Relative humidity at 47 m

Half-hourly relative humidity at 47 m on the CH-LAE tower, 2004-2025, complete. The file is 04_METEO_RH_CORRECTED_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.

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:

04_METEO_RH — builds the product: the merge of the two screenings, the 2012 corrections, the RH > 100 % correction, the identification of the two probe generations and the step between them, and the two reconstruction passes.
02_METEO_TA — the same probe delivers TA, and the January 2016 replacement moved that variable too. What was done about it there is the closest available model for what a homogenised RH would require.
Meteo_Product_Chain — where this product sits in the chain. VPD is computed from it.

Not homogeneous across 21 January 2016

The 47 m probe was replaced on 21 January 2016, together with the logger that read it: a Rotronic MP101A read as an analog voltage gave way to a Campbell CS215 read over a digital line. The exported series steps upwards by 2.1 to 3.3 percentage points of RH at that date, and by 3.9 to 4.1 on the half-hours away from saturation. The earlier era reads too dry.

There is no homogenised column. The step is not one number: it is 3 to 6 percentage points over the drier half of the range and close to nothing near 100 % RH, so no constant correction reaches both ends. A correction that varies with the humidity was built, tested and rejected, because both eras drift internally by more than the step between them and no fixed transfer is right for any particular year. Anything crossing that date must be restricted to one probe generation using FLAG_RH_T1_47_1_SOURCE, or must carry the step as an uncertainty.

Comparisons within one era are unaffected. See What happened in 2016.

Columns

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

Table 1: Columns of 04_METEO_RH_CORRECTED_2004-2025.

column	unit	description
`RH_T1_47_1`	%	Relative humidity, corrected and gap-free. Bounded to `[0, 100]`. Not homogeneous across 2016-01-21.
`FLAG_RH_T1_47_1_MISSING`	-	Whether the value was measured at 47 m, and if not, which sensor it was reconstructed from.
`FLAG_RH_T1_47_1_SOURCE`	-	Which probe generation stands behind the record.

`FLAG_RH_T1_47_1_MISSING`

Filter on == 0 for genuine 47 m measurements. The series is complete, so every other code marks a value transferred from another instrument. Nothing in this product is modelled: a reconstructed value is another measurement of the same quantity, corrected by a monthly offset.

Table 2: Provenance codes and their record counts.

code	meaning	records	share
0	measured at 47 m	369,079	95.7 %
3	reconstructed from NABEL at 49 m on the same tower	15,729	4.1 %
4	reconstructed from MeteoSwiss Lägern, 2.5 km away	920	0.2 %

Codes 1 (never measured) and 2 (removed as faulty) are defined in the notebook but do not occur in this file: the two references between them cover every gap.

Code 3 carries approximately 3 % RH of uncertainty, established by leave-one-year-out validation against the years both sensors measured. Code 4 is weaker still, since Lägern is 2.5 km away and 400 m lower, and is concentrated in 2019 (759 records) and 2024 (107 records), after the NABEL sensor ends.

`FLAG_RH_T1_47_1_SOURCE`

Table 3: Probe generations, their record counts and the period each covers.

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, probe 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

Codes 2 and 3 mark records that no probe at this height stands behind. Code 2 is the changeover outage: the tower recorded nothing for 20 days while the logger was rebuilt, and those records are reconstructed from NABEL. Select FLAG_RH_T1_47_1_SOURCE <= 1 where the probe generation has to be certain, and == 0 or == 1 to stay inside one of them.

Coverage

The tower measurement begins on 2004-09-20 10:45. Everything before that is reconstructed from NABEL. From 2005 onwards every year is at least 92.8 % measured.

Table 4: The years in which measured coverage is not near-complete. Every other year of 2005-2025 is at least 99 % measured.

year	measured	why
2004	28.0 %	the 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, filled from MeteoSwiss

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; a tower power-supply failure in late July and August, and storm damage in late October and November, left records that could not be repaired. The clock error was corrected, the other two periods were removed, and all of it was reconstructed from NABEL.

What happened in 2016

The probe and its acquisition system were replaced together. The logger programs of the two eras name them directly: six CR10X programs spanning 2004-2006 all measure a Rotronic MP101A on a single-ended analog channel at the same conversion, and the CR1000 program installed at the changeover measures a Campbell CS215 over SDI-12. Air temperature and relative humidity arrive from that one probe in one sensor call, which is why the same date splits both variables.

The change moved the series, and the move is on the tower. Three humidity series span the changeover: the tower probe, a NABEL probe at 49 m on the same mast, and MeteoSwiss Lägern 2.5 km away. Both comparisons that contain the tower probe step upwards at January 2016 by roughly ten times what they move in an ordinary year, and the comparison between the two references does not move at all.

Table 5: The change in each pairwise difference across January 2016, in percentage points of RH, against what the same comparison does from one year to the next in an ordinary year.

comparison	change across 2016	ordinary year	away from saturation
tower − NABEL 49 m	+3.32	0.29	+4.14
tower − MeteoSwiss Lägern	+2.12	0.22	+3.85
NABEL 49 m − MeteoSwiss Lägern	−0.35	0.61	−0.24

The 100 % ceiling hides part of it. Relative humidity cannot exceed 100 %, and this record sits at a median of 82.9 % RH with a large share of half-hours at saturation. Where both probes are already against that bound the step has nowhere to go, so the whole-record figure understates it: on the half-hours where nothing is near the ceiling the step is 1.2 to 1.8 times as large. Measured against the humidity itself, it is 3.2 to 6.3 percentage points below 85 % RH and at most 1.1 above 95 % RH.

The probe that was replaced had drifted. Independently of any reference, 1,931 half-hours of the analog era read above 100 % RH, physically impossible and the signature of a capacitive humidity element that has aged. Their share grows through that era and reaches 3.6 % of 2015, the last full year before the replacement. The SDI-12 era has none.

The step is present at every hour, and is larger at night (+4.3 % RH against NABEL) than in daylight (+2.2 % RH). That rules out the radiation-shield mechanism that accounts for part of the TA step, which acts only in sunshine.

The maintenance record adds little: GIN files no device at the 47 m T/RH location for the whole of the exported period, the January 2016 visit report carries no device tag, and the CS215 is named at this logger only in a wiring repair six years later. The identification rests on the logger programs and on the data.

Known limitations

The record is not homogeneous, and cannot be made so with a constant. Any statistic crossing 21 January 2016 (period means, trends, year rankings, anomalies, threshold-hour counts) contains a step of 2 to 4 percentage points of RH. Because the step depends on the humidity itself and vanishes at saturation, no single offset removes it, and none is applied. The level-dependent transfer such a column would need was built and measured, and it does not hold: re-estimated from either half of its own eras it changes by as much as the correction itself, and summer and winter disagree by about as much again. Applying it would also push 19 % of early-era records past 100 %.
Both eras drift internally, by more than the step between them. Against MeteoSwiss Lägern the earlier era moves approximately −0.31 percentage points of RH per year and the later era approximately +0.50, so each era travels further over its own length than the 2016 jump. A trend computed entirely inside one era therefore still contains sensor movement. The drift is not corrected: it is not attributed to either station, and after the NABEL sensor stops in 2018 there is no third humidity series at this site to attribute it with. Correcting it towards MeteoSwiss would replace the measurement with the reference.
Which era is correct is not settled here. The evidence shows that the earlier probe drifted and that the later one reads moister; it does not establish an absolute accuracy for either. The product is the measurement with the probe generation named beside it.
VPD inherits this step, and it is the dominant term in it. 07_METEO_VPD is computed from the homogenised TA and from this series, so its temperature term is put on one level and its humidity term cannot be. What remains is a step of about 17 % of the mean VPD. The RH step is largest in the drier half of the range, which is where VPD is largest. See Vapour pressure deficit at 47 m.
Reconstructed records are estimates, not measurements. Code 3 carries approximately 3 % RH of uncertainty against a sensor-to-sensor agreement of about 2 % RH in summer; code 4 more. January to September 2004 is reconstructed throughout and is the least constrained part of the record.
Values near 100 % are corrected, not restored. A drifting probe loses resolution close to saturation. Where the analog era read above 100 % a daily offset was subtracted and the remainder clipped, which removes the impossible values but does not recover the information they stood for. Analyses of saturation frequency should expect the two eras to differ for this reason as well as for the step.
The NABEL reference is not independent after July 2018. From August 2018 the ingested RH_NABEL_T1_49_1 column reproduces the tower column exactly. Nothing in this file is derived from those months, since the notebook cuts the reference there, but a user comparing this product against that NABEL column directly would be comparing it against itself.