The second important parameter used in the calculation of the energy balance is the downward longwave irradiance. This parameter is produced by all SAFs, it is investigated in the same way with the same indicators than the surface solar irradiance.
It is not possible to apply a simple and reliable quality control on the downward longwave irradiance like for the shortwave irradiance, where clear sky conditions form a physical upper limit. Nevertheless, the CFIsat method [Dürr 2006] described in a previous section can be applied to evaluate the quality of the measurements. The evaluation of the CFIsat parameter needs the knowledge of the dry bulb temperature and the water vapour con-tent of the atmosphere. Therefore, it can only be applied on 4 of the 8 data sets. An example is given on Figure 29 for the data from Payerne. Following the method from Dürr, clear sky conditions should have a CFIsat value below 0%, and cloudy conditions higher then 50%. This quality check pointed out some measurement artifacts on the Figure 28 Surface solar irradiance model-measurements mean bias difference for the 4 months and the 3 SAFs, as average over all stations.
University of Geneva - 21 - Pierre Ineichen Table IV Statistical comparison of the SAFs products with the ground measurements, by sta-tions, SAFs, all stations average, all SAFs average and overall comparison.
ground average 329 330 330 329
nb 2786 2296 2784 7866
abs. difference 17 20 19 19
MBD -15 -13 -10 -13
RMSD 23 27 23 24
ground average 328 327 329 328
nb 2472 4364 2539 9375
abs. difference 22 23 21 22
MBD 5 2 3 3
RMSD 28 33 27 30
ground average 325 323 325 324
nb 2636 2211 2665 7512
abs. difference 16 21 19 19
MBD -15 -18 -13 -15
ground average 309 308 310 309
nb 2760 2238 2736 7734
abs. difference 15 16 19 17
MBD -2 -2 5 0
ground average 314 315 314
nb 2290 2739 5029
abs. difference 25 23 24
MBD -15 -15 -15
RMSD 28 28 28
ground average 307 307
nb 3814 3814
abs. difference 24 24
MBD -11 -11
RMSD 31 31
ground average 322 318 322 320
nb 10654 17213 13463 41330
abs. difference 18 22 20 20
MBD -7 -8 -6 -7
RMSD 23 29 26 27
land SAF
Downward longwave irradiance [W/m2]
osi SAF climate SAF all SAFs
All months All sky conditions
Camborne 50.22°N
5.32°W
Solar elevation >= 10°
Lyon
University of Geneva - 22 - Pierre Ineichen Figure 30 Modelled downward longwave irradiance versus the corresponding ground measurements for the sta-tion of Cabauw.
Figure 31 Relative frequency of occurence of the downward longwave irradiance at the station of Carpentras and for the land SAF.
Figure 29 Cloud Free Index saturation (CFIsat) calculated from ground measurements versus clearness index for the sta-tion of Payerne.
University of Geneva - 23 - Pierre Ineichen for the station of Camborne and the land SAF.
Figure 34 Downward longwave irradiance model-measurements mean bias difference versus the solar elevation angle for the sta-tion of Camborne and the land SAF.
Figure 33 Downward longwave irradiance model-measurements mean bias difference versus the cloud cover for the station of Car-pentras and the climate SAF.
University of Geneva - 24 - Pierre Ineichen Figure 35 Downward longwave irradiance model-measurements mean bias difference versus the cloud top height for the station of Toravere.
data from Cabauw, where not all the data were acquired with a shaded pyrgeometer. A correction was then applied (see Appendix).
The observation of the model-measurements scatter plots for the downward longwave irradiance show a general tendency to underestimate this parameter (Figure 9.1 to a-9.8) as it can be seen for example on Figure 30 for the measurements acquired in Cabauw. It can also be outlined from the Figures in the annexe that there are some strange patterns (i.e. Figure a-9.2) that cannot be explained or isolated by a parameter dependance study. These can be attributed either to ground measurements imprecisions or to model biases (or both).
The model tendency to underestimate is corroborated by the relative frequency of occurence illustrated on Figure 31, where it can be seen that the shape of the model is correctly reproduced but shifted to the left. The complete set of Figures is given in the appendix (Figure a-10.1 to a-10.8).
The validation statistics are given on Table IV for all the stations, SAFs and overall perfor-mance. The bias seems to be station dependent but is less than 5%. There are no differences between nightime and daytime statistics as illustrated in the appendix on Charts a-1.5.
The model dependence with geometric and climatic parameters is illustrated on Figure 32 to 34, where no particular effect could be pointed out. If specific patterns are present (like for example on Figure 34 for the station of Camborne with the solar elevation angle), these are not systematic for all the stations or SAFs, neither dependent on the classification based on the sky conditions as shown in the appendix on Chart 1.3 and a-1.4. These effects are certainly artifacts (it has to be noted that the classification for different sky conditions takes into account only daytime values).
On the other hand, a systematic dependence with the cloud top height is clearly present for all the stations. An example is given on Figure 35 for the station of Toravere. But here again, if the trend is present for all stations, it is difficult to quantify because of the dispersion. It has also to be noted that the cloud top height is not a ground measurements but also a satellite product.
In conclusion, despite the fact that a quality control is difficult to apply and that some strange pattern are present on the graphs, the modelled downward longwave irradiance
University of Geneva - 25 - Pierre Ineichen is of good quality, with a small negative bias and a root mean square difference of less than 8%.
Except a slight dependence with the cloud top height, no particular effect was found. If some effects can be seen on specific graphs, they are station or SAF dependent.