Some more analysis of u/d lwir and clouds

Total / opaque cloud vs Temperature
No slope on the opaque cloud but a definite dip when teperatures are between 16 and 22C
The following plots limit RH to 20 to 40%. Day refers to time that cloud can be measured Night to when cloud is not measured.

D/U LWIR vs Temperature (night values - 0-100% cloud)

Both upward and downward LWIR linear proportional to temperature

D/U LWIR vs Temperature (day values cloud 0-100%)
Very similar to night - slopes are a bit different.

D/U LWIR vs Temperature (day values but limiting cloud to 20 to 40%) (note change in humidity limits

D/U LWIR vs Humidity Temp 22-24C cloud 40-50%
By constraining the temperature to a 2C band The temperature effects on IR are minimised whilst still returning a reasonable number  of results. Note that the ULWIR falls with increasing cloud but the DLWIR rises by 100w/sq m 

D/U LWIR vs Opaque Cloud cover temp 22-24C RH 35-40%
Temperature RH are constrained to minimise these effects. The DLWIR increases by approx 80 w/sqm
I think these last two plots conclusively prove relative humidity and cloud cover have a positive effect on the downward long wave ir (ir increases if cloud and/or RH increase)
Now how do you do this for CO2?

Data available from:



  1. Re: Now how do you do this for CO2?

    Can you do multiple MODTRAN runs for various levels of CO2 in your atmosphere, with your zenith angle set to either 0 or 180 degrees (I think the convention is zenith angle), using 8-12 microns (or whatever) for your band?

    The varied CO2 levels correspond to the opaque cloud cover, the two zenith angles give downward and upward respectively (if I have correctly remembered the convention).

    I am no expert on MODTRAN, but this seems like it would be a tractable modeling approach.

  2. Unfortunately the data available from nrel does not give the dlwir ulwir for specific wavelengths - an overall w/sqm figure is all that is available.

    Available data is shown here

    There are UV bands but not LW IR

    The idea with these plots was to remove "models" so hated by some. So it was just the measured values of interest.