Sunrayce 93 was a solar-powered car race from Arlington, TX to Minneapolis, MN involving 34 collegiate teams from North America during 20-26 June 1993 (figure 1 and table 1). This was the second cross-country solar car race in the United States; the first being the 1990 GM Sunrayce USA, of a similar format, from Orlando, FL to Warren, MI. Solar car racing is a relatively new sport, the first official race being the Swiss Tour De Sol in 1985. The success of this yearly event inspired Hans Tholstrup, the first person to cross a continent in a solar car (The Quiet Achiever - from Perth to Sydney, Australia in 1982) to devise the World Solar Challenge, a cross-continent international solar car race from Darwin to Adelaide, Australia (figure 2). It was first run in 1987, and weather turned out to play a key role in that race, won by General Motor's Sunraycer. Sunraycer got out to a large lead of about 300 km on the first 2 days of racing, and then heavy showers and thundershowers formed over a large area of the route to their north on day 3. All cars were caught in these except for Sunraycer, which extended its lead even more, and ended up winning the race by a total of over 24 racing hours. The poor conditions on day 3 substantially diminished the solar flux, and some cars were unable to continue down the road at times due to the torrential rains and flooding. In subsequent cross-country solar car races, the weather and its advantageous use has played a key role in determining their outcomes also. This article is a record of the role of weather - particularly solar radiation - and it's forecast on the performance of Maize & Blue (figure 3), University of Michigan's (U of M) winning entry in Sunrayce 93.
Sunrayce 93 was sponsored by the United States Department of Energy (DOE), and students at each college or university wishing to participate were required to submit a proposal to its race committee for approval. Ours was written by Furqan Nazeeri, our project manager. 36 teams were entered in the race. The cars were to be built mainly by students. Although there were some limitations on fund-raising and technical support, students were encouraged to pursue all avenues they deemed helpful (Most team's racing uniforms closely resembled those of professional auto racers!). The race crews were to include students, at least half being undergraduates; and a faculty advisor and/or staff advisor was also recommended. Professional help was not allowed on the race crews, although honesty compels me to state that our team had one (for our MagneTek motor), as several others did. The teams were required to participate in a regional qualifier, at which their cars were scrutineered and raced on a track. To qualify, a car was required to average at least 20 mph over 50 mi (about 32.2 km/h over 80.5 km), over a 6 hour period; only counting their fastest laps. Actually, exceptions were made to 2 teams who came very close to the requirements. The average speeds of the cars also determined pole positions for the actual race. Drexel University won the Eastern Qualifier and California State University at Los Angeles (Cal State LA) the Western one. The cars were scrutineered once more a day before the race from Arlington to Minneapolis began.
During the race, the cars all traveled on one route, mostly on secondary highways, and were timed each day. These routes often passed directly through cities, and thus the logistics and tactics of getting through these effectively was important. Charging of car batteries (limited to 5 kWh capacity) through its solar array was allowed at all times the cars were out of impound - approximately 0545-1930 (all times are Central Standard Time, unless otherwise mentioned). Racing time was from 0800 until 1730 or the day's finish. There were numerous ways of being penalized; ranging from running a red light (traffic infraction) to changing batteries (vehicle infraction) to profanity (conduct infraction). A daily observer was assigned to each team's Chase vehicle (figure 4), to be in full view of the race car at all times when racing. Also, race officials were scattered along each day's route to help us and observe. Each day's finishing position determined a vehicle's starting position on the next day (penalty time not included), and the cars would start a minute apart. Thus, daily finishing position was somewhat important regarding vehicle passing. Although courtesy rules were set for passing, the required 3 vehicles from one team (Lead, Race, and Chase) had to pass those from another, which was difficult in some situations. The final standings of the race were determined by total elapsed racing time plus penalty time. Table 2 shows the final standings of Sunrayce 93, courtesy of Electronic Data Systems Corporation.
Over 100 students from many disciplines (mainly engineering) at U of M made significant contributions to the construction, fund raising, and racing of Maize & Blue. The weather group for our team consisted of 3 of the 24 crew members (table 3). These are Mark Kulie, Jeff Zoltowski, and Joseph Bartlo. Joseph researched solar radiation under various sky conditions, local variability and behavior of clouds, and other topics related to solar energy, solar cells, and solar car racing. Solar flux is calculated from a forecast of turbidity, total cloud shadowing, and average cloud transmittance using the algorithm described in Appendix A. The diffuse solar flux was modeled as being anisotropic; but for simplicity and speed in calculations, it was separated into beam and isotropic components, as an approximation to aerosol, molecular, and cloud scattering.
In preparation for the race, Mark and Joseph participated in a race route survey, organized by DOE, from 21-27 June 1992. This was almost exactly one year prior to Sunrayce 93. The race was chosen around summer solstice to maximize solar energy availability, and its north-to-south route to accommodate most solar arrays, which tilt slightly backward. During daylight hours, cloud, temperature and humidity observations were taken every 10 min. Continuous global solar radiation (GSR) readings, and direct and filtered GSR readings were also taken when stopped. Information was obtained on possible tree and building shadowing, surrounding albedo, and elevation changes. This helped Mark to gain experience in the estimation of solar flux. From March to May of 1993, Jeff and Joseph made practice forecasts at least 3 days a week (for Ann Arbor), of the main weather elements that affect solar car racing. An analysis showed that the relative importance of these elements on the energy involved in racing Maize & Blue are as follows: solar radiation, .70; wind, .22; temperature, .06; air pressure, .02. Obviously, solar energy is collected by the solar array and converted into electrical energy. Wind affects the aerodynamics of the car, and is particularly important in determining the drag force on it. Temperature affects many components of the car and comfort, but mainly the conversion efficiency of the solar cells. For every 6 °C increase in cell temperature under a given solar flux, energy collection by the solar array decreases by about 1%. Weather forecasts were also made during vehicle testing in San Angelo, TX in June prior to the race. The details regarding all of the above endeavors are omitted here, not being of great importance to the purpose of this paper. They can be provided to all who are interested.
During racing, each of the weathermen had specific tasks, focusing on what they were most proficient at. Mark showed valuable observational skills during the race route survey and has an interest in maps, so he was mostly involved with observing, monitoring local weather conditions, and navigation. Jeff showed proficiency at forecasting the large-scale, long-term (2-7 day) weather. Thus, he mainly focused on providing forecasts for future days. He also offered useful input on the forecasts for the afternoon/evening charging periods, being on site during the day. Joseph focused on short-term weather situations and provided the 20-minute forecasts of weather elements input into the race simulator. The simulator was run by Eric Slimko, who excellently chose the velocities at which Maize & Blue was raced. At specific times, Joseph discussed racing with Eric and Andy Walberer (our strategists), especially regarding the effects weather may have on it. One of Joseph's talents is the ability to quantify a weather forecast. His goal was to forecast as specifically as possible the magnitude of weather elements and their times of occurrence; trying always to stay as far as possible from vague terms such as "partly, possibly, chance of, should clear soon, etc.". Only one thing will happen, and ideally, that is what should be forecast. Perhaps the major reason why the weather operations worked excellently was that all of us were utilizing our best talents most of the time.
The weathermen for Maize & Blue took the positions of Scout and Spy in the race caravan (figure 5). The 2 forecasters focusing on the local and short-term conditions, Mark and Joseph, were in the Scout vehicle. Jeff, who focused on the large scale and long-term conditions was in the Spy position. This actually would more properly have been called Remote, or something similar, since Jeff really did not Spy on anyone and was remotely located. It is a fun name to have, though. Jeff also had the advantage of being near the finish site, which was important to our forecast for the afternoon/evening charging periods, during which we obtained most of our energy for racing.
In the Scout vehicle (a Mercury Villager mini-van), Harry Yates was the driver. He also did a fine job on logistics and ergonomics. Mark sat in the front right seat, and thus had access to the front and near side windows and sunroof for sky observation. He was also the navigator. Joseph sat in a bench seat in the back of the vehicle, the area of which was surrounded by tinted windows. This was not so undesirable, since Joseph has a small forecasting bias toward clear conditions that he is working to remedy. Thus, the sky always appeared darker to him than it actually was. On the center left side of the vehicle, Harry inserted a wooden bench on which went a Capital Data laptop computer, a Hewlett-Packard paint jet printer, and mobile telephone communications from Ameritech. Thus, there was space in the back of the Villager to comfortably work on maps, data, and the computer; and even to bring in images from Weather Services International (WSI) and print them out in color.
For the Spy position, Jeff drove a luxurious Lincoln Mark VIII, to get swiftly from one day's 'destination to the next. During the racing periods he was at or around the finish; most of the time in a hotel room, bringing in maps and data from WSI and providing updates to Scout and Chase. He was equipped with equipment similar to Scout's. At the end of a racing period, Jeff met with Mark and Joseph briefly, before heading out to the next day's finish in the afternoon or early evening.
The schedule for weather operations was open-ended. At first a specific one was drawn up, but for practical considerations, it was never actually used. However, it was adhered to in a general sense. Perhaps it accomplished its purpose that way. It is very difficult to stick to a schedule of operations in forecasting for any sporting event, since the weather often determines the times and order in which things occur and should be done.
To begin each race day, Jeff was up bright and early at about 0300 to start obtaining weather information on WSI. In the meantime, Mark and Joseph were up between 0400 and 0430 (thanks to Bill Cosnowski) to do the same, interpret local weather conditions, and prepare for setup of the solar array for the morning charging period. Joseph chose the setup locations for all charging periods in the areas that were allowable, and decided on the orientation of the solar array at times when it was questionable as to which direction would maximize solar energy collection. Chad Mentzer routinely oriented the solar array. Between 0530 and 0700, there was communication between Scout and Spy regarding our weather forecast and conditions near the finish site. At 0700, Joseph provided the numerical forecast to Eric in Chase for race simulation and reference. This consisted of 20-minute forecasts of solar flux, wind, and temperature. Air pressure, being a lesser effect, was only mentioned if it seemed important to the overall energy use in racing on a particular day. An estimate of solar flux amounts during the following morning charging and racing periods was included. This was given relative to the typical clear sky solar flux (CSSF) that would be expected at specific times and locations. Jeff also gave verbal input, mainly regarding the charging periods (by design, he had the advantage of being on-site) and longer-term forecasts.
During the race day, the progression of the weather and its comparison to our forecast was monitored by those in Scout and Chase. We provided updates occasionally to confirm that the forecast was still fine or that it needed to be changed. Mark's observational skills were very valuable in monitoring the weather situation, and allowed Joseph to mainly concentrate on weather changes. Meanwhile, Jeff provided updates to those in Chase regarding charging and future day's forecasts as he obtained new information on WSI. To accomplish the above accurately, and to fulfill our other duties, Scout generally stayed 3-40 km ahead of the race caravan during the day; most of the time 15-30 km. Occasionally, Scout was as far as 65 km ahead. Before the race, it was planned to send local hourly observations and even an occasional weather map from Spy to Scout via cellular phone. This was only possible, however, when Scout was stationary, and even then, only some of the time. Thus, Scout stopped the vehicle to bring in hourly observations, satellite and radar images, and other charts from WSI on occasion. In a few instances with a very strong cellular phone signal, WSI information was obtained while moving in Scout. Jeff verbally read weather observations via direct cellular communication on quite a few occasions. Stephen was responsible for excellently arranging the communications systems, as well as for the aerodynamic design of Maize & Blue.
Scout reached the finish at least 15 min. ahead of the caravan on all days so that Joseph could choose the afternoon/evening charging site. Most of the time, Jeff was waiting there also (and in some cases already had a good site chosen). We then exchanged thoughts, and Mark and Joseph obtained any useful maps that Jeff brought in for work on the next day's forecast. Then Jeff left at the first opportunity to the next day's location. Mark and Joseph provided ideas on the next day's weather to Andy Walberer and Eric during the evening. Lesley Camblin did a splendid job in seeing that food, lodging, and other logistic concerns were well taken care of for the weather group.
Below is a day-by-day discussion of our activities during Sunrayce 93, the weather seen, and its influence on racing Maize & Blue. Among the reasons this is done is to illustrate specific topics and situations that come up during a race, and how we dealt with them. Tables 4-9 (4 5 6 7 8 9) list daily forecasts made at 0700 of clouds, wind, and temperature, the resulting forecast solar fluxes, and measured GSR. Table 10 lists the times and positions of top competitors after each day of racing. Figures 6-11 (6 7 8 9 10 11) graphically compare forecast to measured GSR. GSR readings were taken with a newly-calibrated Epply Black-and-White (B&W) radiometer mounted on top of Scout, and an Epply Precision Spectral Pyranometer (PSP) mounted on top of Chase. It was done this way so that the strategists in Chase could see the more rapid changes in solar flux that the PSP could indicate as opposed to the B&W one, which has a longer response time. The PSP was calibrated against the newly-calibrated B&W radiometer. Although the PSP was always near Maize & Blue, readings taken from it are less reliable. Thus, the readings taken from Scout are published here. The performance of the B&W radiometer was excellent, especially considering it was on top of a quickly-moving vehicle for 7 days straight. However, some adjustment of the readings may have to be imagined due to the variable location of Scout relative to Maize & Blue, for verification purposes. Appendix B includes vehicle positions and their probable influence on how representative GSR readings taken at given times and locations are. Maps for the race routes of days 1-6 (figures 12-17 - 12 13 14 15 16 17), which illustrate cloud cover and significant weather systems, can be used for reference.
Weather forecast
Our focus of attention prior to this day was a tropical storm over the Gulf of Mexico, which was approaching southern Texas. To what extent this would affect the weather from Arlington and north on day 1 was uncertain to us. Also, a "cold front" was forecast on the National Meteorological Center (NMC) prognostic charts (figure 18) to push into southern Oklahoma by the evening of day 1 and to continue southward, keeping Ada in clouds; and then to become stationary over northern Texas early on day 2. As it turned out, the time and amount of clearing, with intrusion of drier air from the NW, was the major forecast issue.
The early morning was cloudy with light rain at Arlington, becoming moderate at times, from 0400-0620. The heaviest rain fell from 0545-0610. The whole region was generally cloudy, with some clear areas to our NW. From 0630-0700, there were a few temporary breaks in the low Sc overcast (see table 11 for cloud type abbreviations). Jeff reported some clearing around Tulsa, and thought it would clear in Ada. My forecast (Table 4) called for clearing to begin by about 940, becoming mostly clear by 1020 (near Whitesboro). Scattered Cu were forecast to then develop into the early afternoon, and dissipate toward sunset as dry air moved in aloft from the NW. Since I thought the dry front would be well to our south by the next morning, I called for 90% of CSSF for the morning charging period of day 2 and 85% of CSSFduring racing.
Forecast verification
The racing conditions from Arlington to Whitesboro remained generally cloudy, with only a few bright spots in the Sc overcast. Approaching Whitesboro, we saw GSR readings as high as 600 W/m² from Scout, but they were lower again by the time Maize & Blue arrived. Just after leaving the media stop (at 1105), we saw one GSR reading of 937 W/m² during a break in the clouds before we ran into some low St clouds and some light (and sometimes moderate to heavy) drizzle showers. This seems to occur frequently before morning low clouds dissipate. Solar noon for this day and location was 1229, so it was still morning. After the drizzle, Sc with bright spots were seen before the clouds broke up significantly at 1200, when we were leaving Madill. When they did, it occurred very quickly. By the time we reached Ada, there was 30% Cu cover with very little shadowing. The Cu were most numerous to the SE, and they were moving from the SSE during the afternoon, but also seemed to dissipate somewhat as they moved toward our location. Cu coverage maximized to 45% at 1515, with about 35% shadowing. The Cu then dissipated quite rapidly, and I do not believe there was any shadowing after 1610, although I was having a nice spaghetti dinner from 1710-1800 courtesy of Sunrayce.
Updates and race tactics
After almost an hour of racing (at 0855), I informed Chase that I no longer expected clearing at Whitesboro, but later. The numerical forecast was updated at 0900 and routinely at the media stop, when personal communication between those in Scout and Chase was possible. The short-lived clearing at Whitesboro caused me to still believe that general clearing would occur earlier than it did. From 1100-1200, we all watched and waited (like birds on a wire) for the clearing to occur, which finally did happen. At Ada, I informed those in Chase that there was very little shadowing, and that they should hurry up and "get over here", since I expected even more solar flux for charging than originally forecast. I did this to make sure that we would maintain our speed (see table 10), so long as there were no mechanical problems with Maize & Blue; in case they were entertaining any ideas of slowing down to save charge.
Weather forecast
The skies started out overcast once again, but this time there was no rain, and the low St clouds were actually thin enough to occasionally see a shade of blue sky through. An upper level dry/ridge pattern was developing, indicating the possibility of clearing sooner than on day 1. However, this was not to be. In fact, day 2 was our worst day of forecasting. The previous evening's sounding near Oklahoma City (figure 19) did show abundant low-level moisture, indicating the possibility of the low cloudiness to persist throughout most of the morning, if it were only paid more attention to. No upper level cold air pools appeared to be embedded in the ridge aloft and no strong surface troughs, convergence lines, or other features were seen which would indicate the formation of any organized storm systems.
Thus, my forecast called once again for low overcast early, giving way to mid-morning clearing. Cu were forecast to develop in the late morning, and to dissipate in the late afternoon. For day 3, I boldly forecast mostly clear skies once again, calling for 75% of CSSF during the morning charging period and 90% during racing, regardless of the first 2 mornings' conditions.
Forecast verification
At 0800, the skies cleared up significantly, when Scout was only about 15 km out of Ada. In retrospect, this was to be expected, since shades of blue were seen. The average sky cover from 0800-0910 was about 65%. Yet, this was only temporary, and the skies became overcast once again; although the diffuse solar flux through the Sc was about 7/5 higher than on day 1. The skies remained quite cloudy through Shawnee, but from 1100-1200, the Sc overcast dissipated somewhat - enough to see the sun through half of the time, and relatively thick Cs above them. Maize & Blue must have seen this later than we (see Appendix B). Blocking of sunlight by the Cs may be responsible for such persistence of the Sc. Skies appeared clearer to the NE, where we were headed. Jeff reported that it was brighter in Tulsa, though still had not cleared up. Mark and I were looking for clearing ahead, and by 1140 (where Scout was), we saw the first clearing to the extent that I could not look directly at the sun. Still, the disc of the sun was not completely visible until 1200. From 1200-1315, the sky cleared up nicely. However, after a minimum in Cu cover of 25% at 1315, more started to form. In hindsight, this was due to a small upper level trof which had developed over Oklahoma during the day (figure 20). At 1430, Mark noted some Cs and Ci to our SW, moving in from the WSW. This is likely to have been the result of nearby afternoon convection. Between 1430 and 1630, we were either under the broken Cs/Ci shield or a large Cu cloud. Although the Cu covered only about 40% of the sky, it caused about 70% shadowing over us! At 1720, a band of thin Cs moved in to shadow the sun, but all of the Cu had dissipated by 1730. The Cs moved east of the sun by 1825, and no further shadowing was seen.
The large local amount of afternoon Cu shadowing at the Tulsa Fairgrounds may have been due to an urban effect. The lower tropospheric winds were from the SE, and the Cu were preferentially forming about 8-15 km to our SE and drifting directly over us! Many times during the afternoon, large Cu were overhead, blocking the sun, with clearing along and above most of the horizon. Some of the Cu clouds covered up to 25% of the sky overhead.
Again, the trend of clearing was forecast correctly; and again it occurred later than expected. However, this forecast was not as useful as the one for day 1, since the afternoon/evening charging period was significantly more cloudy than forecast. Thus, the overall amount of solar energy was much less than expected.
Updates and race tactics
I thought we may have to provide a radical update due to the temporary clearing right away. Although being prepared to do so, I held off; and good thing too, since skies subsequently became more cloudy than expected. By 0910, I had already informed Chase of this prospect (so that they should only expect 50-60% of the original amount of solar energy forecast until 1100). My update at Shawnee called for clearing to begin at 1050, with scattered cloudiness by 1200. The clearing was still forecast too early, but by 1200, most of the Sc had dissipated into scattered Cu. From 1150-1205, we compared our radiometer readings from Scout to the solar array power seen from the telemetry in Chase, at their request. They were concerned that the array was not converting as much energy as expected. The GSR averaged 712 W/m² during the period, and the average solar array power was 720 W. My solar energy collection algorithm predicted 794 W under these conditions. The solar array was perhaps the only part of the car that did not perform up to its expectations, providing only about 80% of the power that was originally thought to be possible. The sky had finally cleared, and the only thing to do by then was to hope it remained so for charging; and to try to influence the clouds to move away from the sun, as I did at 1605 at Ada the previous day.
Weather forecast
Finally, it was almost completely clear in the morning when we arose! There was significant cloudiness over south central Kansas and western Oklahoma at 0500 (figure 21). An isolated thundershower was visible above the southeastern horizon when we arose at 0430, and was slowly drifting northward toward the race route (figure 22). Mark showed some concern of it crossing the route as we started racing, but I thought it would dissipate mainly into broken alto and cirroform clouds by then. As it turned out, we were both right (or wrong, to a pessimist). It did dissipate to the extent that the leading Ac blocked the sun only periodically as it drifted across the sun from 0610-0700. The low and middle level clouds could be seen gradually drifting to the N, as some higher scattered Ci did so to the SE. The remnants of the thundershower dissipated more as they moved over the route by 0800.
The upper level ridge was strengthening on the large scale, and the previous evening's sounding at Monet (figure 23) showed promise of a clearer day than the previous 2, since the lower tropospheric air being advected in from the SE was drier than before. There was no reason to expect any organized storm systems, although the possibility of a scattered shower or thundershower was considered. So my forecast called for scattered alto and cirroform clouds from 0810-0850 (mostly remnants of the Cb cloud), clearing to scattered Ci clouds until diurnal scattered Cu would begin to form around 1020. These were not expected to develop a great deal, as a maximum coverage of only 25% was expected, along with occasional scattered Ci and Cs. A mostly clear charging period was forecast for the morning of day 4, with 80% of CSSF expected. 65% of CSSF was expected during the subsequent racing period, with mostly cumulus cloud types expected late as a cold front approached from the W.
Forecast verification
As we began to travel along the route, broken alto and cirroform clouds drifted overhead, as expected. However, at 0800, new Cu began to form. Most of these were behind us at first, but they developed and moved over and several miles ahead of us. These Cu probably used the moisture from the dissipated Cb when the solar radiation became intense enough to cause deep convective currents. Some Cu became quite large (Cg). The maximum time of cloudiness was 0835, just before we began to outrun them, at which time we had a total sky coverage of 65%, 45% of which were Cu or Cg. Afterward, we saw only scattered Cu and Ci, the Cu occasionally becoming large and towering and significantly covering the sky. The afternoon became quite clear, and the majority of clouds seen were scattered to broken cirroform ones which moved in from the WSW near Fort Scott.
An interesting thing to note during the late afternoon was a Ct above the western horizon as the sun was setting into it. It was dissipating as the sun was setting toward it (about 25° above the horizon), and the sun stayed above it; only occasionally being blocked by thin Ci and Cs until sunset. The forecast on this day was the best of all, the solar flux oscillating somewhat about the forecast amounts all day.
Updates and race tactics
As we headed out in Scout, Stephen informed us that Maize & Blue was under clouds, which were becoming quite thick and numerous. They were even concerned of thundershowers developing, and requested our estimation of the behavior of the developing Cu and Cg. After stopping to observe, Mark and I estimated that the clouds were moving from 200° at 30 km/h. Thus, according to our heading (see figure 8), they were coming from behind and to our right. They were also developing quite rapidly out ahead of their direction of movement. Still, with development included, their net velocity along the route appeared only to be about 45 km/h. Nothing that Maize & Blue could not outrun going 80 km/h. So we did!
As noted above, we got out of most of these by 0850. Some towering Cu developed at 1010, and I radioed back to inform the Lead vehicle to be prepared for a thundershower. The only other relatively cloudy period was seen around Miami. TCU ALQDS was reported in a special observation from JLN at 1019. The sky cover reported was 27SCT 180BKN 250BKN. This led me to believe the skies would be somewhat cloudier than forecast during the next 1-2 hours, of which I informed Chase. I ended up changing that forecast again, making it even clearer than the original, as we approached Joplin and saw no Ct, but some scattered large lumps of Cg which did shadow the sun for as long as 6 min. as we traveled at 90 km/h in Scout. The 3 GSR forecasts provided differed by only 230 W/m2 at most, and the actual solar flux was in between them, much closer to the final one.
Weather forecast
Overnight, I had been bringing in weather data and analyzing satellite and radar images, and forecast information. These indicated a large cluster of storms forming over central Nebraska at 2300 on 22 June (figure 24). They were expanding and moving southeastward slowly. By 0100, Ci anvils extended into Iowa and just about covered Nebraska. By 0200, they edged into extreme northwest Missouri, and much of south central and southeast Nebraska was under thundershowers. The high cloudiness reached Kansas City by 0400, and most of northeast Kansas was cloud-covered, with thundershowers far to the N. High clouds were approaching Fort Scott slowly.
The morning charging period was mostly clear with a few thin Ci and Cs (figure 25). It was generally clear to our E and NE. Jeff reported only high clouds at Saint Joseph (there were no hotel accommodations at Cameron for us), but said they were thicker to the NW, and most of the clouds appeared to be moving S; to our W. However, the surface winds were from a southeastward direction, as they were in the entire region. Also, it was reasonably moist in the lower troposphere (figure 26). Higher aloft, it remained relatively dry. Moist air being advected into a highly unstable atmosphere (figure 27) seemed to be a good indication that convection would form.
Thus, I forecast broken Cs and Ci cloudiness in the early-mid morning, with Cu developing around Kansas City. These were forecast to be broken to overcast in the early afternoon, and to diminish afterward with some alto types of clouds, since the approaching cold front was still well to our W. Since the front was expected to move through Des Moines around sunset on day 5, I expected 50% of CSSF during the morning charging period, and 65% during racing.
Forecast verification
As Scout left Fort Scott, the Cs clouds thickened, and formed a uniform veil over the sky. These continued to thicken slightly as we headed N, causing the GSR to remain relatively constant as the sun rose from 0800-1000, although the sun was always quite visible. The GSR was lower than forecast, but it was due to a greater sky coverage than expected - the transmittance of these clouds was forecast accurately. By 1000, the Cs thinned significantly, to the point where they were barely visible. The cloud cover reported at MKC at 0845 was 43BKN 100OVC, and at 0945 was 30SCT 100OVC, but the high clouds throughout the morning were definitely Cs, and they were extremely thin over downtown Kansas City. GSR readings of 700-800 W/m² were seen when there was no shadowing from the scattered Cu.
As we left the media stop, skies quickly clouded up, apparently from our W, for the most part. It seemed as though the Cs veil had briefly got ahead of the major storm system, which was well into northwest Missouri at this time. At the outskirts of Kansas City, dark Ct were overhead and to our W, with the GSR down to 240 W/m². As Scout continued ENE, we saw a rainshower for 5 min., which was about 8 km ahead of Maize & Blue. However, they never experienced it - apparently it had cleared the route by then. After this, some small breaks were seen in the clouds, but it was clear that we would not outrun anything on this day. After briefly reaching 300 W/m², the GSR plummeted again to around 100 W/m² from 1240-1320 as the final line of thundershowers moved through (figures 28 & 29). These stayed to our south, but we remained under alto types of clouds the rest of the afternoon, with a few breaks and general thinning of them seen in the early evening.
The solar flux was overestimated during the entire day, except for approaching and at Kansas City; much more so for the afternoon than the morning. The skies did not clear up as much as expected during the charging period. Yet, I feel it was a good forecast, since the qualitative idea of clouding at Kansas City was correct, on a day when other forecasts and opinions I heard called for mostly clear skies. Although the forecast error for one 20-minute period was 513 W/m², the forecast was more useful than that for day 2, since it helped us to successfully accomplish our racing objectives.
Updates and race tactics
This morning was interesting, because several people were reasoning with me, saying that it was clear to the N and E, and that they did not see any prospect of cloudiness. True, to the untrained eye, it did not appear that the cirroform clouds were advancing and thickening. Eric was considering choosing a gear ratio that would allow Maize & Blue to travel 72 km/h at full-throttle (which is most efficient). He figured it would be best to try to go out fast, hoping that our nearest competitors, California Polytechnic @ Pomona (Cal Poly @ Pomona) and George Washington University (GWU) would follow us. He said that we could then "run them into the ground" with our superior aerodynamics (aerodynamic drag force on the car is proportional to the velocity squared), using the analogy of a spacecraft taking off into orbit. That is true, but I warned him that one must have enough fuel to reach escape velocity; and it appeared to me, with the moisture and winds as they were, that we may not have it in the form of solar energy past Kansas City. He considered this and trusted my judgment, choosing a gear ratio that would allow Maize & Blue to go 61 km/h at full throttle. He also said that "if it didn't cloud up, we would lose Sunrayce". It turned out that we started the day at 61 km/h, and ended up at 46 km/h (in the open road), finishing with about a 20% charge stored in the batteries. So things worked out pretty well for us.
Some of the other competitors were less fortunate. Cal State LA finished first on day 4, and they were expected to be our main competitor. Yet, they lost over 3 hours due to motor problems on day 1, so they were just trying to get back into contention. Cal Poly @ Pomona hung with us, but they drained their batteries considerably more to finish about 18 min. behind us for the day. GWU took it out fast and were ahead of us most of the time. However, we were fairly certain that their batteries were completely drained when we saw them climbing a hill at less than 15 km/h heading toward the finish. We got impatient with this about 3 km from the finish, passing them in a no passing zone, and took a 10 min. penalty in addition to a 20 min. infraction taken on day 3. Still, we were only about 10 minutes behind Cal Poly @ Pomona overall.
A situation occurred that was a bit embarrassing for me, too. As stated above, the sky was clearing significantly approaching the media stop. At the location, some Cu were seen to be developing to our NE, at the outskirts of the city, but it was hazy and hard to tell exactly how much. Jeff was probably on the way to Cameron, and I was not able to contact him on his cellular phone at the time. And although WSI was extremely helpful, the Cu convection did not show up very well on the satellite images. I think that by making the map background and graphics look so neat, they often do so at the expense of detailed cloud portrayal. Since it was so bright, I decided to increase the forecast solar flux amounts at the media stop. About 4 min. after we left, I had to call Chase and tell them to forget about the update, since Ct were moving in and rapidly developing at the outskirts of Kansas City, and to expect lower solar flux amounts than forecast for the afternoon - about half as much!
Weather forecast
Overnight before day 5, a new storm system began to develop over South Dakota and Nebraska. This was in association with a strong upper level low pressure center over southern Alberta with trofs extending to the S and SE (figure 30). The storm area associated with this one was even larger than the previous night's. With the closer approach of a cold front, this system seemed less likely to decay, as I wondered if the previous one would. At 2000 on 23 June, the last clouds with the previous system were leaving our region, and the ones with the new system entering (figure 31). By morning, it became clear that there would be no escape from the cloudiness and rain. The charging period was actually partly sunny until 0615, at which time strong thundershowers formed; first to our NW, and then to our SW. From 0650-0740,thundershowers were experienced, with some brief, very small hail around 0700. The main forecast issue was the timing of the passage of the front and the expected clearing behind it. Clear to scattered skies were already seen by 0600 at Grand Island and Norfolk in Nebraska. By pure advection, the last rain would pass our route by 1100-1200, and the clearing slightly later. However, the heavy cloud area was large (figure 32), and it was questionable as to how the system would continue to develop.
Mark, Jeff, and I all had different opinions on this. We thought it might clear up anywhere from 1200 to after sunset. When I provided the forecast, it called for clearing around 1330. 2 periods of rain were also expected - one in the morning (which we saw when I was providing the forecast), and another rain band to move through around 1200-1230 with the passage of the cold front. In between these, I expected the sky to remain cloudy, but to brighten up somewhat as As clouds would tend to dissipate during mid-morning. The charging period was expected to be mostly clear. For day 6, I forecast clear conditions, with 100% of CSSF for the morning charging period, and 90% of CSSF during the following racing day.
Forecast verification
The morning thundershower ended just about as we started racing. From 0820-0940 the sky became significantly brighter, especially to our S, but still overcast. Since the Scout vehicle went about 35 km ahead of Maize & Blue early on, they probably saw somewhat higher solar flux amounts than us during this period. On a few brief occasions, over 500 W of solar array power was reported. The sky then darkened somewhat (there was an As layer above low, broken Sc), and brightened again from 1050-1120. The skies clouded more as the frontal rain band (which took an awfully long time to develop) approached, and came through just as we left Lineville. Light rain began and quickly ended.
Afterward, some light rain began at 1230. Skies were very dark also, with GSR readings of 80-160 W/m² near solar noon, which implies a cloud transmittance of about .12. The heaviest rain was not yet occurring, and when it did at 1307 (with the strong, gusty winds from the WNW at 7-11 m/s), the GSR was actually a little higher at almost 200 W/m². Some breaks in the clouds occurred after 1400, but no real significant clearing until 1530, when the broken Ac started to dissipate rapidly. By 1700 most clouds were E of the sun, and no shadowing was seen after 1715; only a few thin Ac and cirroform types overhead and to our E.
The forecast on day 5 was very good, although the time of clearing was estimated about 2½ hours too early. The forecasts for the subsequent charging periods were nearly perfect.
Updates and race tactics
By the morning of day 5, it became clear that the main racing issue of the day (other than the ever-present one of reliability) was which cars could do anything close to resembling racing as the charge on all of their batteries dwindled. A solar car can handle one cloudy day quite well, but 2 of them make traveling very difficult. It became an endurance test, and with our more efficient system (MagneTek motor, which was developed by Bob Colewell and Steve; Michelen tires, worked on by Kevin Cain; efficient electrical system, mostly worked on by Ketan Patel; efficient and reliable mechanical systems, mostly worked on by Andy Warner and Andy Carmody; and clean aerodynamics), we were the last car left on the road. Everyone could see the weather conditions were poor, so there really was not much to say about that; at least early on. However, Eric said that we would not have made the finish if the skies had not cleared up by 1630. The extra solar energy was just enough to push Maize & Blue over the final few remaining hills on the route. If the truth is to be told, none of our forecasts were perfect except for YHWH's, which cleared the skies just in time to allow us to finish.
Thus, we spent most of our time on reconnaissance. About 30 minutes before reaching the media stop, I suggested that we time Cal Poly @ Pomona's car and inform the Chase crew of their average speed. This turned out to be 7.0 mi (about 11.3 km) in 28 min., 47 sec., which is 14.6 mi/h (23.5 km/h). The racing was very slow on this day. Upon hearing that, Andy Walberer (who worked on strategy and batteries) was confident that we could pass them after the media stop. We did, and at 1330, Floyd Harris informed me by the side of the road that we were the last car out racing. Actually, Iowa State was too, but only after replacing their batteries (and thus taking a large penalty!). The extended cloudy period drained the batteries of all teams who normally depended on afternoon sunshine to supplement them. During the first 2 days of racing, many teams did this. At 1350, we asked those in Chase what we should do, and Betesy White said to "go out and find 250 W/m²"; and it was only a few minutes later that we did, as if she did not ask for it, but ordered it. Actually, I think Cal Poly @ Pomona went back out later, but the bottom line (one of Andy's favorite expressions) was that we obtained a lead of about 105 min. Ecstasy broke loose at the charging site when we all saw each other, the sun shining in all its glory; and Maize & Blue, the only car to finish day 5 under its own power!
Weather forecast
This day was much simpler to forecast for than the first 5. However, there was a secondary trof to our NW, behind the previous days' low pressure cell. How much cloudiness, if any, that would be caused by it was the major forecast issue of day 6. The trof was still relatively far to the N, and it appeared that it would mainly affect northern Minnesota.
The morning started out clear, with only scattered Ci, Cs, and Ac in our vicinity (figure 33). A stack of Ci and Cs were to the NW, moving ENE. So long as the alto and cirroform types of clouds would not develop greatly, the time and amount of Cu development would affect the solar flux the most. Wind was also a factor on this day. Behind the storm system, winds were strong form the W to WNW and gusty. They were about 16 kt at 10 m during the morning charging period, but the area was quite sheltered, so it was not very noticeable then. Since many parts of the race route are elevated, with large expanses of open farmland west of the road (and at this time, large ponds due to flooding), the crosswinds would be quite strong.
Since the air was dry, and the NMC's Nested Grid Model forecast relative humidities were low (generally from 30-50% in all layers throughout the region), I figured that we should see the first Cu overhead around 1020, and they should cover about 10% of the sky by 1100, reaching a maximum of about 25% before beginning to dissipate around 1430. Cirroform cloud types were forecast to be widely scattered. The winds were forecast to be from 280-300 deg at 4-7 m/s while racing. The wind speed at the height of the Maize & Blue (nose height of 60 cm) was estimated to be about half of that at 10 m. The winds were forecast excellently by Des Moines National Oceanic and Atmospheric Administration (NOAA) Weather Radio, who forecast westerly winds of 10-13 m/s, gusting to 20 m/s. No surface maps were available on WSI with Windows, and we had little time to plot maps and draw isobars. For day 7, the morning charging period was expected to be clear, with 105% of CSSF (clearer than a typical clear sky), and 100% of CSSF during racing.
Forecast verification
The skies did remain clear throughout the entire morning, with a stack of Ci and Cs clouds remaining ahead of us all day (apparently its motion along the route was very nearly the same as ours!). NOAA Weather Radio reported winds from the W at 11 m/s at 0800. The only shadowing during the morning was from roadside trees, which did decrease the GSR by about 2-3%. Tiny Cu were seen at 1000, but literally only a few were near us, and none shadowed the sun. They were more abundant and not as small along the horizon to the north. Jeff went searching for them 80 km to the NW of Albert Lea.
At 1220, we finally approached the first Cu to the N and W. Most of them stayed W of us and others missed shadowing the sun. I do not know how well Maize & Blue fared with the first Cu, but even being 20 km ahead of them, we may have seen shadowing later than they. Almost miraculously, we were not shadowed until many Cu were behind us and they sky was about 20% covered at 1252. After this, numerous shadows were seen, as the Cu coverage maximized to 45% at about 1445. After this, the Cu dissipated quite rapidly, and a few narrow bands of Cs and As crossed the sun briefly.
The winds followed the NOAA forecast closely, and thus were about 5-9 m/s, gusting to 15 m/s at car height.
Updates and race tactics
Since the weather was fine for solar car racing, our major task in Scout at the beginning of racing was to time our nearest competitors over the first 20 mi (about 32.2 km) of racing, and report this to those in Chase, at their request (table 12). We used the Sunrayce 93 Route Book and our odometer to pinpoint the location as accurately as possible, and then waited for cars to pass. I am surprised at how slow Cal Poly @ Pomona started out, since they were 105 min. behind us.
There was not much to update early in the day, since the skies were so clear. I asked Harry to stop the Scout vehicle in an open area so that I could check the wind. It was about 8 m/s on my face, and about 6 m/s at car height. An anemometer is perhaps the most useless instrument one can take along on a solar car race. First, if it is mounted to a moving vehicle, an accurate reading is next to impossible, since the vehicle's speed is so difficult to measure and will dominate the calculation. Second, the wind varies so much temporally and spatially, even under the steadiest of winds, that a rough estimation is probably about as good as anything else. Third, a solar car can be equipped with a cruise control that allows one to drive according to speed, power, or battery current. Ours did not have all of these features, though. Thus, a car can automatically adjust its speed to account for variations in the wind. The driver must keep the car stable, though, and Deanna and Andris did a fine job of that.
At the media stop, there were still no clouds in the immediate vicinity, but only on the distant northern horizon. I could not contact Jeff to find out how conditions were at Albert Lea, nor access WSI at the time. Thus, I decided to call for clear skies the rest of the day. When it became certain around 1245 that there was no escape from the developing Cu, I said to use my original forecast. This called for cloudier conditions than seen up until this time, but was okay for simulating the remainder of the day's racing.
When we got to the charging site, I found a spot that I thought was the best, and we were not certain whether it was allowable. Some racers on the other teams told me it was not, but I was not satisfied with that. I was very friendly to the other teams, and most of them were to me also - actually, more so than our team; but I could also tell when they did not want us to benefit. So we asked several Sunrayce officials regarding this, and the spot turned out to be okay. We were the furthest back from the western tree line at the Albert Lea Fairgrounds, and this was helpful to maximize charging. Actually, I tried to get the solar array moved more to the S so that the sun would set in a hole in the tree line, but it was as far as we could go. The sun actually would have set on the N side of the tree opening at sunset at 1958, but we had to stop charging at 1930 to get Maize & Blue into impound without penalty. So as it turned out, the sun was centered in the opening between trees then. I was happy to have been able to estimate the sun's future position so well, only seeing its position when we arrived and using a cheap compass and accurate calculations of solar position (Appendix A).
Weather forecast
There was not much forecasting to be done on this day. The skies started out mostly clear, with only a few scattered Ci and thin Ac clouds to the NW. Rainshowers were over extreme northern Minnesota, under the secondary trof, but they never came near the southern part of the state. I expected skies to be mostly clear once again, with only a few scattered Ci or very thin Ac. A numerical forecast was not even given, since Maize & Blue had plenty of energy to finish the race. The winds were once again expected to be strong; 5 m/s from the WNW (at solar car height). Thus, we would get a slight tail wind as we headed to Minneapolis.
Forecast verification
The Ac developed more than expected from 0830-1030 as they drifted eastward and grew overhead, but we got through them just fine. The maximum sky coverage due to these and the few scattered Ci that were present was 45% at 930. The average cloud amount during racing was about 30%. Still, the GSR was about 90% of CSSF. Skies had almost completely cleared by the time we finished at 1149, which was probably due to the secondary trof passing to our east.
Updates and race tactics
There was no need to provide any updates. The cloudiness did not become extensive, and diminished the solar flux even less. I saw Maize & Blue coming to the simulated finish in the Minneapolis Zoo early enough to catch up and join our team across the finish line to celebrate our victory!
Solar energy is typically forecast or estimated over periods of years, seasons, months, and days. DOE compiles climate-related statistics on solar energy availability, and the National Weather Service provides daily forecasts of GSR at Model Output Statistics sites. Dennis Baker once included GSR as a forecast variable in weather laboratory activities at U of M. Very seldom are the small temporal and spatial variations paid attention to, since in most activities or applications, they are generally unnoticeable. Edward Ryznar and Donald Portman have discussed with me the importance of local solar flux on optical scintillation. I do not pretend that others have not applied forecasts such as ours to some activities in the past. For example, it has long been a practice to forecast the total amount of boundary layer heating to aid in the forecast of the amount and character of cumulus convection for soaring. Yet the types of analyses and forecast shown above are very useful in solar car racing. Local variations in cloudiness affect our activities to a comparable magnitude as the overall weather characteristics in the vicinity. I witnessed this during our first evening charging period while forecasting for Sunrunner, U of M's entry in the 1990 GM Sunrayce USA. A Cb cloud began to form to our W at 1630 Eastern Standard Time (EST). Evening charging was only allowed from 1730-1930 EST in that race. By that time, the Cb had grown quite large and its anvil sunk and covered the whole general area around the sun; yet most of the sky was clear. This significantly decreased our solar energy collection during charging, and when I was brought to the charging site, I noticed our solar array was pointed toward the clear sky, slightly ESE of zenith. I suggested to our strategist Dave Bell that it may be better to point the solar array to the general direction of the sun, and he asked that I verify that the best direction was chosen. My solar flux measurements confirmed this. All other teams had their arrays pointed toward the sun.
Local variations in solar flux also play an important role in racing. It can be shown that velocity variations under partly cloudy skies can optimize vehicle performance using silver-zinc batteries. The efficiency of silver-zinc batteries is relatively constant as a function of current draw. One can slow down their solar car in the clear areas, collecting more solar energy than at a constant velocity. The velocity variations will cause a greater kinetic energy loss, due to an increased aerodynamic drag, but solar energy gain can overcome this. As an example, under 50% Cu shadowing for typical summer conditions in the United States near solar noon, traveling 45.3 mi/h (72.9 km/h) under clouds and 35.8 mi/h (57.6 km/h) under clear areas, averaging 40 mi/h (about 64.4 km/h), uses approximately the same energy amount as traveling at a constant speed of 39.0 mi/h (62.8 km/h) (figure 34). However, this is true only if the tactics are run perfectly. Slightly less of an advantage will typically be gained. Using lead-acid batteries, as we were restricted to in Sunrayce 93 for cost, availability, and familiarity reasons, this tactic would not work well, since their efficiency decreases rapidly as a function of current draw. Obviously, much more current must be drawn from the batteries traveling fast under clouds (without much help from the solar array) than traveling slow in clear areas (during which the batteries may actually be charging!).
An interesting phenomenon which I have observed in working on solar car projects is the augmentation of solar flux under scattered cloudiness at times when the sun is near a cloud edge; particularly under thin, whitish-based cumuliform clouds and at a high solar elevation. Hourly solar radiation averages have been reported as high as 1340 W/m². The first time I met George Ettenheim, he asked me under which conditions I would expect the most solar radiation, and then told me it was under a thin altocumulus overcast. He exaggerated by saying "overcast" and perhaps meant broken-scattered, stating that the sun only stays behind the darkest areas of cloud for a short period of time. Otherwise, solar radiation can be greater than that under a clear sky; especially when the sun is clearly visible. These events are usually short-lived, but very good for solar car racing! In an analysis of 23 years of Australian solar radiation data (from the National Climate Center, Australian Bureau of Meteorology) for October and November along the World Solar Challenge route, the greatest 30 min. average solar flux measured was 1278 W/m² at Darwin from 1200-1230 Local Mean Time on 11 November 1974. The highest single reading of GSR I have measured is 1357 W/m² at Dearborn Heights, MI (solar radiation capital of the world - ha ha), under a 25%-covered Cu humilis sky at 1136 EST on 9 July 1991. For reference, my estimated horizontal extraterrestrial solar flux at this time is 1216 W/m²!
Regarding the role of weather specialists in a solar car race, it is too restrictive to force them to deal exclusively with weather and its forecast in all cases. We have been blessed with other talents also, and the team will benefit greatest if we use them all. This is true for everyone. A number of cases are illustrated above which show how we contributed to the racing of Maize & Blue. In some situations (such as when we timed Cal Poly @ Pomona), we were not doing "meteorology" at all! Energy, and its efficient and timely use, is always the most important aspect of racing, and human energy can even improve one's forecasts. Thus, rather than just "forecasting the weather", a weather specialist on a race or similar activity should always keep their goal in mind and tailor their forecast to achieve it.
Working on the U of M's Sunrunner and Maize & Blue teams has been a very enjoyable and
rewarding experience for me. In these pages, I have tried to relate what issues were faced
as a weather specialist in Sunrayce 93, and how they were dealt with. Hopefully, some insights
into the behavior and forecast of solar energy are also apparent here. I encourage participation
in this type of activity to anyone who is fortunate enough to have the opportunity to do so!
Text and embedded graphics are copyright of Joseph Bartlo, though may be used with proper crediting.