For the Hgrid element, it would be nice to have a masking feature. This feature is particularly useful when plotting, for example masking the land area. In tri-plot, simply by setting nan value does not work well, and only option seems to be applying a mask.

matplotlib.tri module seems to have some relevant functionalities.

For SCHISM (as well as ADCIRC), the tidal boundary is setup using a list of frequencies. These frequencies are naive, e.g., no idea about the time. The reference time information is provided through the equilibrium argument which is typically computed through tide_fac.f. tide_fac.f program also computes the nodal factors according to the start time and total runtime.

One common way is to use this tide_fac program to compute the nodal arguments, etc., and update a pre-existing bctides.in. Another way is to compute the nodal argument locally, for example using utide, and update the aguments. For any of these cases, it would be nice to have a Tidefac object which contains the computed nodal factors and equilibrium argument.

In addition to that, a method to read output from the tide_fac program could be also useful.

Currently the Hgrid.subset_nodes() is implemented as following, which has an undesired behaviour -

Here, a logical selection is used based on if the index is present in the dataset or not. This results in a ordered indexing of the asked nodes. For example, if nodeid=[1, 4, 3] is passed, it will return the xy in the order of [1, 3, 4]. However, more often then not the nodes are required at their asked ordering, e.g., [1, 4, 3]. Additionally, if a not is out of bound, it will not raise any error under current implementation, which is a misfortune.

The proper implementation should be a direction slicing, like self.xy[nodeid-1, :]. Why -1? Because in hgrid convention the indexing starts from 1, not 0 (python indexing).

After the introduction of the pyschism toolbox by schism-dev, the functionality of this toolbox is mostly available through the official toolbox. jamal919/pyschism repository also releases the pip repo for the schism-dev/pyschism. The logical step at this point is to pivot the toolbox.

I have come up with an idea of rewriting the various parts of the toolbox, particularly the analysis in a functional pattern to be able to continue using the toolbox. Hence, to revamp it is necessary to have a new name, structure of the functions.

Dr. Wu, in SCHISM group, posted a script for acquiring HYCOM data from hycom's DODs server. It would be nice to be implement similar downloader in the webdata module.

https://github.com/wenfanwu/get_hycom_online/blob/main/get_hycom_online.m

JTWC deck files often contains transitions, for example in cyclone Ana (https://www.emc.ncep.noaa.gov/gc_wmb/vxt/DECKS/bsh072022.dat).
The additional fields in the transition line looks something like this - "0, , 0, 0, 0, 0, TRANSITIONED, shC32022 to sh072022, "

The current reader pycaz.cyclone.jtwc.read_jtwc can not read such file.

Xarray DAP access seems to be failing, for not reproducible reason with an "OSError [Errno -73] NetCDF: Malinformed or inaccessible DAP2 DATADDS or DAP4 DAP response". Running the script multiple times, often fix it, but that is a hit or miss. For example, during writing this issue, I had run the script 5 times to get it work finally!

Previously I have experimented with pydap to download the data, and my experience is it always works, or throws Error that are legible. However, one issue with pydap is I could not find a way to download a subset, rather I had to download the full data and make a subset afterwards. Needs further investigation, and potentially a backup backend function based on pydap to download the data if xarry fails a preselected (say 3) times.

Currently in Gr3/Hgrid, the nodes can be accessed through gr3.nodes property. While this can take a slicer, the access is not really nice. More access option is necessary to access only the data, access the lon-lat positions etc.

Namelist file handing, nml update, modify feature using f90nml toilered for using with SCHISM, WWM.

Demonstration notebook for gfs.py and hwrf.py classes, to download web data.

Any JTWC track file with empty line throws an error saying -

Traceback (most recent call last):
  File "build_sflux.py", line 24, in <module>
    track = read_jtwc('track_rsmc.csv')
  File "/home/khan/MEGA/Codes/pyschism/schism/io.py", line 76, in read_jtwc
    ncyclone = int(fields[1].strip())
IndexError: list index out of range

Possible fixes are checking for empty lines and exclude them.

As the extent of the Gr3/Hgrid is often needed to data preparation, it would be nice to have an extent method. The signature could be the following -

def extent(self, buffer: int) -> tuple/named tuple of  (East, West, South, North)

Current data.py provides functionalities to download data from NOAA-GFS DAP server, and NOAA-HWRF model website. These two classes are very different and needs separate location.

In the spirit of updating the code base, it would be nice to break this file into two separate classes, and save them into a module called webdata. Each file inside can handle scraping data from a single source.

The NOAA-HWRF output can be moved to a file called hwrf.py, and GFS can be moved to gfs.py.

On JTWC, radial info values can be zero if the radial info is not defined, or the part of the radial quadrant is on land (set to zero). Current implementation of fradinfo interpolates over theta. Where the radinfo becomes 0, radinfo becomes 0, and thus rmax calculation raises issues.

Fix could be selected interpolation by filling the gaps with non-zero mean value.

Affected file cyclone.py

The current implementation of Gr3 and Hgrid is currently sort of complicated. They are implemented as classes, not data classes. There should be better option to access these datasets, particularly more applicable for functional approach.

In Hgrid, it would be nice to have a Land/Ocean boundary generation feature based on the depth value. Currently the options to generate the boundary segments are either through SMS/ADCIRC during mesh generation, or using xmgredit after mesh generation. matplotlib.tri module could be useful for this purpose.

Readme file needs to be updated before Version 1 released.

Currently the import statements import local functions as local files, e.g., .module notation. It is better to update it so that the imports are typical python module structure, e.g., from pycaz.tide import *

Handling altimetry related data handling got a bit difficult due to various data format changes, e.g., reprocessing and introduction of groups inside netCDF files See here for Jason-3. Particularly I felt it during handling Sentinel-6 data.

Sentinel-3 data is also now available through EUMETSAT data store. Notebooks are needed to explore those too.

The scripts folder contains multiple self-contained files which does only one thing. To kick-start the project, it would be ideal to first organise these scripts into the library itself. The scripts may serve the checkpoints for further developments, or testing units.

To extend the idea further, the scripts might be transformed into notebooks for demonstration purposes.

bctides is the boundary condition file used in SCHISM. It is a mandatory input file.

Similar to the dictionary like object used for Hgrid and Gr3, a similar object structure with helper functions, and useful built-in methods can be added to the library.

Following are the essential feature needed.

• Bctides data object
• read_bctides function
• update_bctides function

SCHISM takes meteorological inputs as so called sflux files.

Sflux has a predefined structure, and a predefined way of naming. The objective of the sflux object would be the following -

• Provide a interface for creating 3 types of sflux object : air, prc, and rad
• They will hold basic sflux related information, but will not process data
• Will provide method for writing the data into files
• Also will provide method to write the sflux_inputs.txt

Current schism/sflux.py is supposed to provide these functionalities, but currently it is not working due to a missing field called 'grid'. Needs re-implementation.

Several imports are missing in schism.py. They are following -

• datetime and timedelta from datetime
• pandas
• netcdf4

Tide filters are convolving array for removing tidal signal from hourly timeseries. These are quite efficient and used by organizations like PSMSL and SONEL to remove tide and compute monthly and yearly means. It would be nice to have these filters into the package.

Bctides class in the old implementatino was initialised with empty list values. It is an extremely dangerous practice, and should never be used. As empty list which is passed as an argument has a pre-defined reference, which means, if the read command is called twice or more time it will append to the OLD referenced empty list (which is not empty after the first call). Should be updated, either by totally removing such impelemtation, or using a copy.deepcopy command when assigning to the class variables.

A dedicated tide module is needed to take into account multi-dimensional functionalities - solving, interpolate, reconstruction etc. The current folder also includes multiple unnecessary files - like schism.py etc that needs to be removed.