Parcels

Nevertheless, there are also a few minor updates in this v1.0.5
- Support for the `**` operator to take power in JIT mode (424)
- Fixed a bug where `runtime` was sometimes not correctly set when deferred-loading fields (422)

Note that installation instructions have changed. After creating the `conda` environment, users should now run `python setup.py install`

- Support for `FieldLists` (393), which allows Kernels like `AdvectionRK4` to work on multiple `Fields` at once. See also the [tutorial](http://nbviewer.jupyter.org/github/OceanParcels/parcels/blob/master/parcels/examples/tutorial_FieldLists.ipynb)
- Changing the plotting routines from `Basemap` to `cartopy` (401). Note that this means you will need to uninstall `Basemap` and install `cartopy`, as they are conflicting
- Interpolation of velocities on C-grids. This also means that the angle file is not needed anymore in `FieldSet.from_nemo()` (394)
- Much smaller output files by controlling the NetCDF chunk size (366)
- Displaying a `Progressbar` for long (> 10 seconds) runs of `ParticleSet.execute()` (381 and 418)
- Possibility to initialise custom `Variables` directly in `ParticleSet.from_list()` (397)
- Numerous bug fixes

As always, please let us know if anything isn't working as expected.

**Note that this release has some installation issues, so should not be used. Use v1.0.5 instead**

- Fixed bug in Windows install (355)
- Spatially varying Brownian (340) and exponential variate (333) diffusion
- Added option for two-dimensional histograms in `plotParticleTrajectories()` (367)
- Arguments for `Field.from_netcdf()` have changed to match `FieldSet.from_netcdf()` (364)
- Minor other bugs fixed

- Parcels is now also Python3-compatible
- No need anymore for `FieldSet.advancetime()`, as advancing of time for large datasets is now dealt with under the hood. For datasets that have more than 3 time snapshots, Parcels runs in 'defer_load' mode, where the actual reading of NetCDF files is only performed when required. This means that longer lists of filenames in `FieldSet.from_netcdf()` (and `FieldSet.from_nemo()`) don't require more memory. If you still require the entire dataset to be loaded in one go, you can use `full_load=True` as an option to `FieldSet.from_netcdf()`.
- The`ParticleFile` class now always writes particle data in `array` format. See also http://oceanparcels.org/faq.html#outputformat
- The website has been revamped at http://oceanparcels.org, with repository at https://github.com/OceanParcels/oceanparcels_website

- Changes to the arguments of `ParticleSet.execute()`: (289)
- the `interval` argument has been renamed `outputdt` and should now be set when creating the `ParticleFile` object
- the `show_movie` argument has been split into a `moviedt` argument (to set the frequency of the animation) and a `movie_background_field` argument that determines which background field to show (either a `Field` object, or the string `vector`) or the default `None` for no background field
- `endtime` cannot be a `timedelta` object anymore. Only valid formats for `endtime` are a `datetime` object or a `double`. If you want to give a `timedelta` object, use the `runtime` argument
- Except for `dt`, all arguments controlling intervals should always be positive, regardless of whether you run in forward or backward mode. Hence, to change the direction of a run, the only thing to change is to negate `dt`
- Renaming of old `FieldSet.from_nemo()` to `FieldSet.from_parcels()` and new `FieldSet.from_nemo` to handle Curvilinear NEMO grids (285)
- Adding of a `timer` class to profile CPU time (288)
- Adding of option `ParticleFile(..., write_ondelete=True`) to write only particle data when particle is deleted (290)
- Adding option to write `Kernels` directly as C-functions, for `JITParticles` (278)