FYI, @vishnuthilakan, @Yohanna-Villalobos, @AugustThomasson, @cdgomezo:
I updated the master to include changes I did for my CH4 inversions, in order to create a new LUMIA release (https://github.com/lumia-dev/lumia/releases/tag/egusphere-2024-3122). It might have introduced some incompatibilities with your existing config files:
Emission uncertainties
There's now two options for the uncertainty structure:
- linear: the uncertainty of a state variable $\mathbf{x_i}$ is proportional to the sum of the absolute value of the emissions/fluxes controlled by that variable (e.g., if $\mathbf{x_i}$ is daily, but the fluxes are hourly, then $\mathbf{\varepsilon_{x_i}}$ will be proportional to the sum of the absolute value of the hourly fluxes. This was the normal LUMIA behaviour until now.
- file: the uncertainty structure is directly read from a file.
This needs to be set in the optimize.emissions.{tracer}.{category}.error_structure section of the YAML files:
optimize:
emissions:
tracer_XX:
category_YY:
error_structure:
type: # Should be "linear" or "file"
file: # Needed only if "type" is set to "file" (then it needs to be the path to the correlation file).
annual_uncertainty:
horizontal_correlation:
type:
correlation_length:
temporal_correlation:
type:
correlation_length:
optimization_interval:
optim_scf: # optional, see section below
See https://github.com/lumia-dev/lumia/blob/master/run/egusphere-2024-3122/LUMIA-Lpri.yaml for a full example.
Optimization type
The standard so far in LUMIA is to solve for an offset to the prior emissions. There should now be the possibility to solve for a scaling factor instead. This is controlled by optimize.emissions.optim_scf. It's turned off by default, and it's not 100% tested.
FYI, @vishnuthilakan, @Yohanna-Villalobos, @AugustThomasson, @cdgomezo:
I updated the master to include changes I did for my CH4 inversions, in order to create a new LUMIA release (https://github.com/lumia-dev/lumia/releases/tag/egusphere-2024-3122). It might have introduced some incompatibilities with your existing config files:
Emission uncertainties
There's now two options for the uncertainty structure:
This needs to be set in the
optimize.emissions.{tracer}.{category}.error_structuresection of the YAML files:See https://github.com/lumia-dev/lumia/blob/master/run/egusphere-2024-3122/LUMIA-Lpri.yaml for a full example.
Optimization type
The standard so far in LUMIA is to solve for an offset to the prior emissions. There should now be the possibility to solve for a scaling factor instead. This is controlled by
optimize.emissions.optim_scf. It's turned off by default, and it's not 100% tested.