lagtraj Lagragian simulations trajectories

Producing a Lagrangian forcing

There are three steps to making forcing profiles with lagtraj:

1. Download a domain for a given date-range (small for Eulerian simulations, big for Lagrangian)

2. Produce trajectory

3. Extract forcing profiles along the trajectory

0. Getting started

lagtraj stores everything (source data, definitions for how domains, trajectories and forcings are set up) in a data directory (by default this will be data/ relative to where lagtraj is invoked). The directory structure is as follows:

data
├── domains
│   ├── eurec4a_circle_eul.yaml
│   └── eurec4a_circle_eul_data
│       ├── an_model_2020-01-01.nc
│       :
│       └── fc_single_2020-01-03.nc
├── forcings
│   ├── eure4a_20191209_12_eul.yaml
│   └── eure4a_20191209_12_eul.nc
└── trajectories
    ├── eure4a_20191209_12_eul.yaml
    └── eure4a_20191209_12_eul.nc

You can either make your own domain/forcing/trajectory definition (these are stored in yaml-files) by creating a meta.yaml file in the relevant directory or use the ones that lagtraj comes with by running the following command:

$> python -m lagraj.input_definitions.examples

The following domain/forcing/trajectory definitions are currently included
with lagtraj:

lagtraj://
 ├── domains
 │   ├── eurec4a_north_atlantic
 │   └── eurec4a_circle_eul
 ├── forcings
 │   └── eurec4a_20191209_12_eul
 └── trajectories
     ├── eurec4a_20191209_12_lin
     └── eurec4a_20191209_12_eul


To use for example the `eurec4a_north_atlantic` domain definition
for downloading domain data run lagtraj.domain.download as follows:

    $> python -m lagtraj.domain.download lagtraj://eurec4a_20191209_12_eul 2020/01/01 2020/01/08

1. Making source data available

lagtraj is based around making all data required for interpolation, integration and forcing calculation being available before trajectories are integrated. This was done to reduce the number of data requests required to the data storage backends (e.g. ECMWF), but does mean that the expected extent that a trajectory will reach must been known before performining a trajectory integration, otherwise lagtraj will issue a warning when the edge of the available domain is reached.

Either create your own domain definition in data/domains/<domain_name>.yaml and run

$> python -m lagtraj.domain.download <domain_name> [start date (yyyy-mm-dd)] [end date (yyyy-mm-dd)]

Or use one of the domain definitions included with lagtraj (e.g. eurec4a_north_atlantic

$> python -m lagtraj.domain.download lagtraj://eurec4a_north_atlantic [start date (yyyy-mm-dd)] [end date (yyyy-mm-dd)]

2. Producing a trajectory

Once you have downloaded the required domain data you can either create a trajectory definition in data/trajectories/<trajectory_name>.yaml and run

$> python -m lagtraj.trajectory.create <trajectory_name>

Or use one of the domain definitions included with lagtraj (e.g. eurec4a_20191209_12_lag

$> python -m lagtraj.trajectory.create lagtraj://eurec4a_20191209_12_lag

3. Producing forcing profiles

$> python -m lagtraj.forcing.create <forcing_name>

Implementation details

Required utilies:

• download ECMWF data

• conversion of physical variables

• interpolation (do we need Steffen interpolation?)

• smoothing

• plotting (to some extent)

• generating HighTune formatted output netCDF files

• timing (can use tqdm for this)

Algorithmic approach:

a) Download all data needed (across entire domain, single/model levels+). Currently this creates daily files, but we may aim for the minimum number of requests from ECMWF (However, grib to netcdf conversion via the cds api can only deal with files of up to about 10GB).

b) Split these into daily files

Considerations:

• may need whole domain for plotting

• how many download requests can we make with ECMWF's data server? One big request seems better than many small ones