First results

[jbusecke]

We have the first results! I encourage everyone to dig into the data!

This notebook should be a good starting point!

[jbusecke]

The full notebook to produce the results in the cloud store can be found here

[paigem]

@jbusecke and I discussed some more results today! 🎉

Current figures:

• air-sea fluxes of CM2.6 dataset
  • 20 years
  • using two aerobulk-python algorithms: ecmwf and ncar

Key observations so far:

Sanity test/validation

The longterm mean fields look very similar across algorithms for all variables. Example of sensible heatflux full long term (20yr) mean. From left to right: ecmwf, ncar, and the difference between the two

Heat/moisture fluxes:

Latent Heat Flux

Evaporation

Sensible Heat Flux

1. Generally, small-scales reinforce the mean fields (i.e. they increase the negative magnitude of the large-scale field)
2. In most locations, the small-scale contributions are very small (a few percent points).
3. There are differences between algorithms, which we will need to explore more. But there are common features of more significant contributions of the small-scales to the large-scale fields in the Western boundary and near-equatorial regions (exceeding 5% and sometimes 10% depending on the variable and algorithm).

Momentum fluxes:

Zonal Momentum Flux

Meridional Momentum Flux

1. The story is a bit more complicated than for the heat fluxes.
2. In most locations, the small-scale contributions are very small (a few percent points).
3. The small-scale contribution is completely opposite in the two algorithms. In ecmwf, the small-scales amplify the large-scale pattern globally, while ncar shows a damping (positive values in large-scale are general negative in the small-scale contribution) of the large-scale pattern globally.

General comments:

• The relative small-scale contribution (4th column) shows bands of high magnitudes which results from dividing by near-zero values, and so should not be interpreted as meaningful.
• The above results generally hold across seasons. There is a large difference in the northern hemisphere, and some variables show marginal differences in the summer and winter times compared to the fall and spring.

Next steps

• Determine if the major differences (particularly in the momentum fluxes) are limited to a single algorithm or model
  • Compute fluxes using more algorithms
  • Compute fluxes on NCAR data
• Assess the influence of absolute vs. relative winds

[dhruvbalwada]

Is the few percent contribution of the small scales when you are considering the 20 year mean flux? What do these fluxes and contributions look like for single time snapshots?

[jbusecke]

I think these will all be superseeded by #47 eventually. Wondering if we should keep the issue open or close?