Document created: 27 April 03
Air University Review, January-February 1976

Weather Probability Forecasts
A Cost-Saving Technique in Space 
Launch and Missile Test Operations

THE Space and Missile Test Center (SAMTEC), an operational component of the Air Force Systems Command, manages the Western Test Range (WTR), which extends from the launch head at Vandenberg AFB, California, to the Indian Ocean. Ideally located for its mission between Point Sal and Point Conception on the rocky central California coast, SAMTEC can launch polar orbiting vehicles southward and ballistic missiles westward without overflying populated areas during the critical first few minutes of flight. The facilities of the Western Test Range, which has been designated by the Secretary of Defense as a National Range, are utilized by DOD agencies, NASA, and other DOD-approved range users for the conduct of aerospace and related test programs.

Ballistic missiles launched from Vandenberg are normally targeted so that the re-entry vehicles (RV's) re-enter the atmosphere and impact at designated, highly instrumented target areas on the Kwajalein Missile Range (KMR), operated by the U.S. Army in the Marshall Islands, or on the SAMTEC-operated Canton Island complex in the Phoenix Islands. Land-based instrumentation at those locations is frequently augmented by mobile sensors (ship-borne and airborne) managed by the Air Force Eastern Test Range (AFETR) and the Pacific Missile Range (PMR), operated by the U.S. Navy.

The USAF operational missile fleet is an all-weather system that can accomplish its wartime objectives under virtually all weather conditions. In a test environment where the performance of every component of an aerospace system must be verified and documented, such is not the case. Although the degree of weather influence varies greatly from mission to mission, the weather factor is present for every launch operation conducted on the WTR. On one end of the scale are launches that have only weather sensitivities imposed by Range Safety to insure a safe launch; at the other end are complex R&D ballistic missile launches with a mix of uprange, midrange, and downrange weather constraints. For these latter missions, in particular, the weather factor is extremely important. Activation of all facilities and sensors necessary to support a complex launch operation must begin several hours before scheduled launch time. If an operation is scrubbed late in the countdown after activation of supporting activities, thousands of dollars (in some cases hundreds of thousands) in range costs have been expended with no payoff. Not all scrubs are due to weather, and not all can be anticipated and avoided, but it was primarily to avoid costly "weather scrubs" due to failure to meet mandatory weather criteria that SAMTEC began using weather probability forecasts in making decisions to activate the range and continue a countdown.

A look at the many weather constraints and criteria involved in an operational test or R&D situation is requisite to an examination of the system operation.

Weather influences on SAMTEC launches are manifested in four ways. First, there are the range safety considerations. A safety hazard can exist from falling debris if the launch vehicle is destroyed near or shortly after launch, either accidentally or as a result of a destruct command from the Missile Flight Control Officer. An additional safety hazard arises if certain missiles are destroyed near the ground: diffusion of toxic gases into a populated area. This hazard exists primarily with liquid-fueled rocket engines found in the Titan boosters and with certain upper-stage engines. The weather parameters important in these cases are upper air winds that determine debris fallout patterns and, in the case of toxic gas diffusion, low-level vertical profiles of wind temperature.

Second, and closely related to the safety problem, is that of the vehicle's guidance capability and structural integrity during the powered flight through the lower portion of the atmosphere. Under certain conditions of high wind speeds and strong vertical wind shear, the capability of some vehicles to remain on a predetermined trajectory may be exceeded by the atmospheric forces. Structural damage may also result from excessive wind shears.

The third type of constraints has to do with the effects of weather on mandatory range sensors, from launch through re-entry and impact of the RV'S in the target area. Optical tracking from landbased and airborne instruments is mandatory for many ballistic missile operations. Obviously, this requires a cloud-free (clear) line of sight from the sensor to the target. Another example is the possible degradation of RV impact scoring. At both KMR and Canton, splash detection radars determine the precise RV impact location by detecting the plume of water from the RV splash. Heavy precipitation in the vicinity of impact, or between the radar and the plume, will seriously limit the capability to score. 

Finally, some ballistic launches have as a primary or secondary objective a requirement to encounter or avoid certain weather conditions in the re-entry corridor that are independent of any sensor requirements. These weather-sensitive launches are normally associated with tests to determine the capability of experimental RV nose tip ablative material to survive re-entry through a hostile weather environment, such as ice crystal clouds or heavy precipitation. Under severe conditions, it is conceivable that an experimental nose tip may erode to the point of RV disintegration prior to reaching detonation altitude. Although the survivability characteristics of a new RV can be examined in laboratories and by theoretical models, some actual data must be obtained under operational environmental conditions in order to validate models and laboratory experiments. Thus we seek to launch certain ballistic missile systems into prescribed weather conditions covering a wide range from "good" to "bad."

Weather forecasts for most military operations are usually presented as categorical statements of what the weather will be in terms of cloud cover, visibility, precipitation, winds, severe weather, etc. When the forecast is presented, there is normally also a discussion of the synoptic weather pattern, trends, and the possibility that the weather may differ from the categorical forecast. What the decision-maker hopes to get from this information is a determination that weather will or will not hinder the accomplishment of mission objectives. In many cases a simple statement of the probability that specific mandatory weather criteria will be met is more useful. This single probability number contains everything important about the weather, its effect on a particular aspect of the mission objective, and the confidence of the meteorologist in his forecast in a particular situation. This number is much simpler to include in any objective decision-making algorithm applied to the mission.

For example, many USAF operations require a clear line of sight from one point to another. This requirement exists for tactical bombing, close air support for ground forces, tactical reconnaissance, and, as noted previously, optical tracking in certain missile tests. Clear line of sight is a function of cloud amount and cloud thickness at each level, as well as the angle of view. A categorical forecast of clouds issued in conventional terminology for this requirement does not answer the basic question of what is the probability that the target can be seen. The meteorologist can, however, objectively translate his forecast into probabilistic terms and give the decision-maker a single number to use in making his choice of "go" or "no go".

An important ingredient in tailoring of probability forecasts to optimum decision-making requires the meteorologist to be intimately informed of the mission objectives and knowledgeable about the manner in which the weather can degrade the attainment of those objectives. And the decision-maker has to understand how to interpret the probability forecast. For example, one common requirement for many ballistic programs is the need to view the re-entry form an optically instrumented aircraft. If the forecaster is aware that the aircraft cannot climb above 30,000 feet and he expects dense overcast cirrus clouds above that altitude, he will give a low probability that the objective will be met. If, on the other hand, he knows that the aircraft can operate at 50,000 feet, he will give a high probability of meeting mandatory objectives. The decision-maker in this example has to be confident that the meteorologist has been provided with the altitude capability of the aircraft.

Almost all ballistic launches require scoring of the RV impact locations, and the capability of splash detection radars to score is adversary affected by showers in the impact area. However, experience has taught the SAMTEC staff meteorologists that scattered lines or areas of showers moving through the area normally do not prohibit scoring opportunities at some time during a 3- or 4-hour launch window. In this case the forecaster will give a high probability that the splash detection radars will be able to score the impact, but the decision-maker must be aware that later decisions to hold the launch for a scoring opportunity may be necessary.

At SAMTEC, weather briefings given at number to use in making his choice of Readiness Reviews are short and to the point, consisting primarily of a statement of the probability that mandatory criteria will be satisfied during the scheduled launch window. An example of a typical briefing slide for a "moderately constrained" ballistic mission is shown in Figure 1. Here the uprange upper wind speed and wind shear limitation in the 30- to 40-thousand-foot altitude range are related to booster performance. Since the 155-knot wind speed and 35.5-knot-per-l000-foot vertical wind shear occur much less than 1 percent of the time and only in conjunction with major wintertime storms, the forecaster confidently gave a high probability that these constraints would not occur. The midrange criteria involved mandatory optical coverage by ARPA Maui Optical Site in Hawaii. The cloud cover was expected to consist of scattered high thin clouds that would not seriously limit optical coverage. Downrange, two mandatory criteria (no rain and no convective buildups) related to RV performance, and two were for weather-dependent sensor requirements. For this mission the forecast was favorable, and the decision to enter the countdown was easy. If, for example, the midrange probability had been given as .10 because of an active weather system near Hawaii, the decision-maker would request an update just prior to decision time. If still unfavorable, he would scrub the operation and reschedule for a later date.

Figure 1. Probability of acceptable weather

Readiness Reviews are normally held out 24 hours prior to launch. Whenever constraints are difficult to meet, additional probability forecasts are issued prior to each decision to commit resources. An example of how these procedures are used can be found in the Minuteman Natural Hazards Program.

The Minuteman Natural Hazards Program involved six launches of specially designed re-entry vehicles from Vandenberg into predetermined and precisely defined weather conditions in the Kwajalein impact area. A prime objective was to evaluate RV performance and to relate that performance to meteorological conditions encountered by the RV's. The success of each of the six flights was essential for accomplishment of the overall program objective. Since each test could be conducted only during the occurrence of certain weather conditions in the impact area, test planners were faced with the problem of how to determine the best times to schedule range support. To activate the ranges and begin the missile count down at random times or repeatedly on a day-by-day basis would have resulted in an enormous expense and would have required virtually full-time dedicated support by range resources that were required to support other programs. We therefore relied on weather probability forecasts to limit activation of SAMTEC and other support range resources to those times when there was a reasonable likelihood that the necessary weather conditions would occur.

To provide the quality of forecast information required, the Air Weather Service deployed a Defense Meteorological Satellite Program (DMSP) mobile readout terminal and a TPQ-11 cloud-detection radar set to Kwajalein during the tests to augment existing instrumentation. Additionally, a forecaster was deployed from Vandenberg AFB by the SAMTEC staff meteorology office.

A weather team, composed of Minuteman spa representative, the SAMTEC meteorologist, and other technical consultants, prepared probability forecasts and go/no-go recommendations, which were briefed to key test personnel at Vandenberg via telephone conference at critical decision points. This usually occurred 12 hours before the scheduled launch time. Based upon the weather team recommendations, test planners would either activate the resources of all ranges involved in the test and begin the countdown or reschedule the test and plan to evaluate the situation at the critical decision point for the new launch window.

For these tests, conventional weather forecasts were of limited value. The critical parameter upon which decisions were based was a Weather Severity Index (WSI), a complicated function of the liquid water content, ice water content, and ice crystal structure in the re-entry corridor. This number could be directly related to expected re-entry vehicle performance. The WSI was extremely variable spatially and temporally because of the frequent convective shower activity in the area. Probability forecasts were issued based on the predominant cloud features expected, the amount and duration of expected convective activity, and the forecaster's confidence. Conventional meteorological data, plus data from instrumented aircraft and special radar equipment, were used to verify the forecasts.

It became apparent very early in the test series that the "bad" weather requirements (those requiring large WSI values) would be difficult to satisfy. It became necessary to "threshold" at a very low probability value in order not to miss an opportunity. For the high WSI launches, go decisions were generally made if the probability of success was greater than about 20 percent. In retrospect, when the tests were successfully concluded, it was determined that the high WSI values occurred only 10 percent of the time, so the 20 percent threshold value was reasonable. (This is an important concept for those who use probability forecasts to understand. If the climatological expectancy of the desired weather is small and the mission urgency is high, thresholds should be set low so as not to miss an opportunity. If, on the other hand, the weather is easy to obtain and/or the mission urgency is not great, thresholds should be set high so as not to waste resources. The decision-maker should set these threshold values in advance, although real-time adjustments will sometimes be necessary.)

Two innovative procedures were developed as a result of SAMTEC'S involvement with the Minuteman Natural Hazards Program. Since the high WSI requirements were hard to satisfy and since there was normally a large spatial variation in WSI values, a real-time retargeting capability was developed that allowed targeting of the RV'S into any one of three widely separated impact areas as late as 20 minutes before launch. Using the sampling aircraft and radar, the weather team could then make last-minute recommendations as to which target was to be used.

The second innovation was implemented to use the range resources more efficiently. Again, since high WSI values ("bad" weather) were hard to find and low WSI values ("good" weather) were relatively easy to find, SAMTEC adopted the practice of scheduling high and low WSI missions for the same time and then making a decision based on the probability forecasts 6 to 12 hours prior to launch as to which mission would be activated. With this procedure, seven "good" weather launches were completed during a period when "bad" weather launches had the highest priority, and SAMTEC avoided a potentially serious scheduling problem that could have seriously impacted other range users' programs.

Over a period of 14 months, the six Minuteman Natural Hazards Program launches were successfully completed with 13 actual countdowns, 7 of which were terminated prior to launch because the weather criteria could not be satisfied. If we assume that without special weather probability forecasts and reasonably established threshold values the ranges would have been activated each time a launch was scheduled and scrubbed late in the count if criteria could not be met, 31 attempts would have been required. A documented value analysis has shown that by avoiding 18 unsuccessful countdowns, through the use of weather probability forecasts a net cost avoidance of $3,200,000 in range support costs was achieved for this program.

AT SAMTEC it has long been apparent that both weather and the use weather services play significant roles in our day-to-day decisions. The use of probability forecasts instead of categorical forecasts is a change introduced only recently. After almost one year’s experience with probability forecasting, the consensus is that it has definitely been a step forward. Although converting to probability forecasts does not change the basic forecasting accuracy, it does provide a means of getting directly to the crux of the weather problem associated with a particular operation. Based on SAMTEC experience, probability forecasts are more closely related to the real state of the science in that probability forecasts verify much better than categorical forecasts on a "normalized" scale. To put it another way, with probability forecasts the uncertainty is automatically included in the forecast, and meteorologists seem to be able to quantify their uncertainty very well. This can be very important for decision-makers.

To illustrate this point, the SAMTEC staff meteorologist prepared the verification graph shown in Figure 2. The verification figures include all probability forecasts issued from July through December 1974. Probability forecasts cannot be individually verified (except in the very special cases when the probability is exactly 0 or 1.0). They must be evaluated statistically. Ideally, if a probability of .03 is assigned for the occurrence of an event on 100 separate occasions, the event should occur 3 times. The 3 forecasts for which the event did occur should not be considered "busts." Therefore, the dashed diagonal line in Figure 2 represents perfection. Because of the relatively small sample size, the forecasts have been grouped into the probability ranges indicated. The circles indicate the observed percentage frequency of occurrence of the criteria covered by the forecasts in that probability range. Note that throughout the forecast range there is little indication of excessive optimism or pessimism in the forecasts, although some bias toward pessimism is evident in the .75 to .89 range. All in all, the forecasts are very reliable, i.e., for a forecast of say .60, over a period of time, the event will occur 6 out of 10 times.

A user of weather forecasts who demands a categorical forecast from a meteorologist is often asking more than the meteorologist is capable of delivering—in fact is asking him to degrade his capability. In the case of the .60 forecast, for example, he is asking that this be changed to 1.0. Obviously, 6 out of 10 will verify, but 4 will not. By asking for and receiving a categorical forecast, this uncertainty may not be conveyed, and a significant piece of information is not available to the decision-maker.

USING PROBABILITY forecasts does not solve all of the commander's problems, but it puts the job of making the tough decision where it belongs—with the operator, not the meteorologist. On those occasions when the probability is very high or very low (and as can be seen in Figure 2, these will be most of the instances), the decision is easy and the forecast approaches a categorical statement. For those situations where the probabilities are in the middle range, the decisions are tough, and this is where the other factors of mission urgency, costs of failure to achieve objectives, probabilities of meeting objectives at a later date, etc., become most important. The operator is in the best position to weight these various factors. If he had insisted on a "yes" or "no" forecast for these middle-range cases, then to some extent he has delegated the tough decision to the meteorologist or else he in effect ignores the forecast. Neither of these alternatives is desirable. This kind of conflict can be avoided by using probability forecasts and at the same time placing the real decision-making responsibility where it belongs--on the man responsible for the mission.

Major General Herbert A. Lyon, USAF (Ret), (M.S., Purdue University; M.S., George Washington University) was Commander, Space and Missile Test Center, Vandenberg AFB, California, until his recent retirement. After flying training in 1943, he flew C-46s in campaigns from New Guinea to the Philippines and Okinawa, then was Commander, 5th Combat Cargo Squadron, during the Japanese occupation. He has since served in Air Proving Ground Command; Arnold Engineering Development Center; DCS/Development, Hq USAF; Air Force Systems Command; and as Vice Commander, Space and Missile Systems Organization. General Lyon is a graduate of Army Command and General Staff School and Industrial College of the Armed Forces.

Lieutenant Colonel Lynn L. LeBlanc (M.S., Massachusetts Institute of Technology; Ph.D., Texas A&M University) is Commander, Detachment 30, Sixth Weather Wing (MAC), and staff meteorologist for the Space and Missile Test Center, Vandenberg AFB, California. Since he entered military service in 1957, his assignments have been in forecasting at Randolph AFB, Texas, and Tan Son Nhut AB, Vietnam; to staff duties with DCS/Operations, Hq Air Weather Service; and as Chief Systems Analysis and Design Branch, later Special Projects Branch, Air Force Global Weather Central. Lieutenant' Colonel LeBlanc has been selected for promotion to colonel.

Air & Space Power Home Page | Feedback? Email the Editor