Photosynthetic photon flux density at 47 m
Half-hourly incoming photosynthetic photon flux density at 47 m on the CH-LAE tower, 2004 to September 2026, gap-filled to completeness. The file is 03_METEO_PPFD_IN_GAPFILLED_2004-2026.09 (parquet and CSV): 398,832 records on a continuous 30-minute index TIMESTAMP_MIDDLE (middle of the half-hour, UTC+01:00, no daylight saving), from 2004-01-01 00:15 to 2026-09-30 23:45.
The interactive dashboard summarises the product on one page.
Method, evidence and checks:
03_METEO_PPFD_IN: builds the product (merge of the two screenings, 2012 corrections, nighttime offset, gap-filling), identifies the sensor and tests it at both hardware changes.RADIATION_SENSOR_CONTINUITY: compares four radiation sensors and places the decline described below on this sensor rather than on either reference.- Meteo Product Chain: this product depends on
SW_IN, which is built first.
Against both of its independent references the 47 m quantum sensor reads about 3 to 3.5 % lower in 2026 than in 2021, and about 6 to 7 % below its 2006-2010 level. The fall has not continued since 2024: 2025 and 2026 lie within 0.4 % of 2024.
The decline is not corrected (see Known limitations). A trend or period comparison that spans the last years of the record contains it. Comparisons within a year, and the use of single half-hours, are unaffected.
There is one value column and no _HOMOGENIZED column: one instrument over the whole record, read through the same conversion by two acquisition systems, with no step at either hardware boundary.
Columns
Every record carries a value, and the flag is defined at every record.
03_METEO_PPFD_IN_GAPFILLED_2004-2026.09.
| column | unit | description |
|---|---|---|
PPFD_IN_T1_47_1_gfXG |
µmol m-2 s-1 | Incoming photosynthetic photon flux density, gaps filled. Complete, non-negative, exactly zero at night. |
FLAG_PPFD_IN_T1_47_1_ISFILLED |
- | Whether the value was measured, and if not, which method produced it. |
FLAG_PPFD_IN_T1_47_1_ISFILLED
Filter on == 0 for measured records. Every other code marks a filled value.
| code | meaning | records | share |
|---|---|---|---|
| 0 | observed | 382,228 | 95.8 % |
| 1 | daytime gap, filled by the XGBoost model | 8,374 | 2.1 % |
| 2 | daytime gap, filled by the timestamp-only fallback model | 600 | 0.2 % |
| 3 | nighttime gap, set to zero | 7,543 | 1.9 % |
| 4 | short daytime gap, filled by linear interpolation | 87 | 0.02 % |
Code 3 is not an estimate: the sun is below the horizon and the value is zero. Codes 0 and 3 together are 97.7 % of the record.
The 16,604 filled records fall in 159 runs. Most runs are single half-hours (126 of 159), but eight runs of a day or longer hold 16,301 of the filled records: the period before the measurement begins and a few outages. Filtering on the flag therefore removes mainly whole periods. Seventeen of the 23 years (2026 to September) have fewer than a hundred filled records, and seven have none.
Coverage
The tower measurement begins on 2004-09-20 10:45. Everything before that is modelled.
| year | measured | why |
|---|---|---|
| 2004 | 28 % | the record begins on 20 September |
| 2012 | 93 % | logger clock error, power-supply failure and storm damage |
| 2016 | 94 % | 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 could not be repaired. The clock error was corrected, the other two periods were removed, and all of it was gap-filled. The windows are the same as in SW_IN and TA, because the fault was in the logger and the power supply. The January 2016 outage runs from 1 January to 21 January 14:15, when the replacement logger came online.
How the gaps were filled
Daytime gaps are filled by one XGBoost model over the whole record; unlike SW_IN and TA there is no period split, because all three drivers span the full record. The drivers are the gap-filled SW_IN from notebook 01, potential radiation computed from the site coordinates, and MeteoSwiss Lägern global radiation. The tower pyranometer is the strongest driver (same height, same mast, r2 0.99), but it is itself gap-filled, and PPFD_IN gaps tend to coincide with SW_IN gaps because both sensors are on the same logger. Lägern, 2.5 km away, covers that case with more scatter. Code 2 marks the few half-hours where neither radiation driver was available. Nighttime gaps are set to zero. A reference is used only as a driver and never replaces a measured value.
Known limitations
- The sensor lost response between 2022 and 2024, and it has not recovered. It falls against the co-located tower pyranometer and against MeteoSwiss Lägern together, by 2.86 % and 3.45 % between 2021 and 2026, and reaches 5.63 % and 7.13 % below its 2006-2010 level in 2026. The change over the last year (+0.4 % and +0.3 %) is within the ordinary year-to-year change of these ratios (0.6 % and 0.7 %), so one stable season does not yet show that the response has settled. Both references moving together places the change on this instrument, and
RADIATION_SENSOR_CONTINUITYreaches the same conclusion from four sensors. It is not corrected: it develops over years, so there is no boundary at which to apply a correction, and no maintenance record identifies a cause, so a correction would be a rescaling towards a reference. The likely cause is the sensor itself after twenty years in the field (soiling, or ageing of the diffuser and detector). Allow for several per cent in analyses that span the last years of the record, or use a reference-normalised quantity. Gap-filled records inherit the state of the sensor at the time and neither add nor remove the bias. - A comparison against
SW_INacross 2013 shows that sensor’s drift, not this one’s. The 47 m pyranometer rises by about 3 % against its references from 2013 (see Incoming shortwave radiation at 47 m). Any ratio of the two products, such as a photon-to-energy ratio over the whole record, contains that change. - The MeteoSwiss gap-filling driver changed level in October 2010.
SW_IN_LAE_MSsteps by about 5 % on 6 October 2010, when that station’s radiation instrumentation was rebuilt. The fills are not affected, because a driver supplies the state of the sky. Do not read a difference between this product and MeteoSwiss Lägern as evidence about the tower across that date. - Nighttime values are exactly zero, not measurements. The sensor reads a small positive offset at night: about 4.2 µmol m-2 s-1 before the 2016 logger replacement and 2.5 µmol m-2 s-1 from then to the end of 2021. From 2022 the
diivescreening has already removed it. The offset was removed per day before gap-filling. For instrument noise or the offset itself, use the screened database series. - 2004 is 72 % modelled. Everything before 2004-09-20 10:45 rests on the
SW_INproduct, which is itself reconstructed for that period. These records are a fill built on a fill and carry the uncertainty of both.
The January 2016 changes
They did nothing measurable to this series. Two things happened that month: the tower logger was replaced, which moved TA_T1_47_1 by 1.3 °C, and the fieldbook records a new Kipp & Zonen PAR sensor installed on 8 January 2016.
That sensor is not this column. The CR1000 logger program reads two incoming-PAR instruments at 47 m and names both. PPFD_IN_T1_47_1, exported here, is a Delta-T sunshine sensor, whose signal is already scaled to µmol m-2 s-1 and is read at a multiplier of exactly 1. The Kipp & Zonen PAR LITE installed that January is PPFD_IN_T1_47_2, read at the reciprocal of its own sensitivity, and its record begins on 2016-01-21 14:15. The maintenance record alone cannot make this distinction, because it files both instruments under the same operation tag.
The Delta-T sensor was installed in August 2004 and appears in every surviving logger program since, at multiplier 1 and offset 0 in both the CR10X and the CR1000 era. The maintenance record shows no replacement and no calibration in twenty-two years, only two cleanings, in December 2016 and April 2021.
Across 2016 the ratio of this sensor to the tower pyranometer changes by +0.4 % and to MeteoSwiss Lägern by +0.5 %, both less than in an ordinary year. A change confined to the tower acquisition system could move both tower sensors and leave their ratio flat, but it could not move their ratio against a station 2.5 km away. Only the nighttime offset changed, and the per-day correction removes it on both sides of the date. The mst and diive screenings meet at the end of 2021 without a step either. Notebook 03 asserts all of this on every run.