Air pressure at 47 m

Half-hourly atmospheric pressure at 47 m on the CH-LAE tower, 2005-2025, exported as measured. The file is 05_METEO_PA_2005-2025 (parquet and CSV): 356,448 records on a continuous 30-minute middle-timestamp index (named TIMESTAMP_MIDDLE in the file) in local time (UTC+1, no daylight saving), from 2005-09-02 00:15 to 2025-12-31 23:45.

The interactive dashboard summarises the product on one page: coverage and provenance, seasonality, distributions, extremes and trends, with a table view behind every chart. It is standalone and works offline.

Method, evidence and checks:

05_METEO_PA — builds the product: the unit harmonisation of the two screenings, the 2012 corrections, the identification of the barometer from the logger programs, the check of the absolute level against the barometric formula, and the two hardware boundaries.
Meteo_Product_Chain — where this product sits in the chain. No other meteo product is computed from it.

One column, no flag, and gaps left in

Unlike SW_IN, TA and PPFD_IN, this product is not gap-filled. There is a single value column, no flag column, and about 1 % of the record is missing: a non-null value is a measurement and a NaN is a real gap. Code that assumes a complete series has to handle the gaps explicitly.

There is also no _HOMOGENIZED column, although the record does change level at the January 2016 acquisition change. The step is small and is described under Known limitations, together with what to do about it.

Columns

Table 1: Columns of 05_METEO_PA_2005-2025.

column	unit	description
`PA_T1_47_1`	kPa	Atmospheric pressure at 47 m, measured and corrected. Gaps are `NaN`.

The unit is kPa, the FLUXNET convention. The database stores the two screenings of this measurement in different units, kPa before 2022 and Pa after, because the two loggers wrote different units. The notebook harmonises them from the magnitude of the data and refuses to proceed if that verdict disagrees with the unit tag the database carries.

Measured values range from 89.45 to 95.54 kPa, with a median of 93.31 kPa.

Coverage

352,438 of the 356,448 records carry a value, 98.9 %. The record begins on 2005-09-02, when the archive of this logger begins; the barometer itself was already in service, but no earlier data are held. There is no PA for 2004 and for the first eight months of 2005, and none is reconstructed, so the index of this product starts later than that of SW_IN, TA and PPFD_IN.

Gaps are short except where a logger was out of service. Of 415 separate gaps, 399 are one hour or shorter; the longest is the 20.6-day outage of January 2016 during which the tower logger was replaced.

Table 2: The years in which measured coverage is not near-complete. Every other full year of 2006-2025 is at least 99.3 % measured.

year	measured	why
2005	99.4 % of 2 Sep to 31 Dec	the record begins on 2 September
2009	97.0 %	outages
2010	98.9 %	outages
2012	94.1 %	power-supply failure and storm damage
2016	94.3 %	the January outage during which the logger was replaced
2019	95.7 %	outages

The 2012 faults are the largest interruption inside the measured period. A logger clock error shifted one block of August 2012 by 15.5 hours and was corrected; a tower power-supply failure in late July and August and storm damage in late October and November left values that were biased rather than missing, and those periods were removed. The removal windows were derived from this variable’s own disagreement with the independent hut barometer rather than copied from the other products, so they do not coincide exactly with the windows used for SW_IN, TA and PPFD_IN. Anything that assumed the four products exclude identical periods should use each product’s own missingness instead.

What the sensor is, and what the absolute level rests on

The instrument is a Vaisala PTB101B barometer. Every logger program that has ever measured it, the six CR10X programs of 2004-2006 and the CR1000 program installed in January 2016, converts its signal with the same multiplier and the same offset, spanning 600 to 1060 hPa, which is that barometer’s own range. The maintenance record does not name the instrument once in twenty-one years: it has never been replaced, calibrated or cleaned as far as the record goes, and it is not in the device inventory.

Pressure is the one meteo variable here whose absolute level can be checked against physics rather than only against another sensor, and it passes. The barometric formula, with nothing fitted to the tower data, predicts the difference between this barometer and each reference from that reference’s own pressure, temperature and humidity. It places the barometer within 5 m of its nominal elevation using MeteoSwiss Lägern (2.5 km away at 845 m, about 110 m above the barometer) and within 9 m using the NABEL barometer at the hut, and the two references agree with each other to 4 m. It also reproduces the annual wave in the difference between two barometers at different heights, 87 Pa measured against 79 Pa predicted, the two cycles correlating at 0.99. A unit error, a wrong offset or a lost factor of ten would fail this by orders of magnitude.

What the January 2016 acquisition change did

It moved the level by about 0.09 kPa, and this is the one hardware change in the record that did something measurable.

The barometer and its conversion did not change. What changed is how the logger read the signal: the CR10X excited the sensor and measured it single-ended, while the CR1000 measures it differentially. A single-ended measurement carries whatever offset sits between the sensor’s ground and the logger’s, and a differential measurement does not.

The change is placed on the tower rather than on a reference by the pattern of three barometers. Both differences involving the tower move across 2016, by −0.0884 and −0.0888 kPa, while the difference between the two references moves by −0.0001 kPa. Which era is correct is settled absolutely rather than by preference: the later era sits on the barometric prediction against both references and the earlier era sits about 0.1 kPa above it. The record before 21 January 2016 therefore reads approximately 0.09 kPa high.

The step is bracketed by the 20.6-day outage that ends on 21 January 2016 and cannot be dated more finely. The logger program on disk is dated 18 January 2016 and the maintenance record names that same program file, so the record, the program and the maintenance log agree on the date independently of one another.

The December 2021 change of screening software, which coincides with a hardware change for some other variables, does nothing here: the difference against MeteoSwiss changes less across that boundary than it does in an ordinary year.

Known limitations

The record is not homogeneous across 21 January 2016, and is not corrected. The values before that date read approximately 0.09 kPa high, as described above. No homogenised column is provided, for three reasons: the step is one part in a thousand of the value and under ten metres of air, so air density and the flux quantities computed from it change by the same one part in a thousand; a second, rescaled column would change the set of columns this product and the merged product ship; and a residual drift of the same size, described in the next point, would remain after any homogenisation. Analyses that need one absolute level across the whole record should subtract 0.09 kPa from the records before 21 January 2016, the later era being the one to keep. Analyses within a year, or of pressure changes rather than pressure levels, are unaffected.
A slow drift of about 0.1 kPa over twenty years cannot be attributed. Beyond the two dated events, the difference between this barometer and MeteoSwiss Lägern falls by approximately 0.1 kPa across the record. Slightly more than half of that falls after mid-2018, when the second barometer at the site stops, so there is no witness that can say which of the two instruments is moving. It develops over years rather than stepping at a date, so there is no boundary at which a correction could be applied.
The MeteoSwiss Lägern pressure series changed level in October 2010. The same rebuild of that station’s instrumentation that moved its radiation record moved its pressure record: PA_LAE_MS steps by approximately 0.06 to 0.07 kPa relative to both barometers at this site, which is what places the change there. It does not affect this product, which never uses Lägern to correct a value, but a difference between this product and MeteoSwiss Lägern must not be read as evidence about the tower across that date.
The NABEL hut barometer is not an independent record after July 2018. The database series PA_NABEL_H1_2_1 runs nominally to the end of 2018, but from 3 July 2018 its values simply repeat the tower series to five decimal places. The notebook detects this and truncates the reference there. Anything else reading that series should do the same: comparing the tower against it over those months returns perfect agreement and means nothing.
The record does not begin until September 2005. Anything joining this product to SW_IN, TA or PPFD_IN, whose index starts on 2004-01-01, has to cope with a shorter series rather than assuming a common period.