Select DataTree nodes based on inherited dimension

[zan-xhipe]

I have weather data with (time, latitude, longitude) dimensions. We construct a DataTree from the data where each node contains the mask for a weather feature. So a node has dimensions (latitude, longitude) with a single value for the time coordinate for the timestep where the feature occurs.

<xarray.DataTree> Group: / │ Dimensions: (time: 8, latitude: 201, longitude: 561) │ Coordinates: │ * time (time) datetime64[ns] 64B 2025-01-01T06:00:00 ... 2025-01-03 │ level float64 8B 950.0 │ * latitude (latitude) float64 2kB -20.0 -20.25 -20.5 ... -69.75 -70.0 │ * longitude (longitude) float64 4kB -80.0 -79.75 -79.5 ... 59.5 59.75 60.0 │ Data variables: (12/21) │ div (time, latitude, longitude) float32 4MB ... │ zg (time, latitude, longitude) float32 4MB 579.5 579.4 ... 417.9 │ hus (time, latitude, longitude) float32 4MB ... │ ta (time, latitude, longitude) float32 4MB <Quantity([[[ 11.937... │ ua (time, latitude, longitude) float32 4MB <Quantity([[[ -6.453... │ va (time, latitude, longitude) float32 4MB <Quantity([[[ 1.910... │ ... ... │ nvo (time, latitude, longitude) float32 4MB -0.09545 ... -0.3291 └── Group: /feature ├── Group: /feature/0 │ Dimensions: (latitude: 201, longitude: 561) │ Coordinates: │ time datetime64[ns] 8B 2025-01-01T06:00:00 │ level float64 8B 950.0 │ * latitude (latitude) float64 2kB -20.0 -20.25 -20.5 ... -69.5 -69.75 -70.0 │ * longitude (longitude) float64 4kB -80.0 -79.75 -79.5 ... 59.5 59.75 60.0 │ Data variables: │ mask (latitude, longitude) float64 902kB nan nan nan ... nan nan nan ├── Group: /feature/1 │ Dimensions: (latitude: 201, longitude: 561) │ Coordinates: │ time datetime64[ns] 8B 2025-01-01T06:00:00 │ level float64 8B 950.0 │ * latitude (latitude) float64 2kB -20.0 -20.25 -20.5 ... -69.5 -69.75 -70.0 │ * longitude (longitude) float64 4kB -80.0 -79.75 -79.5 ... 59.5 59.75 60.0 │ Data variables: │ mask (latitude, longitude) float64 902kB nan nan nan ... nan nan nan ├── Group: /feature/2 │ Dimensions: (latitude: 201, longitude: 561) │ Coordinates: │ time datetime64[ns] 8B 2025-01-01T12:00:00 │ level float64 8B 950.0 │ * latitude (latitude) float64 2kB -20.0 -20.25 -20.5 ... -69.5 -69.75 -70.0 │ * longitude (longitude) float64 4kB -80.0 -79.75 -79.5 ... 59.5 59.75 60.0 │ Data variables: │ mask (latitude, longitude) float64 902kB nan nan nan ... nan nan nan ... ├── Group: /feature/9 │ Dimensions: (latitude: 201, longitude: 561) │ Coordinates: │ time datetime64[ns] 8B 2025-01-02T12:00:00 │ level float64 8B 950.0 │ * latitude (latitude) float64 2kB -20.0 -20.25 -20.5 ... -69.5 -69.75 -70.0 │ * longitude (longitude) float64 4kB -80.0 -79.75 -79.5 ... 59.5 59.75 60.0 │ Data variables: │ mask (latitude, longitude) float64 902kB nan nan nan ... nan nan nan ├── Group: /feature/10 │ Dimensions: (latitude: 201, longitude: 561) │ Coordinates: │ time datetime64[ns] 8B 2025-01-02T18:00:00 │ level float64 8B 950.0 │ * latitude (latitude) float64 2kB -20.0 -20.25 -20.5 ... -69.5 -69.75 -70.0 │ * longitude (longitude) float64 4kB -80.0 -79.75 -79.5 ... 59.5 59.75 60.0 │ Data variables: │ mask (latitude, longitude) float64 902kB nan nan nan ... nan nan nan └── Group: /feature/11 Dimensions: (latitude: 201, longitude: 561) Coordinates: time datetime64[ns] 8B 2025-01-02T18:00:00 level float64 8B 950.0 * latitude (latitude) float64 2kB -20.0 -20.25 -20.5 ... -69.5 -69.75 -70.0 * longitude (longitude) float64 4kB -80.0 -79.75 -79.5 ... 59.5 59.75 60.0 Data variables: mask (latitude, longitude) float64 902kB nan nan nan ... nan nan nan 

I want to select all the features within a time range.

When I try dt.sel(time=np.datetime64('2025-01-01')) I get

KeyError: "not all values found in index 'time'. Try setting the `method` keyword argument (example: method='nearest')." Raised whilst mapping function over node with path '.' 

Adding method='nearest' I get.

ValueError: Dimensions {'time'} do not exist. Expected one or more of FrozenMappingWarningOnValuesAccess({'latitude': 201, 'longitude': 561}) Raised whilst mapping function over node with path 'feature/0' 

I tried using dt.filter(lambda x: x.is_leaf and x.ds.where(x.time==np.datetime64('2025-01-01'))) but I just get back all the nodes.

I tried putting dates in the path, but this only helps for the simplest cases, like a single day or month. This doesn't work if I want to get all the nodes for a three month period.

Is there any way I can achieve this with DataTrees?