Cloud Cover Over Kosovo

Clouds over Yugoslavia and the Adriatic Sea (NASA)

Cloud Cover Over Kosovo

John S. Bohlson, Leslie O. Belsma, Bruce H. Thomas

In response to an Air Force request for better weather forecasting in the Balkans, Aerospace developed a higher-resolution cloud-analysis prototype that provided more accurate cloud-cover information in support of operations in Kosovo.

Earth's cloud cover frequently affects the outcome of modern combat because sophisticated aircraft (and weaponry such as laser-guided missiles and night vision sights) do not operate reliably in the presence of clouds. Knowing the state of the cloud cover can determine the success of reconnaissance missions, and it is the most critical factor in the accuracy of humanitarian airdrops. To be of value, cloud data must be current and accurate, but such data can be difficult to obtain, especially in areas where access is limited by military or political restrictions (see sidebar, Weather and Warfare).

Brig. Gen. Fred Lewis, director of Air Force Weather, recognized in 1998 a need for improved cloud data to support military operations in the Balkans. Gen. Lewis wanted a cloud-analysis model that could do a better job at analyzing and forecasting clouds than either the Air Force Weather Agency (AFWA) model then in use or even the extensive upgrade under development at the time.

The grid is an array of points superimposed on a map of Earth's surface. Observations of clouds are not taken at grid points, but at irregularly spaced points. Nephanalysis is the process that interpolates cloud data observations to the points on the grid. The distance between adjacent grid points on the AFWA polar stereographic whole-mesh reference grid is 381 kilometers at 60 degrees latitude. All finer-resolution grids are defined relative to this whole-mesh reference grid. For example, the distance between points on an 8th-mesh grid is 48 kilometers; on a 16th-mesh grid, 24 kilometers; and on a 64th-mesh grid, only 6 kilometers.

Because cloud cover data at 64th mesh presents much finer detail than data at 8th mesh, data at 64th-mesh is called fine-grid data. Data from a lower-resolution grid, such as that based on an 8th mesh, is called coarse-grid data.

The Aerospace Corporation stepped in to develop a prototype system that leapfrogged over the planned AFWA cloud model upgrade to deliver automated cloud-analysis products (such as amount of cloud cover or cloud classification) at a much higher resolution. This prototype became the basis for a cloud-analysis system that was then put into operation in less than 60 days in the spring of 1999 to improve weather support for the war in Kosovo. The improved resolution allowed forecasters to provide more accurate cloud-cover predictions to the battlespace planners and pilots.

Cloud-Analysis Model

The Air Force Weather Agency, at Offutt Air Force Base, Omaha, Nebraska, has used automated cloud-analysis models to generate quantitative information on clouds since 1970. The earliest AFWA three-dimensional cloud-analysis model, the 3-D Neph, used space-based cloud-cover imagery from Defense Meteorological Satellite Program (DMSP) satellites. The current model, RealTime Nephanalysis, known as RTNeph, combines ground-based observations with data from DMSP and the Television and Infrared Observation Satellite (TIROS) of the National Oceanic and Atmospheric Administration to produce worldwide cloud analyses at a 48-kilometer resolution. It computes the number of cloud layers, the percentage of cloud coverage, and the height of the base and top of each layer on a 48-kilometer grid. A cloud-forecast model uses the analyses to produce cloud forecasts at this same grid resolution.

U.S. Air Force Drawing of a DMSP Block 5D-2 spacecraft.

DMSP Cloud Data

Defense Meteorological Satellite Program (DMSP) satellites have been providing worldwide cloud imagery for national programs since 1966. The Air Force Weather Agency uses data from three-dimensional cloud analyses in developing computer cloud-forecast models for the military. Data from DMSP satellites formed the cornerstone of the Aerospace protoype cloud model. The U.S. Air Force has launched more than 30 DMSP satellites. The constellation includes at least two sun-synchronous polar-orbiting satellites flying at about 800 kilometers above Earth, with one satellite orbiting in early, and the other in late, morning.

Unlike other meteorological satellites, DMSP provides imagery at the edge of its 3000-kilometer swath that nearly matches the quality of imagery directly below the satellite. The primary sensor, the operational line scan, collects cloud imagery in a visible and a long-wave-infrared band. The operational line scan calibrates, indexes, and stores the data for transmission. During daylight, the fine-mode resolution of the visible-band data is 0.62 kilometers, and the resolution of the infrared-wavelength data is 2.8 kilometers. Fine data is collected on a regional basis up to a quarter obit. On-board smoothing is used to decrease the data rate (and therefore resolution) to provide data for the entire orbit. The operational line scan also has a unique capability that allows it to gather visible-light data at night at a 3.5-kilometer resolution with as little as one-quarter-moon illumination. Additional satellite sensors measure atmospheric vertical profiles of moisture and temperature and a variety of space environmental parameters.

DMSP has proved to be a valuable tool in scheduling and protecting military operations. The last of the Block 5D-2 series of satellites was launched April 4, 1997. The Block 5D-3 series, the first of which was launched in December 1999, accommodate larger sensor payloads and feature a larger power supply, more on-board memory, and increased battery power that will extend the life of the satellites from the current four years to five.

Improved Resolution Enhances Forecasting

The Air Force Space and Missile Systems Center is developing the upgrade to AFWA's current cloud-detection and forecast system. The new system, known as CDFSII, will increase the resolution of the AFWA cloud analyses and forecasts from an 8th mesh (48-kilometer) grid to a 16th mesh (24-kilometer) grid. Also, by combining data from multiple weather satellites, it will improve cloud detection in stressing conditions such as low clouds and fog, thin cirrus clouds, and tropical clouds.


Image of cloud cover over the Balkans, April 15, 1999, generated by the current AFWA low-resolution cloud-cover-analysis model with grid points 48 kilometers apart.	Clouds over the Balkans on the same date in a higher-resolution image generated by the new CDFSII cloud-cover analysis model. Grid points are 24 kilometers apart.	The highest-resolution image of the same area on the same date generated by the Aerospace prototype model, with grid points 6 kilometers apart.

When completed in December of this year, the new CDFSII system will

provide a multiple-satellite data-acquisition system that combines the high spatial resolution of DMSP imagery with multispectral data from TIROS
merge the global coverage of these polar-orbiting systems with the frequent refresh available from an international constellation of geostationary weather satellites
perform multiple-satellite-specific cloud detection using science algorithms from SERCAA (Support of Environmental Requirements for Cloud Analysis and Archives)
use clustering techniques to accomplish cloud layering and typing and then combine these independent satellite cloud data records using an optimal interpolation scheme
feed the cloud analyses into a single-cloud forecast model to deliver short-term (12-hour) and long-term (48-hour) forecasts

The AFWA polar stereographic whole-mesh reference grid. The distance between the grid points is 381 kilometers at 60 degrees north or south latitude.

Battlespace Weather Forecasting

Although CDFSII will improve forecasting beyond that provided by the current system at AFWA, Gen. Lewis decided that an even finer-scale automated cloud-analysis and forecasting system was needed to support operations, specifically the "weather function," in the Balkans. Under a new centralized support concept, an Operational Weather Squadron was activated in Germany. The Weather Squadron performs the weather function continuously during the intelligence preparation of battlespace in a series of steps that converts weather data into intelligence and communicates it to users. The weather officer collates weather information collected throughout the battlespace, combines it with weather data received from weather flight observers and forecasters and from higher headquarters, and then generates the weather forecasts.

The Aerospace Prototype

In response to Gen. Lewis's call for high-resolution cloud analyses to better support the Weather Squadron, Aerospace developed a prototype cloud-analysis model at a fine-scale grid, increasing the resolution to 6 kilometers (64th mesh) from the 24 kilometers planned for the CDFSII system.

Aerospace developed the prototype by modifying the CDFSII SERCAA algorithms to improve computing speed. Faster processing enabled use of DMSP "fine mode" data, which is higher resolution than the DMSP "smooth mode" data used in CDFSII.

The Aerospace cloud-analysis prototype represents the first quantitative use of DMSP fine-mode data in a meteorological model. Fine data is not available worldwide, so the prototype model produces the 64th-mesh fine grid regionally using the fine data and then combines it with a worldwide analysis at the coarser CDFSII grid using DMSP smooth-mode data.

The new model began generating cloud analyses as a prototype in January 1999 at the Aerospace Environmental Application Center facility at AFWA. The Kosovo conflict prompted the Air Force to issue a request to put this prototype system, along with high-resolution forecasting capability, into operation within 60 days. Within two weeks of the Air Force request, Aerospace wrote code to post fine-grid cloud data over Kosovo from the prototype system as images on the Air Force Weather Information network, which provides data to weather forecasters in Europe. Color-coding was added to highlight aspects of the cloud mask.

Aerospace cloud-cover analysis prototype image of the Mediterranean area, April 6, 1999. A visual representation of the gridded cloud data (cloud mask) can be used by the weather officer to quickly assess cloud conditions. Tools added to the prototype system enable color-coding of the data to highlight aspects of the cloud mask.

From Laboratory to Battlespace

Aerospace, AFWA, Atmospheric Environmental Research, and Sterling Software (the CDFSII contractor) worked together to develop an operational system based on the Aerospace prototype. The prototype had to be ported from the laboratory environment to a 24-hour-a-day operational capability, expanded to include multispectral TIROS algorithms, and combined with the forecast model. A capability to "tune" the algorithms on a regional basis to produce a better cloud analysis was added.

Tuning is not as critical for the TIROS algorithms because additional channels allow for more cloud-discrimination tests, but because DMSP has one visible and one infrared channel, a single-test cloud-detection algorithm that is very sensitive to threshold settings is used. Aerospace found that using the same thresholds in the DMSP algorithms for each satellite was inadequate because of differences in DMSP smooth- and fine-mode calibration. Fine data typically exhibits a 5-degree Kelvin cold bias, so a bias term was added to the fine-mode infrared threshold values that greatly improved the fine-grid (64th-mesh) results.

A major task in the transition of the prototype to an operational forecast system involved modifying the CDFSII coarse-grid cloud-forecast model to run at the higher resolution (6 kilometers). Code was written to overwrite the coarse-grid values with fine grid for all grid points touched by one satellite swath of DMSP fine data, which is typically an eighth or less of an orbit. The forecast model then advected the resulting fine-grid cloud analysis with high-resolution winds from an operational mesoscale model to produce the cloud forecast.


Cloud forecast image of the Balkans on April 6, 1999, derived from the CDFSII 24-kilometer analysis.	Cloud forecast image of the Balkans on April 6, 1999, derived from the Aerospace prototype 6-kilometer analysis.

The team completed the system in less than 60 days, and Gen. Lewis specifically commended Aerospace for this support to the nation's warfighters. The operational implementation of the Aerospace prototype cloud model improved weather forecasting for operations across the Balkans by providing more accurate cloud-cover data to the air tasking, order-planning, and execution processes. This system is still in use to support regional high-resolution cloud forecasting.