Mean +/-stdev of
2019: 56.6 +/- 38.4 ug/m³   

Mean +/- stdev (from yearly means)
1998 to 2019: 41.3 +/- 7.8 ug/m³
2002 to 2019: 40.8 +/- 8.6

Trends 1998 - 2018: 38.32+0.28*x (almost flat)
           2002 - 2018: 34.98+0.69*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.

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

181 common day direct sun readings for 2019 at Uccle and Diekirch:
Diekirch                    : 314.20 +/- 38.95
Uccle (Brewer 178): 330.38 +/- 41.24

After 3 years of slight thinning, the TOC is again increasing.

Trendlines (start year is x = 1):
1998 to 2019:                          (+4 DU/decade)
2002 to 2019:                          (- 8 DU/decade )

Uccle  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 178 is the reference: 1.0499 (R² =0.81)




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].

Attention: The instrument for measuring CO2 (API Teledine E600) has been replaced by a Vaisala GMP343 sensor the 27 Jun 2017.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 2019: 434.4 +/- 31.7 ppmV

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

Trends :
2002 to 2019: 1.19 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 6 years, and gives the readings of Diekirch (DIK), Mauna Loa (MLO) and Hohenpeissenberg (HPB) from 2014 to 2019; 2019 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

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 un-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 the yearly amplitude of the sinus fit is 8.48 ppmV (a total swing of ~17 ppmV, to be compared to about 15 ppmV at the HPB station [48]).

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

.

Mean and stdev of the year 2019 (from monthly averages):
Diekirch:  11.96 +/- 6.35 
Findel:     10.68 +/- 6.59   

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

1998 to 2019 :  10.59 +/- 0.73 °C                
2002 to 2019 :  10.69 +/- 0.77  °C
2010 to 2019 :  10.90 +/- 0.96 °C (last decade)

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 2019 (be aware that 2018 was a very strong El Nino year!):
meteoLCD:   +0.070 °C/year
Findel:          +0.048 °C/year           

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

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

DTR = daily max - daily min temperature

Mean and stdev of the year 2019 (from monthly averages):
Diekirch: 9.43 +/- 3.25
Findel:    8.08 +/- 2.55

Mean DTR at Diekirch:
1998 to 2019:  8.63
2002 to 2019:  8.75
2010 to 2019:  8.83  (last decade)

For 1998 to 2018: 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]

Trends  from 2002 to 2018:
(2016 was a very strong El Nino year!):
Diekirch:    +0.095 °C/year since 2002
Findel:       +0.086
Germany:  +0.084   [46]
NAO:         +0.058

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 the strong 2016 El Nino year! [51]); the correlations between all the DJF temperature series and the NAO_DJF normalized index is statistically significant (=0.59) at the 5% level.

Mean moist enthalpy of 2019: 31.98 +/- 15.34 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 2019:  28.65 +/- 1.90 kJ/kg

Trend is slightly positive from 2002 to 2019; the 2016 "monster" El Nino extended into 2017 .

1998 - 2019 mean +/- stdev: 693.9 +/- 129.9 mm 
2002 - 2019 mean +/-stdev:  657.0 +/- 102.4 mm
 

The negative trend from 1998 to 2018 seems spectacular: -82 mm/decade, caused by the very high values of 2000 and 2001.
The 2002-2019 period has a pratically 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 - 2019 a 5.14 years period (~62 months, R² = 0.45); in the model x = 0  corresponds to 2002), with a mean value of 649 mm and an amplitude of 95 mm; the phase shift of 1.04 rad is close to 1/6 period. All these values are close to those of the preceding year. 
[6] 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.

Values of total solar energy of the year 2019:
Diekirch: 1172.6 KWh/m²
 

1998 to 2019 mean +/- stdev: 1115.7 +/- 48.8 kWh/m2
2002 to 2019 mean +/- stdev: 1115.7 +/- 52.8 kWh/m2

Trends from 1998 to 2019 is nearly flat, and relative important (+9.11 kWh*m^-2*y^-1) for last decade. [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]

Values of sunshine hours of the year 2019:
Diekirch:     1776 hours (215m asl) (from pyranometer)
Findel:        1975 hours (365m asl, (from Campbell- Stokes)
Trier:          1935 hours (279m asl) [40] [53] (Petrisberg)
Maastricht: 1926 hours (114m asl) [53]

Very small negative trends:
1998 to 2019:  - 12 hours/decade
2004 to 2019:  - 21 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 (in July and August the excess of Findel readings was highest).

All 4 stations give totals that practically always vary in the same manner (synchronous increase and decrease). But notice how the readings in 2018 between the 2 groups (Diekirch, Maastricht) and (Findel, Trier) are quite different (blue arrow)!

The trend over 2002 - 2019 is slightly positive, the trend line from 1998 to 2019 should be taken with a grain of salt, as the 1998 readings seem abnormally low.
 

.

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.

 

mean +/- stdev:
1999 to 2019: 55.28 +/- 4.29 kWh*m^-2*y^-1
2002 to 2019: 55.55 +/- 4.26
2010 to 2019: 56.22 +/- 4.74  (last decade)

Trend from 2002 - 2018 is nearly flat.
 

^{The 2 independent measures of UVB and UVA doses all point to a very slight increase since 2002, conforming to the increase of total solar irradiance}
(trend is +0.88 Kwh*m^-2*y^-1)

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

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.

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 0.7°C/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!

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.

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 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!

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.

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).

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 (https://airvisual.com/luxembourg/diekirch/diekirch/meteolcd 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 meteolcd.wordpress.com [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.