Incoming longwave radiation at 47 m

Half-hourly incoming (downwelling) longwave radiation at 47 m on the CH-LAE tower, September 2005 to September 2026, exported as measured. The file is 06_METEO_LW_IN_2005-2026.09 (parquet and CSV): 368,976 records on a continuous 30-minute index TIMESTAMP_MIDDLE (middle of the half-hour, UTC+01:00, no daylight saving), from 2005-09-14 00:15 to 2026-09-30 23:45. 2026 holds January to September only.

The interactive dashboard summarises the product on one page.

Method, evidence and checks:

06_METEO_LW_IN: builds the product: merge of the two screenings, 2012 corrections, identification of the instrument from the logger programs and the maintenance record, and the measurement of the calibration change described below.
01_METEO_SW_IN: the shortwave channel of the same radiometer. It establishes the instrument’s identity, dates its replacement, and shows what the January 2016 and December 2021 hardware changes did to the shortwave side.
02_METEO_TA: the air temperature product, the only diagnostic driver of notebook 06.
Meteo Product Chain: where this product sits in the chain.

The record changes level on 7 June 2016

This series is not homogeneous. Until 7 June 2016 the pyrgeometer signal was converted with the calibration factor of the pyranometer in the same instrument, 10.03 instead of 12.83 µV/W/m². On that date the logger program was corrected. Everything recorded before it reads too low.

The error is proportional to the difference between the value and \(\sigma T_a^4\), so it is largest under a clear sky and close to zero under low overcast: about 8 W m^-2 in the middle of the distribution and up to about 23 W m^-2 on a clear, dry night (17 % of the nighttime difference before the change, at its median and at its 2nd percentile).

No correction and no homogenised column are provided (see Known limitations). Filter on FLAG_LW_IN_T1_47_1_SOURCE before comparing the two eras, and treat any trend across mid-2016 as containing this change.

Columns

Table 1: Columns of 06_METEO_LW_IN_2005-2026.09.

column	unit	description
`LW_IN_T1_47_1`	W m^-2	Incoming longwave radiation, as measured. Contains `NaN`.
`FLAG_LW_IN_T1_47_1_SOURCE`	-	Which instrument and which calibration factor produced the value. Defined at every record.

There is no ISFILLED flag, because nothing is filled: a non-null value is a measurement and a NaN is a real gap.

The values are downwelling longwave irradiance, median 316.15 W m^-2, range 135.45 to 441.78 W m^-2. A pyrgeometer’s voltage measures the net exchange between the sky and the instrument body, roughly -50 W m^-2 at this site; the downwelling flux is that net signal plus the instrument’s own emission. That term is already included in every value. Reading the raw converted voltage as the flux would be wrong by about 370 W m^-2.

`FLAG_LW_IN_T1_47_1_SOURCE`

Table 2: Provenance codes and their record counts over the 364,073 measured records.

code	meaning	period	records	share
0	CNR4, its own calibration factor	from 2022-01-07	82,828	22.8 %
1	CNR1 at 12.83 µV/W/m², the pyrgeometer’s own factor	2016-06-07 to 2021-12-14	96,037	26.4 %
2	CNR1 at 10.03 µV/W/m², the pyranometer’s factor; reads low	before 2016-06-07	184,104	50.6 %
3	changeover, era undetermined	2021-12-15 to 2022-01-06	1,104	0.3 %

Code 3 marks the three weeks between the installation of the CNR4 and the arrival of its own constants in the logger program, during which the new instrument was read through the old instrument’s factor. Neither the maintenance record nor the data can assign this interval to either side.

Coverage

The measurement begins on 2005-09-14 13:15, when the pyrgeometer channel was wired to the tower logger. There is no longwave record before that date and none is reconstructed. 364,073 of the 368,976 records carry a value (98.7 %).

Of 998 gaps, 977 are one hour or shorter and eight exceed a day. The longest, 20.6 days from 1 to 21 January 2016, is the site-wide outage during which the tower logger was replaced; PA has the same gap.

Table 3: Years in which measured coverage is below 95 %. Every other year of 2006-2025 is at least 95 % measured, and January to September 2026 is complete.

year	measured	why
2012	92.3 %	logger clock error, power-supply failure and storm damage
2016	94.3 %	the January outage during which the logger was replaced

In 2012 a logger clock error shifted one block of August by 15.5 hours, and a tower power-supply failure in late July and August and storm damage in late October and November left records that were removed. LW_IN shows no sign of either fault in its own values, and it has no independent reference that could reveal a level error. The removal therefore rests on the documented logger-wide fault and is the conservative choice.

Known limitations

The record is not homogeneous across 7 June 2016. Values before that date read low by about 17 % of the difference between the value and \(\sigma T_a^4\), over about half the record. This is not corrected. Undoing it requires splitting each stored value into the instrument’s own emission and its net signal, which needs the radiometer body temperature. This product does not read that channel, and its treatment in the processing that produced the earlier values cannot be verified from the available files. Rescaling with air temperature as a proxy would put an unverified model into the exported numbers. The SOURCE flag marks the affected era instead.
The record also changes across December 2021, and the cause is not established. The radiometer was replaced (CNR1 to CNR4) on 14 December 2021, and the screening software changed from mst to diive at the turn of the year. The two cannot be separated, and there is no reference against which to test either. The clear-sky part of the signal changes by about 11 %, about half the size of the June 2016 change, which accounts for most of the 6.5 W m^-2 between the two periods once their temperature difference is removed. This is equally consistent with a change of instrument, a change of screening, and a gradual change in the lower atmosphere, which moved the same statistic in the same direction from 2019 to 2024 and reversed through 2025 and the dry year 2026. Nothing is corrected. Allow for a change of this order across that date; the SOURCE flag separates the two instruments.
The January 2016 acquisition change did not move this series. It moved TA_T1_47_1 by 1.3 °C and is easily confused with the June date five months later. The logger program installed at the changeover applies the same conversion as the programs before it, and the four months between the two 2016 dates behave like an ordinary year.
The series contains gaps and is not gap-filled. Code that assumes a complete series will fail or propagate NaN.
There is no independent reference for this variable at the site. MeteoSwiss Lägern has no longwave channel, NABEL has no pyrgeometer, and the below-canopy sensor installed in 2024 measures a different quantity. Every statement above rests on the physical relation between longwave radiation and air temperature rather than on a second instrument, so the evidence is weaker than for SW_IN or TA. Where it does not identify a cause, notebook 06 says so.
Summer 2026 was unusually dry and clear. At MeteoSwiss Lägern, relative humidity was 95 % or higher in 0.7 to 5.5 % of the half-hours of July to September 2026, against 4 to 38 % in the same months of 2019-2025. With so few overcast nights, the overcast ceiling of nighttime apparent emissivity (its 95th percentile), one of the two checks in notebook 06, falls to 0.93 to 0.95 in those months, below the same months of every earlier year. It falls in the same way when computed with the Lägern air temperature, so it reflects the weather, not the sensor.
Absolute accuracy is limited by the applied calibration. Both CNR1 eras convert the pyrgeometer with a single factory constant and no correction for the dome. The two radiometer channels of the successor instrument differ by 28 % in sensitivity, the scale of the error removed in June 2016. What remains is the ordinary field uncertainty of an unventilated pyrgeometer, which is larger than for the shortwave channel of the same instrument.