Data trends at meteoLCD: 1998 to 2020

Trends computed from yearly averages at meteoLCD, Diekirch, Luxembourg.
Graphs may be freely copied and used, under the condition to cite:
MASSEN, Francis: Data trends at meteoLCD, 1998 to 2020.

Older trends are here!

Attention: in all trend equations (y = a+b*x) the variable x represents the year, with x = 1 for the first year in the trend period.

An Addendum 4 has been added to report on the PM (fine particle) measurements (04-Apr-20)


Most important conclusions from 1998 (2002, 2008) to 2020:

1. Significant increase in sunshine duration since 2008 by 167 hours*decade-1
2. Local temperatures show warming of 0.10C/y since 2002 (2018 was a strong El Nino year)
3. Diurnal temperature range (DTR) trend since 1998 is positive (= no anthropogenic warming fingerprint ).
4. The winter trend since 2002 shows a warming of +1.3 C*decade-1 ; this positive trend is also shown by the winter NAO index (+0.6 C*decade-1)
5. Since 1998 the ground O3 trend is positive,
 from 2002 to 2020 the total thickness of the ozone layer slightly decreases by 9.3 DU*decade-1
6. Local CO2 mixing ratio continues to increase at about 4.8 ppmV per year; the asymptotic background is close to that measured at Mauna Loa.
7. The trend of the biologically effective yearly UVB dose is distinctly positive from 2008 to 2020 (like solar irradiance and sunshine)
8. The trend of the UVA dose is distinctly positive from 2008 to 2020 (as eff. UVB, solar irradiance and sunshine hours)
9. Precipitation (rainfall) shows a sinusoidal pattern of close to 5 years period.
10. Energy content of moist air (enthalpy) shows a positive trend
11. The fine particles PM2.5 and PM10 concentrations are very low, and are well synchronized with those at Beidweiler.
NO/NOx measurements have been definitively stopped at the end of 2017.

Ground Ozone [ug/m3]
("bad ozone")

Mean +/- stdev of
2020: 52.9 +/- 35.9 ug/m3 

Mean +/- stdev (from yearly means)
1998 to 2020: 41.8 +/- 8.0 ug/m3
2002 to 2020: 41.5 +/- 8.8

Trends 1998 - 2020: 37.31+0.37*x
             2002 - 2020: 30.60+0.78*x           
(be careful: API400 values may be too low!)

Please note the succession of 3 different instruments; after the definitive breakdown of the Teledyne API400, the old O341M sensor from Environnement SA was used again and finally replaced by a CAIRSENS O3&NO2 in 2018. As NO2 values are below the minimum of this instrument, readings can be taken as O3 only. A comparison has shown an excellent concordance with the official Beckerich  station, whose data are used for calibrating the Cairsens readings..


See [1] [2] [3] [17]

Total Ozone Column [DU]
("good ozone")

Mean and stdev of the year 2020: 310.1 +/- 31.6 DU
minimum : 208.5 (18 Nov)
maximum: 426.5 (28 Jan)

183 common day direct sun readings for 2019 at Uccle and Diekirch:
Diekirch                    : 309.6 +/- 30.9
Uccle                         : 327.3 +/- 34.2

Since 2008 there is a negative trend of -14.4 DU/decade.

 has a slight positive trend of +3.6 DU/decade for the 1998-2019 period, and a similar of +3.9 DU/decade from 2010 to 2019.(see also [16])










Calibration multiplier to apply to the Diekirch DU data [55] and [56]
if Uccle Brewer(s) are the reference: 1.0584 (R2 =0.83)

See [4] [8] ([8] shows strong positive trend starting 1990 for latitudes 45-75 North, Europe): [27] give +1.32 DU/y at the Jungfraujoch for 1995-2004.
See also recent EGU2009 poster [16].

CO2 mixing ratio in ppmV 

Attention: The instrument for measuring CO2 (API Teledyne E600) has been replaced by a Vaisala GMP343 sensor the 27 Jun 2017. The jump from 2017 to 2018 seems implausible high, so a zero bias should be considered possible!

Mean and stdev of the year 2020: 440.8 +/- 30.7 ppmV

The 1998-2001 data are too unreliable to be retained for the trend analysis.

Trends :
2002 to 2020: 1.
4 ppmV*y-1
2018 to 2020: 4.8 ppmV*y-1

The sharp plunge in 2013 should be taken with caution; there also was a change in the calibration gas the 21 Jan. 2014.







The second picture zooms on the last 3 years, and gives the readings of Diekirch (DIK), Mauna Loa (MLO)  from 2018 to 2020; and Hohehpeissenberg (HPB) from 2018 to 2019 as 2020 data are not yet available for HPB. Note the very different elevations! Mauna Loa has no vegetation at all, Diekirch and HPB similar grass and forests.

The yearly trends are for this period are

219m asl
+4.8 (Vaisala GPM343)  ppmV/year  
Manua Loa
3397m asl, no vegetation
+ 2.56 ppmV/year
[34] [57]

Hohenpeissenberg (HPB)
977m asl, forests
+ 3.63 ppmV/year [48] [57]

Be careful with the Vaisala readings, as the Vaisala GPM343 might not give the same accuracy as the former API! These readings also are given for local atm. pressure and non-dried air!




The CO2 data (monthly averages) show the summer-time lows, which reflect the impact of variable seasonal photo-synthesis (see here). A simple 12 month periodic sinus pattern was also found in 2014 and 2015. Actually, as shown in addendum 3, the CO2 lowering intensity of wind speed seems to be an important modifier of this pattern, possibly masking the effect (or better: the non-effect) of photosynthesis. This happened in 2016 and 2017.

This year 2020 the yearly amplitude of the sinus fit is 8.56 ppmV (a total swing of ~17.1 ppmV, to be compared to about 12.4 ppmV at the HPB station for 2019 [48]).
The R2 is rather poor, due to the low readings in Feb and March, but the summer minimum clearly stands out.


See the end of addendum 3 for a picture of CO2 versus windspeed.



Air temperature [C]

Mean and stdev of the year 2020 (from monthly averages):
Diekirch:  12.58 +/- 5.83 
Findel:     11.28 +/- 6.20 

Mean Diekirch temperatures (+/- stdev) from yearly averages:

1998 to 2020 :  10.68 +/- 0.83 C                
2002 to 2020 :  10.79 +/- 0.87  C

The sensor location has not been moved since 2002! Sensor is a PT100 (see comments in 2015_only.xls); new 4-20mA amplifier (with calibration) installed the 4th May 2016.

Trends  from 2002 to 2020 (be aware that 2018 was a very strong El Nino year!):
meteoLCD:   +0.101 C/year
Findel:           +0.056 C/year           

Latest Global temperature anomaly trends for same period 2002-2018:
UAH (satellite, LTT):       + 0.0123C/year   [45]
RSS (satellite, LTT)   :    + 0.0172C/year
CRU (Hadcrut4krig v2):   + 0.0164C/year

Highest decadal Central England warming trend from 1691 to 2009: +1.86C/decade for 1694-1703!
See also [15] (which may be obsolete)

Diurnal Temperature 
Range (DTR)  [C]

DTR = daily max - daily min temperature

Mean and stdev of the year 2020 (from monthly averages):
Diekirch: 9.40 +/- 3.75
Findel:    8.24 +/- 2.88

Mean DTR at Diekirch:

1998 to 2020:
2002 to 2020:  8.78

For 1998 to 2020: all trends are positive, the 24hmin trend is lower than the 24hmax trend.

A fingerprint of climate warming is that daily minima increase more rapidly than daily maxima so that the DTR trend should become negative. This fingerprint does not exist here (and neither at the Findel station, see 2nd plot).

The BEST observational data set for Luxembourg [29] stops at 2013. For our latitude of 50 North, BEST shows a positive DTR trend for the period 1988 to 2011, whereas theCMIP5 multi-model mean gives a similar but negative trend... so much for the concordance between climate models and observations! (graph here).

See [5], [13] and [30]

Winter temperatures [C]

Values of winter DJF temperature of the year 2020:
Diekirch: 6.04
Findel:     4.37
NOA:        1.27
 NAO normalized index [47]

The trends show warming winters since 2002 to 2020, with the warming probably caused by the NAO.

Trends  from 2002 to 2020:
(2016 was a very strong El Nino year!):
Diekirch:    +0.130 C/year since 2002
Findel:       +0.111
Germany:  +0.100   [46]
NAO:         +0.063

The plot shows the mean temperatures from December (of previous year) to February. It also shows in magenta the NAO index for the months Dec to Feb

The North Atlantic Oscillation clearly influences our winters (but note the exception for 2019! [51]); the correlations between all the DJF temperature series and the NAO_DJF normalized index are statistically significant (=0.60, 0.54, 0.64) at the 5% level.

Enthalpy of moist air in kJ/kg

Mean moist enthalpy of 2020: 32.35 +/- 13.13 kJ/kg

See [24] on how the energy content of moist air is calculated. Several authors, (e.g. Prof. Roger Pielke Sr.) insist that air temperature is a poor metric for global warming/cooling, and that the energy content of the moist air and/or the Ocean Heat Content (OHC) are better metrics.

Mean yearly moist enthalpy values are mostly very close, but they may change from zero up to 60 KJ/kg during a year. Moist enthalpy can not be calculated for temperatures <= 0 C.

mean +/- stdev from 2002 to 2020:  28.84 +/- 2.03 kJ/kg

Trend is positive from 2002 to 2020; the 2016 "monster" El Nino extended into 2017 .


Total Yearly Rainfall [mm]

Values of rainfall (precipitation) of the year 2019:
Diekirch:  622.2 mm
Findel:     788.7 mm

1998 - 2020 mean +/- stdev: 682.2 +/- 127.8 mm 
2002 - 2020 mean +/-stdev:  655.1 +/- 99.8 mm

The negative trend from 1998 to 2018 seems spectacular: -80 mm/decade, caused by the very high values of 2000 and 2001.
The 2002-2020 period has a near zero trend!








Clearly precipitation shows an oscillation pattern, so linear trends should be taken with precaution (or simply seen as non-sensical).

 A good model for the Diekirch data is a sinus function: the calculation (Levenberg-Marquart algorithm) suggests for the interval 2002 - 2020 a 5.21 years period (~62 months, R2 = 0.44); in the model x = 0  corresponds to 2002), with a mean value of 651 mm and an amplitude of 91 mm; the phase shift of 1.13 rad is close to 1/5 period. All these values are similar to those of the preceding two years, but note that the year 2018 is an outlier!
gives short term periods of 6 to 7 years for the Western European region.

The oscillatory rainfall pattern is a good example how foolish it is to apply linear regressions to data when these are harmonic, something the media, activist groups and many politicians often do without much thinking.

Solar energy on a horizontal plane

Values of total solar energy of the year 2020:
Diekirch: 1203.5 KWh/m2

1998 to 2020 mean +/- stdev: 1119.5 +/- 51.1 kWh/m2
2002 to 2020 mean +/- stdev: 1120.3 +/- 55.1 kWh/m2

Trends from 1998 to 2020 is small, and relative important (10 kWh*m-2*y-1) for last 13 years. [52]

(see Addendum 1 for calculations of radiative forcing and solar sensitivity, addendum 2 for detecting a solar influence on temperature and moist enthalpy)

Helioclim satellite measurements show ongoing solar dimming over Luxembourg for 1985 to 2005  [33] (see graph)
[14] finds 0.7 Wm-2y-1 for West-Europe 1994-2003 , meteoLCD +1 Wm-2y-1 for 1998-2003.See also [9]

Sunshine hours
(meteoLCD values derived from pyranometer data by Olivieri's method)

Values of sunshine hours of the year 2020:
Diekirch:      1843 hours (215m asl) (from pyranometer)
Findel:          2030 hours (365m asl, (from Campbell- Stokes)
Trier:            1951 hours (279m asl)
[40] [53]
Maastricht: 2017 hours (114m asl) [53]

1998 to 2020:   +10.6 hours/decade
2008 to 2020:   +167 hours/decade !

The decline from 2015 to 2017 is clearly visible in the German PV electricity production, but the negative trend reverses for the very sunny year 2018 [58].

See paper [23] by F. Massen comparing 4 different methods to compute sunshine duration from pyranometer







The 2nd graph shows the plots of the four above-mentioned stations. It should be noted that meteoLCD (Diekirch) is located in a valley, Findel, Trier and Maastricht airport on top of a plateau. The Findel totals are much higher than those of the other stations, which certainly is also partially caused by the use of the Campbell-Stokes instrument known to give too high readings.

All 4 stations give totals that practically always vary in the same manner (synchronous increase and decrease).

The trends of all 4 stations are strongly positive since 2008: those of Diekirch (+167 h/decade), Findel (+188h/decade) and Maastricht (+166 h/decade) are practically the same, whereas Trier-Petrisberg shows an astoundingly high trend of +304h/decade), which should be accepted with some precaution!









These strong positive trends probably suffice to explain the warming since 2008:

- all correlations between mean temperature are yearly sunshine hours are positive for the 4 stations, and these correlations are all statistically significant at the alpha = 0.95 level: Diekirch 0.71, Findel 0.61, Trier  0.69, Maastrich 0.58

- see the last multi-graph figure for the temp-versus-sunshine relationship at the 4 stations meteoLCD, Findel, Maastricht and Trier-Petrisberg.


Biologically eff. UVB dose on a horizontal plane in kWh/m2

Erythemal UVB dose of the year 2020: 0.153 kWh/m2

mean +/- stdev:
1999 to 2020: 0.133 +/- 0.009   eff. kWh*m-2y-1
2002 to 2020: 0.135 +/- 0.008
2008 to 2020: 0.135 +/- 0.009

The trend over 2002 - 2020 is slightly positive, the trend line from 2008 to 2020 distinctly positive, in concordance with solar irradiance and sunshine hours.


See [10] and [22] (poster finds slight positive trend in June (+2%) and negative trend in August (-1%), no trend for other months, for period 1991 to 2008.


UVA dose on a horizontal plane in kWh/m2 

UVA dose of the year 2020: 58.56 KWh/m2
(some intermittent problems with internal temperature stabilization of the sensor;
the influence seems minimal, so all readings have been kept)

mean +/- stdev:
1999 to 2020: 55.43 +/- 4.24 kWh*m-2*y-1
2002 to 2020: 55.70 +/- 4.20
2008 to 2020: 55.68 +/- 4.55 


The 4 independent measures of  solar irradiance, sunshine hours, eff.UVB and UVA doses all point to a strong increase since 2008.

NOx, NO and NO2 concentration in ug/m3

(End of measurements useable for trends in 2013. Measurements stopped in 2017).

Attention: only 78% of possible measurements available due to sensor downtime!
Comparison between yearly averages at Diekirch, Luxembourg-Libert and Vianden (rural) [39]; Luxembourg and Vianden values derived by inspection from graphs):


NOx NO2 NO maximum of daily avg. NO2
Diekirch 28    23 5   62
Luxembourg-Libert   ~ 50   ~ 82
Vianden   ~ 10   ~40

The NOx/NO measurements by the AC31M instruments from Environnement SA have been stopped the 30 December 2017. The AC31M has reached its end of life.


see [11] which gives ~30% reduction from 1990 to 2005 for the EU-15 countries.

FINE PARTICLES (PM2.5, PM10) in ug/m3

(new paragraph)

For most of the year 2020 meteoLCD had 3 different LLS particle sensors working, albeit not all for the full year:

The "standard" sensor is the AIRVISUAL from IQAIR (Switzerland) whose PM2.5 readings are continuously uploaded into the iQAIR cloud and are accessible here. The PM10 and PM2.5 readings are also stored in the internal memory of the sensor at a rate of 1 measurement every 15 minutes.
The other two sensors are a Airmaster Pro (AM7) from China and a PurpleAir sensor, model SD II; both store their readings on an internal SD card, and have a RTC (real time clock) which is only set at the start of the sensor. The measuring rate is about every 4 seconds for the AirMaster and about every 20s for the Purpleair (but for reasons unknown one one reading at midnight was stored in the PurpleAir SD card). Look here and here for 2 papers comparing these different instruments and her [59] for a longer blog comment on one year long particle measurements at meteoLCD.

In this trends report we will only use the Airvisual data and those of the official BEIDWEILER station (data downloaded from discomap [60]).

Mean and standard deviations from monthly averages:


PM 2.5 ug/m3 PM10 ug/m3
Diekirch 3.9 +/- 3.9 4..1 +/- 4.13
Beidweiler 6.6 +/- 6.8 19.3 +/- 19.1

The PM 2.5 readings are acceptable close, but the Airvisual PM10 readings are far too low. The Airmaster worked only from April to September. Here the averages for all sensors for these 6 months. The synchronicity of all sensors is good, and the correlation coefficients are all significant at the 95% level.

  Airmaster Airvisual PurpleAir Beidweiler
PM 2.5 7.8 4.0 10.6 6.6
PM 10 8.8 4.2 13.2 19.3

The second graph shows that the Airmaster Pro and PurpleAir PM10 are reasonably close during the April to September period.

All PM readings at Diekirch have been corrected for humidity by dividing the raw readings by the growth-factor GF = 1 + (0.25*RH2)/(1 - RH), where RH is the relative humidity neasured by the internal sensor (RH: 0...1). See [59]


1 Europe's Environment 4th AR (2007) Fig. 2.2.3
2 EPA: Ozone trends.

Jonson et al: Can we explain the trends in European ozone levels? Atmos. Chem. Phys. Discuss., 5, 59575985, 2005.

4 Ozone trends at Uccle
5 Rebetez, Beniston: Analyses of the elevation dependency of correlations between sunshine duration and diurnal temperature range this century in the Swiss Alps. 1998.
6 R.G. Vines, CSIRO: European rainfall patterns. International Journal of Climatology, vol.5, issue 6, p. 607-616.
7 (15 Jan 2009).
8 J.W. Krzyscin, J.L.Borkowski: Total ozone trend over Europe: 1950 - 2004. ACPD, 8, 47-69, 2008.
9 NASA: Solar Physics: The Sunspot Cycle.
10 de Backer et al: (temporarly unavailable)
11 EEA: Emission trends of NOx 1990 - 2005
12 L. Motl: . Dec.2009
13 K. Makovski: The daily temperature amplitude and surface solar radiation..Dissertation for the degree of doctor of sciences. ETHZ 2009.
14 A. Ohmura: Observed long-term variations of solar irradiance at the earth's surface. Space Science Reviews (2006) 125: 111-128
15 J. van Oldenvorgh: Western Europe is warming much faster than expected. Clim.Past. 16Jan.2009
16 Van Malderen, De Backer, Delcloo: Revision of 40 years of ozone measurements in Uccle, Belgium. Poster, EGU2009, Vienna.
17 EEA: Air pollution by ozone across Europe during summer 2009
18 Climate4 you: Global temperature trends
19 Lindzen & Choi: On the determination of climate feedback from ERBE data (GRL, 2009)
20 Scafetta, N.: Empirical analysis of the solar contribution to global mean air surface temperature change. Journal of Atmospheric and Solar-Terrestrial Physics, 2009 (doi:10.1016/j.jastp.2009.07.007)
21 Massen, F., Beck, E. :Accurate estimation of CO2 background level from near ground measurements at non-mixed environments
in: Leal, W., editor: The Economic, Social and Political Elements of Climate Change
Climate Change Management, 2011, Part 4, 509-522. Springer. DOI: 10.1007/978-3-642-14776-0_31
22 De Backer & Van Malderen: Time series of daily erythemal UVB doses at Uccle Belgium. Poster, July 2009.
23 Massen, F.: Sunshine duration from pyranometer readings, 2011
24 Massen F.,  Calculating moist enthalpy from usual meteorological measurements (July 2010) and Calculating moist enthalpy revisited (Sep. 2010)
25 CDIAC: Online Trends
26 UAH MSU data:
27 Vigouroux et al:  Evaluation of ozone trends from g-b FTIR observations. Atmos. Chem. Phys., 8, 68656886, 2008
28 UNEP, Scientific Assessment: Stratospheric Ozone and Surface Ultraviolet Radiation
30 K. Makowski: The daily temperature amplitude and surface solar radiation. Dissertation ETH Zrich #18319, 2009
32 Tim Osborne: NAO data (
33 Helioclim satellite measurements of solar irradiation at
36 Fung: A Hyperventilationg Biosphere (Sep. 2013)

Fraunhofer Institut Stromproduktion :

38 EEA interacive map.
39 Portail de l'Evironnement:
40 Wetterkontor:
41 Amplitude of the atmosphere's seasonal CO2 cycle . CO2science
42 Fung et al. : The changing carbon cycle at Mauna Loa observatory
43 Zeng et al.: Agricultural green revolution as a driver of increasing atmospheric CO2 seasonal amplitude (2014)
44 Zeng: The changing CO2 seasonal cycle (presentation, 2014)
45 Sceptical Science Trend Calculator :
46 Kmpfe, Kowatsch: Winter 2014/15 in Deutschland: Erneut zu mikd - warum ?
47 NOAA Climate Prediction Center:
48 Hohenpeissenberg Klimagase:
49 solar irradiance Uccle: ( this link is defunct:
50 Francis Massen: Ozone, change is the norm!
51 El Nino's and La Nina's years and intensities:
52 Solar cycle progression :
53 Meteo data from all DE and NL stations: (click on Archiv)
54 Massen, F. : First Radiation Amplification factor for 2016
55 Massen.F, Zimmer M.: Comparing the year 2016 Total Ozone Column measurements at Uccle and Diekirch (pdf)
56 MASSEN F, ZIMMER M., THOLL R., HARPES N.: Comparing the year 2017 Total Ozone Column measurements at Uccle and Diekirch (pdf)
57 WDCGG (World Data Center for Greenhouse Gases):  (registration mandatory to access data)
59 Francis Massen: One year of fine particle measurements by Airvisual Pro at meteoLCD (at blog, published 23-Jan-2020)
60 Discomap data retrieval for EEA data:



Addendum 1
2014 update!
Lindzen & Choi [19] define the non-feedback climate sensitivity as ΔT0 = G0*ΔF, where G0 = 0.25 Wm-2 and ΔF is the change in radiative forcing. A change in solar irradiance of -0.82 kWh*m-2y-1  (decade 2005 to 2014) corresponds to ΔF = - 820/8760 = -0.09 Wm-2 and should yield a cooling of  ΔT0 = -0.25*0.09 = -0.02 K (or C).per year. The meteoLCD measurements give a cooling of 0.0057 Ky-1, about 3 times less.

Scafetta [20] defines a climate sensitivity in respect to changes in solar radiation by k1s = ΔT/ΔF and finds k1s = 0.053. Our data for the decade 2005 to 2014 give ΔT/ΔF= - 0.0057/(-0.09) = 0.06, a value close to that of Scafetta!.

Summary for the 2005 to 2014 decade:
meteoLCD (observations) ΔT/ΔF = 0.06 K/(Wm-2y-1) ΔT       = - 0.0057 Ky-1
ΔF       = - 0.09 Wm-2y-1
ΔT/ΔF =   0.06
Lindzen & Choi G0 = 0.25 ΔT = - 0.0200 Ky-1 (computed)
Scafetta k1s  = 0.053 ΔT = -0.0050 Ky-1 (computed)


Addendum 2
2018 update!


It makes for an interesting exercise to compare the influence of mean yearly solar forcing on moist enthalpy and air temperature for the 17 years period 2002 to 2018.

Both air temperature and moist enthalpy are positively correlated to changes in solar forcing ( = mean solar irradiance). The Pearson correlation between mean solar irradiance and moist enthalpy is 0.73 and is significant at the p = 0.05 level, whereas the correlation between mean solar irradiance and temperature is 0.42 (not significant).

A change of 1 Wm-2 of mean solar irradiance would cause a (big!) average heating of 0.5 C per decade and a change of 0.9 kJ/kg of moist enthalpy per decade.

Possibly taking into account some lag (as for instance 4 months for temperature lagging solar forcing) would change these numbers.

Our temperature measurements give a heating of 0.7C/decade for the same period (Findel shows 0.5/decade), which is close to the correlation given if the solar irradiance was the unique warming influence!



Addendum 3
A short analysis of the seasonal CO2 pattern in 2014.

The mean monthly CO2 data show an oscillatory pattern which can be modeled by a 6 month period sine wave. This is not consistent with the commonly admitted explication that the summer lows and winter highs are a fingerprint of changing photosynthesis, which should lead to a single annual sinus wave (as in 2013).

The 6 month period is essentially caused by the low Jan, Feb and Dec values, and is replaced by the usual 12 period if these months are omitted.

The right figure shows the monthly mean CO2 and monthly mean wind speeds. Clearly low wind goes with high CO2, independent of the seasons (significant correlation R = -0.86 !)

The next figure gives the CO2 mixing ratios versus the monthly mean wind speed; the usual exponential model beautifully describes this pattern. The horizontal asymptote of 395.5 ppmV should correspond to the background CO2 level, as shown in [21].

There is some debate about the (global) changes of the seasonal CO2 amplitude, which seems to increase due to global greening [41], agricultural green revolution [43], changing air trans-continental circulation [42] and possibly other unknown factors. Look also at the presentation [44].

Locally it seems that the effects of higher/lower wind speeds and photosynthesis are difficult to untangle. If we restrict our data to those days where the mean wind speed is less than 1, the correlation between CO2 and wind speed is lower (-0.76) but still significant.

Curiously all the papers studying this seasonal amplitude problem seem to ignore the influence of changing wind speeds.



The same analysis for 2015

Here again higher wind speeds usually go together with lower CO2 levels (notice the exception on March!), but the monthly mean values do not follow the usual model well.

If we take all 17520 individual measurements, the picture becomes clearer, and we find that our "bumerang" model follows reasonably well the overall pattern. The horizontal asymptote suggest a background CO2 level of about 389 ppmV, which seems a bit low.



The same analysis for 2016 (wind speed from cup anemometer)

The high wind speeds lower the January , February (and December) values which normally should be higher; so the "usual" sinus pattern with a trough during the summer months is mostly absent.


Using all CO2 measurements of the year, we find again our boomerang pattern; the usual model has a better R2 than in 2015, but the asymptotic value of 383 is definitively too low!


CO2 versus wind speed for 2017 (wind speed by cup anemometer):

The Mauna Loa average CO2 mixing ratio for 2017 is 406.6, which would suggest that our asymptotic value of 392.3 is too low. If we use only the measurements by the new Vaisala GMP343 sensor, the asymptotic value becomes 395.7.


CO2 versus wind speed for 2018 (wind speed by cup anemometer, 17520 data points):

The Mauna Loa average CO2 mixing ratio for 2018 is 408.5, so our asymptotic value of 406 is quite close. The R2 of the model (the goodness of the fit) is also quite acceptable: R2 = 0.50. All parameters are significant at the 5% level (alpha = 0.95).


CO2 versus wind speed for 2019 (wind speed by cup anemometer, 17520 data points):

The Mauna Loa average CO2 mixing ratio for 2019 is 411.44, so our asymptotic value of 411 is practically the same! The R2 of the model (the goodness of the fit) is also quite acceptable: R2 = 0.52. All parameters are significant at the 5% level (alpha = 0.95).


CO2 versus wind speed for 2020 (wind speed by cup anemometer, 17568 data points):

The Mauna Loa average CO2 mixing ratio for 2020 is 414, so our asymptotic value of 417 is very close, keeping in mind that CO2 levels increase slightly with latitude!
The R2 of the model (the goodness of the fit) is also quite acceptable: R2 = 0.56. All parameters are significant at the 5% level (alpha = 0.95).


Addendum 4 
Fine particle measurements at meteoLCD

An Airvisual Pro sensor from iQAir has been operational for the full year 2019. Besides temperature, humidity and CO2, this sensor also measures PM2.5 and PM10 concentrations in ug/m3. All data are stored in a cloud managed by iQAir ( shows only the PM2.5 concentrations; a private dashboard holds all data). The measuring principle for the PM is LLS (Laser Light Scattering). Ambiant air is sucked into the measuring chamber by a small fan running continuously; there is no drying nor correction for pressure variations. It has been found that the most important correction for this type of sensor is the humidity correction, as above 70% RH condensing water on the aerosol particles inflates the count and the reported mass. At meteoLCD we divide the raw data by a growth factor GF =a+(b*RH**2)/(1-RH) with a=1 and b = 0.25

Read the article "One year of fine particle measurements by Airvisual Pro at meteoLCD" on the blog [ref. 59]

The next two plot show the PM2.5 readings per day, together with the corresponding values of the official measuring station at Beidweiler, which uses a Horiba sensor.

The correspondence shown on the first plot is excellent; the second plot suggest to divide the Diekirch data by 0.9572, if Beidweiler is considered as the reference.

The PM10 readings of the Airvisual Pro are very close (too close!) to it's PM2.5, and as such considerably too low if compared to the Beidweiler readings; we have no explanation for this for the moment.








file: meteolcd_trends.html


09 Mar 2020: Update to include 2019 data finished; not all addendum's updated to 2019

04 Apr 2020: Added Addendum 4 with the fine particle measurements of 2019

20 Jan 2021: Started update to 2020 data

05 Feb 2021: Update to 2020 finished.