LWC of CLEO initialisation is too small compared to observations

[nilsnevertree]

The liquid water content used in CLEO is to low compared to the observations by the ATR from the cloud composite dataset.

This is due to the interplay of two things:

1. Overestimation of LWC by the double log-normal fit which we use, becuase we omit 0 values in the fit for larger droplets
2. Underestimation of the LWC when used in CLEO

[nilsnevertree]

There are multiple important things which need to be considered while fitting the cloud composite input data

• The fit needs to be performed on the normalized distributions! Then the real concentration can be derived from by multiplying with the bin width
• For the rain radii, the bins are too small. A coarser bin width reduces variance strongly, but making the measurements differ stronger from the distro meters resolution
• For the initialised SDs, the droplet radii need to be limited to the largest measured droplet! With this, no abnormally large droplets will be collected.

We can find, that a single LnNormal fit gives reasonable results to this. If we consider variance of the data to be $\sigma = 0.01 \cdot y$.

The gamma distributin is harder to fit to the input data, but can give good results for individual clouds.

[nilsnevertree]

Really good results can be seen when using:

• coarsen by 3 from 95e-6 m onwards
• Fit Double LnNormal to MSD
• NOT considering variance.
• r_min as min(radii bins) and r_max as 1.5 * max(radii bins)

Results in :

• correlation : 0.95
• mean difference: 0.03 g m-3
• mean rtol : 15%

Bottom left panel

[nilsnevertree]