Cloud Flow
Visualizing Long-Term, Large Scale Atmospheric Processes
View MP4 videos

Cloud Flow is part of an ongoing project intended to provide a visual index and perceptual experience of long-term, large scale atmospheric and environmental processes. The goal is to give the viewer a palpable impression of the climatic processes in which we are immersed. These works will hopefully provide a sense of the dynamism and even violence of atmospheric processes and perhaps instill a greater respect for the global events which are being altered by human activity. Cloud Flow is related to the Landscape Temporal Variations, but the time frame is expanded from days to years, and the field of view is expanded from localities to continents. The first series of Cloud Flow videos compress a year of weather satellite imagery into a five minute animation. The series depicts the contiguous states of the U.S., showing much of the North American continent. A related body of work, Storms, animates images of weather radar returns over the same time period and expanse. The field of view of these pieces manifests my interest in events with an expansive spatial extent, such as the Reentry Project. Cloud Flow continues my interest in clouds and related atmospheric processes, in common with most of the Sky/Ground works and many of my videos.

Viewpoints of the two GOES satellites
Two satellites sit in geostationay orbits over the equator
able to image these views of the earth.

I am interested in the landscape sublime and how it might be manifested through contemporary environmental sensing and imaging systems. Since initial descriptions of the sublime by Burke and Kant, intense weather and storms have been considered to belong to the realm of the sublime. Nineteenth century landscape painters often attempted to capture the sublime in the landscape as manifested by powerful storms, perhaps most notably Turner. In the nineteenth and twentieth century, as urbanity expanded, humanity became desensitized to the landscape sublime as it was superseded by the technological sublime. Now, with environmental catastrophe threatening, the landscape sublime has begun to reemerge in our consciousness. In this work I seek to provide a modern technological window into the landscape sublime.

Lat/Long aligned image from GOES East
Florida is visible in the center just to the left.

These videos exhibit an extreme temporal compressnion of a year into minutes. I first become interested in compressing landscape time as a means to visualize proposed land art pieces which would evolve very slowly over time. The Cumulus work would generate clouds in geometric patterns but those patterns might only be discernable over durations of hours. I began to do time-lapse studies of clouds and the landscape, and became fascinated with the insights available through this process of temporal modification.

Unlike most of my video work, these projects employ found images. Cloud Flow is an animation of infrared images from the GOES (Geostationary Operational Environmental Satellite) weather satellites. The companion works of Storms employs synthetic imagery constructed from data from the NEXRAD network of ground-based weather radars. In Cloud Flow, there is a camera-like optical system which creates the image, while in Storms the image is constructed from sensor data. There is no optical imaging involved. In both cases, images were retrieved from an expansive archive of weather-related data and imagery maintained by the Iowa Environmental Mesonet, a partnership of Department of Agronomy at Iowa State University, the Department of Hydro Science and Engineering at University of Iowa, the Iowa Depart of Transportation and the National Weather Service, supported by grants from the National Science Foundation. This information is in the public domain and stretches back to the late nineteenth century. It is commonly used by researchers in the earth sciences.

I began working with the NEXRAD images for Storms in 2008. in 2010 I discovered the GOES images and began the Cloud Flow project. The first series in Cloud Flow is "CONUS, 2006-Present. CONUS is a technical term used by the U.S. Department of Defense, General Services Administration, NOAA/National Weather Service, and others and has been defined both as the CONtinental United States, and as the 48 CONtiguous states.

Lat/Long aligned image from GOES West
California and the Baja Peninsula are visible on the right.
The curved arc is the result of the perspective transformation.

Things To Look For

While it may be difficult to observe long-term climatic events by comparing multiple years, on a smaller time scale a number of significant phenomena can be seen. The images are formed in the infrared, which enables cloud images to be visible during the night. The infrared is sensitive to temperature. The rhythmic pulsing of the ground is the result of diurnal heating and cooling of terrain. During the winter months the production of "lake effect" snow is quite visible, as the Great Lakes actually appear to be steaming. During the summer months the production of convective cumulus, particularly in the Southwest and Midwest is very visible, as thermals form from differential ground heating. Throughout the year, lee clouds can be seen to form east of the Continental Divide, triggered by the prevailing Westerly winds. What is most noticeable is the constant swirls of high and low pressure fronts, with highs circulating clockwise and lows counter-clockwise. During the late summer, hurricanes can be seen to swirl up the East coast, sometimes entering the interior to gradually die out.

The Ethics of Employing Found Imagery

In an unrealized curatorial project I focused on the work of artists who create works entirely from found imagery, very often from various forms of surveillance systems. Satellite imagery certainly must be considered as surveillance imagery, whether from the latest classified ground observation systems to the more benign weather satellites employed here. In an age of image saturation some artists have eschewed the creation of new images, considering this a form of pollution, and instead concentrate on editing or harvesting from the plethora of images in which our culture is immersed. While this might be considered a from of iconoclasm, it is an idea which merits thoughtful consideration. The filmmaker Manu Luksch has authored a Manifesto for CCTV Filmmakers, which advocates methods for employing surveillance footage as a medium. In the early 1980s Peter Fend formed the group OECD which began to exhibit footage selected from earth resources monitoring satellites such as LandSat in support of environmental concerns. German media artist Thomas Koner has produce a number of haunting videos using footage from traffic or surveillance cams. British net art duo Thomson and Craighead's networked installations assemble "Short Films About Nothing" from webcam footage and web audio, providing an absurdist and humorous response to the proliferation of web images. Jill Magid succeeded in turning surveillance cameras on herself to create an intimate and disturbing self-portrait, convincing British police to train their web cams on her as she moved about Liverpool. All of these artists have produced compelling works without directly producing their own imagery. The use of imagery from surveillance systems not only avoids adding to the our cultural glut of imagery, but also draws attention to the ubiquity of surveillance in our society, whether it be of shoppers at the mall or the weather patterns of the globe.

Combined GOES East/West Image
Most composite images possess this visible splice point.
Both brightness and time differences are visible.


The creation of these animations involves many operations on the original satellite imagery. First, an image of CONUS must be composited from views from two satellites, GOES East and West. An image of the US from the East coast to about -111 longitude is provided by GOES East, shown above. To obtain the remainder of the western US the Goes West satellite image must be used. NOAA (the Nation Oceanic and Atmospheric Administration) provides images from GOES which are transformed so that their edges align with lines of latitude and longitude. This transformation is required because of the oblique view of North American seen by the satellites. These transformed images can be downloaded from the NOAA website. Mesonet obtains these images, and combines and crops them to make a complete CONUS image. These are the images I downloaded from Mesonet.

Each video frame of the Cloud Flow animations, at a rate of thirty frames per second, represents one hour. A year is therefore comprised of 8760 frames (8784 for leap years) and this many frames must be downloaded to produce the animation for a year. I wrote a PERL script to perform this download, which typically takes about a day to complete for one year of images. The operation is complicated by the fact that many images are missing due to data communication errors, so these missing frames must be tallied and handled correctly. Two GOES images are available for every hour, so if the first is missing the second is used.

Dropouts in Goes West Image
The GOES West image in this composite has substantial dropouts.
These occur in about one percent of images and must be manually retouched.

Once the images are downloaded the effects of combining two different satellite images must be handled. A typical composite image has a pronounced splice point. Not only are the view points different, resulting in different brightness levels, but the time of the image can differ also. I wrote a stitching algorithm in C++ which minimizes the brightness differences and introduces some x axis blurring to smooth the time difference.

About 1% of the images have drop-outs, consisting of dark bands, where the satellite data link was interupted during the image transmisstion. These are manually retouched, which can take several days of work. Finally, the image is resampled to conform to the HD aspect ratio and a border added so that the image aligns with the Storms images. I added a small timer in the lower left of the images showing the year, month, date, and hour, to provide a temporal reference. About two percent of hourly images are missing and these are simply skipped over, resulting in a running time for each animation which differs and is slightly less than the four mintes, fifty three seconds which the above frame counts would produce.

Finally, after all these operations are completed, the image sequance is encoded as a 1080 HD video. The web videos are encoded as SWFs at reduced resolution. The years 2006 through 2010 are currently complete and work on 2011 to the present is currently in progress. Mesonet only has images from 2006 to the present but work is underway to obtain prior years imagery directly from NOAA. The intent is to eventually have every year from 2000 to the present.

Villanueva, NM