Dirk Baker of Campbell Scientific set up separate temperature-measuring instrumentation at the official Death Valley/Furnace Creek weather station in May, 2021. The purpose was, in part, to compare measurements with the current, official Campbell Scientific (CS) temperature equipment at Death Valley through the summer of 2021. It was later decided to extend the comparison through the summer of 2022 at least. Baker has provided a couple of excellent write-ups which detail the CS equipment in use at the Death Valley station. A more-recent article (dated July 29, 2022) provides an analysis on how several different sensors, aspirated (active) and not aspirated (i.e., passive), compare. This blog entry is meant to draw attention to Baker’s study and to highlight (and maybe explain) some of the important and interesting comparisons, especially during the record-hot period during July of 2021.
Baker’s first article which describes the setup is here (June 21, 2021)
And the follow-up article is here (Death Valley, a Year Later/July 29, 2022)
Link to the nitty-gritty findings page, as presented as a poster at an AMS conference
My photo from July 10, 2021, (above) shows the Death Valley official instrumentation on the left and Baker’s newer (temporary) instrumentation on the right. (Another good photo of the instrumentation, by Baker, is in his AMS poster.) The official Campbell Scientific instrumentation (on the left) is 100 percent automatic and includes the anemometer and wind vane, the rain gage, and the passive radiation shield which houses the temperature sensor and the hygrometer (CS 215L). The temporary equipment (on the right) includes active and passive HygroVUE 5 sensors/shields, which were developed as a replacement for the now discontinued CS215L sensors. The active shields are the ones with the wider covers on top.
The first article by Baker (from June 2021) describes the new instruments and the advantages of using an aspirated radiation shield. (In my opinion, it is rather strange and unfortunate that the official DV station does not incorporate an aspirated thermometer shield. There may be a good reason for this…perhaps sufficient electrical power to run an aspirated fan was/is not available at the station site. ) Since the current Death Valley station site is somewhat poorly ventilated compared to earlier decades (due to an increase in development around the Visitors Center building), and since the station is above bare ground, summertime maximums on sunny afternoons tend to drift too high as it is. The “passive” radiation shield only accentuates, or maybe a better word is “exacerbates,” the inadequate ventilation situation. According to Baker: “At times of low or no wind…air can become trapped in a passive shield that also causes a bias.” The maximum temperature reports from Death Valley/Furnace Creek, especially on the hottest days with light wind, run about 1-3 degrees (F) too high, in my estimation Another long-winded way of saying this is:
Death Valley maximums in summer tend to trend a degree (F) or two, or maybe even three degrees too high, especially when afternoon wind is light. This is when compared to the expected maximums based on the climatology and meteorology, and compared to maximums at surrounding stations, and when compared to adequately-ventilated locales not too far from the Furnace Creek station. The reason for this “high bias” is the somewhat wind-protected station location and the non-aspirated radiation shield at the Furnace Creek Visitors Center. Here is my write-up in 2020 on the Death Valley weather station ventilation status.
Baker’s “A Year Later” update contains the link to the poster presentation at the AMS annual meeting in January, 2022. Baker leaves little doubt that the current official Death Valley station instrumentation has a bias towards warm temperature measurements. He addresses the importance of station siting, saying: “Carefully selecting the location (or siting) for a weather station and mounting the sensors at or near two meters above ground help address the question of representativeness.” (I find it interesting that Baker veers a little a bit into this weather-station siting guidance. Perhaps he shares my concerns with the exposure and ventilation issues at the Death Valley site.)
The title of Baker’s poster at the AMS conference was “Measurement and Uncertainty in Death Valley Temperatures,” and its first “key point” is
“July 2021 CS215 high not supported by other sensors.”
(The CS215 instrument is the official Death Valley temperature-measuring sensor and passive radiation shield combo, manufactured by Campbell Scientific. This is the one that provided the record maximum temperatures of 130F in the summers of both 2020 and 2021.)
That lead-off key point by Baker does not exactly drip with confidence, does it?! This throws cold water, so to speak, on the “official” high temperature of 130F at DV on July 9, 2021, and perhaps also onto the 130F on August 16, 2020. Might the NWS wind up knocking a degree or two off of these 130F maximums?!
Baker discusses the “uncertainties” associated with determining temperature with such electronic equipment. Some sensors are very sensitive and quick to respond to temperature change, some not so much. There are different methods and different timespans to consider. Should the “current” temperature be based on a one-minute average? A 5-minute average? How about the daily maximum temperature? Should it be the highest of all of the 1-minute temperatures? Or the highest of all of the instantaneous readings? I asked Baker how the current Death Valley temperature is determined, and this was his response (which can be found in the comments section of his first article):
——
Hi Bill,
Thanks for your comment and I apologize for the delay in getting back to you! I wanted to confirm with my collaborators that they were ok with me sharing that information.
The NWS station samples the sensors once per minute. The maxima and minima for the hourly reporting are based simply on the numerical maximum and minimum for a given hour and the time at which those occur. There is no additional processing.
That said, the sensor on that station is the CS215 which, according to its specifications, has a time constant of 120 seconds or less. Assuming (for the purposes of conversation) that it is exactly 120 seconds for that specific sensor; theoretically, it would take about 360 seconds for it to fully equilibrate to a step change in temperature. You probably already know this, but it is worth pointing out in a public post.
Whether that is truly equivalent to smoothing or averaging is probably debatable, but it will dampen the sensor’s response to short-duration fluctuations.
Best,
Dirk
So, it appears that the hourly and daily max and min temps at the current DV station are based on samples taken every minute. It would appear that any quick and meaningful temperature fluctuations between the one-minute samples might be missed. The “time constant of 120 seconds” means that, in the event of a temperature change, the instrument will have a 63 percent response in 120 seconds or less time in air moving at one meter per second. Don’t quote me on this, but I think that this means that if the actual AND instrument-measured temperature is 110.0F, and if the actual temperature suddenly increases to 113.0F and it stays at 113.0F, and if air is moving past the sensor at one meter per second, then by 120 seconds the sensor will show at least 63 percent of this three-degree F change (which in this example would be about two degrees F, or very nearly 112.0F). I think we can presume that if there is a decent breeze blowing that these responses to temperature will be faster. Even though the temperature sensor is quite sensitive, it is not super-sensitive to temperature change, and very quick “blips” in temperature up or down are masked to some extent. When combined with the one-minute readings (as opposed to one-second readings, for example), the temperature measurements by the CS215 are “dampened.” If the instrument’s time constant were faster, and if readings were taken more often than every minute, then somewhat higher daily (or hourly) maximums and somewhat lower daily (or hourly) minimums would be the result. We are generally not talking about more than a degree (F) here either way on most days, I suspect, but it is just something to keep in mind with these electronic stations. For comparison, the official ASOS stations at our airports in the United States base the daily max and min temperatures on 5-minute averages. Raw temperature is measured each minute by the ASOS instrumentation (in an aspirated shield!), and the ASOS one-minute temperature display provides the average of the five previous raw temperature measurements. These one-minute temperatures (based on the 5-minute averages) are then used to determine daily max and min temperatures. If the day’s hottest official ASOS one-minute temperature is 100F, for example, and if it were derived from five raw measurements of 99, 103, 99, 100 and 99, then the day’s high temperature winds up at 100F. Nobody will ever know that on the measurement for one particular minute that the ASOS measured a 103F, unless an observer on hand took a peak at the raw data!
My temperature measurements with the sensitive Kestrel instrument (both while walking around and while driving) on hot and sunny summer afternoons in Death Valley have revealed that hot “bubbles” are commonplace. The air temperature at shelter level may quickly vary by about 2-4 degrees F (as measured with the sensitive handheld Kestrel) as the bubbles come and go. This is one reason why I am harping on this sensor sensitivity and frequency-of-measurements stuff with the electronic equipment. For a station in a humid area with plenty of grass and vegetation around the station, these instrument and measurement details are probably not of much concern. But at a desert station, and especially a desert station with ventilation issues, it is a good idea to know these things!
Since not all sensors, not all weather stations, not all radiation shields, not all temperature sensor-level heights, not all exposures, not all ground covers, etc., etc., etc., etc., are the same…well, you get the idea, I hope. Temperature measurement with these electronic stations might not be as straight-forward or simple as it seems.
July 9, 2021 Maximum of 130F at Death Valley
Baker’s most recent article looks closely at the data for July 9, 2021, when the official DV/FC station recorded a 54.4C/130.0F maximum. The maximum occurred at 2354Z, or 3:54 p.m. PST. It occurred during a period of about 15-20 minutes when the temperature was about two degrees C higher than during the remainder of the afternoon. Baker calls this the “spike.” From about 1:00 p.m. PST to about 6 p.m. PST, temperature was mostly 50-52C, except during the spike. Again, these were the measurements by the sensor in the passive shield on the official Death Valley equipment. Below is the graph by Baker for the afternoon and evening of July 9. It compares the readings with the adjacent temporary sensors. The CS215 line is red and the times shown are PST.
Added above is the hourly temperature, humidity and wind data for the Death Valley/Furnace Creek station. The times are PDT, so 5:00 p.m. PDT = 4:00 p.m. PST.
And, what the heck, let’s blow up the July 9 afternoon section.
The three thermistor beads in an active (aspirated) radiation shield (black line in the graph) were thought to be the best reference for comparison purposes. From Baker:
“My thinking was that the beads—mounted in an active shield—would provide the highest accuracy measurement of air temperature against which to compare the rest of the measurements. This thinking was not wrong, but there are some key caveats.”
The “beads” have a very short time constant, just 7 seconds! And since the beads were in an aspirated shield, they were superb at picking up the short-lived temperature changes (the hot bubbles!) during the afternoon. The temperature uncertainty of these beads at these high temperatures is negligible, just plus or minus 0.05 degrees C. The uncertainty of the CS215 sensor is 0.9 degrees C for temperatures from 50C to 60C. A close examination of the graph for the beads from about 1 p.m. to 6 p.m. PST shows dozens of rather rapid temperature spikes and drops of one to three degrees C magnitude. From 1 p.m. to 3 p.m. PST the beads show approximately 22 instances of temperature spikes to 51C or more, and 28 spikes to 50C or lower. There are ten spikes to 49C or lower (the lowest is about 48.2C/119F) and about six spikes to 52C or higher (the highest is about 53.2C/128F). From 1 p.m. to 3 p.m. PST the HADS data page for the official station shows that temperature ranged from 122.0F to 126.5F (50.0C to 52.5C). Obviously, there is a significant “dampening effect” with the official station instrumentation compared to the beads in the aspirated shield.
Timeframe, 1 p.m. to 3 p.m. PST/Death Valley weather station
Temperature range with the official instrumentation: 122.0F to 126.5F/50.0C to 52.5C
Temperature range with the “beads” sensors: 119F to 128F/48.2C to 53.2C
From 3 p.m. PST (1500) to about 3:45 p.m. (just prior to the pronounced spike to 130F with the official temperature sensor) the measurements with the sensitive beads showed about a half dozen swings from 50C to 52C (122.0F to 125.6F). In one short timeframe of only 4-6 minutes (around 3:15 p.m. PST) the beads showed a change from about 49C to 53C to 50C (120F to 127.5F to 122F). During this same timeframe the CS215 sensor indicated little change, around 1 degree F at most. Note that the CS215 sensor plot rides along towards the top of the spikes on the beads plot. In the 10-minute running average chart provided by Baker, the CS215 temperatures from about 1 p.m. to 3:45 p.m. PST are decidedly warmer than the beads. The difference is close to 2 degrees F (1.1 degrees C). Could this discrepancy be due to the “uncertainty” factor with the CS215 sensor? If the weather turned cooler for the day at 3:45 p.m. PST, then the CS215 max temp for the day would have been 126.5F (rounded up to 127F) and the beads max temp would have been about 128.5F. The greater sensitivity of the beads makes a big difference! Though the CS215 ran about two degrees F warmer on the 10-minute averages compared to the beads, the beads came out ahead by two degrees F on the max temp through 3:45 p.m. PST.
Baker’s “spike,” when the CS215 jumped up to 130.0F/54.4C at 3:54 p.m. PST, commenced around 3:45 p.m. PST. The HADS dataset for the official station shows 129.5F/54.2C at 4:00 p.m. and 129.7F/54.3 at 4:02 p.m. The one-minute chart (the red line!) shows a drop back down to 126.5F/52.5C a few minutes later. (The CS215 did not register a temperature warmer than 127.0F/52.8C for the remainder of the day.) The one-minute wind data chart shows speeds of primarily 0.5 to 1.2 meters per second (just 1 to 3 mph) during the period from about 3:45 to 4:05 p.m. PST. The HADS dataset indicates a SSE wind direction at 4 p.m. PST, but the speed is calm.
Amazingly, and perhaps alarmingly, the 10-minute average running temperature chart shows the CS215 some 2.0 to 2.5 degrees C warmer than the values provided by the beads during the timeframe of the spike. The highest 10-minute value for the CS215 is close to 53.8C/128.8F, and at the same time the beads show only 51.5C/124.7F. These are approximations, eyeballed off of Baker’s chart.
During the approximately 10-minute timeframe when the CS215 shows 53.5C to 54.5C/128F to 130F (off of the one-minute chart), the beads are coming in with a much larger range of temperature, about 49.2C to 53.2C (121F to 128F). Yes, when the beads were at 121F, the CS215 was no lower than 128F. The remainder of the daylight hours shows the CS215 running 1-2 degrees C warmer than the beads, for the most part. After sunset, the CS215 appeared content to run about one degree C cooler than the beads.
On the following afternoon, the official high (with the CS215) was 129.4F. The temperature differences between the CS215 and the beads during the afternoon were not quite as large as on the 9th, likely due to the breezier conditions.
Taking a quick look at the other sensors for the afternoons of July 9 and 10…
The “109” sensor in the passive shield (only 0.05 deg C uncertainty and a fast time constant of 20-30 seconds) yielded temperatures that were fairly close to the official (CS215) equipment during the afternoons, with generally slightly higher upward spikes due to the greater reaction times (less dampening). However, during the “spike” on the 9th, this 109 sensor remained close to one full degree C lower than the CS215. On the 10th, the 109 sensor managed a high temperature slightly greater than the 129.4F with the official CS215.
The HygroVUE 5 (time constant <130 sec and 0.4C uncertainty factor at 50-60C) sensor in the passive shield was generally 0.5 to 1.0 degree C lower than the CS215 on the 9th, but a full two degrees C lower during the spike timeframe. The HygroVUE gave a maximum of only 127F on the 9th, and close to 129F on the 10th (compared to 130.0F and 129.4F for the CS215). Why would the CS215 be so much warmer than the HygroVUE sensor, in the same passive shield? Again, during the “spike” timeframe, wind was very nearly calm. The two sensors were at the same height and only a few feet apart. About 0.5C/1.0F of the difference on the max temp on the 9th MIGHT be attributed to the “uncertainty” factor, as the CS215 is at 0.9C and the HygroVUE is at 0.4C. The other 1.0C/2.0F difference is less obvious, but I can’t help but go back to the proximity of the black solar panel and electrical box to the CS215 on the official instrumentation. Here is a pic from August 2020:
The black solar panel is rather obscured in this view, but the picture in Baker’s poster shows it nicely. The official temperature sensor for the most important weather station in the world (in my biased opinion!) is in a passive radiation shield that is about six inches from a white electrical box and maybe 12 inches from a solar panel. And, it is a couple of inches from a gray metal support. Do you think that these devices might get super hot in the Death Valley summer sun during the afternoon? Do you think that, during calm winds or during a very light drift from the east, that heated air along the box and the panel and the metal support might manage to drift into the nearby radiation shield? The setup by Baker would appear to be much less susceptible to such spurious effects.
Not surprisingly, Baker’s HygroVUE sensor in the active/aspirated shield showed lower temperatures than the same sensor in the passive shield during the afternoons of the 9th and the 10th. The differences on the 10-minute average chart between these two were around 0.4C to 0.7C for the most part, based on my estimation. The maximum temperature on the 9th with the aspirated HygroVUE was only 52.5C/126.5F, or 3.5 degrees F lower than the 130.0F on the CS215 in its unaspirated shield! I agree with Baker: the 130.0F maximum off of the CS 215 is not supported! On the 10th, when the CS215 had 129.4F, the aspirated HygroVUE reached a max temp of about 53.6C, or around 128.5F. The better breeze on the 10th would have reduced any enhanced and spurious heating off of the box/panel/supports near the CS215.
SOME CONCLUSIONS!
When Death Valley reached 130F in August, 2020, and again in July, 2021, the hottest maximums occurred a day prior to the “peak” of the event (the “peak” defined as the day when most other stations in the Death Valley region had their maximum temperatures for the event, and when upper-air temperatures from the Las Vegas sounding were highest from about 850mb to 700mb). The reason for the one-day early Death Valley maximums are due to wind, station exposure, and perhaps the nature of the instrumentation of the official DV station. Wind is generally lighter during the afternoons in Death Valley as a summer heat wave is building as compared to the very peak of the event and subsequent afternoons. Thus, on July 9, 2020, a day before the hottest air of the event was over the region, wind was very light, and the official station ticked off a few extra degrees F upward thanks to the combination of the poor ventilation through the station site, the bare ground, the passive/unaspirated radiation shield, and perhaps the sensor’s proximity to the electrical box and solar panel and gray metal support.
The “fix” should not be too difficult. Get the sensor in an aspirated radiation shield and further away from metal stuff that is in direct sunlight! And if you are really serious about getting the best maximum temperatures at Death Valley which compare fairly to those in earlier decades, move the station site to a spot with zero obstructions to the south, such as the final station site for Greenland Ranch, just east of Highway 190!
Should the readings of 130F in the summers of 2020 and 2021 be reduced a degree or two or three F? Yep.
Leave a Reply
You must be logged in to post a comment.