In situ surface pressures measured at 2 s intervals during the 150 sol Phoenix mission are presented and seasonal variations discussed. The lightweight Barocap®/Thermocap® pressure sensor system performed moderately well. However, the original data processing routine had problems because the thermal environment of the sensor was subject to more rapid variations than had been expected. Hence, the data processing routine was updated after Phoenix landed. Further evaluation and the development of a correction are needed since the temperature dependences of the Barocap sensor heads have drifted after the calibration of the sensor. The inaccuracy caused by this appears when the temperature of the unit rises above 0°C. This frequently affects data in the afternoons and precludes a full study of diurnal pressure variations at this time. Short-term fluctuations, on time scales of order 20 s are unaffected and are reported in a separate paper in this issue. Seasonal variations are not significantly affected by this problem and show general agreement with previous measurements from Mars. During the 151 sol mission the surface pressure dropped from around 860 Pa to a minimum (daily average) of 724 Pa on sol 140 (Ls 143). This local minimum occurred several sols earlier than expected based on GCM studies and Viking data. Since battery power was lost on sol 151 we are not sure if the timing of the minimum that we saw could have been advanced by a low-pressure meteorological event. On sol 95 (Ls 122), we also saw a relatively low-pressure feature. This was accompanied by a large number of vertical vortex events, characterized by short, localized (in time), low-pressure perturbations.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 During the Phoenix mission air temperatures were measured using thermocouples situated at three levels on a 1 m deck-mounted mast. Pressure was measured with a system based on the Vaisala Barocap®/Thermocap® technology housed on the deck of the lander, about 1 m above ground level [Taylor et al., 2008]. Analyses of the temperature data are reported by Davy et al. . These measurements were made at 0.5 Hz and ran almost continuously through the landed mission apart from short daily breaks for data transfers. Winds were determined at the top of the mast, roughly 2 m above ground level, with a telltale indicator and SSI camera system [Gunnlaugsson et al., 2008; Holstein-Rathlou et al., 2009]. The telltale was imaged more frequently than anything else on the lander but wind data are generally limited to a few periods on each sol or Martian day and there are limited nighttime observations. Temperature and pressure data throughout the mission are shown in Figure 1. The diurnal temperature data were generally very similar from one day to the next throughout the first 60 sols of the mission with daily maxima of around −30°C and minima of −80°C. Winds also showed a fairly regular diurnal pattern [Gunnlaugsson et al., 2008].
 Surface pressure measurements have been made before on Mars, by both Viking Landers [Hess et al., 1980] and by the Pathfinder Lander [Schofield et al., 1997]. Pressures on those missions were measured by Tavis magnetic reluctance diaphragm sensors. Barocaps® were used on the Phoenix Lander, at least in part because of mass considerations. Harri et al.  discuss the performance of earlier Barocaps® in comparison with the Tavis sensors. Meteorologists on Earth have long used surface pressure as a forecasting tool, with falling pressures being a sign of the potential approach of inclement weather. Networks of surface pressure measurements, adjusted to a common geopotential level, provide the basic surface pressure maps with isobars, highs and lows and fronts that are a basic meteorological tool. While “weather” on Mars is somewhat different from Earth and has much lower pressures, a basic pressure measurement remains an essential component of any meteorological measurement station. In addition to pressure variability on the “synoptic” time scale of a few days, the Mars pressure data have significant seasonal variations due to part of the atmosphere freezing out onto the polar surfaces. There is also a much stronger (relative) diurnal signal compared to Earth due to thermal tides. Short-term (of order 20 s) fluctuations in pressure can also be used to detect the passage of vortices and dust devils.
 For the pressure measurements on Phoenix it was found that the original procedure used to calculate calibrated readings from the flight data did not work optimally because of the thermal environment of the measurement system and an adjustment (PCOR1) has been applied. There were also uncertainties illustrated by differences between pressures calculated from the primary and secondary Barocaps® (PCOR2). The present paper will focus on these data processing procedures and on the seasonal variations in pressure. There is a fairly regular diurnal, tidal pattern superimposed on a seasonal variation and also less regular semidiurnal variations and variations associated with weather systems. We presently have concerns about the accuracy of our pressure data on diurnal and semidiurnal time scales and can only present some preliminary results. There are also short time scale (∼20 s) pressure drops associated with the passage of dust devils or other vortices. These are discussed by Ellehøj et al. . They are on a short (20 s) time scale that is not affected by the longer time scale pressure adjustments that we are concerned with here. The uncertainties mentioned above have also no effect on seasonal time scales as any diurnal inaccuracies will be small and average out.
 In the text below times will be given primarily as fractions of sols expressed in Mars Local Mean Solar Time at the landing site (68.21°N, 234.24°E) with sol 0 as midnight on the date of the Phoenix landing (25 May 2008). The corresponding solar longitude, Ls = 76.2. In terms of sols measured from the beginning of the Martian year at Spring Equinox, Ls = 0, add 164 to the Phoenix sol.
2. Pressure Data Processing
 A raw pressure reading was calculated every 2 s from the outputs of the Barocap® and Thermocap® sensor heads by the MET data acquisition system onboard the spacecraft. There were three Barocaps® (B1, B2, and B3) and two Thermocaps® (TC1 and TC2) flown on the lander. The primary pressure sensor head (B1) is an LL (not an acronym, just a designation) type Barocap® while secondary Radiosonde Pressure Sensor Number 1 (RSP1) Barocaps® (B2 and B3) were used for housekeeping checks at 512 s intervals. All Barocaps and Thermocaps were mounted on the same printed circuit board (PCB) inside the pressure sensor within a 2 cm × 3.5 cm area. This was housed within the closed box formed by the PCB and the upper Faraday shield and connected to the atmosphere with a narrow tube (Figure 2). The tube comes out of the Upper Payload Electronics Box (PEB) and is orientated pointing to the East. While an ideal arrangement would have been connection to the atmosphere via a small orifice in a horizontal surface, the winds from the East were generally of order 4 ms−1 [Holstein-Rathlou et al., 2009] and dynamic pressure errors are estimated to be <0.2 Pa. The two Thermocaps® consistently reported essentially the same temperatures of the printed circuit board.
 Vaisala Barocaps® are capacitive, absolute pressure sensors manufactured by silicon micromachining [Harri et al., 1998]. When the pressure changes the silicon diaphragm bends and changes the vacuum gap leading to a capacitance change. Pressure computations are dependent on the temperature of the Barocap®. This is monitored by adjacent Thermocap® sensors. These sensors consists of platinum wires separated by a glass–ceramic dielectric. The sensor's capacitance is a strong function of temperature. The formulae used to compute temperature and pressure from the capacitances are commercially sensitive and access to them is limited. The sensor was calibrated by the Finnish Meteorological Institute (FMI) and calibration checks were made by MacDonald, Dettwiler and Associates Ltd. (MDA).
 Calibrated pressure readings are calculated after the data are transmitted to Earth. The procedure used for this (PCOR1) is presented below. The reason why the raw pressure readings calculated onboard differ from the actual, calibrated readings is associated with a lag of the actual Barocap® temperature relative to that reported by the Thermocap® because of differences in thermal contact.
 After Phoenix landed it appeared that also a second adjustment (PCOR2) may be needed because there appear to have been slight changes in the temperature dependence of the Barocap® outputs between calibration on Earth and operation on Mars. The LL (B1) and RSP1 (B2, B3) Barocaps® respond differently, and it was these differences that made us aware of the problem. The changes in the temperature dependence of the Barocaps® had apparently occurred before the last spacecraft test in Martian conditions, the Assembly, Test, and Launch Operations (ATLO) landed configuration test. However, the data of this test had not been provided to FMI due to International Traffic in Arms Regulations (ITAR) restrictions. Hence, the calibration coefficients used in the calculation of raw pressure onboard the spacecraft had not been updated. The ATLO data have recently become available and further analyses are being undertaken using it. We estimate that a reliable adjustment procedure will be available in early 2010.
2.1. Thermal Lag Adjustment (PCOR1)
 Calibrated pressure readings depend on the temperature of the B1 pressure sensor head, Tb. The raw pressure readings are however calculated using the temperature measured by the adjacent TC1 temperature sensor head, at temperature TT. The pressure adjustment PCOR1 is needed because the B1 sensor head (see Figure 2) has a weaker thermal contact to the Printed Circuit Board (PCB) than TC1. Because of this a temperature gradient is formed between these components if the PCB temperature rises or falls. During rising temperature the Barocap® stays colder than the Thermocap® and during falling temperature the Barocap® stays warmer. Hence the raw pressure readings calculated onboard differ from the actual, calibrated pressure in rapidly changing temperature. Raw pressure values of the Barocap® B1 are high in rising temperatures and low in falling temperatures. The difference between the calibrated and raw pressure vales Δp has been shown experimentally to be
where ∂p/∂T is the partial derivative of the calibration equation. The value of ∂p/∂T is 5.34 ± 0.3 Pa K−1 over the temperature and pressure ranges experienced.
 This effect was known already at the design phase of the pressure sensor. However, the MDA and FMI teams had only limited information about the thermal environment of the sensor at that time. It was decided to not include the correction in the onboard software but to adjust the data afterward if needed. An original PCOR1 procedure was developed as a precaution for the worst case: rapidly changing temperature. After Phoenix landed it appeared that the actual thermal environment was worse than the expected worst case. The temperature was not only changing rapidly but there were also fast changes in the temperature gradient due to a nearby heat source. Information on a relocation of the heat source had not been provided initially due to ITAR restrictions.
 FMI had performed a “temperature gradient response test” during the pressure sensor assembly level test campaign. The PCOR1 procedure is based in part on the results of that test. In the test it was found out that the Thermocap® (TC1) temperature sensor head has a strong thermal coupling to the PCB. Hence, the temperature measured by TC1 is practically the same as the temperature of the PCB. Further, it was found out that the Barocap temperature Tb can be calculated from the Thermocap temperature TT using a lumped system model analogous to Newton's law of cooling so that
Here λ is the time constant of the temperature changes. The value was determined to be 78.7s.
 The adjustment is obtained by first solving the thermal conductivity equation (2) for Tb and then calculating the difference between the calibrated and raw pressure vales using equation (1) above. Thermocap temperature readings TT are needed to solve the thermal conductivity equation. Ideally we would have TT data at the same 2 s rate as other quantities but unfortunately the usage of data transfer capacity had already been frozen when it was appreciated that these would be needed and they were not transmitted to Earth. However, TC1 data were included in the 512 s “housekeeping” information. Thus, in general, we only have the TT data sampled once every 512 s that were transmitted to Earth. In the original PCOR1 procedure it was assumed that dTb/dt would be nearly constant during the 512 s measurements intervals. After Phoenix landed it appeared that this assumption was not always valid due to the nearby heat source. Therefore the procedure had to be updated. There are two versions of the updated procedure: one developed by the York University group and another developed by FMI. In these updated procedures the 512 s data is interpolated, using a natural cubic spline, to obtain simulated 2 s TT data. Then the thermal conductivity equation above is solved using the interpolated TT data. The updated procedures have been verified using the original calibration data.
 Interpolating the temperature values is the most problematic part of the whole procedure and is a source of errors. When the nearby heat source is turned on or off the temperature change rate changes faster than can be properly represented with the 512 s measurement interval. Hence interpolating temperature readings during the intervals when the heat source is turned on or off with the required accuracy is challenging. Various schemes were tried for interpolations but the natural cubic spline appears to perform well and is used here. The York University group use a simple cubic spline application while the FMI/MDA adjustment scheme, used to produce the calibrated pressure data in the PDS archive, uses a slightly modified spline interpolation method which includes use of estimated TT values added at local extrema.
 Solving (2) above also requires initial conditions for Tb at the first time point after power-up t0. It was found that a satisfactory estimate could be obtained using the time constant for the process with
An estimate for the temperature change rate dTT/dt is calculated by differentiating the spline function used to interpolate TT. Even if this estimate is not always accurate the impact on the data is small as the effects of initial conditions are transient and decay with time constant λ ≈ 80 s. Practically only the first 512 s interval after power-up is affected by the inaccuracy of the initial condition.
 Although the MET data acquisition had been set up in a way that data transfer capacity prevented us from obtaining the TT data continuously we were able to obtain them on three occasions (on sol 74–75; sol 81–82 and sol 88). We have used these to test the efficacy of our interpolation scheme and its impact on the calibrated pressures. Part of the 2 s TT data obtained from (Phoenix) sol 81/82 are shown in Figure 3a, together with the curve interpolated from data sampled only once every 512 s. Differences between interpolated estimates and 2 s TT temperatures for the three periods when the 2 s data are available are generally less than 0.04 K which will correspond to a pressure error of order 0.2 Pa, although in extreme cases errors of 0.1 K can occur. The root mean square error between interpolated estimates and 2 s TT temperatures, σT, and calibrated pressure based on these two sets of temperature, σp, are 0.0414 K and 0.1117 Pa for sol 74–75; 0.0182 K and 0.0466 Pa for sol 81–82 and 0.0164 K and 0.0394 Pa for sol 88.
 As noted above, heating and cooling phases of the PCB lead to lags in the temperature of the B1 relative to the TC1 and cause errors in the raw pressures. The “saw tooth” pattern in the raw pressure data in Figure 3b is a characteristic indicator of this. PCB heating and cooling rates are of order ± 0.003 Ks−1. Assuming, as in equation (3) that the Barocap® and Thermocap® temperatures could change at approximately the same rate during steady heating or cooling this yields a temperature difference of order 0.3 K. The corresponding difference between raw and calibrated pressure readings is of order 1.5 Pa. At sol 81.93 (the time of the maximum cooling rate) the adjustment is indeed approximately +1.5 Pa. The differences between pressures calculated with the measured 2 s TT data and with the interpolated TT values are generally less than 0.2 Pa and appear to be highest during periods of change from heating to cooling or vice versa. We can use these results based on our limited 2 s TT data to estimate the errors associated with interpolation.
 Errors associated with the use of equation (2) are harder to assess but we can use the data of the temperature gradient response test to indicate the uncertainty in the value used for λ. Our best estimates of these errors is 0.2 Pa and adding this to the temperature interpolation error allows us to estimate a possible error for each data point. FMI and MDA took a more conservative approach in the production of the “official” calibrated pressure data posted to the Planetary Data System (available at http://pds-atmospheres.nmsu.edu/) and some data were omitted. For example the first 512 s intervals after power-up were left out because of the effect of the initial condition. In our study of the limited 2 s TT data the magnitude of the errors in the calibrated data caused by Barocap®/Thermocap® temperature difference is a maximum of 0.5 Pa. In other parts of the data pressure “peaks” with magnitude up to 1 Pa, apparently caused by the imperfect TT interpolation, are detected close to local TT extrema. The data sequences where this kind of “error peak” occurs are omitted in the data calculated with the FMI/MDA adjustment program. Our current view is however that all data can be used for analyses of diurnal and seasonal variations with or without the omitted or “banned” data because the length of the error peaks is short (<512 s).
 The raw pressure data can be used to identify the passage of dust devils and dustless vortex features (usually several per sol, near midday and early afternoon) since the time scale of errors in the raw data is longer than the time constant of Barocap temperature changes λ (circa 80 s). The vortex passages have a shorter time scale, generally of order 20 s and are reported in detail by Ellehøj et al. . The temperature induced relative pressure errors in the raw data on time scales of order 20 s are small, <0.1 Pa. On the other hand error peaks in the calibrated data could lead to misinterpretations. Because of this the raw data should be used in vortex identification. Tests of the vortex identification scheme using raw and calibrated data indicate no significant differences in the detection of the short time scale negative pressure perturbations used to identify vortices [Ellehøj et al., 2009].
2.2. Temperature Dependence Correction (PCOR2)
 The Barocap® sensor heads are sensitive to temperature. As explained in section 2.1 this temperature dependence is compensated for using the output of a Thermocap® temperature sensor. The calibration constants used in this compensation were calculated using the data of the assembly level calibration tests, performed over 2 years before the start of the mission. If the temperature dependences of the Barocap® sensors have changed after the calibration tests then the readings of the sensor heads might be temperature-dependent despite the compensation. These possible temperature dependences are studied by comparing the data of the three Barocap® sensors to each other. There are small (<3%) differences in the overall pressure levels detected by the different Barocap®s. When studying the temperature dependence on a diurnal time scale these differences in pressure levels are compensated for by normalizing the pressure curves before comparing them to each others. This normalization is done by substituting the first reading of a data set (for example the data of one sol) from all measurement points in that data set.
 For most of the time these normalized pressure variations detected by all Barocaps® are the same with a 2 Pa precision. However, during time intervals when the Thermocap® temperature exceeds 0°C the normalized pressure curve of Barocap® 1 differs noticeably from the normalized pressure curves of the other two Barocaps®. At time points when the Thermocap® temperature exceeds 0°C the reading of Barocap® 1 usually starts to rise and the readings of Barocaps® 2 and 3 start to fall. As a result the pressure difference between B1 and B2 increases by about 5 Pa relative to the value at 0°C. Our conjecture is that the temperature dependences of all Barocaps® have changed and because of this the reading of Barocap® 1 may be too high and the readings of Barocaps® 2 and 3 may be too low in temperatures >0°C. This is shown in Figure 4.
 In the data shown in Figure 5 there seems to be two pressure maxima per sol in the B1 data, occurring circa 0830 and 1530 local Mars time. However, the later maximum occurs during the time when Thermocap® temperature is >0°C. Hence it might be that there is actually only one maximum per sol or the later maximum may be lower than it seems to be in the 2 s data measured with B1. From Figure 5 we see also that the noise of the data of B2 and B3 is much higher than the noise of the Barocap® 1 data. The reason for this is that B2 and B3 are of a different type than B1. This supports the use of data from B1 for vortex identification.
 Detailed studies on the temperature dependence phenomenon are going on. The data of the ATLO tests are being used in this study. It has been found that it is possible to develop a procedure for correcting the errors caused by the temperature dependence. Preliminary results show that that the difference between the reading of B1 compared to B2 and B3 is caused for the most part by changes in B2 and B3. It is possible that B1 has not changed at all. Results of tests performed with the pressure sensor Flight Spare Model support this. A preliminary estimate is that less than 1 Pa of the 5 Pa increase in the difference between B1 and B2 when temperature rises from 0°C to +25°C is caused by B1. A physical explanation for the changes in that the RSP1 Barocaps® (B2 and B3) has also been found. When a RSP1 Barocap® is heated to sterilization temperature the temperature dependence of the sensor changes. The temperature dependence also continues to slowly change after the heating. Component-level Barocap® tests have revealed that this behavior is typical to RSP1 Barocaps® but not to LL Barocaps® and is associated to the materials used in these components. A heating test was performed on the pressure sensor before calibration and this seem to have triggered a slow calibration change in the RSP1 Barocaps. The RSP1 Barocaps® had been formally qualified for the Phoenix mission but because of schedule restrictions long-term stability had not been studied. FMI has decided not to use RSP1 Barocaps® in future missions.
 No temperature dependence correction (PCOR2) has been used in this study as the correction procedure is not yet available. For the study of the seasonal pressure variation the effect is small, and as noted above the time scale is too long to affect the vortex and dust devil investigation. The temperature dependence related uncertainties are however a serious limitation as far as studies of diurnal and semidiurnal pressure perturbations are concerned and we will only present limited results from sols on which TC1 remained <0°C throughout the sol. These occasions were rare as TC1 rose above 0°C for several hours on most afternoons, as in Figure 5a.
 We anticipate that the magnitudes of the measured diurnal variations are approximately correct since the magnitudes of the pressure variations detected by all Barocaps® are approximately the same (Figure 5b).
2.3. Accuracy, Time Response, Resolution, and Stability of the Pressure Sensor
 The initial (SOL0) accuracy of the pressure sensor is defined as the maximum difference between the reading of the sensor (B1) and true pressure during the first sols on Mars. We estimate that the SOL0 accuracy is <6 Pa at times when the Thermocap® temperature is <0°C and <11 Pa at times when the Thermocap® temperature is >0°C. These estimates are based on the data measured during the interplanetary cruise and comparison of the readings of the different Barocaps® during the first sols on Mars. The Atmospheric Science Theme Group (ASTG) requirement for sol 0 accuracy was ±10 Pa.
 In the time scale from seconds to minutes the most important characteristics of the sensor are time response and resolution. Although the sensor itself has rapid response (<1 ms) there are delays associated with the pressure tube and the dust filter (Figure 2). The test data shows that the response time of the sensor is approximately 3 s. The effect of this response time is that the pressure drops associated with vortex passages are likely to be underestimated. This is discussed in more detail by Ellehøj et al. . The resolution of the pressure sensor is limited by the noise in the 2 s pressure data. The peak to peak value of this noise is about 0.1 Pa. The ASTG requirement for resolution was 0.5 Pa.
 On diurnal and semidiurnal time scales the most important characteristic of the sensor is the temperature dependence discussed in section 2.2. On seasonal and synoptic time scales the most important characteristic is stability. Here we define stability as the difference between the accuracy at the beginning and the end of the mission. In other words stability is the maximum drift of the calibration of B1 during the mission. We estimate that the stability of the pressure sensor is <8 Pa. This estimate is based on the observation that the difference between the readings of B1 and B2 changes by about 8 Pa during the mission. Comparison of the data of B1 to B3 gives similar results. The magnitude of this change is the same at all temperatures which means that the levels of the readings of all sensor heads may have changed but the temperature dependences have not. As described above the RSP1 Barocaps® (B2 and B3) are more unstable than the LL Barocap® B1. Hence we can safely assume that the change in the reading of B1 compared to B2 and B3 is mostly caused by B2 and B3. From this we can estimate that the calibration of B1 has changed less than 8 Pa during the mission.
3. Long-Term Pressure Variations
 The Viking Landers provided over three years of in situ pressure data which have been carefully analyzed by Tillman  and Tillman et al. . Viking observed pressures during Year 1 which were strongly influenced by major dust storms but Years 2 and 3 appear characteristic of “Years without Great Dust Storms.” Annual cycles show two local maxima, and two minima. The global minimum occurs near Ls 148 while there is a preceding local maximum at about Ls 45. The pattern corresponds to deposition and sublimation from the two polar caps with the global minimum occurring at the time when net deposition to the South polar cap has ceased. Tillman et al.  model the annual cycle with a five component spectral model which can be expressed in the form of a mean, an annual component, and four harmonics:
Here pressures are in hPa and the phase S is in sols (either Phoenix sols or sols starting at Ls 0). The phase shift Si is converted to sols from the Ls values given by Tillman whose phase shifts ϕ0i relate to the time when the spectral component is zero and increasing. The Mars year is taken as 668.6 sols. The Viking landing sites (22.27°N, 312.05°E and 47.67°N, 134.28°E for Landers 1 and 2, respectively, at elevations −3637 m and −4495 m relative to the Mars MOLA 2000 aeroid) were closer to the Equator than Phoenix. The Phoenix landing site is at 68.22°N, 234.25°E at an elevation −4126 m and in Table 1 and Figure 6 we have adjusted the Viking amplitudes to the Phoenix elevation before averaging.
Table 1. Five Component Spectral Model Components From Tillman et al.  for Two Viking 1 Years and One Viking 2 Year Without Global Dust Stormsa
Viking 1 Year A
Viking 1 Year B
Viking 2 Year A
Phoenix at −4126 m
To adjust to the Phoenix MOLA elevation (−4126 m) Viking 1 amplitudes are increased by a factor 1.0455 and Viking 2 values reduced by a factor of 0.967 assuming a scale height of 11 km. Values used for Phoenix are averages of the adjusted amplitudes (hPa) and averaged phases. Phases (Si) in parentheses in the final column are the equivalent phase values in Phoenix sols.
Mean at site
Fundamental Phase (Ls)
0.717; 222.8 (279)
1st harmonic Phase (Ls)
0.592; 32.4 (−97)
2nd harmonic Phase (Ls)
0.110; 354.1 (494)
3rd harmonic Phase (Ls)
0.063; 330.2 (450)
4th harmonic Phase (Ls)
0.016; 320.3 (432)
 In general, Figure 6 shows fair agreement between the Phoenix measurements and Tillman's model, and thus with Viking data. There is an offset of order 10 Pa and the mean slopes dp/dt are slightly different but year to year variations are to be expected. The York Mars GCM [see Moudden and McConnell, 2005] pressures are also in fair agreement in the earlier part of the mission but diverge after sol 100 (Ls 123).
 The time scale in Figure 6 is extended beyond the Phoenix data to show the annual minimum from Tillman's model. Annual minimum pressures reported by Viking 1 were around 683 Pa and Viking 2 were near 740 Pa. Adjusted to the Phoenix elevation these are 714 Pa and 716 Pa, respectively. Both occurred near Ls148. The Pathfinder pressure minimum, at 3682 m MOLA elevation was approximately 670 Pa (available at http://mars.jpl.nasa.gov/MPF/science/atmospheric.html) at Ls 148 (Pathfinder sol 20). The corresponding minimum pressure at the Phoenix elevation is extrapolated as 698 Pa. The Phoenix pressures show a local minimum (daily averaged) pressure of 724 Pa which occurred near Ls 143 (sol 140), rather earlier than Viking and Pathfinder values.
 Phoenix lost power on sol 151 and it is not clear whether the minimum that we saw is due to a strong local or regional weather feature or the annual cycle. It is possible that the pressure started to drop again after sol 150/151 but, sadly, we were unable to observe that. Phoenix never (so far) recovered after the power loss event on sol 151. The observed pressure minimum was a much sharper feature than those from Viking 1 or Pathfinder but increased irregularity in the pressure data were also present in Viking Lander 2 data [see, e.g., Tillman, 1988, Figure 1]. While loss of data beyond sol 151 limits our interpretations we can see from the σp curve in Figure 6 that we appear to have seen a “Tillman Transient,” of type T1 in Tillman's notation, in the data. Similar features were identified in the Viking Lander data [Tillman, 1988] and attributed to an interference between the thermal tides and propagating Kelvin waves which appears to occur repeatedly at this time of the Martian year [see Wilson and Richardson, 2000; Wilson and Hamilton, 1996]. The magnitudes of the Phoenix σp values both before and within the Tillman Transient periods are considerably lower than those recorded by the Viking Landers where background values were approximately 10 Pa for VL1 and 5 Pa for VL2 [Tillman, 1988]. This is consistent with the expected decrease in the amplitude of the thermal tides with latitude.
 Diurnal cycles in pressure will be discussed briefly below but a more extended evaluation must await the resolution of the thermal effects on calibration coefficients (PCOR2). The pressure data also reveal occasional “synoptic” features associated with meteorological events but, as noted by Tillman et al.  the significant “weather” only starts at the end of the summer. There is a “high”-pressure departure from the steady decline around Phoenix sol 25 and a low-pressure system at around sol 95. Ellehøj et al.  highlight the much increased dust devil and vortex activity at that time.
4. An Example of Diurnal Pressure and Temperature Variations
Figure 7 shows PCOR1 corrected pressure data for the period from sols 142–145. As noted above the comparison of data of the different Barocap®s confirms that the response of the sensor is practically temperature-independent as long as the temperature inside the sensor is less than 0°C, and this was satisfied during this period. It is also at the time when σp has increased – the Tillman Transient effect.
 Air temperature data (from the upper thermocouple, T1) show a regular diurnal cycle from a maximum of about −40°C to a nighttime minimum of near −95°C. Midsol conditions are characterized by large turbulent fluctuations while evenings are calm. It is not entirely clear what causes the nocturnal fluctuations in temperature but winds are then often from an Easterly direction and pass over the lander before reaching the mast. Heimdal crater is also to the East, about 20 km away and could generate wave disturbances. Spectral analysis of the pressure records is planned after the method for correcting the temperature dependence error has been developed. Note the cyclic heating pattern in the nighttime TC1 data which was one of the factors necessitating the implementation of PCOR1.
 The diurnal and semidiurnal pressure variations are assumed to be caused by a mix of solar tides and Kelvin waves. According to the AMES [Wilson and Hamilton, 1996] and York University [Moudden and McConnell, 2005] Mars GCMs the pressure maximum should occur near 1300–1400 local Mars time at the Phoenix site at this time of year. This matches with the Phoenix measurements. Normalized pressure ranges are slightly less than 1% (amplitude <0.5%) which is consistent with the GCM model output although absolute values differ slightly. On these sols there is reasonable agreement of diurnal amplitude and phase between the GCM simulations of pressure and the measurements. Agreement was less satisfactory during sols 136–139 and but it does appear that a significant local meteorological event may have occurred at around that time as suggested by the pressure data in Figure 1. Further investigation and more comparisons with GCM output are planned once PCOR2 is finalized.
 Meteorological measurements that would be relatively simple on Earth proved a little more challenging on Mars but we were successful in acquiring 150 sols of 2 s pressure data. There are short breaks on most days for data transmission and two longer interruptions (approximately 1 sol) when data were not retrieved. The procedure used in the calculation of calibrated readings was updated so that it worked in the actual thermal environment of the sensor. Further evaluation of effects when the pressure sensor warms above 0°C is needed and a correction will be developed in order to study diurnal time scale variations of pressure associated with thermal tides and weather systems. The seasonal variation of diurnally averaged pressures shows a general reduction throughout the mission due to the freezing out of atmospheric CO2 on the southern pole but loss of power at the end of the mission prevented us from acquiring sufficient data to properly determine the timing and depth of the annual pressure minimum. Short time scale signatures of small-scale vortices of the type associated with dust devils were clearly present in the pressure data and are reported by Ellehøj et al.  in the current issue.
 The Phoenix mission as a whole is led by Peter Smith of University of Arizona. We are grateful to him and other members of the Phoenix team for the opportunity to participate. Funding for Canadian participation in Phoenix has been provided by the Canadian Space Agency. We are particularly grateful to Jim Tillman for his incisive comments on our work and especially for bringing to our attention the “Tillman Transient” behavior as a possible explanation of the pressure variations at the end of the mission.