Skip to content

Commit

Permalink
Merge pull request #10 from pp-mo/rearrange_sections
Browse files Browse the repository at this point in the history 
Tidy + rearrange existing sections
  • Loading branch information
@pp-mo
pp-mo authored Jan 26, 2023
2 parents 795c4af + 61dc28a commit 3446bcd
Show file tree
Hide file tree
Showing 13 changed files with 2,858 additions and 3,867 deletions.
12 changes: 5 additions & 7 deletions ...book_shadow_copies/Mesh Tutorial Intro.md → ...book_shadow_copies/Mesh_Tutorial_Intro.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -14,13 +14,8 @@ jupyter:

# LFRic + Iris tutorial : Unstructured meshes

<!-- #region tags=[] -->
---

Some important initial setup (always do first) ..
<!-- #endregion -->

## Important : use + stability of notebooks
## Important Preliminary : use + stability of notebooks

A good deal of the content relies on code which is still experimental.
We must expect that there are various outstanding problems, and things sometimes crash.
Expand Down Expand Up @@ -49,6 +44,9 @@ whenever something seems to be taking a long time with no result...
* [01 - Load and Examine some LFRic data](./Sec_01_Load_and_Examine.ipynb)
* [02 - Mesh concepts and Meshes in Iris](./Sec_02_Meshes.ipynb)
* [03 - Plotting and Visualisation](./Sec_03_Plotting.ipynb)
* [04 - Regional Extraction](./Sec_04_RegionExtraction.ipynb)
* [05 - Regridding and UM data comparison](./Sec_05_Regridding.ipynb)



## Reference : terminology (probably, a separate linked glossary ??)
Expand All @@ -65,7 +63,7 @@ whenever something seems to be taking a long time with no result...


<!-- #region -->
# **TOPICS LIST?**
# Work To Do : **TOPICS LIST**

A draft list of topics for discussion.
NOTE : all these basically need re-casting as interactive sections lead by task questions.
Expand Down
143 changes: 62 additions & 81 deletions notebooks/.notebook_shadow_copies/Sec_01_Load_and_Examine.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -14,128 +14,109 @@ jupyter:

# Section 1 : Loading and Examining some LFRic data

Let's dive right in by taking a look at some output file contents.

**TODO: data needs to be available somewhere sensible**

```python tags=[]
from pathlib import Path
datadir = Path('/scratch/sworsley/lfric_data')
datadir.exists()
```
First run some preliminary Python setup, imports etc ...

```python
%ls -1l {datadir}
```

```python
# !ncdump -h /scratch/sworsley/lfric_data/latlon_surface.nc | head -n 100
```
# import the top-level Iris package
import iris

```python
!ncdump -h /scratch/sworsley/lfric_data/20210324T0000Z_lf_ugrid.nc | head -n 100
# import local routines handling access to some test data
from testdata_fetching import lfric_filepth
```

## Relationship to existing Iris usage

Much the same ...
<!-- #region -->
## Iris unstructured loading
Let's dive right in by taking a look at some mesh content.

"Unstructured" data can be loaded from UGRID files (i.e. netCDF files containing a UGRID-style mesh).
This is just like normal Iris loading, except that we must *enable* the interpretion of UGRID content,
roughly like this ...

```python
import iris
iris.FUTURE.datum_support = True # avoids some irritating warnings
```

```python
# From these, grab one UM and one LFRic datafile, which roughly correspond
with PARSE_UGRID_ON_LOAD.context():
cube_list = iris.load(path [, constraints])
# ..and/or..
single_cube = iris.load_cube(path [, constraints])
# ..and/or..
selected_cubes = iris.load_cubes(path, cube_constraints)

um_filepth = datadir / '20210324T0000Z_um_latlon.nc'
lfric_filepth = datadir / '20210324T0000Z_lf_ugrid.nc'
```

<!-- #region tags=[] -->
## Just for reference : some UM data, and what that looks like in Iris ...
**Exercise : first import the `PARSE_UGRID_ON_LOAD` object from iris.experimental.ugrid.load**
<!-- #endregion -->

```python
um_cubes = iris.load(um_filepth)
from iris.experimental.ugrid.load import PARSE_UGRID_ON_LOAD
```

```python
print("n(UM-cubes) = ", len(um_cubes))
print("first 10 cubes ...")
um_cubes[:10]
```
<!-- #region -->
---

```python
um_rh = um_cubes.extract_cube('relative_humidity')
um_rh
```
The variable `lfric_filepath` is already set up, pointing to a suitable test file.

<!-- #region -->
### NOTE: loading a single cube
You could instead load a single cube directly from the file.
```python
um_rh = iris.load_cube(um_filepth, 'relative_humidity')
```
This is in fact rather faster, from a file like this with lots of data-variables (i.e. diagnostics).
<!-- #endregion -->
**Exercise : Load all data from `lfric_filepath`, with the UGRID loading enabled, and print the first 10 cubes.**
Use the plain 'load' method, as shown above.
NOTE : ***expect this to take a few seconds to complete.***

<details><summary>Sample code solution <b>click to reveal</b></summary>

```python
with PARSE_UGRID_ON_LOAD.context():
cubes = iris.load(lfric_)

cubes[:10]
```
</details>
<!-- #endregion -->

## What's in the LFRic files ?
```python
# ... space for user code ...

Let's start with a quick look at a dump of the file
-- but not actually all of it, as there are ***dozens*** of disagnostic variables ...

with PARSE_UGRID_ON_LOAD.context():
cubes = iris.load(lfric_filepth)

```python
!ncdump -h {lfric_filepth} | head -n 120
cubes[:10]
```

The mesh metadata alone can be better viewed using the "ugrid checker" program, which knows how to interpret it ...
( NOTE: this is a public utility, also designed here in AVD : see https://github.com/pp-mo/ugrid-checks#readme )

```python
!ugrid-checker -sqe {lfric_filepth} | head -n 24
```
**NOTEs :**
* putting just `cubes` at the end triggers notebook printing output
* this also means you can click on each cube to "expand" it into a detail view -- try it
* the effect of `print(cubes)` is different -- try it

---
<!-- #region -->
## Loading a single cube
You can instead load a single cube directly from the file.
This is considerably _faster_ in this case, since the whole file contains ~100 data-variables (i.e. diagnostics).

Let's not bother any more with that : Instead, we can load it into Iris which does a reasonable job of interpreting the mesh-structured data.
**Load just the cube named `relative_humidity_at_screen_level`, from the same file, and show that.**
Hint : it's nicer to use the `load_cube` function

<details><summary>Sample code solution <b>click to reveal</b></summary>

```python
# Load the LFRic single datafile

from iris.experimental.ugrid.load import PARSE_UGRID_ON_LOAD
# Note the use of the special context. This is basically because the Iris mesh functionality is still 'experimental'
with PARSE_UGRID_ON_LOAD.context():
lfric_cubes = iris.load(lfric_filepth)
```
lfric_rh = iris.load_cube(lfric_filepth, "relative_humidity_at_screen_level")

```python
print("n(LFRic-cubes) = ", len(lfric_cubes))
print("first 10 cubes ...")
lfric_cubes[:10]
lfric_rh
```
---

**NOTEs :**
* putting just `cubes` at the end triggers notebook printing output
* the effect of `print(cubes)` is different -- try it
</details>
<!-- #endregion -->

```python
lfric_rh = lfric_cubes.extract_cube('relative_humidity_at_screen_level')
with PARSE_UGRID_ON_LOAD.context():
lfric_rh = iris.load_cube(lfric_filepth, "relative_humidity_at_screen_level")

lfric_rh
```

---
Or, just to show a faster selective loading ...

```python
with PARSE_UGRID_ON_LOAD.context():
lfric_rh = iris.load_cube(lfric_filepth, 'relative_humidity_at_screen_level')

lfric_rh
```

## What you initially notice about "mesh cubes"
Expand Down Expand Up @@ -173,14 +154,14 @@ print("cube.mesh_dim() = ", lfric_rh.mesh_dim())
<br>

```python
#-------------------------------
# Utility Function
###-------------------------------
### Utility Function
#
def is_meshcube(cube):
return cube.mesh is not None

#-------------------------------
# Testing ...
### Testing ...
#
from iris.tests.stock import realistic_3d
nonmesh_cube = realistic_3d()
Expand Down
Loading

0 comments on commit 3446bcd

Please sign in to comment.