The 1913 maximum temperature world record at Death Valley: should we expect a new record in the near future?

Abstract

In accordance with record high global mean temperatures new local maximum temperature records have been observed in many places worldwide over the last few years. The question arises, what is the highest 2-m temperature on earth to be expected in the near future due to global warming? The investigation focuses on the temperature time series of Death Valley, California, the present world record holder of maximum temperature. A critical review is given on this maximum temperature world record of 134°F or 56.7°C, set on 10 July 1913. Different evaluations like comparison with neighbor stations, comparison with 20th century re-analyses and measures of dispersion are carried out. They all show that this record is highly questionable, because no physical mechanism is known, which could explain such a statistical outlier of the maximum temperature on a local scale over desert regions. A new low-pass filter technique of time series is being used to determine the long-term climatological temperature trend between 1911 and 2023 in Death Valley in the frame of global change, which is quite impressive. Finally, the probabilities of the occurrence of certain temperature thresholds in the near future are derived by utilizing a general extreme value distribution. It is shown that the probability of a new temperature world record of 135°F or 57.2°C in the next few years, which would make the present—albeit questionable—record obsolete, is still very low, despite global warming.

Introduction

For a long time, the maximum temperature world record was attributed to a Libyan Station, El Aziziya, with its 58.1°C (136.5°F), observed on 13 September 1922. A scientific evaluation showed that this record obviously was false and consequently it was refused by the World Meteorological Organization WMO [1]. Since then, the new record holder is the Furnace Creek Ranch at Death Valley in California, USA (furthermore denoted by DV). With its 134°F (56.7°C) it is dating back more than 100 years to 10 July 1913 [2]. At the time of the record the location was called Greenland Ranch. Many publications were dealing with the extreme climate of DV and its remarkable temperature record [3–8]. It is surprising, that no other place on earth including DV recorded a similar high temperature since then, despite generally increasing temperatures and many new records in other regions in the frame of global warming [9, 10]. There have been several investigations carried out concerning the plausibility of the Death Valley world record but no final decision on its validity has been made neither by the US National Weather Service nor by the WMO so far [11, 12]. A most extensive and sound investigation with many documents and photographs has been delivered by WT Reid, who finds that the record of July 10, 1913, due to a comparison with neighbor stations is more than questionable [13].

In the present paper additional scientific arguments are collected to receive an answer on the plausibility of the 1913 world record of DV. Furthermore, the question is addressed on the statistical probability, that a certain temperature threshold will be reached or exceeded in the near future. In section “Exploitation of the DV temperature time series” a careful statistical exploitation of the temperature time series of DV is presented. In section “Consideration of trends in the time series” temperature trends of the temperature time series of DV are investigated and an adjustment of the series to the actual climate is carried out. In section “Homogeneity of the DV time series”, inhomogeneities of the DV series are investigated by comparing the series with data of neighboring stations and with NOAA/CIRES 20th century reanalysis and the European Center for Medium Range Weather Prediction ERA20C reanalysis. In section “Plausibility of the 1913 heat world record in Death Valley”, a conclusion on the plausibility of the present valid maximum temperature world record is given. Section “The probability of a new maximum temperature record at DV”6 deals with the probability of a new temperature world record at DV in the near future. Estimates are presented if and when the temperature at DV will reach or exceed the 131°F, 132°F or 55°C, 56°C respectively and higher thresholds in the future.

Exploitation of the DV temperature time series

The series of the daily maximum temperature (furthermore denoted by TX) data of DV has been retrieved from NOAA’s National Centers for Environmental Information website [14, 15]. The temperature is given in full °F so that the transformation to °C contains an uncertainty of ±0.28°C. This uncertainty is, furthermore, expressed by a “∼”. The DV temperature time series contains several data gaps, which must be considered, when computing statistics. In the whole 113-year time series from 1 January 1911 to 31 December 2023, 106 one-day and longer data gaps exist in 49 years, especially during the first few decades. During the first half of the period (1911–1966) 35 years are incomplete, during the second half (1967–2023) only 14. The available data cover anyhow 98.4% of all possible data. One- and two-day gaps have been filled by the author through linear interpolation, so that the basis for the following statistical evaluation covers a data availability of 98.6%. Unfortunately, even after filling the short gaps a considerable number of years (18) is still affected by longer data gaps, so that no proper time series of annual mean values can be derived. The number of available data for single calendar days in the whole series varies between 105 in early January to the full 113 in most of the summer days. For each calendar day of the year, except for February 29 in leap years, the average TX was calculated as well as the highest maximum (daily maximum maximorum, furthermore denoted by TXX) has been retrieved from the whole available time series from 1911 to 2023. If we compare the absolute heat record of 134°F ∼ 56.7°C in DV on July 10, 1913, with the other daily maxima, we are surprised to see that no other maximum in the whole long history until today except the 131°F ∼ 55.0°C on July 13, 1913, three days after the record—exceeded the 130°F ∼ 54.4°C level so far, see Fig. 1.

The time series have been smoothed with a MEan Value Spline Smoothing (MEVSS) filter [16, 17]. This method outperforms other smoothing methods like Gaussian, LOESS or LOWESS. MEVSS is obtained by a variational procedure. The smoothest possible curve, defined by the smallest possible sum of squared second order time derivatives/differences is determined under the (strong) constraint, that the mean value of observed values within certain time windows is conserved. The choice of smoothing time windows is somewhat arbitrary, but it is obvious here to take 30-year intervals (30-year MEVSS) for climatic trends, because a 30-year interval is recommended by WMO to define the short-term climate. The absolute minimum of smoothing windows to smooth an annual cycle is four (seasons). The smoother the input data, the shorter the windows can be chosen. Hence, for the smoothing of long term (113 year) mean daily TX and TXX series of an annual cycle monthly mean values (30-day MEVSS) have been chosen. One advantage of the method is the conservation of minima and maxima for sinusoidal time series. Another important advantage is that a reasonable smoothing of non-periodic time series may be carried out from the very beginning until the very end of the series. The MEVSS curves also allow to compute the trend-adjusted Variance of the data for series with a trend. The (squared) standard deviation (STD) of the data around the MEVSS curve is obtained by computing the squared difference between the observed and the MEVSS data for each calendar day and smoothing them again by a MEVSS. Because of the stronger relative scatter of the squared differences of TXX as compared to the mean TX, a larger smoothing window is recommended for the squared differences. We have chosen a 60-day window for the MEVSS of the squared differences in Fig. 1. The square root of this MEVSS values of the daily TXX finally gives its STD. It is largest in winter (approx. 2.5°F or 1.4°C, not shown) and smallest in late summer, with approx. 1.5°F or 0.8°C. This is due to more frequent and intense air mass changes in winter than in summer at DV.

Usually, values in a time series exceeding the mean plus/minus three times the STD are characterized as outliers. Therefore, the three times STD curves are also plotted in Fig. 1. In contrast to the mean summer TX, which all lie below the respective + 3 STD curve, two striking outliers appear for TXX. The first one on 10 July 1913 with the world record and the second on 16 August 2020 with 130°F ∼ 54.4°C. Does that mean that the climate of Death Valley is permitting gross outliers? We shall try to get an answer on that with different statistical approaches.

Instead of looking at daily maximum temperatures like in Fig. 1, we can now look to annual maximum temperatures (annual maximum maximorum, furthermore denoted by TXXX). Fortunately, only in five years with longer data gaps in summer, values are unavailable. In years with one- and two-day data-gaps in summer the missing values have been checked with the aid of reanalysis data, if they eventually could have contained the annual maximum, but none were found to be high enough [18]. The time series of TXXX between 1911 and 2023 is shown together with the 30-year MEVSS curve and the 30-year MEVSS + 3 STD curve in Fig. 2. The STD of TXXX has been computed as the root mean square of all differences TXX minus the corresponding MEVSS values. For comparison the series of mean summer (Jun-Aug) temperatures at DV and the North American temperature deviation from the 20^th century mean are plotted [10, 15]. We see again that the 134°F ∼ 56.7°C in 1913 appears as an exceptional peak in the series. One may wonder, why the maximum temperature world record occurred just at a time, when North America observed the lowest mean summer temperatures of the series. The 134°F ∼ 56.7°C of 1913 represents a pronounced outlier like in Fig. 1 with a + 4.2 STD difference to the smooth curve. The second outlier of Fig. 1 in 2020 is now not at all an outlier, because the smooth MEVVS curve of TXXX and hence also the +3 STD curve shows a strong trend within the series. When we compare the trends of TXXX and $TX-$ of DV in Fig. 2 we find a quite good agreement. After relative high values during the first few years of the series there is a negative trend towards a minimum around the middle of the 20th century. After that an increasing positive trend is leading to the highest value of the whole series at its end—the sign of global warming. The trend in the second half of the TXXX and $TX-$ series is in good agreement with the North American mean summer temperature but in the first half the trends are opposite. The higher temperatures in the first decades of the DV series disagree with the whole of North America.

Consideration of trends in the time series

Whereas until the 70-s of the last century the mean North American as well as global (climatological) temperature changes were marginally. Since then, a substantial increase of the mean and extreme temperatures has taken place, so that trends must be considered for statistical evaluations [19]. The climate change modifies the frequency distribution of observed data. Figure 3 shows the cumulative frequency distribution of all summertime TX in the 10-year period of 1956–5 and of 2013–2. The form of the distribution itself has not changed but the level has increased dramatically by roughly +4°F or 2.2°C. This means that e. g. a temperature of 122°F ∼ 50.0°C or more was observed only in 4% of all summer days in the middle of the last century, whereas today more than 15% lie above this threshold. The shift of the frequency distribution also means that the probability of a certain maximum temperature of the mid-20th century corresponds to the probability of a 4°F or 2.2°C higher temperature today. If we want to apply extreme value statistics to observational data in times of climate change, we must consider this change and de-trend the data.

The de-trending of the data can be carried out by the following procedure: Due to the existence of longer data gaps we must first compute MEVSS curves between 1911 and 2023 for each calendar day of a year and from that compute MEVSS curves from 1 January to 31 December for each year of the series. This leads to a 2-dimensonal representation of the smooth (climatological) temperatures on each calendar day within the given series of years. In Fig. 4 such a 2-dimensional display for the smoothed TX of DV is shown. The increase of the mean (climatological) TX in midsummer is remarkable. Whereas the mean TX in the second half of July 1960 was below 116°F ∼ 46.7°C, it climbed to nearly 120°F ∼ 48.9°C in 2023, in perfect agreement to Fig. 3. If we cut the diagram in Fig. 4 horizontally at a certain year, we get the mean (climatological) annual temperature variation for that specific year. If we cut the diagram at a certain calendar day vertically, we get the mean (climatological) trend of TX for that specific day. By averaging these daily curves over a month/season/year, we easily get monthly/seasonal/yearly trends of TX.

If we want to de-trend the observed TX of DV and adjust it to the present climate (2023) we compute the difference between all observed TX and their smooth (climatological) value (according to Fig. 4) on each day of the series. Then we add this difference for each day of each year to the daily MEVSS values of the year 2023. The so adjusted series can be evaluated as if the climate was constant during the whole period. If we now compute the same series like in Fig. 1 but from the adjusted values, we receive Fig. 5. The curves have shifted towards higher values. The value of the 134°F ∼ 56.7°C outlier of 10 July 2013 would be an outlier of an incredible 136.7°F ∼ 58.2°C under the present (2023) climate, with an even larger deviation of + 4.4 STD. A quite similar deviation of the DV absolute record of + 4.5 STD is given by [11]. The value of the 16 August 2020 outlier in Fig. 1 is now just a little higher (130.4°F ∼ 54.7°C), well below the limit of + 3 STD, because the climatological temperature change between 2020 and 2023 was only that small. The discrepancy between the two outliers in Fig. 1 and Fig. 5 is caused by the non-stationary climate. Hence, in the whole adjusted series of daily absolute maximum temperatures of DV only the 10 July 1913 outlier remains. A value with a deviation of + 4.4 STD needs a good argument to be accepted. If we look at the course of the smoothed daily extremes in Fig. 5, it seems, that the extremely high TXX values around 10 July—mostly originating from the year 1913 - have an impact on the curve. If we compute Fig. 5 under omittance of the year 1913 (not shown), the hump around 10 July vanishes and the 10 July 1913 value lies at an incredible +6.6 STD above the smooth MEVSS curve.

We know of explainable outliers in temperature series of stations exceeding the +3 STD. One remarkable example is the Canadian temperature record of 49.4°C = 120,9°F at Lytton, BC on 29 June 2021 [20]. The series of TXXX for that location is shown in Fig. 6. It shows that this record was an outlier of +4.4 STD. The reason for this outlier can be explained by an unprecedented heat dome which extended to British Columbia in summer 2021 due to upper tropospheric wave resonance effects which are favored by climate change [21]. A mechanism like in Lytton cannot be made responsible for an outlier at Death Valley, because a heat dome is sitting every summer over the US-Southwest.

Homogeneity of the DV time series

It is common practice to compare time series of neighbor stations to prove, if there is an inhomogeneity in one of the series. Stations should not lie too far apart and should possibly belong to the same climatic regime. The closest station to DV at the beginning of the 20^th century was located at Independence, CA (furthermore denoted by IN) in the Owens Valley, which lies on the dry leeside of the Sierra Nevada, some 124 km (77 miles) Northwest of DV. The nearest station to the Southwest of DV was 161 km (100 mi) distant at Jean, NE (furthermore denoted by JE), close to Las Vegas. Like DV the two other stations lie in a desert area with very dry and hot summers. The big difference between the stations is their elevation. Whereas DV lies in a depression with a negative elevation of −51.2 m (at the time of the heat record in 1913), IN lies at an elevation of 1181.1 m and JE at 872.9 m. Fortunately, height differences do not play an important role for a correlation of summer time TX over desert areas, because under such conditions the lower troposphere shows a very deep well mixed layer with a dry adiabatic lapse rate at the time of the daily temperature maximum. The potential temperature stays nearly constant with height up to the mid troposphere. Close to the ground super-adiabatic lapse rates may occur during the day but this should be the case at all locations and hence not affect the difference of the potential temperatures between two stations. Figure 7 shows an example for a deep mixed layer with a constant potential temperature up to more than 5000 m above Las Vegas on June 30, 2013, in the early afternoon (00 UTC) [22]. A deep mixed layer has a further advantage for comparing neighbor stations: A change of the station temperature means that the whole well mixed layer temperature must change accordingly, which needs lots of energy. The heating of a dry adiabatic column of air from 1000 hPa up to 500 hPa with a mass of roughly 5000 kg m⁻² by just 1°C (1.8°F) consumes an energy of approximately 5 MJ. This is equivalent to a heating of the air column by roughly 700 W m⁻² for the duration of two hours. A deep mixed layer acts like a thermostat for the surface temperature. This is also the reason, why the daily variability of the summertime TXX is considerably lower than in winter. When the stable boundary layer in the morning after a clear night has been burnt off by insolation, the further temperature increase over deserts becomes very slow. Furthermore, even small spatial differences of a deep mixed layer temperature create large pressure gradients. They immediately cause thermally driven circulations (breezes) which try to equalize these differences. As the depth of the mixed layer is somewhat smaller over higher terrain, the heating will be a little more effective there and consequently a small increase of the potential temperature with station elevation will generally occur. Nevertheless, the best spatial correlation of TX on earth will occur in desert areas on summer days.

If we reduce all TX of the nearby stations dry adiabatically to DV we must add 11.4°C (20.6°F) for IN and 9.0°C (16.2°F) for JE. The three temperature-series for summer 1912 are shown in Fig. 8. It is not a surprise that the TX of all three stations is highly correlated. The correlation coefficients between DV and IN and between DV and JE are quite high (see Table 1). The somewhat lower coefficient for JE might be caused on one hand by the larger distance from DV and on the other hand by the fact, that some cooler air intruding from the North, which show a clear signal at IN and somewhat weaker at DV hardly made it to Jean in summer 1912 (e.g. on 10 June, 22 June, 3 July and 18 August). To avoid a too strong influence of the short cool spells at IN and DV, the median instead of the mean value of all summer (1 June to 31 August) TX has been used for comparison. The median of the difference of the reduced TX between DV minus IN was -0.6°F or −0.3°C and between DV minus JE −3.2°F or −1.8°C. As expected, DV was a little (potentially) cooler than both other stations due to their higher elevation. We must generally admit that the temperature readings at that time could contain some errors, because they were usually not taken by well-trained meteorological professionals [13]. Therefore, it is interesting to see how the statistical comparison in the area looks like for modern observing technology. Unfortunately, both stations IN and JE are not operational anymore and also the station location of DV has changed slightly since the early 20th century (new elevation is −59.1 m). Now there is a station at Oak Creek (furthermore denoted by OC), very close to IN at an elevation of 1479.8 m. Instead of JE a station at Las Vegas (Las Vegas forecast office, furthermore, denoted by LV) at an elevation of 619.7 m can be used for the comparison. In summer 2020 the correlation between DV and OC was r = 0.92 and between DV and LV r = 0.94, much higher than the 1912 and 1913 values. Hence, we can infer that the quality of the manual observations in the early 20th century was below today’s standard. The median difference of the daily reduced TX at DV minus OC in summer of 2020 was −3.5°F or −1.9°C and DV minus LV −4.2°F or −2.3°C (see Table 1). These numbers confirm that the spatial correlation of summertime TX in the area is excellent, and that the potential temperature increases on average slightly with elevation.

The more we are surprised, when we look at the diagram for 1913 (Fig. 9). Despite of still similar high correlation coefficients r(DV-IN) = 0.86 and r(DV-JE) = 0.74, the difference of the median value DV minus IN now gives a positive value of +5.9°F or +3.3°C and DV minus JE is only slightly negative with −0.7°F or −0.4°C. Other stations in the area like Tonopah, NV, 110 mi to the North of DV or Logandale, NV, 216 km (134 mi) to the East and even stations like Flagstaff, AZ and Yuma, AZ some 500 km (300 mi) to the SE and S of DV confirm these findings that between 1912 and 1913 there is an inhomogeneity in the DV series. The daily maximum (potential) temperature in 1913 at DV was often several degrees higher as compared to the surrounding. The comparison for TX on July 10, 1913, shows even more extreme differences. On this day with the 134°F ∼ 56.7°C at DV the reduced TX at IN was only 124.6°F (51.4°C) and 125.2°F (51.8°C) at JE. Also, the reduced TX at all other stations in the area stayed much below DV.

We have a new possibility to check, if there is some inhomogeneity in the DV series between 1912 and 1913. Long term re-analyses, e. g. NOAA-CIRES-DOE Twentieth Century Reanalysis, furthermore, denoted by 20CR, are now available, which allow a further comparison [23–25]. We can look at the closest grid points surrounding DV and see, that the potential temperatures in the lower troposphere are extremely well spatially correlated in summer so that the values of the closest grid point have been used without interpolation to the exact location of DV. The reanalysis does not provide 2 m above ground maximum temperatures but values at certain pressure levels every 6 h. The 00 UTC time, which corresponds to 4 pm PST (the day before) has been selected, as it is next to the common time of the maximum temperature at DV. To avoid an influence of the difference between the real and model topography, we have taken temperature values of the reanalysis in the free troposphere at 850 hPa. The correlation coefficient between DV and 20CR is r = 0.72 for 1912 and r = 0.74 for 1913 (see Table 1). The median difference DV minus 20CR gives −1.1°F or −0.6°C for 1912 and +4.3°F or +2.4°C for 1913. Like for the comparison between DV and IN or JE also 20CR tells us that the average TX in DV in 1913 was a few degrees higher as compared to the reanalysis than in 1912. It is important to note that for the reanalysis no observed temperatures were used, so that no interdependence between observed and reanalyzed temperatures exists. On July 10, 1913, the temperature of the reanalysis reduced to DV was only 121.0°F or 49.4°C, full 13°F or 7.2°C lower than DV’s record. If we assume a super-adiabatic surface layer at DV of +5°F or 2.8°C this implies that the true maximum temperature according to the 20C reanalysis was 126°F ∼ 52.2°C rather than the documented 134°F ∼ 56,7°C. The ECMWF—ERA20C reanalysis shows similar temperatures at the nearest grid point to DV [26]. The median temperature difference DV minus ERA20C in 1912 was −4.5°F or −2.5°C and +2.4°F or +1.3°C in 1913. Again, the average maximum temperature of DV in 1913 was much higher as compared to the reanalysis than in 1912. The reduced temperature of ERA20C on July 10, 1913, was even 14.6°F or 8.1°C lower than at DV, which implies an estimated maximum temperature at DV of 124°F ∼ 51.1°C. Hence, the maximum temperature series of DV has a clear discontinuity between 1912 and 1913.

Plausibility of the 1913 heat world record in Death Valley

We have now a series of findings to scrutinize the validity of DV’s maximum temperature world record:

• The 134°F ∼ 56.7°C represents a considerable outlier (+4.0 STD) in the series of observed absolute daily temperature maxima (TXX) in the summer between 1912 and 2023

• The 134°F ∼ 56.7°C represents the only and considerable outlier (+4.2 STD) in the series of annual maximum temperatures (TXXX) at DV between 1912 and 2023

• The summer temperature series (TX) of the first few decades of the 20th century in DV show a different trend than the average North American series

• The temperature maximum of 134°F ∼ 56.7°C corresponds to a exorbitant high value of 136.7°F or 58.2°C when adjusted to the actual climate (2023)

• The 134°F ∼ 56.7°C represents the only and considerable outlier (+4.4 STD) in the series of trend adjusted absolute daily temperature maxima (TXX)

• The correlation between TX at neighbor stations and DV in the early years of the 20th century was quite a bit lower than in the 21st century. This is an indication that the quality of the early observations was lower than today.

• A comparison of the daily temperature maxima at DV with surrounding stations shows that in the summer of 2013 the maximum temperatures at DV were too high by approximately +4.5°F or 2.5°C

• A comparison of the 134°F ∼ 56.7°C temperature maximum on 10 July 1913 at DV with surrounding stations shows that the temperature at DV was too high on that day by approximately +9°F

• A comparison of the 134°F ∼ 56,7°C temperature maximum on 10 July 1913 at DV with reanalysis data shows that the temperature at DV was too high on that day by approximately +9°F if we assume a 5°F or 2.8°C super-adiabatic surface layer

• The generalized extreme value function (see Fig. 10) fits the data much better if we omit the year 1913

• The extreme value statistics (see section “The probability of a new maximum temperature record at DV”) omitting values of 1913 and adjusted to the climate of 1913 tells us that the highest possible temperature at DV at that time was 131.8°F (55.4)°C.

• The 500 hPa reanalysis field on 10 July 1913 shows an anticyclone over the US Southwest with a geopotential height of 5940 m, typical of a heat dome. The value of the 500 hPa geopotential height during pronounced heat domes of recent cases reached roughly 6000 m (e.g. on 16 August 2020 or on 7 July 2021), when 130°F ∼ 54,4°C at DV have been observed. Hence, the 1913 heat dome was not exceptional.

How can this inconsistencies be explained? Looking to the files we learn that between 1912 and 1913 there was no documented change in the location of the station or instrumentation. Also, there was no change in the land-use (irrigation). To avoid a cooling influence due to evaporation or evapotranspiration from vegetation and irrigated areas, the station at DV was always located outside the ranch above natural desert soil. The station was located East of the ranch which means that the advection from the vegetated area due to the dominant winds from South or North, mainly parallel to the valley axis, did not play an important role.

But there was a change of the observer. Oscar Denton, who was the first person to stay several summers in DV—although it was questioned, if humans could survive there during summer—obviously was aware of some former temperature readings by non-standard thermometers, e. g. attached on a wall without screening well above 130°F ∼ 54.4°C [13]. He even told visitors about temperatures “in the shade” up to 145°F ∼ 62.8°C and up to 160°F ∼ 71.1°C “in the sun” [28]. During the first intense heat wave of his time around 10 July 1913 he might have been disappointed by the readings in the meteorological screen and used readings of some wall thermometer instead? Or was it too cumbersome for him to go from the ranch to the screen a few hundred meters outside in the afternoon of an unbearably hot day, when the observation had to be carried out and he made just an estimate? We will probably never know the exact answer.

In principle the 1913 record could also have been caused by a misreading of the thermometer by Denton. 5° or 10° misreading errors are not uncommon during manual temperature observations. In July 1913, however, a sequence of extremely high temperature observations was documented, which makes this hypothesis unreliable.

A radiation error seems to be the most probable reason for the too high summer maxima and especially the world record during Denton’s time. An unscreened thermometer is still attached to a pile close to the entrance gate of the Furnace Creek Ranch, see Fig. 10. Some authors argue that the extreme could have been caused by hot blowing sand [5]. But during strong winds the surface temperature is not too far from the air temperature in the screen and why should this effect only have happened once in more than 100 years? Other hypotheses point toward local effects like a down-rush or a downslope windstorm [3]. This can be physically excluded, because a down-rush in a very deep mixed atmosphere, caused by evaporating precipitation can only lead to a cooling at the surface and forced downslope windstorms (foehn winds) can only develop in a stably stratified atmosphere. It is interesting to note, that super-adiabatic lapse rates between DV and Independence only occur in the first 12 years of the whole time series, mostly during the years of the observer Oscar Denton. In recent years the mean difference in summer between the reduced daily TX at DV minus IN/OC is virtually always negative (between 4 and −8°F or between −2.2 and −4.4°C). The suspicion that the TX in summer at DV in the beginning of the series are too high is also supported by Fig. 4. In the first decades of the 20th century the mean summer TX is higher than in the middle of the century, although the mean TX for the whole year does not show this feature. Therefore, the summer maximum in the first decades of the series (above 116°F ∼ 46.7°C, Fig. 7) could be seen as Oscar Denton’s legacy. It is not our intention to blame the observer for incorrect data, because without volunteers like Oscar Denton, who was obviously not a well-trained meteorological professional but who was willing to withstand periods of unbearable heat—he expressed it as “sweating my life away”—in DV without air condition, we would not have any data from these early times.

Concluding, the record of 10 July 1913 is more than questionable—in perfect agreement to the findings in [8, 11, 13], who have basically investigated the internal data consistency and homogeneity with neighbor stations. Taking all findings on that day, the correct maximum was probably around 125°F ∼ 51.7°C, still a very high value for the first half of the twentieth century. This implies, that the true temperature world record dates to August 16, 2020, with its 130°F ∼ 54.4°C in DV or to 9 July 2021, with the same value, see [29–31]. It is not the author’s duty to decide on the validity of the record, this must be made by responsible persons at the US National Weather and Climate Service and WMO. Moreover, it might be more interesting to look into the future, instead of discussing the past with the many uncertainties and irregularities.

The probability of a new maximum temperature record at DV

Due to the significant warming during the last decades, we should not take the statistics of the original observations in the whole period 1911 to 2020 to derive probabilities of future maximum temperatures. With the aid of Fig. 4 we can deduce the TXXX data de-trended to the “thermal climate of 2023”, which are shown in Fig. 11. The dashed curve represents the corresponding Generalized Extreme Value distribution (GEV), with µ = 126.99°F = 52.77°C, σ = 1.901°F = 1.056°C and ξ = −0.1247 [32], determined by a maximum likelihood method. The parameterµ is the location parameter, σ the scale parameter and ξ the shape parameter. If we take the 136.4°F (58.0°C), which corresponds to the trend adjusted 134°F ∼ 56.7°C world record, we compute a probability of approximately once in 2000 years. Court [5], concluded, that it was a once in 650 years event, based on the first 37 annual maxima. In Fig. 11 also a continuous GEV curve is plotted, based on all TXXX data except the 1913 world record. This curve fits the data in the upper range much better than the dashed curve and is given by µ = 127.08°F = 52.82°C, σ = 1.869°F = 1.038°C and ξ = −0.2505. Taking these values, the adjusted world record lies above the theoretical maximum value of the GEV-function, a further strong argument against the validity of the 1913 record.

The probability that in the years 2024–2033 (with the climate of 2023) 130°F ∼ 54.4 or more is observed at least once is p (T ≥ 130°F ∼ 54.4°C) = 82.0% for the dashed curve and 74.3% for the continuous curve.

Where µ_a is the climate change adjusted location parameter. If we do this exercise and assume an increase of µ_a by 0.109°F = 0.061°C per year we receive a very high probability that 130°F ∼ 54.4°C or more is observed at least once in the coming 10 years p (T ≥ 130°F ∼ 54.4°C) = 92.3% for the dashed curve and 90.0% for the continuous curve of Fig. 10. There is also a quite high probability to reach or exceed the 131°F ∼ 55.0°C level at DV in the next 10 years. For all thresholds in °F and °C the probabilities with and without taking the value of 1913 into account can be found in Tables 2 and 3. There is still a reasonable probability even for the 132°F ∼ 55.6°C threshold, but to reach or exceed the 135°F ∼ 57.2°C, which would make the questionable 134°F ∼ 56.7°C record obsolete, a rather low probability of 4.9% results for the dashed curve and less than 0.1% for the continuous curve. Hence there seems to be enough time for the responsible bodies at the National Weather Service or WMO to have a deeper look into the plausibility of the still valid world record, before a new record will be set.

There is a physical reason why record maximum temperatures in desert regions are increasing with a slower pace than in other climate regions. The radiative forcing of the increased atmospheric greenhouse gas concentration is in the order of a few W m⁻² [33]. A change of the surface temperature of e. g. 200°F ∼ 93.3°C, which has been observed 2001 in DV [34], by just one °F = 0.6°C, however, increases the blackbody emittance by a little more than 6 W m⁻². The minimum temperature is generally seen as more sensitive on the increased greenhouse gas concentrations, hence an investigation of the minimum temperature at DV, which might have set a world record value recently seems to be promising [35]. Recently a paper has been published [36] where the maximum surface temperature during mid latitudes heat waves is estimated by moist static energy considerations. In agreement to our results, they find values a little higher than 55°C (131°F) possible.

In a recent publication it was pointed out that in a changing climate not only the temperature change but also the air humidity change should be considered [37]. The relevant quantity to consider both temperature and humidity is the wet bulb temperature. It was found that the highest observed wet bulb temperatures in some regions of the world (e.g. along the Persian Gulf) increasingly exceed the 90°F ∼ 32.2°C threshold, which is close to the limit of human tolerance. Considering the low humidity in Death Valley (dew point temperature during heat waves is typically in the 30s °F or close to 0°C) the wet bulb temperature rarely exceeds 70°F ∼ 21.1°C. Hence, Death Valley is far from having the “deadliest” climate on earth. Drinking lots of water, tranquility in the shade, sweating and possibly some ventilation allows us to survive heat waves there—albeit miserably—without air conditioning.

Author contributions

Reinhold Steinacker (Conceptualization [lead], Data Curation [lead], Formal Analysis [lead], Funding Acquisition [lead], Investigation [lead], Methodology [lead], Project Administration [lead], Resources [lead], Software [lead], Visualization [lead], Writing—Original Draft [lead], Writing—Review & Editing [lead])

Conflict of interest: None declared.

Funding

None declared.

Conflict of Interest

There are no conflicts of interest.

Data availability

The data is available on request from the author.

References