Merge branch 'rem_header' into 'main'

remove comment See merge request !78

Merge branch 'rem_header' into 'main'
7e896e78 · Snyder, Amelia Marie · 0a172b75 · c9e58582 · 7e896e78
Commit 7e896e78 authored 11 months ago by Snyder, Amelia Marie
--- a/workflows/archive/cprep_create_collection_from_zarr.ipynb
+++ b/workflows/archive/cprep_create_collection_from_zarr.ipynb
@@ -8,7 +8,6 @@
   },
   "source": [
    "# C-PrEP Zarr -> Collection Workflow\n",
-    "## In progress - figuring out how to label bnds dim of dataset\n",
    "\n",
    "This is a workflow to build [STAC collections](https://github.com/radiantearth/stac-spec/blob/master/collection-spec/collection-spec.md) from the zarr assets for the dataset named above. We use the [datacube extension](https://github.com/stac-extensions/datacube) to define the spatial and temporal dimensions of the zarr store, as well as the variables it contains.\n",
    "\n",

 %% Cell type:markdown id:6c10e07b-1e60-4926-af1d-fa75dc78e5d4 tags:

 # C-PrEP Zarr -> Collection Workflow
-## In progress - figuring out how to label bnds dim of dataset

 This is a workflow to build [STAC collections](https://github.com/radiantearth/stac-spec/blob/master/collection-spec/collection-spec.md) from the zarr assets for the dataset named above. We use the [datacube extension](https://github.com/stac-extensions/datacube) to define the spatial and temporal dimensions of the zarr store, as well as the variables it contains.

 To simplify this workflow so that it can scale to many datasets, a few simplifying suggestions and assumptions are made:
 1. For USGS data, we can use the CC0-1.0 license. For all other data we can use Unlicense. Ref: https://spdx.org/licenses/
 2. I am assuming all coordinates are from the WGS84 datum if not specified.

 %% Cell type:code id:201e0945-de55-45ff-b095-c2af009a4e62 tags:

 ``` python
 import pystac
 from pystac.extensions.datacube import CollectionDatacubeExtension, AssetDatacubeExtension, AdditionalDimension, DatacubeExtension
 import xarray as xr
 import cf_xarray
 import os
 import fsspec
 import cf_xarray
 import hvplot.xarray
 import pandas as pd
 import json
 import numpy as np
 import pyproj
 from pyproj import Transformer
 import cartopy.crs as ccrs
 import cfunits
 import json
 import sys
 sys.path.insert(1, '..')
 import stac_helpers
 ```

 %% Cell type:markdown id:9e9cff62-0e60-4e59-8994-aece90c2bbc6 tags:

 ## Explore list of files to incorporate

 %% Cell type:code id:0bfe4a5b-72da-4f0a-93f8-5c899cc2602b tags:

 ``` python
 # get list of zarr stores to create collections for - we will run through one
 # as an example, then iterate through every item in the list
 fs = fsspec.filesystem('s3', requester_pays=True)
 s3_path_list = fs.ls('s3://nhgf-development/workspace/DataConversion/gdp/cprep')
 zarr_name_list = [fp.split('/')[-1] for fp in s3_path_list]
 print(f'NHGF S3 Endpoint\n')
 print(f'zarr stores found: {len(zarr_name_list)}\n')
 print(f'example file path: {s3_path_list[0]}\n')
 print(f'example file name: {zarr_name_list[0]}')
 ```

 %% Cell type:code id:eee2188c-1e17-4d94-ad37-cb791f701852 tags:

 ``` python
 fs2 = fsspec.filesystem('s3', anon=True, endpoint_url='https://usgs.osn.mghpcc.org/')
 osn_path_list = fs2.ls('s3://mdmf/gdp/cprep')
 zarr_list = [fp.split('/')[-1] for fp in fs2.ls('s3://mdmf/gdp/cprep')]
 print(f'OSN Pod Endpoint\n')
 print(f'zarr stores found: {len(zarr_list)}\n')
 print(f'example file path: {osn_path_list[0]}\n')
 print(f'example file name: {zarr_list[0]}')
 ```

 %% Cell type:markdown id:d412ee91-7d5b-494f-9c57-61ee69f5261b tags:

 ## Test an example with one zarr from the list of 297

 %% Cell type:markdown id:67ca098d-b64e-4379-b4db-c55cbf558dac tags:

 ## Collection ID

 %% Cell type:code id:0ed579f2-9be9-4871-bac0-6e589d781c40 tags:

 ``` python
 # name for STAC collection
 # most of them look like this one
 collection_id = os.path.splitext(zarr_list[100])[0]
 # this one fails on decode times
 collection_id = "cprep_tasmax_day_I35txdetrp1-DeltaSD-A38P01K00_rcp85_r1i1p1_I35Land_20060101-20991231"
 ```

 %% Cell type:markdown id:116b5837-8e85-4ae7-964a-803533ded714 tags:

 ## Asset Metadata Input

 %% Cell type:code id:dd6fa323-132a-4794-8c80-576933f547a0 tags:

 ``` python
 # url to zarr store that you want to create a collection for
 zarr_url = f's3://mdmf/gdp/cprep/{collection_id}.zarr/'

 # define keyword arguments needed for opening the dataset with xarray
 # ref: https://github.com/stac-extensions/xarray-assets
 xarray_opendataset_kwargs = {"xarray:open_kwargs":{"chunks":{},"engine":"zarr","consolidated":True},
                          "xarray:storage_options": {"anon": True, "client_kwargs": {"endpoint_url":"https://usgs.osn.mghpcc.org/"}}}
 # description for zarr url asset attached to collection (zarr_url)
 asset_description = "Open Storage Network Pod S3 API access to collection zarr group"
 # roles to tag zarr url asset with
 asset_roles = ["data","zarr","s3"]
 ```

 %% Cell type:code id:e1441cd4-e94c-4902-af46-8f1af470eb6b tags:

 ``` python
 # url to zarr store that you want to create a collection for
 zarr_url2 = f's3://nhgf-development/workspace/DataConversion/gdp/cprep/{collection_id}.zarr/'

 # define keyword arguments needed for opening the dataset with xarray
 # ref: https://github.com/stac-extensions/xarray-assets
 xarray_opendataset_kwargs2 = {"xarray:open_kwargs":{"chunks":{},"engine":"zarr","consolidated":True},
                          "xarray:storage_options":{"requester_pays":True}}
 # description for zarr url asset attached to collection (zarr_url)
 asset_description2 = "S3 access to collection zarr group"
 # roles to tag zarr url asset with
 asset_roles2 = ["data","zarr","s3"]
 ```

 %% Cell type:markdown id:b213b74f-ad17-4774-93b6-3b62be616b45 tags:

 ## Data Exploration

 %% Cell type:code id:b5ee1874-d280-4158-8d56-ecab750b6000 tags:

 ``` python
 # open and view zarr dataset
 fs2 = fsspec.filesystem('s3', anon=True, endpoint_url='https://usgs.osn.mghpcc.org/')
 # ds = xr.open_dataset(fs2.get_mapper(zarr_url), engine='zarr',
 #                              backend_kwargs={'consolidated':True}, chunks={})
 ds = xr.open_dataset(fs2.get_mapper(zarr_url), engine='zarr',
                             backend_kwargs={'consolidated':True}, chunks={}, decode_times=False)
 ds
 ```

 %% Cell type:markdown id:996e60ba-13e4-453a-8534-e62ce747f0fa tags:

 ## Collection Metadata Input

 %% Cell type:code id:482d204d-b5b6-40e5-ac42-55b459be1097 tags:

 ``` python
 # description of STAC collection
 collection_description = ds.attrs['title']
 print(f'collection description: {collection_description}')
 ```

 %% Cell type:code id:df4fdc77-72f5-4cb1-afeb-b7b879d5ede1 tags:

 ``` python
 # license for dataset
 collection_license = stac_helpers.license_picker(ds.attrs['license'])
 ```

 %% Cell type:markdown id:0bc7e9b3-ad62-4b10-a18e-66b7ed2d35dc tags:

 ## Identify x, y, t dimensions of dataset
 May require user input if dimensions cannot be auto-detected.

 %% Cell type:code id:ab91268f-7200-4cb1-979a-c7d75531d2c0 tags:

 ``` python
 dims_auto_extract = ['X', 'Y', 'T']
 dim_names_dict = {}
 for d in dims_auto_extract:
    dim_names_dict[d] = stac_helpers.extract_dim(ds, d)
 print(f"Dimension dictionary: {dim_names_dict}")
 ```

 %% Cell type:markdown id:667cb9e8-c4a5-4d26-8e30-be8c934adc37 tags:

 ## Get crs info

 %% Cell type:code id:4a981eca-93fd-4a9d-bbdf-52d0a99d6bc0 tags:

 ``` python
 crs_var = 'crs'
 ```

 %% Cell type:code id:172d9e6b-de88-4e3a-9184-706110a99659 tags:

 ``` python
 # use pyproj to automatically extract crs info
 crs = pyproj.CRS.from_cf(ds[crs_var].attrs)

 # alternatively, create the appropriate cartopy projection
 # crs = ccrs.LambertConformal(central_longitude=crs_info.longitude_of_central_meridian,
 #                             central_latitude=crs_info.latitude_of_projection_origin,
 #                             standard_parallels=crs_info.standard_parallel)
 ```

 %% Cell type:code id:1a0f2a58-a126-4438-a4c2-5c35653f4a62 tags:

 ``` python
 ds[crs_var]
 ```

 %% Cell type:code id:a702f7ff-6201-4283-8e8c-c23f685d7165 tags:

 ``` python
 crs.to_proj4()
 ```

 %% Cell type:markdown id:a8c3ed37-8564-400b-a7fb-25bd5e43d21c tags:

 ## Create Collection Extent

 %% Cell type:markdown id:69f0d837-68a5-4fed-9a14-5d75cfbb0da4 tags:

 ### Spatial Extent

 %% Cell type:code id:d46805e0-8e94-4ebe-aa01-d9a2d7051459 tags:

 ``` python
 # pull out lat/lon bbox for data
 # coordinates must be from WGS 84 datum
 # left, bottom, right, top

 # Note: try changing around the commented out lines below to get type float rather than a numpy float
 #spatial_bounds = [ds[dim_names_dict['X']].data.min().compute().astype(float), ds[dim_names_dict['Y']].data.min().compute().astype(float), ds[dim_names_dict['X']].data.max().compute().astype(float), ds[dim_names_dict['Y']].data.max().compute().astype(float)]
 #spatial_bounds = [ds[dim_names_dict['X']].data.min().compute().astype(float).tolist(), ds[dim_names_dict['Y']].data.min().compute().astype(float).tolist(), ds[dim_names_dict['X']].data.max().compute().astype(float).tolist(), ds[dim_names_dict['Y']].data.max().compute().astype(float).tolist()]
 spatial_bounds = [ds[dim_names_dict['X']].data.min().astype(float).item(), ds[dim_names_dict['Y']].data.min().astype(float).item(), ds[dim_names_dict['X']].data.max().astype(float).item(), ds[dim_names_dict['Y']].data.max().astype(float).item()]
 print(spatial_bounds)
 print(f'\nspatial_bounds data type: {type(spatial_bounds[0])}')
 ```

 %% Cell type:code id:a9febf0f-d299-4758-a2a8-abed7328698d tags:

 ``` python
 XX, YY = np.meshgrid(ds[dim_names_dict['X']].data, ds[dim_names_dict['Y']].data)
 ```

 %% Cell type:code id:b0846798-c968-4db2-9505-3078327dd358 tags:

 ``` python
 transformer = Transformer.from_crs(crs, "EPSG:4326", always_xy=True)
 lon, lat = transformer.transform(XX.ravel(), YY.ravel())
 ```

 %% Cell type:code id:680aeb26-51cb-468d-8bda-b6e6cc7af611 tags:

 ``` python
 print(f'lower left coordinates (WGS84): {min(lon)}, {min(lat)}')
 print(f'upper right coordinates (WGS84): {max(lon)}, {max(lat)}')
 ```

 %% Cell type:code id:d72e37ab-2b9d-4a15-a502-be79622dd7cb tags:

 ``` python
 # create a spatial extent object
 spatial_extent = pystac.SpatialExtent(bboxes=[[min(lon).item(), min(lat).item(), max(lon).item(), max(lat).item()]])
 ```

 %% Cell type:markdown id:a04c8fca-1d33-43ac-9e2b-62d7be2887f7 tags:

 ### Temporal Extent

 %% Cell type:code id:41a84995-867c-4152-8c57-85e3758bbb77 tags:

 ``` python
 # pull out first and last timestamps
 temporal_extent_lower = pd.Timestamp(ds[dim_names_dict['T']].data.min())
 temporal_extent_upper = pd.Timestamp(ds[dim_names_dict['T']].data.max())
 # temporal_extent_lower = ds.indexes[dim_names_dict['T']].to_datetimeindex().min()
 # temporal_extent_upper = ds.indexes[dim_names_dict['T']].to_datetimeindex().max()
 print(f'min: {temporal_extent_lower} \nmax: {temporal_extent_upper}')
 # create a temporal extent object
 temporal_extent = pystac.TemporalExtent(intervals=[[temporal_extent_lower, temporal_extent_upper]])
 ```

 %% Cell type:code id:1b1e37c4-5348-46ad-abc9-e005b5d6c02b tags:

 ``` python
 collection_extent = pystac.Extent(spatial=spatial_extent, temporal=temporal_extent)
 ```

 %% Cell type:markdown id:20b00e88-5a13-46b3-9787-d9ac2d4e7bd6 tags:

 ## Open up NHGF STAC Catalog and create a collection

 %% Cell type:code id:adf6c59d-58cd-48b1-a5fd-3bb205a3ef56 tags:

 ``` python
 # define folder location where your STAC catalog json file is
 catalog_path = os.path.join('..', '..', 'catalog')
 # open catalog
 catalog = pystac.Catalog.from_file(os.path.join(catalog_path, 'catalog.json'))
 ```

 %% Cell type:code id:e56853f8-1cb1-46ea-b6ee-5c9295b20311 tags:

 ``` python
 #create cprep overall collection
 # note: each dataset has the same spatial extent, so the extent extracted from one dataset is used here
 # 241 ocurrences:
 # min: 2006-01-01 12:00:00
 # max: 2099-12-31 12:00:00
 # 54 occurences:
 # min: 1981-01-01 12:00:00
 # max: 2005-12-31 12:00:00
 # 2 occurrences
 # min: 1970-01-01 00:00:00.000016425
 # max: 1983-11-03 21:13:54.048972992
 overall_temporal_extent= pystac.TemporalExtent(intervals=[[pd.Timestamp('1970-01-01 00:00:00.000016425'), pd.Timestamp('2099-12-31 12:00:00')]])
 overall_collection_extent = pystac.Extent(spatial=spatial_extent, temporal=overall_temporal_extent)
 overall_collection_id = 'cprep'
 collection_extent
 if catalog.get_child(overall_collection_id):
    cprep_collection = catalog.get_child(overall_collection_id)
    print("existing collection opened")
    cprep_collection.extent=overall_collection_extent
    cprep_collection.description=collection_description
    cprep_collection.license=collection_license
 else:
    cprep_collection = pystac.Collection(id=overall_collection_id,
                                   description=collection_description,
                                   extent=overall_collection_extent,
                                   license=collection_license)
    print("new collection created")
 ```

 %% Cell type:code id:f039d8aa-f385-4660-a662-dfe363cc25b2 tags:

 ``` python
 if catalog.get_child(overall_collection_id):
    cprep_collection.normalize_and_save(root_href=os.path.join(catalog_path, overall_collection_id), catalog_type=pystac.CatalogType.SELF_CONTAINED)
 else:
    catalog.add_child(cprep_collection)
    catalog.normalize_and_save(root_href=catalog_path, catalog_type=pystac.CatalogType.SELF_CONTAINED)
 ```

 %% Cell type:code id:7e96811b-95ae-406a-9728-55fc429d4e1f tags:

 ``` python
 # create collection for current zarr
 if cprep_collection.get_child(collection_id):
    collection = catalog.get_child(collection_id)
    print("existing collection opened")
    collection.extent=collection_extent
    collection.description=collection_description
    collection.license=collection_license
 else:
    collection = pystac.Collection(id=collection_id,
                                   description=collection_description,
                                   extent=collection_extent,
                                   license=collection_license)
    print("new collection created")
 ```

 %% Cell type:markdown id:a21c76e8-cd57-4eb5-a33f-7c668a3b3205 tags:

 ## Add zarr url asset to collection

 %% Cell type:code id:094832af-d22b-4359-b0f6-cf687acce5cc tags:

 ``` python
 asset_id = "zarr-s3-osn"
 asset = pystac.Asset(href=zarr_url,
                     description=asset_description,
                     media_type="application/vnd+zarr",
                     roles=asset_roles,
                     extra_fields = xarray_opendataset_kwargs)
 collection.add_asset(asset_id, asset)
 ```

 %% Cell type:code id:0c298d07-f234-4a08-986d-87f4a39e9ae6 tags:

 ``` python
 asset_id2 = "zarr-s3"
 asset2 = pystac.Asset(href=zarr_url2,
                     description=asset_description2,
                     media_type="application/vnd+zarr",
                     roles=asset_roles2,
                     extra_fields = xarray_opendataset_kwargs2)
 collection.add_asset(asset_id2, asset2)
 ```

 %% Cell type:markdown id:f67cd5c9-db33-45c2-bc21-480cd67354f4 tags:

 ## Add datacube extension to collection

 %% Cell type:code id:fc00946d-2880-491d-9b3b-3aeeb4414d6c tags:

 ``` python
 # instantiate extention on collection
 dc = DatacubeExtension.ext(collection, add_if_missing=True)
 ```

 %% Cell type:markdown id:8bdd77a2-7587-485e-afb7-42af3a822241 tags:

 ### Add cube dimensions (required field for extension)

 %% Cell type:code id:120a4914-3302-44a5-a282-0308ac84f040 tags:

 ``` python
 # list out dataset dimensions
 # When writing data to Zarr, Xarray sets this attribute on all variables based on the variable dimensions. When reading a Zarr group, Xarray looks for this attribute on all arrays,
 # raising an error if it can’t be found.
 dims = list(ds.dims)
 print(dims)
 ```

 %% Cell type:markdown id:e7dc357c-91ec-49ae-83e5-400f791f9792 tags:

 #### user input needed - you will need to look at the crs information and create a cartopy crs object after identifying the projection type:
 reference list of cartopy projections: https://scitools.org.uk/cartopy/docs/latest/reference/projections.html

 %% Cell type:code id:1b1d05ff-8e43-44a7-8343-178b112c4ad6 tags:

 ``` python
 # the datacube extension can accept reference_system information as a numerical EPSG code,
 # WKT2 (ISO 19162) string or PROJJSON object.
 # we will use a projjson, as was done by Microsoft Planetary Computer here:
 # https://planetarycomputer.microsoft.com/dataset/daymet-annual-na
 # https://planetarycomputer.microsoft.com/api/stac/v1/collections/daymet-annual-na
 projjson = crs.to_json()
 print(projjson)
 ```

 %% Cell type:markdown id:b98e8e4d-91f5-4b42-8bc4-adc4ebcad264 tags:

 #### user review needed - looks at the steps pulled out and make sure they make sense

 %% Cell type:markdown id:0930906a-6dda-4ef6-b7c8-f3cd26f4bc56 tags:

 **Time**

 %% Cell type:code id:498fc9be-8e51-417b-accd-a5460792f760 tags:

 ``` python
 time_step = pd.Timedelta(stac_helpers.get_step(ds, dim_names_dict['T'], time_dim=True)).isoformat()
 if time_step=='NaT' and ((ds.isel(time=1).time - ds.isel(time=0).time)==1) and (ds.time.units=='days since 1961-01-01 00:00'):
    time_step="P1DT0H0M0S"
 print(f'time step: {time_step}')
 ```

 %% Cell type:code id:990fe69c-1073-44e7-b429-64dde28ab5a5 tags:

 ``` python
 # # debugging for time steps: get all step values and locations
 # time_step = stac_helpers.get_step(ds, dim_names_dict['T'], time_dim=True, debug=True, step_ix=1)
 ```

 %% Cell type:code id:79f23b28-289d-46d2-aa77-f965e99db9cc tags:

 ``` python
 # # debugging for time steps, cont:
 # # please choose one of the index locations printed above
 # # this will print the time steps adjacent to it
 # ix = 3343
 # ds.isel(time=slice(ix-1,ix+3)).time
 ```

 %% Cell type:markdown id:66ef74eb-66d4-432f-be4c-56818579e04a tags:

 **X/lon**

 %% Cell type:markdown id:354bdf96-b5f6-4949-90b7-0c762c70fb78 tags:

 **rounding error in spatial steps**: need to round to 13th decimal to take care of rounding error that comes up in calculating spatial steps

 %% Cell type:code id:3f903b3a-c37d-45b1-ab8d-e89516fbc511 tags:

 ``` python
 x_step = stac_helpers.get_step(ds, dim_names_dict['X'], round_dec=13)
 print(f'x step: {x_step}')
 ```

 %% Cell type:code id:118f7c27-e07b-43a7-a0b5-9f4adcdddbbd tags:

 ``` python
 # # debugging for spatial steps: get all step values and locations
 # x_dim=dim_names_dict['X']
 # x_step = stac_helpers.get_step(ds, x_dim, debug=True, step_ix=1)
 # print(f'\nx dim name (for next cell): {x_dim}')
 ```

 %% Cell type:code id:ce23eee0-8d5f-4e8a-b6fa-13ab276eb660 tags:

 ``` python
 # # debugging for spatial steps, cont:
 # # please choose one of the index locations printed above
 # # this will print the time steps adjacent to it
 # ix = 5
 # ds.isel(lon=slice(ix-1,ix+3)).lon
 ```

 %% Cell type:markdown id:f28d25eb-c43d-4a98-8350-8faf517a90c5 tags:

 **Y/lat**

 %% Cell type:markdown id:f199759e-06c0-4cb0-b821-5f006a837dc7 tags:

 **rounding error in spatial steps**: need to round to 13th decimal to take care of rounding error that comes up in calculating spatial steps

 %% Cell type:code id:0c0b0938-a6ee-4023-b201-75712029f2b5 tags:

 ``` python
 y_step = stac_helpers.get_step(ds, dim_names_dict['Y'], round_dec=13)
 print(f'y step: {y_step}')
 ```

 %% Cell type:code id:8a8ba920-0959-49bf-aed2-08d56c511b2c tags:

 ``` python
 # # debugging for spatial steps: get all step values and locations
 # y_dim=dim_names_dict['Y']
 # y_step = stac_helpers.get_step(ds, y_dim, debug=True, step_ix=1)
 # print(f'\nx dim name (for next cell): {x_dim}')
 ```

 %% Cell type:code id:16a23616-8309-4855-ba75-fcd04704d5e6 tags:

 ``` python
 # # debugging for spatial steps, cont:
 # # please choose one of the index locations printed above
 # # this will print the time steps adjacent to it
 # ix = 5
 # ds.isel(lat=slice(ix-1,ix+3)).lat
 ```

 %% Cell type:markdown id:00a5e041-081d-428d-ac2e-75d16de205e6 tags:

 #### user input needed - you will need to copy all of the dimensions printed below into the dict and fill in the appropriate attributes(type, axis, extent, etc.):

 Please see [datacube spec](https://github.com/stac-extensions/datacube?tab=readme-ov-file#dimension-object) for details on required fields.

 If you have a dimension like "bnds" that is used on variables like time_bnds, lon_bnds, lat_bnds to choose either the lower or upper bound, you can use and [additional dimension object](https://github.com/stac-extensions/datacube?tab=readme-ov-file#additional-dimension-object). We recommend making the type "count" as Microsoft Planetary Computer did [here](https://github.com/stac-extensions/datacube/blob/9e74fa706c9bdd971e01739cf18dcc53bdd3dd4f/examples/daymet-hi-annual.json#L76).

 %% Cell type:code id:4c862f5e-d89a-4cf7-9936-610a0df730ae tags:

 ``` python
 print(dims)
 ```

 %% Cell type:code id:5a443497-67a9-4dce-a8e9-b08d31a88223 tags:

 ``` python
 # create a dictionary of datacube dimensions you would like to assign to this dataset
 # dimension name should come from the coordinates printed above
 # we do not recommend including redundant dimensions (do not include x,y if you have lon,lat)
 # note that the extent of each dimension should be pulled from the dataset
 dims_dict = {dim_names_dict['T']: pystac.extensions.datacube.Dimension({'type': 'temporal', 'description': stac_helpers.get_long_name(ds, dim_names_dict['T']), 'extent': [temporal_extent_lower.strftime('%Y-%m-%dT%XZ'), temporal_extent_upper.strftime('%Y-%m-%dT%XZ')], 'step':time_step}),
             dim_names_dict['X']: pystac.extensions.datacube.Dimension({'type': 'spatial', 'axis': 'x', 'description': stac_helpers.get_long_name(ds, dim_names_dict['X']), 'extent': [spatial_bounds[0], spatial_bounds[2]], 'step': x_step, 'reference_system': projjson}),
             dim_names_dict['Y']: pystac.extensions.datacube.Dimension({'type': 'spatial', 'axis': 'y', 'description': stac_helpers.get_long_name(ds, dim_names_dict['Y']), 'extent': [spatial_bounds[1], spatial_bounds[3]], 'step': y_step, 'reference_system': projjson}),
             'bnds': pystac.extensions.datacube.Dimension({'type': 'count', 'description': stac_helpers.get_long_name(ds, 'bnds'), 'extent': [ds.bnds.min().item(), ds.bnds.max().item()]}),
            }
 ```

 %% Cell type:code id:27e62de2-1143-453f-8f56-cf0a80048158 tags:

 ``` python
 # make sure you added all the right dims
 assert sorted(list(dims_dict.keys())) == sorted(dims)
 ```

 %% Cell type:markdown id:0f277883-a3fd-425f-966a-ca2140d0ef2f tags:

 ### Add cube variables (optional field for extension)

 %% Cell type:code id:e9272931-fc0b-4f2a-9546-283033e9cde8 tags:

 ``` python
 # drop metpy_crs coordinate we have added
 if 'metpy_crs' in ds.coords:
    ds = ds.drop_vars('metpy_crs')

 # pull list of vars from dataset
 vars = list(ds.variables)

 # spec says that the keys of cube:dimensions and cube:variables should be unique together; a key like lat should not be both a dimension and a variable.
 # we will drop all values in dims from vars
 vars = [v for v in vars if v not in dims]

 # Microsoft Planetary Computer includes coordinates and crs as variables here:
 # https://planetarycomputer.microsoft.com/dataset/daymet-annual-na
 # https://planetarycomputer.microsoft.com/api/stac/v1/collections/daymet-annual-na
 # we will keep those in the var list

 # create dictionary of dataset variables and associated dimensions
 vars_dict={}
 for v in vars:
    unit = stac_helpers.get_unit(ds, v)
    var_type = stac_helpers.get_var_type(ds, v, crs_var)
    long_name = stac_helpers.get_long_name(ds, v)
    vars_dict[v] = pystac.extensions.datacube.Variable({'dimensions':list(ds[v].dims), 'type': var_type, 'description': long_name, 'unit': unit})
 ```

 %% Cell type:markdown id:11ad5352-884c-4472-8864-4570a96f66e5 tags:

 ### Finalize extension

 %% Cell type:code id:10141fd4-91d6-491d-878b-02653720891d tags:

 ``` python
 # add dimesions and variables to collection extension
 dc.apply(dimensions=dims_dict, variables=vars_dict)
 ```

 %% Cell type:markdown id:615ca168-75fb-4135-9941-0ef5fe4fd1cb tags:

 ## Add STAC Collection to Catalog and Save

 %% Cell type:code id:e2120a55-3d04-4122-a93f-29afcdb8cb1b tags:

 ``` python
 # # helper to find items of wrong type
 # d = collection.to_dict()
 # print(*stac_helpers.find_paths(d))
 ```

 %% Cell type:code id:4b75791b-6b2d-40be-b7c6-330a60888fb5 tags:

 ``` python
 if cprep_collection.get_child(collection_id):
    collection.normalize_and_save(root_href=os.path.join(catalog_path, collection_id), catalog_type=pystac.CatalogType.SELF_CONTAINED)
 else:
    cprep_collection.add_child(collection)
    cprep_collection.normalize_and_save(root_href=catalog_path+f'/{overall_collection_id}', catalog_type=pystac.CatalogType.SELF_CONTAINED)
 ```

 %% Cell type:markdown id:c734db5d-a9b4-42b4-88fb-50d84493f694 tags:

 ## The big job
 Now that we have tested the workflow, we will implement it in a big for loop for all 297 zarrs.

 %% Cell type:code id:929a90d4-2ae1-480a-aec3-348d19a667d8 tags:

 ``` python
 fs2 = fsspec.filesystem('s3', anon=True, endpoint_url='https://usgs.osn.mghpcc.org/')

 # define folder location where your STAC catalog json file is
 catalog_path = os.path.join('..', '..', 'catalog')
 # open catalog
 catalog = pystac.Catalog.from_file(os.path.join(catalog_path, 'catalog.json'))

 for zarr in zarr_list:
    collection_id = os.path.splitext(zarr)[0]

    # url to zarr store that you want to create a collection for
    zarr_url = f's3://mdmf/gdp/cprep/{collection_id}.zarr/'

    # define keyword arguments needed for opening the dataset with xarray
    # ref: https://github.com/stac-extensions/xarray-assets
    xarray_opendataset_kwargs = {"xarray:open_kwargs":{"chunks":{},"engine":"zarr","consolidated":True},
                              "xarray:storage_options": {"anon": True, "client_kwargs": {"endpoint_url":"https://usgs.osn.mghpcc.org/"}}}
    # description for zarr url asset attached to collection (zarr_url)
    asset_description = "Open Storage Network Pod S3 API access to collection zarr group"
    # roles to tag zarr url asset with
    asset_roles = ["data","zarr","s3"]

    # url to zarr store that you want to create a collection for
    zarr_url2 = f's3://nhgf-development/workspace/DataConversion/gdp/cprep/{collection_id}.zarr/'

    # define keyword arguments needed for opening the dataset with xarray
    # ref: https://github.com/stac-extensions/xarray-assets
    xarray_opendataset_kwargs2 = {"xarray:open_kwargs":{"chunks":{},"engine":"zarr","consolidated":True},
                              "xarray:storage_options":{"requester_pays":True}}
    # description for zarr url asset attached to collection (zarr_url)
    asset_description2 = "S3 access to collection zarr group"
    # roles to tag zarr url asset with
    asset_roles2 = ["data","zarr","s3"]

    # open and view zarr dataset
    try:
        ds = xr.open_dataset(fs2.get_mapper(zarr_url), engine='zarr',
                                     backend_kwargs={'consolidated':True}, chunks={})
        decode_times=True

    except:
        ds = xr.open_dataset(fs2.get_mapper(zarr_url), engine='zarr',
                             backend_kwargs={'consolidated':True}, chunks={}, decode_times=False)
        decode_times=False

    # description of STAC collection
    collection_description = ds.attrs['title']

    # license for dataset
    collection_license = stac_helpers.license_picker(ds.attrs['license'])

    dims_auto_extract = ['X', 'Y', 'T']
    dim_names_dict = {}
    for d in dims_auto_extract:
        dim_names_dict[d] = stac_helpers.extract_dim(ds, d)

    crs_var = 'crs'
    # use pyproj to automatically extract crs info
    crs = pyproj.CRS.from_cf(ds[crs_var].attrs)

    # create a spatial extent object
    spatial_bounds = [ds[dim_names_dict['X']].data.min().astype(float).item(), ds[dim_names_dict['Y']].data.min().astype(float).item(), ds[dim_names_dict['X']].data.max().astype(float).item(), ds[dim_names_dict['Y']].data.max().astype(float).item()]
    XX, YY = np.meshgrid(ds[dim_names_dict['X']].data, ds[dim_names_dict['Y']].data)
    transformer = Transformer.from_crs(crs, "EPSG:4326", always_xy=True)
    lon, lat = transformer.transform(XX.ravel(), YY.ravel())
    print(f'lower left coordinates (WGS84): {min(lon)}, {min(lat)}')
    print(f'upper right coordinates (WGS84): {max(lon)}, {max(lat)}')
    spatial_extent = pystac.SpatialExtent(bboxes=[[min(lon).item(), min(lat).item(), max(lon).item(), max(lat).item()]])

    # pull out first and last timestamps
    if decode_times:
        temporal_extent_lower = ds.indexes[dim_names_dict['T']].to_datetimeindex().min()
        temporal_extent_upper = ds.indexes[dim_names_dict['T']].to_datetimeindex().max()
    else:
        temporal_extent_lower = pd.Timestamp(ds[dim_names_dict['T']].data.min())
        temporal_extent_upper = pd.Timestamp(ds[dim_names_dict['T']].data.max())
    print(f'min: {temporal_extent_lower} \nmax: {temporal_extent_upper}')
    # create a temporal extent object
    temporal_extent = pystac.TemporalExtent(intervals=[[temporal_extent_lower, temporal_extent_upper]])

    collection_extent = pystac.Extent(spatial=spatial_extent, temporal=temporal_extent)

    # create collection for current zarr
    if cprep_collection.get_child(collection_id):
        collection = catalog.get_child(collection_id)
        print("existing collection opened")
        collection.extent=collection_extent
        collection.description=collection_description
        collection.license=collection_license
    else:
        collection = pystac.Collection(id=collection_id,
                                       description=collection_description,
                                       extent=collection_extent,
                                       license=collection_license)
        print("new collection created")

    asset_id = "zarr-s3-osn"
    asset = pystac.Asset(href=zarr_url,
                         description=asset_description,
                         media_type="application/vnd+zarr",
                         roles=asset_roles,
                         extra_fields = xarray_opendataset_kwargs)
    collection.add_asset(asset_id, asset)

    asset_id2 = "zarr-s3"
    asset2 = pystac.Asset(href=zarr_url2,
                         description=asset_description2,
                         media_type="application/vnd+zarr",
                         roles=asset_roles2,
                         extra_fields = xarray_opendataset_kwargs2)
    collection.add_asset(asset_id2, asset2)

    # instantiate extention on collection
    dc = DatacubeExtension.ext(collection, add_if_missing=True)

    # list out dataset dimensions
    dims = list(ds.dims)

    projjson = crs.to_json()

    time_step = pd.Timedelta(stac_helpers.get_step(ds, dim_names_dict['T'], time_dim=True)).isoformat()
    x_step = stac_helpers.get_step(ds, dim_names_dict['X'], round_dec=13)
    y_step = stac_helpers.get_step(ds, dim_names_dict['Y'], round_dec=13)
    print(f'x step: {x_step}')
    print(f'y step: {y_step}')

    # create a dictionary of datacube dimensions you would like to assign to this dataset
    dims_dict = {dim_names_dict['T']: pystac.extensions.datacube.Dimension({'type': 'temporal', 'description': stac_helpers.get_long_name(ds, dim_names_dict['T']), 'extent': [temporal_extent_lower.strftime('%Y-%m-%dT%XZ'), temporal_extent_upper.strftime('%Y-%m-%dT%XZ')], 'step':time_step}),
                 dim_names_dict['X']: pystac.extensions.datacube.Dimension({'type': 'spatial', 'axis': 'x', 'description': stac_helpers.get_long_name(ds, dim_names_dict['X']), 'extent': [spatial_bounds[0], spatial_bounds[2]], 'step': x_step, 'reference_system': projjson}),
                 dim_names_dict['Y']: pystac.extensions.datacube.Dimension({'type': 'spatial', 'axis': 'y', 'description': stac_helpers.get_long_name(ds, dim_names_dict['Y']), 'extent': [spatial_bounds[1], spatial_bounds[3]], 'step': y_step, 'reference_system': projjson}),
                 'bnds': pystac.extensions.datacube.Dimension({'type': 'count', 'description': stac_helpers.get_long_name(ds, 'bnds'), 'extent': [ds.bnds.min().item(), ds.bnds.max().item()]}),
                }
    # make sure you added all the right dims
    assert sorted(list(dims_dict.keys())) == sorted(dims)

    # drop metpy_crs coordinate we have added
    if 'metpy_crs' in ds.coords:
        ds = ds.drop_vars('metpy_crs')

    # pull list of vars from dataset
    vars = list(ds.variables)

    # spec says that the keys of cube:dimensions and cube:variables should be unique together; a key like lat should not be both a dimension and a variable.
    # we will drop all values in dims from vars
    vars = [v for v in vars if v not in dims]

    # Microsoft Planetary Computer includes coordinates and crs as variables here:
    # https://planetarycomputer.microsoft.com/dataset/daymet-annual-na
    # https://planetarycomputer.microsoft.com/api/stac/v1/collections/daymet-annual-na
    # we will keep those in the var list

    # create dictionary of dataset variables and associated dimensions
    vars_dict={}
    for v in vars:
        unit = stac_helpers.get_unit(ds, v)
        var_type = stac_helpers.get_var_type(ds, v, crs_var)
        long_name = stac_helpers.get_long_name(ds, v)
        vars_dict[v] = pystac.extensions.datacube.Variable({'dimensions':list(ds[v].dims), 'type': var_type, 'description': long_name, 'unit': unit})

    # add dimesions and variables to collection extension
    dc.apply(dimensions=dims_dict, variables=vars_dict)

    if cprep_collection.get_child(collection_id):
        collection.normalize_and_save(root_href=os.path.join(catalog_path, collection_id), catalog_type=pystac.CatalogType.SELF_CONTAINED)
    else:
        cprep_collection.add_child(collection)
        cprep_collection.normalize_and_save(root_href=catalog_path+f'/{overall_collection_id}', catalog_type=pystac.CatalogType.SELF_CONTAINED)
 ```

 %% Cell type:code id:ca1bd339-a580-48fe-af4e-c1de7c8cbc48 tags:

 ``` python
 zarr_url
 ```

 %% Cell type:code id:334d826f-df7f-4b9c-82de-1354352945f0 tags:

 ``` python
 ```

 %% Cell type:code id:a60045a8-1496-47e0-9d3d-e4bf02bbfbc6 tags:

 ``` python
 ```

 %% Cell type:code id:1e68c395-5cc2-478e-bfb9-57e2bc22a47c tags:

 ``` python
 ```

 %% Cell type:code id:9cbfd7ae-5eda-46be-a978-7b9e32306cb2 tags:

 ``` python
 ```

 %% Cell type:code id:b868cf72-86a6-4fc0-84a6-58633700e2eb tags:

 ``` python
 ```

 %% Cell type:markdown id:ac25d30a-0a27-4241-a631-522f6dcbe764 tags:

 ## Supplemental - Calendar Exploration based on zarr conversion workflow
 A special handling of the calendars was set up in the [zarr converstion workflow](https://code.usgs.gov/wma/nhgf/geo-data-portal/gdp_data_processing/-/blob/main/workflows/opendap/CIDA/CPREP/cprep_conversion.ipynb?ref_type=heads) for this dataset, so this was further explored to see if it was necessary.

 %% Cell type:markdown id:3877e708-f600-48c2-b3e4-e267dd547c57 tags:

 ### Types of calendars
 There appear to be two types of calendars used in in this dataset's time variable. Let's check that first.

 %% Cell type:code id:e314605d-5676-4add-935e-1ca59e4852d4 tags:

 ``` python
 ds_example = []
 calendar_list = []
 calendar_count = {}
 for f in zarr_list:
    collection_id = os.path.splitext(f)[0]
    zarr_url = f's3://mdmf/gdp/cprep/{collection_id}.zarr/'
    ds = xr.open_dataset(fs2.get_mapper(zarr_url), engine='zarr',
                        backend_kwargs={'consolidated':True}, chunks={},
                        decode_times=False)
    calendar = ds.time.calendar
    try:
        calendar_count[calendar] = calendar_count[calendar] + 1
    except:
        calendar_count[calendar] = 1
    if calendar not in calendar_list:
        ds_example.append(collection_id)
        calendar_list.append(calendar)
 ```

 %% Cell type:code id:0ff931d7-f0dd-4721-97ad-0b65b796db53 tags:

 ``` python
 calendar_count
 ```

 %% Cell type:code id:030a1cfd-f568-4cfe-b7d0-5fae88c1e819 tags:

 ``` python
 calendar_list
 ```

 %% Cell type:code id:7493aa82-e707-4f10-bb22-125661c30c79 tags:

 ``` python
 ds_example
 ```

 %% Cell type:markdown id:c17f24ff-7763-4400-80b3-3d815c2285fd tags:

 ### Julian calendar
 #### What do we get for min/max times if we decode the times upon reading in the data (the xarray default)?

 %% Cell type:code id:03fdf554-c6ba-47a4-b8ce-8993d1cd00e9 tags:

 ``` python
 # Julian
 zarr_url = f's3://mdmf/gdp/cprep/cprep_pr_day_I35prp1-DeltaSD-A32D01K00_rcp26_r1i1p1_I35Land_20060101-20991231.zarr/'
 ```

 %% Cell type:code id:708f2cf5-79ab-49af-8067-de31d0d13ee6 tags:

 ``` python
 # open and view zarr dataset
 fs2 = fsspec.filesystem('s3', anon=True, endpoint_url='https://usgs.osn.mghpcc.org/')
 ds = xr.open_dataset(fs2.get_mapper(zarr_url), engine='zarr',
                    backend_kwargs={'consolidated':True}, chunks={},
                    decode_times=True)
 ds
 ```

 %% Cell type:code id:c6a94627-edb3-43f2-8b79-e45d5fad89cc tags:

 ``` python
 print('min time step:')
 ds.indexes[dim_names_dict['T']].to_datetimeindex().min()
 ```

 %% Cell type:code id:c87e35ad-6fe9-4057-83eb-c5e335a75446 tags:

 ``` python
 print('max time step:')
 ds.indexes[dim_names_dict['T']].to_datetimeindex().max()
 ```

 %% Cell type:markdown id:459848fa-b464-4254-a648-7205c032db01 tags:

 #### What do we get for min/max times if we do NOT decode the times (like was done in the zarr conversion workflow)?

 %% Cell type:code id:601db820-e332-42c2-a90b-24f527bb197b tags:

 ``` python
 # open and view zarr dataset
 fs2 = fsspec.filesystem('s3', anon=True, endpoint_url='https://usgs.osn.mghpcc.org/')
 ds = xr.open_dataset(fs2.get_mapper(zarr_url), engine='zarr',
                    backend_kwargs={'consolidated':True}, chunks={},
                    decode_times=False)
 ds
 ```

 %% Cell type:code id:5ec7044a-c744-4980-813b-fd15e5e06d7e tags:

 ``` python
 z = zarr_url
 start_date = z.split('/')[5].split('_')[7].split('-')[0].split('.')[0]
 end_date = z.split('/')[5].split('_')[7].split('-')[1].split('.')[0]
 start_date = f'{start_date[0:4]}-{start_date[4:6]}-{start_date[6:8]}'
 end_date = f'{end_date[0:4]}-{end_date[4:6]}-{end_date[6:8]}'
 calendar = ds.time.calendar
 if calendar == 'julian':
    cftime_index = xr.cftime_range(start_date, end_date, freq='D', calendar='julian')
    ds['time'] = cftime_index

 elif calendar == 'noleap':
    cftime_index = xr.cftime_range(start_date, end_date, freq='D', calendar='noleap')
    ds['time'] = cftime_index
 ```

 %% Cell type:code id:950168e8-b2c2-4177-add0-efab5794db44 tags:

 ``` python
 ds
 ```

 %% Cell type:code id:faddefa3-cd71-43f9-ba6f-1e886a5a442f tags:

 ``` python
 print('min time step:')
 pd.Timestamp((ds.indexes['time'].to_datetimeindex().min()))
 ```

 %% Cell type:code id:68b10114-3b44-4ed6-9c3f-7882e33b04f3 tags:

 ``` python
 print('max time step:')
 pd.Timestamp((ds.indexes['time'].to_datetimeindex().max()))
 ```

 %% Cell type:markdown id:61bf449c-bd99-427d-a7d1-1eed51d2da75 tags:

 ### noleap
 #### What do we get for min/max times if we decode the times upon reading in the data (the xarray default)?

 %% Cell type:code id:0a486983-7643-4f6e-b68f-3f6084074210 tags:

 ``` python
 # noleap
 zarr_url = f's3://mdmf/gdp/cprep/cprep_pr_day_I35prp1-DeltaSD-A12D01K00_rcp26_r6i1p1_I35Land_20060101-20991231.zarr/'
 ```

 %% Cell type:code id:14f03b46-04fd-4a78-b5cf-bcc3331d7a30 tags:

 ``` python
 # open and view zarr dataset
 fs2 = fsspec.filesystem('s3', anon=True, endpoint_url='https://usgs.osn.mghpcc.org/')
 ds = xr.open_dataset(fs2.get_mapper(zarr_url), engine='zarr',
                    backend_kwargs={'consolidated':True}, chunks={},
                    decode_times=True)
 ds
 ```

 %% Cell type:code id:f225f473-adde-443e-b8c0-6d477639b40d tags:

 ``` python
 print('min time step:')
 ds.indexes[dim_names_dict['T']].to_datetimeindex().min()
 ```

 %% Cell type:code id:e20aaa0e-218c-4dfe-8cae-a2ca3854233d tags:

 ``` python
 print('max time step:')
 ds.indexes[dim_names_dict['T']].to_datetimeindex().max()
 ```

 %% Cell type:markdown id:50a860f5-98b4-48b0-b872-63bc5dc271ae tags:

 #### What do we get for min/max times if we do NOT decode the times (like was done in the zarr conversion workflow)?### What do we get for min/max times if we do NOT decode the times? - noleap

 %% Cell type:code id:737458bc-cba5-4c93-83bf-963cbb635f24 tags:

 ``` python
 # open and view zarr dataset
 fs2 = fsspec.filesystem('s3', anon=True, endpoint_url='https://usgs.osn.mghpcc.org/')
 ds = xr.open_dataset(fs2.get_mapper(zarr_url), engine='zarr',
                    backend_kwargs={'consolidated':True}, chunks={},
                    decode_times=False)
 ds
 ```

 %% Cell type:code id:c1ae19f2-5713-4ca8-8299-a566c15b29ae tags:

 ``` python
 z = zarr_url
 start_date = z.split('/')[5].split('_')[7].split('-')[0].split('.')[0]
 end_date = z.split('/')[5].split('_')[7].split('-')[1].split('.')[0]
 start_date = f'{start_date[0:4]}-{start_date[4:6]}-{start_date[6:8]}'
 end_date = f'{end_date[0:4]}-{end_date[4:6]}-{end_date[6:8]}'
 calendar = ds.time.calendar
 if calendar == 'julian':
    cftime_index = xr.cftime_range(start_date, end_date, freq='D', calendar='julian')
    ds['time'] = cftime_index

 elif calendar == 'noleap':
    cftime_index = xr.cftime_range(start_date, end_date, freq='D', calendar='noleap')
    ds['time'] = cftime_index
 ```

 %% Cell type:code id:b7cc4b04-5b08-4bd0-bd49-6c9ba37f4df9 tags:

 ``` python
 ds
 ```

 %% Cell type:code id:df3038fa-a03b-41dc-a029-773a8fbf496d tags:

 ``` python
 print('min time step:')
 ds.indexes[dim_names_dict['T']].to_datetimeindex().min()
 ```

 %% Cell type:code id:e3bada84-0d24-42db-b345-6d657c40ba06 tags:

 ``` python
 print('max time step:')
 ds.indexes[dim_names_dict['T']].to_datetimeindex().min()
 ```

 %% Cell type:markdown id:fb491fae-4681-4b1a-aef6-95e4bbb764f1 tags:

 ### Conclusion:
 The workflow is overwriting the calendar values so that instead of being marked in the 12:00 hour, they are marked with the 00:00 hour. There doesn't seem to be different handling for Julian vs. noleap calendars. We will use the data's true data values and let xarray decode the times for us, rather than using this shifted timestamp.