Outgoing shortwave radiation at 47 m

Half-hourly outgoing (reflected) shortwave radiation at 47 m on the CH-LAE tower, 2005-2025, exported as measured. The file is 12_METEO_SW_OUT_2005-2025 (parquet and CSV): 355,872 records on a continuous 30-minute index, TIMESTAMP_MIDDLE (middle of the half-hour), in local time (UTC+01:00, no daylight saving), from 2005-09-14 00:15 to 2025-12-31 23:45. The first measurement is at 2005-09-14 11:15.

The series ends on 31 December 2025. It is part of FP2026.1 and of the merged meteo product, where it is NaN from 1 January 2026; the extension to 30 September 2026 is in progress.

The interactive dashboard summarises coverage, provenance, seasonality, distributions, extremes and trends. Its instrument card shows how much of each year came from the CNR1, the changeover window and the CNR4.

Notebooks:

12_METEO_SW_OUT: builds the product. It merges the two screenings and shows that they are on one scale, applies the 2012 corrections and the nighttime offset, and identifies the instrument and conversion constant behind each part of the record.
SW_OUT_T1_47_1: the screening of the current era.
01_METEO_SW_IN: the incoming channel of the same radiometer. SW_OUT / SW_IN is the surface albedo, which is the only diagnostic available for this product and the basis of every albedo figure below.
Meteo_Data_LW_OUT: the outgoing longwave channel of the same instrument, with the same December 2021 boundary.
Meteo_Product_Chain: where this product sits in the chain.

One value column, and which flag matters

The file has two flags for one value column, because the record has two boundaries on different dates.

FLAG_SW_OUT_T1_47_1_INSTRUMENT marks a real break. The radiometer was replaced on 14 December 2021, and the two instruments convert this channel differently. Filter on this flag before comparing values across that date. Drop code 2, the weeks in which the new instrument was still read with the old instrument’s constant, before using December 2021 or early January
FLAG_SW_OUT_T1_47_1_SOURCE is provenance only. It names the screening that produced each value. The two screenings agree to 0.03 % on two years of simultaneous records, so trends and multi-year means may be taken across the boundary between them. For the same reason there is no _HOMOGENIZED column.

Nothing is gap-filled, so the value column contains NaN and every non-null record is a measurement. Nighttime values are exactly 0, which is a physical value, not a gap.

What the values are

The values are upwelling shortwave irradiance above the canopy. The stand reflects roughly a tenth of the incoming shortwave, so the values are an order of magnitude smaller than SW_IN: the daylight median is 22.0 W m^-2, the daylight mean 32.4 W m^-2, and the largest half-hour 236.2 W m^-2. Over the whole record, nights included, the median is 0.2 W m^-2.

Dividing by SW_IN gives the surface albedo. Over the 157,444 half-hours with an incoming flux above 20 W m^-2 and a measured reflected flux, the median ratio is 0.112. Thirty-five of them (0.022 %) give a ratio above 1.05. They occur at low incoming flux, and 28 fall between December and March, as expected from snow on the upward-facing sensor and a low sun.

Columns

Table 1: Columns of 12_METEO_SW_OUT_2005-2025.

column	unit	description
`SW_OUT_T1_47_1`	W m^-2	Outgoing shortwave radiation, as measured and corrected. Contains `NaN`; exactly `0` at night.
`FLAG_SW_OUT_T1_47_1_SOURCE`	-	Which screening produced the value. Defined at every record.
`FLAG_SW_OUT_T1_47_1_INSTRUMENT`	-	Which radiometer produced the value, and with which conversion constant. Defined at every record.

There is no ISFILLED flag, because nothing is filled. PA, LW_IN and LW_OUT are exported the same way.

`FLAG_SW_OUT_T1_47_1_SOURCE`

Table 2: Screening codes and their record counts. Shares are of the 351,612 measured records.

code	meaning	records	measured	share
0	diive screening, from 2020	70,048	70,048	19.9 %
1	MeteoScreeningTool, shown to be on the same scale	282,776	281,564	80.1 %
2	neither screening supplied this half-hour	3,048	0	0 %

Code 2 is not the same as a gap. A half-hour that a screening supplied and this product then removed (for example a record inside the 2012 fault windows) keeps that screening’s code and is NaN. Code 2 means that no value was ever supplied.

`FLAG_SW_OUT_T1_47_1_INSTRUMENT`

Table 3: Instrument codes and their record counts. Shares are of the measured records.

code	meaning	period	records	measured	share
0	CNR4, this channel with its own constant of 14.38 µV/W/m²	from 2022-01-07	69,840	69,760	19.8 %
1	CNR1, one constant of 10.03 µV/W/m² for all four channels	to 2021-12-14 13:15	284,907	280,729	79.8 %
2	changeover: CNR4 installed, CNR1 constant still in the logger program	2021-12-14 13:45 to 2022-01-06 23:45	1,125	1,123	0.3 %

There is no code for either 2016 date. The tower logger was replaced in January 2016 and a radiometer calibration factor was corrected in June 2016, and neither moved this channel (see The two 2016 dates).

Coverage

Of the 355,872 records, 351,612 have a value (98.8 %). Most gaps are short: of 452 gaps, 431 are one hour or shorter and seven exceed a day.

Table 4: The five longest gaps.

from	to	days	cause
2016-01-01	2016-01-21	20.6	site-wide outage during the tower logger replacement
2019-11-04	2019-11-19	15.8	station outage
2012-10-28	2012-11-09	13.0	removed in this product: storm damage and stressed cables
2012-07-29	2012-08-10	12.6	removed in this product: tower power-supply failure
2009-08-05	2009-08-15	9.8	station outage

Four years are below 98 % measured: 2012 at 92.7 %, 2016 at 94.3 %, 2019 at 95.7 % and 2009 at 96.9 %. The 2012 figure results from this product: 1,212 records were removed inside the two windows of documented logger-wide fault, as in notebooks 01 to 03 and 06.

Two corrections change values. The August 2012 clock error is corrected by the same +15.5 h shift as the other tower variables. Unlike LW_IN, this variable can confirm the shift, because reflected shortwave follows the sun and its correlation with potential radiation shows whether the block is on the solar clock. The nighttime zero offset is removed per day over the whole record, because only the newer screening had removed it; without this step the first sixteen years would carry an offset that the last four do not.

The two screenings

The two screenings read different inputs: the MeteoScreeningTool read delivered CSV files, the diive screening reads the raw record. They could therefore have converted the same voltages with different constants. They overlap for two full years on the same instrument, which allows a direct comparison.

On 14,542 well-illuminated half-hours that both report, spread over 24 calendar months, the factor between them is 0.99970 (0.03 %), and the ratio of their standard deviations is 0.99955. The smallest conversion difference known at this site is 3.5 %, 118 times larger. The two eras are therefore joined without rescaling, and no _HOMOGENIZED column is exported.

The two 2016 dates

The tower logger was replaced in January 2016. In June 2016 a new logger program changed a radiometer calibration factor from 10.03 to 12.83 µV/W/m²; this is the change that notebook 06 finds in LW_IN. Had that factor reached the outgoing shortwave channel, the albedo would have stepped by about a fifth, because only one of the two constants in the ratio would have changed.

It did not. The season-controlled albedo moves by −1.0 % across both 2016 dates, compared with −5.3 % for a placebo pair of periods within the earlier era and −21.8 % expected from a factor change on this channel alone. The record is homogeneous across both dates. The product notebook asserts this, so a re-run that finds such a step fails instead of exporting it.

The December 2021 radiometer exchange

The CNR1 was replaced by a CNR4 on 14 December 2021, and the conversion changed with it. The CNR1 logger programs convert all four channels with a single constant, so at least three of the four were converted with a constant that is not their own. The two shortwave channels of the CNR4 each have their own constant, and these differ by 3.5 %.

The size of the resulting step cannot be determined from this product. The season-controlled albedo differs by 4.2 % between the two instruments, but the same statistic differs by 10.0 % between two periods of the CNR4 era in which nothing changed. The growing-season albedo of the stand varies from year to year by more than the instrument difference, so the CNR1’s channel mismatch cannot be separated from it with the years available. Nothing is corrected: that would require the CNR1’s two detector sensitivities, which were not recorded at this site.

Code 2 isolates the weeks between the installation of the CNR4 and the update of its constants in the logger program. Compared with MeteoSwiss Lägern, which measures the incoming channel independently, the changeover window is 8.3 % away from an ordinary winter. If the CNR1 constant had still been in force, it would be 38.5 % away; these are 2.1 and 7.5 interquartile widths of the between-winter scatter. This is evidence against the old constant having been in force, but not proof that the window is correct. Seven winters can rule out an error of tens of per cent, not a small one, and the test applies to the incoming channel, whose constant in the same program was a different number. The window keeps its own code either way.

Known limitations

The record is not homogeneous across 14 December 2021, and the size of the step is not known. Filter on FLAG_SW_OUT_T1_47_1_INSTRUMENT before comparing values or albedos across that date, and treat any trend across it as containing an unquantified instrument change.
The 24 days of the changeover have an unverified conversion. Code 2 marks them. The MeteoSwiss comparison argues against the old constant, but does not establish which constant was applied.
The CNR1 era carries an unquantified channel mismatch. All four channels were converted with one constant, although each detector has its own factory calibration. Because both shortwave channels were divided by the same number, it cancels in their ratio, so the CNR1-era albedo does not depend on the absolute scale. The real difference between the two detectors does not cancel: the reflected flux of that era, and its albedo, include the difference between the two sensitivities. It is a fixed scale error within the era, not a break, so comparisons within the era are not affected.
The series contains gaps and is not gap-filled. Code that assumes a complete series will fail or propagate NaN. Nothing else measures what this canopy reflects: MeteoSwiss Lägern and NABEL measure incoming shortwave only, and the site’s other outgoing shortwave sensor is below the canopy. If a complete series is needed, fill it downstream, and do not train one model across the December 2021 boundary.
There is no independent reference. Every quantitative statement above uses the ratio to the incoming channel of the same instrument. This ratio cancels anything that scaled both channels together (a supply voltage, a dirty dome), and therefore cannot detect it.
The 2012 windows were removed without evidence from this series. They rest on the documented logger-wide fault, as in 01 to 03 and 06. The albedo can detect a fault in the reflected channel alone, not one that moved both channels of the instrument together, as a power-supply problem does.
Coverage counts nighttime zeros as measurements. In the diive era the screening set every nighttime value to zero, including gaps. This is physically correct, but from 2020 a nighttime record is a known value rather than a delivered measurement, and the coverage of those years is not strictly comparable with that of the earlier era.
A candidate removal window in July 2024 is still in the data. The screening notebook lists 2024-07-14 to 2024-07-19 for removal, but the removal was not committed. The albedo over the window is 2.6 % below the surrounding weeks, which is within how those weeks differ among themselves, and the maintenance record names no fault in this channel. The window is neither removed nor flagged; it is to be settled in the screening notebook.
The October 2010 rebuild of the MeteoSwiss Lägern station changes nothing in this file, but it steps that station’s incoming shortwave by about 5 %. A comparison of this product with Lägern across that date includes the step. The changeover test above uses only winters well after it.