two intertwined narratives. Forming Weather (象) examines how inscription stabilises atmospheric continuity, rendering it legible and communicable. Weathering Form (氣) traces how weather’s unbound nature exceeds this capture, forcing recalibration and producing new openings at the very points where closure is attempted. This recursion foregrounds uncertainty as a generative friction, where systems confront the limits of their own observation.

Each day, the weather arrives as a broadcast. Stills from Forming Weather Weathering Form, 2026.

Modern weather forecasting is a product of communicational infrastructure, consolidated under coordination imperatives rather than disinterested scientific inquiry. It emerged at the point where weather stopped being a local experience that could be managed through situated discernment and occasional instruments. It became a coordination challenge across an expanding maritime logistics. What mattered was not ignorance of the atmosphere, but the lack of a synchronised means to align dispersed observations into a shared present. In 1859, the Royal Charter storm exposed the catastrophic limit of isolated assessment: the failure lay in the absence of a time-aligned frame that could hold many readings together. Maritime trade, naval power, and wider regimes of risk management demanded anticipations that could circulate ahead of ships, and in advance of the storms that imperilled them. Telegraphy served as the technical catalyst that made this demand implementable. It enabled dispersed readings of meteorological observations to be gathered near-simultaneously, compared, and reissued as warnings across a network. For the first time, observations could outpace the moving front they described, and prediction became a regime of time management.

WWII military plane over Clouds for weather and tactical forecasting.

As forecasting extended beyond the maritime circulation networks, weather forecasting hardened into a procedural language: it had to remain consistent across expanding organisations, extended decision hierarchies, accelerated reporting cycles, and high-frequency issuance, so that many actors could read the same atmospheric state and act within the same time frame. It is mobilised through military apparatuses, where the First World War recast the atmosphere as a tactical volume for ballistics and chemical warfare. The “front”, a term borrowed from the carnage of the trenches, transformed the weather map from a static description into a field of fluid confrontation. It is standardised through aviation logistics, where the rise of international flight demanded that the three-dimensional flux of the air be inscribed into universal protocols. To manage the high-speed movement of aircraft, the weather image evolved into a technical grammar of isobaric surfaces and synchronised icons. It is integrated into risk measurement through standardised indices, where industrial agriculture and insurance markets convert atmospheric anomalies into financial interfaces, attaching weather traces to premiums, contracts, and seasonal expectations. It is disseminated through mass media, where newspapers and broadcasting reformatted forecasts for public address. Warnings and outlooks became communicational devices, synchronising everyday readiness as a shared cadence rather than expert-only interpretation. In such transformations, weather charts and coded reports shifted from interpretive aids to decision inputs: they parameterised readiness, routing, scheduling, and exposure within a single anticipatory frame. By standardising dispersed readings, they make complex phenomena circulate, comparable, and commensurable.

ENIAC produced the world's first successful computer-based numerical weather forecasts, 1946. U.S. Army Photo.

Hand-drawn charts and situated interpretation could only function with sparse stations, slower publication rhythms, and shorter decision-making chains. As observing networks expanded and reporting cycles tightened, forecasting required a form that could be ingested, recomputed, and standardised within a single procedural cycle. Numerical weather prediction emerged when the inscription of weather had to pass through centralised processing without losing coherence. In 1950, at Princeton’s Institute for Advanced Study, Jule Charney, Agnar Fjörtoff, and John von Neumann used ENIAC to compute a 24-hour forecast on the 500 mb surface by numerically integrating a simplified vorticity equation, a demonstration often positioned as the point at which this approach became workable. What matters here is the change in temporality: prediction shifts from interpretive extrapolation to time-advancement. The future is produced as a sequence of discrete steps, each iteration carrying the present forward under specified rules, so forecasting becomes a workflow that can be rerun, evaluated, and inspected, rather than a singular act of reading. Weather, in this framework, appears as a gridded state field; forecasting becomes the controlled propagation of that field through discrete time. The demand for analysis and later data assimilation follows directly, since observations do not arrive as a coherent initial field unless they are organised into one.

Typical instantaneous array of computed total clouds, The GISS Model of the Global Atmosphere,1974. Stills from Forming Weather Weathering Form, 2026.

Constructing such a coherent field required standard communication protocols and a unified temporal framework. The postwar expansion of global observing and data-exchange systems made this integration possible. Over time, the forecasting workflow was embedded in a global knowledge infrastructure composed of sensors, satellites, observing networks, data centres, communications standards and international institutions – what Paul N. Edwards describes as “A Vast Machine." Within this apparatus, global weather is not given in advance. It is assembled as a stable field through successive passages of selection, formatting, exchange, checking, and recomputation, so that a volatile atmosphere can be held long enough to be shared, compared, and acted upon. Under the WMO’s World Weather Watch, the Global Observing System gathers readings from land stations, ships, aircraft, and satellites; the Global Telecommunication System relays them as near-real-time exchange; and the Global Data-processing and Forecasting System concentrates computation and product generation in specialised centres, which then redistribute outputs back through the network. The workflow operates recursively, reorganising observations into a coherent state of the present, immediately projecting that analysis forward. The infrastructure calibrates itself through this cycle, providing the time-aligned, globally commensurable initial field without which numerical forecasting cannot sustain its rapid pace of correction and renewal.

15th October, 1987, Daily Weather Report, Met Office.

From the telegraph onward, forecasting has depended on technologies that align dispersed perception into a shared present. The first weather satellite image of Earth, transmitted by the TIROS-1 satellite, introduced a new way of sensing, turning meteorological observation into televisual and cinematic: the atmosphere was no longer only diagrammed and coded from ground-based readings, but an image captured from the outside. The technological gaze of the satellite did more than expand the field of vision; it exposed a mismatch between inherited classificatory habits and a new regime of seeing: satellite imagery foregrounded textures, vortices, and scale effects that ground-based taxonomies were not built to stabilise. The paradigm shift was not that older classifications were rendered obsolete, but that the sensorium had changed. Communication reconfigured the weather for the public. Television reformatted forecasts into addressable sequences, turning outlooks into scheduled spectacles and shared temporal cues. Mobile interfaces compressed this address into symbols, numbers, alerts, and minute-level updates, relocating weather sensorium from the sky to the screen. In each passage, mediation does not sit outside meteorology as packaging. It operates as a condition of coherence, continuously redrawing what is legible, circulating, and actionable.

Earth-2 Climate Digital Twin Platform. NVIDIA. GTC 2024.

The shift toward hyper-fidelity forecasting is accompanied by the rise of artificial intelligence, which moves forecasting away from the first principles of mathematical equations toward pattern recognition. It reads the atmosphere as accumulated data, predicting the future by identifying statistical correlations across vast historical archives. While this enhances predictive accuracy and efficiency, it consolidates the forecasting process into a sequence of internal correlations, in which the complex links between included data and the variables left outside remain separate. Yet, viewed through this historical trajectory—from the telegraph to the neural network—it has never been about knowing the atmosphere in its own inaccessible essence. It has always been about communication and coordination. The forecast is a mediation that binds atmospheric volatility to the requirements of maritime logistics, industrial agriculture, aviation scheduling, energy load planning, disaster response, and insurance underwriting. In this sense, the forecast must shift from a medium of communication into a mode of configuration: it dictates the modalities by which observations are ingested, the architectures through which computation is organised, the channels of distribution, and the mechanisms by which outputs are integrated into decision-making chains. The appeal of the digital twin lies in its ability to package simulation as an interactive surrogate of reality, particularly in scenario modelling for extreme events, where forecasting is absorbed into broader processes of analysis and decision-making. Critically, high-fidelity models also risk fostering an illusion of seamlessness, which makes mediation easier to overlook. The smoother the output becomes, the more easily the simulated coherence is mistaken for the essence itself. This hyper-fidelity does more than describe atmospheric states; it shapes the very cognitive and strategic frameworks through which we anticipate and act. An irreconcilable gap persists between the limits of technology and the volatility of the atmosphere. What exceeds the system, including rare extremes, out-of-distribution dynamics, unresolved correlations, and the unassimilated residues, returns as signals within the very cycle that seeks to erase it, forcing a constant state of correction, recalibration, and reconfiguration of our apparatus. Forecasting stays oriented toward a future that is never fully present, sustained through ongoing correction at the edges of delay, excess, and misalignment. Weather, in this sense, is not what is predicted, but the generative constraint of the very procedures that attempt to stabilise it. extended reading

Anderson, Katharine. "The Weather Prophets: Science and Reputation in Victorian Meteorology." History of Science 37, no. 2 (1999).

Bjerknes, Vilhelm. “Weather Forecasting as a Problem in Mechanics and Physics”. Meteorologische Zeitschrift 21 (1904): 1–7.

Edwards, Paul N. A Vast Machine: Computer Models, Climate Data, and the Politics of Global Warming. Cambridge, MA: The MIT Press, 2010.

Latour, Bruno. “Visualization and Cognition: Drawing Things Together.” In Knowledge and Society: Studies in the Sociology of Culture Past and Present, vol. 6, edited by Henrika Kuklick, 1–40. Greenwich, CT: JAI Press, 1986.

Lynch, Peter. “The Origins of Computer Weather Prediction and Climate Modeling.” Journal of Computational Physics 227, no. 7 (2008): 3431–3444.

Met Office. “D-Day: The Most Important Weather Forecast in History.” Accessed February 13, 2026.

Morris, Bill. “WMO Weathered the Cold War, but Can It Survive Capitalism? After 150 Years of International Cooperation, Meteorology’s ‘Vast Machine’ Is Adapting to Private Weather Forecasting.” Eos(American Geophysical Union) , 2023.

Moro, Jeffrey. “Systematizing Technique: Grid Techniques for a Planet in Crisis – The Infrastructures of Weather Prediction.” Amodern. 2026.

Pasquinelli, Matteo. “Three Thousand Years of Algorithmic Rituals: The Emergence of AI from the Computation of Space.” e-flux Journal, no. 101 (June 2019).

Völter, Helmut. Wolkenstudien. Cloud Studies. Études des nuages. Edited by Jörn Dege and Mathias Zeiske. 2nd ed. Leipzig: Spector Books, 2014.

World Meteorological Organization. “World Weather Watch.”Accessed February 13, 2026.

Wood, Brian Kuan. “We Are the Weather.” e-flux Journal, no. 45 (May 2013).