Vapour pressure deficit at 47 m

Half-hourly vapour pressure deficit at 47 m on the CH-LAE tower, 2004-2025, complete. The file is 07_METEO_VPD_2004-2025 (parquet and CSV): 385,728 records on a continuous 30-minute middle-timestamp index (named TIMESTAMP_MIDDLE in the file) in local time (UTC+1, no daylight saving), from 2004-01-01 00:15 to 2025-12-31 23:45.

VPD is not measured. It is computed from air temperature and relative humidity by formula, so its properties are inherited from those two products rather than from an instrument of its own.

Method, evidence and checks:

07_METEO_VPD — computes the product, and measures what the January 2016 acquisition change leaves in it.
02_METEO_TA — the temperature input, and the correction applied to it.
04_METEO_RH — the humidity input, and why no corrected version of it exists.
Meteo_Product_Chain — where this product sits in the chain.

Not homogeneous across 21 January 2016

Air temperature and relative humidity at 47 m come from one probe, replaced together with its logger on 21 January 2016. Both inputs change level at that date.

The temperature input is corrected: this product uses the homogenised TA, so its temperature term is on one consistent level across the whole record. The humidity input is not corrected, because no defensible correction for it exists, see Relative humidity at 47 m.

What remains is a step of roughly 17 % of the mean VPD at that date, measured against MeteoSwiss Lägern. Any statistic crossing 21 January 2016 contains it. Use FLAG_VPD_T1_47_1_SOURCE to stay inside one probe generation. Comparisons within an era are unaffected by the step, though see the limitation on drift below.

Columns

Every one of the 385,728 records carries a value, and both flags are defined everywhere.

Table 1: Columns of 07_METEO_VPD_2004-2025.

column	unit	description
`VPD_T1_47_1`	kPa	Vapour pressure deficit, computed and gap-free. Not homogeneous across 2016-01-21.
`FLAG_VPD_T1_47_1_ISFILLED`	-	What the record was computed from.
`FLAG_VPD_T1_47_1_SOURCE`	-	Which probe generation stands behind the record.

Values range from 0 to 4.345 kPa, with a median of 0.179 kPa. VPD is zero at saturation and cannot be negative.

There is deliberately no _HOMOGENIZED column. That name would assert the record is comparable across January 2016, and it is not: only one of the two inputs can be put on a single level.

What it is computed from

VPD is the difference between the saturation vapour pressure at the air temperature and the actual vapour pressure, so it follows from TA and RH alone. The Magnus coefficients are those used by ReddyProc.

Table 2: The two inputs.

input	column used	from
air temperature	`TA_T1_47_1_HOMOGENIZED_gfXG`	notebook `02`
relative humidity	`RH_T1_47_1`	notebook `04`

The temperature input is the homogenised column, and the choice affects more than accuracy: the two pre-2016 input errors act on VPD in opposite directions. The earlier TA reads about 1.3 °C too cold, which lowers saturation vapour pressure and pulls VPD down; the earlier RH reads a few percentage points too dry, which pushes VPD up. A VPD computed from the uncorrected temperature therefore reports a 2016 step about three times smaller, not because that series is better, but because one error masks the other. The masking depends on the particular temperature and humidity and is not a property to rely on, so the product uses the best available estimate of each input and states the residual step openly.

`FLAG_VPD_T1_47_1_ISFILLED`

Filter on == 0 for records computed from two genuine 47 m measurements. Every record is computed by the formula, none is modelled, but a computed record need not rest on two measurements.

Table 3: Provenance codes and their record counts.

code	meaning	records	share
0	measured `TA` and measured `RH`	369,056	95.7 %
5	measured `RH`, gap-filled `TA`	23	0.01 %
6	`RH` reconstructed from NABEL at 49 m	15,729	4.1 %
7	`RH` reconstructed from MeteoSwiss Lägern	920	0.2 %

Code 6 outranks 5 because it is the larger error: RH enters the formula linearly and a reconstructed value carries approximately 3 % RH of uncertainty. Codes 1, 2 and 4 are reserved for modelled values and do not occur.

`FLAG_VPD_T1_47_1_SOURCE`

Carried over unchanged from notebooks 02 and 04, which describe the same probe and whose flags are identical record for record.

Table 4: Probe generations and their record counts.

code	probe and acquisition	period	records
0	Campbell CS215 on SDI-12	2016-01-21 14:15 to 2025-12-31 23:45	174,356
1	Rotronic MP101A, single-ended analog	2004-09-20 10:45 to 2015-12-31 23:45	197,739
2	acquisition changeover, generation undetermined	2016-01-01 00:15 to 2016-01-21 13:45	988
3	before the tower record begins	2004-01-01 00:15 to 2004-09-20 10:15	12,645

Coverage

The series is complete. What varies is how much of each year rests on two measurements rather than on a reconstructed humidity.

Table 5: The years in which fewer than 99 % of records are computed from two measurements. Every other year of 2005-2025 is at least 99 %.

year	computed from two measurements	why
2004	28.0 %	the tower humidity record begins on 20 September
2009	97.2 %	scattered outages
2012	92.8 %	logger clock error, power-supply failure and storm damage
2016	94.3 %	the January outage during which the probe and logger were replaced
2019	95.7 %	an outage after the NABEL reference ends

Known limitations

The record is not homogeneous across 21 January 2016. The residual step is approximately 17 % of the mean VPD, against a year-to-year variation of about 2 % in the same statistic. It is what the uncorrectable humidity step leaves behind after the temperature term has been put on one level. Restrict anything crossing that date to one probe generation using FLAG_VPD_T1_47_1_SOURCE, or carry the step as an uncertainty. Computing VPD from the uncorrected TA instead would report a step about three times smaller, because the two input errors have opposite signs on VPD and one masks the other. That is a smaller number, not a better product: this one uses the best available estimate of each input.
Both humidity eras also drift internally. Against MeteoSwiss Lägern the earlier era moves approximately −0.31 percentage points of RH per year and the later era approximately +0.50, so a VPD trend computed entirely inside one era still contains sensor movement. Over the length of each era this drift amounts to more than the 2016 step itself. It is not corrected, because it is not attributed to either station and, after the NABEL sensor stops in 2018, there is no third humidity series at the site to attribute it with.
The step is largest where VPD is largest. The humidity error is biggest in the drier half of the range and vanishes near saturation, so warm dry afternoons, the conditions that produce high VPD, carry more of it than the record-wide figure suggests.
Reconstructed humidity propagates. 4.3 % of records rest on a humidity transferred from another instrument rather than measured at 47 m. Filter on FLAG_VPD_T1_47_1_ISFILLED == 0 where that matters.
The temperature correction has its own limits. The homogenised TA puts the two eras on one level; it does not make either absolutely accurate. Both 47 m sensors sit in passive radiation shields that heat in sunlight, and what 02 removes is the difference between the two shields’ responses, not the heating itself.