HYSETS

Overview

HYSETS is a hydrological dataset for North America. North American hydrological dataset with extensive coverage across Canada and parts of the United States.

Dataset Information

• Region: North America
• Module: hydrodataset.hysets
• Class: Hysets

Features

Static Attributes

Static catchment attributes include: - Basin area - Mean precipitation - Topographic characteristics - Land cover information - Soil properties - Climate indices

Dynamic Variables

Timeseries variables available: - Streamflow - Precipitation
- Temperature (min, max, mean) - Potential evapotranspiration - And additional variables depending on the dataset

Usage

Basic Usage

from hydrodataset.hysets import Hysets
from hydrodataset import resolve_data_path

# Initialize dataset
data_path = resolve_data_path("hysets")
ds = Hysets(data_path)

# Get basin IDs
basin_ids = ds.read_object_ids()
print(f"Number of basins: {len(basin_ids)}")

# Check available features
print("Static features:", ds.available_static_features)
print("Dynamic features:", ds.available_dynamic_features)

# Check default time range
print(f"Default time range: {ds.default_t_range}")

# Read timeseries data
timeseries = ds.read_ts_xrdataset(
    gage_id_lst=basin_ids[:5],
    t_range=ds.default_t_range,
    var_lst=["streamflow", "precipitation"]
)
print(timeseries)

# Read attribute data
attributes = ds.read_attr_xrdataset(
    gage_id_lst=basin_ids[:5],
    var_lst=["area", "p_mean"]
)
print(attributes)

Reading Specific Variables

# Read with specific time range
ts_data = ds.read_ts_xrdataset(
    gage_id_lst=basin_ids[:10],
    t_range=["1990-01-01", "2020-12-31"],
    var_lst=["streamflow", "precipitation", "temperature_mean"]
)

# Read basin area
areas = ds.read_area(gage_id_lst=basin_ids[:10])

# Read mean precipitation
mean_precip = ds.read_mean_prcp(gage_id_lst=basin_ids[:10])

Working with Multiple Data Sources

If the dataset provides multiple sources for variables:

# Request specific data source
ts_data = ds.read_ts_xrdataset(
    gage_id_lst=basin_ids[:5],
    t_range=["1990-01-01", "2000-12-31"],
    var_lst=[
        ("precipitation", "era5land"),  # Specify source
        "streamflow"  # Use default source
    ]
)

API Reference

hydrodataset.hysets.Hysets

Bases: HydroDataset

HYsets dataset class extending RainfallRunoff.

This class provides access to the HYsets dataset, which contains hourly hydrological and meteorological data for various watersheds.

Attributes:

Name	Type	Description
region		Geographic region identifier
download		Whether to download data automatically
ds_description		Dictionary containing dataset file paths

Source code in hydrodataset/hysets.py

class Hysets(HydroDataset):
    """HYsets dataset class extending RainfallRunoff.

    This class provides access to the HYsets dataset, which contains hourly
    hydrological and meteorological data for various watersheds.

    Attributes:
        region: Geographic region identifier
        download: Whether to download data automatically
        ds_description: Dictionary containing dataset file paths
    """

    def __init__(
        self, uri: str, region: Optional[str] = None, download: bool = False
    ) -> None:
        """Initialize HYsets dataset.

        Args:
            uri: Path to the data directory
            region: Geographic region identifier (optional)
            download: Whether to download data automatically (default: False)
        """
        super().__init__(uri)
        self.region = region
        self.download = download
        # cloud path: aqua_fetch cannot read S3, use cache_*_to_zarr instead
        if str(uri).startswith("s3://"):
            return
        self.aqua_fetch = HYSETS(uri)

    # OSS relative paths (folder HYSETS); default source for every dynamic
    # feature is ERA5, stored in one big NetCDF
    _STATIC_REL = "HYSETS/HYSETS_watershed_properties.txt"
    _ERA5_NC_REL = "HYSETS/HYSETS_2023_update_ERA51.nc"
    # AquaFetch HYSETS.static_map
    _STATIC_RENAME = {
        "Drainage_Area_km2": "area_km2",
        "Centroid_Lat_deg_N": "lat",
        "Slope_deg": "slope_degrees",
        "Centroid_Lon_deg_E": "long",
    }
    # AquaFetch HYSETS.dyn_map resolved to cleaned names (nc dynamic_features
    # coord uses the raw keys; "total_runoff" has no mapping and is ignored)
    _DYN_RENAME = {
        "10m_u_component_of_wind": "windspeedu_mps",
        "10m_v_component_of_wind": "windspeedv_mps",
        "2m_dewpoint": "dptemp_c_mean_2m",
        "2m_tasmax": "airtemp_c_2m_max",
        "2m_tasmin": "airtemp_c_2m_min",
        "discharge": "q_cms_obs",
        "evaporation": "evap_mm",
        "snow_density": "snowdensity_kgm3",
        "snow_evaporation": "evap_mm_snow",
        "snow_water_equivalent": "swe_mm",
        "snowfall": "snowfall_mm",
        "snowmelt": "snowmelt_mm",
        "surface_downwards_solar_radiation": "solrad_wm2",
        "surface_downwards_thermal_radiation": "lwdownrad_wm2",
        "surface_net_solar_radiation": "solradnet_wm2",
        "surface_net_thermal_radiation": "lwnetrad_wm2",
        "surface_pressure": "airpres_hpa",
        "surface_runoff": "q_mm_obs",
        "total_cloud_cover": "cloudcover",
        "total_precipitation": "pcp_mm",
    }

    def read_object_ids(self) -> np.ndarray:
        if self._is_cloud():
            fs = self._make_s3fs()
            uri = str(self.data_source_dir).rstrip("/")
            with fs.open(f"{uri}/{self._STATIC_REL}".removeprefix("s3://")) as fh:
                idx = pd.read_csv(fh, index_col="Watershed_ID", sep=",",
                                  usecols=["Watershed_ID"]).index
            return np.array(sorted((str(i) for i in idx), key=lambda x: int(x)))
        return super().read_object_ids()

    def cache_attributes_to_zarr(self) -> None:
        import zarr

        fs = self._make_s3fs()
        uri = str(self.data_source_dir).rstrip("/")
        with fs.open(f"{uri}/{self._STATIC_REL}".removeprefix("s3://")) as fh:
            static = pd.read_csv(fh, index_col="Watershed_ID", sep=",")
        static.index = static.index.astype(str)
        static = static.rename(columns=self._STATIC_RENAME)
        static.columns = self._clean_feature_names(list(static.columns))
        static = static.loc[:, ~static.columns.duplicated(keep="first")]
        # p_mean is derived from the precipitation timeseries (matches local)
        static["p_mean"] = self._p_mean_from_precip(static.index)

        zarr_name = self._attributes_cache_filename.replace(".nc", ".zarr")
        out, opts = self._zarr_path_and_opts(zarr_name)
        ids = static.index.tolist()
        n = len(ids)
        root = zarr.open_group(out, mode="w", storage_options=opts, zarr_format=2)
        for col in static.columns:
            vals = static[col].values.astype(str) if static[col].dtype == object else static[col].values
            arr = root.create_array(col, shape=(n,), chunks=(n,), dtype=vals.dtype)
            arr[:] = vals
            arr.attrs["_ARRAY_DIMENSIONS"] = ["basin"]
        basin_arr = root.create_array("basin", shape=(n,), chunks=(n,), dtype=str)
        basin_arr[:] = ids
        basin_arr.attrs["_ARRAY_DIMENSIONS"] = ["basin"]
        root.attrs["coordinates"] = "basin"
        self._write_zarr_units(root, "static")
        print(f"Attributes zarr written to: {out}")

    def cache_timeseries_to_zarr(self, batch_size: int = 200) -> None:
        import zarr

        fs = self._make_s3fs()
        uri = str(self.data_source_dir).rstrip("/")
        nc_path = f"{uri}/{self._ERA5_NC_REL}".removeprefix("s3://")

        stations = self.read_object_ids().tolist()  # Watershed_IDs as strings
        n = len(stations)
        all_times = pd.date_range(self.default_t_range[0], self.default_t_range[1], freq="D")
        nt = len(all_times)
        times_ns = all_times.asi8

        cleaned_var_lst = []
        for info in self._dynamic_variable_mapping.values():
            for s in info["sources"].values():
                if s["specific_name"] not in cleaned_var_lst:
                    cleaned_var_lst.append(s["specific_name"])

        zarr_name = self._timeseries_cache_filename.replace(".nc", ".zarr")
        out, opts = self._zarr_path_and_opts(zarr_name)
        chunk_t = min(nt, 365)
        chunk_b = min(batch_size, n)
        root = zarr.open_group(out, mode="a", storage_options=opts, zarr_format=2)
        if "basin" not in root:
            for vn in cleaned_var_lst:
                arr = root.create_array(vn, shape=(n, nt), chunks=(chunk_b, chunk_t),
                                        dtype="float64", fill_value=np.nan)
                arr.attrs["_ARRAY_DIMENSIONS"] = ["basin", "time"]
            time_arr = root.create_array("time", shape=(nt,), chunks=(chunk_t,), dtype="int64")
            time_arr[:] = times_ns
            time_arr.attrs["_ARRAY_DIMENSIONS"] = ["time"]
            time_arr.attrs["units"] = "nanoseconds since 1970-01-01"
            time_arr.attrs["calendar"] = "proleptic_gregorian"
            basin_arr = root.create_array("basin", shape=(n,), chunks=(n,), dtype=str)
            basin_arr[:] = stations
            basin_arr.attrs["_ARRAY_DIMENSIONS"] = ["basin"]
            prog = root.create_array("_progress", shape=(n,), chunks=(n,), dtype="int8", fill_value=0)
            prog[:] = 0
            prog.attrs["_ARRAY_DIMENSIONS"] = ["basin"]
            root.attrs["coordinates"] = "basin time"
        progress = root["_progress"]

        # Open the 13 GB ERA5 NetCDF over S3 with h5py directly.  xr.open_dataset
        # would eagerly scan the metadata of all 14425 variables on open (one per
        # station), which over S3 means tens of thousands of tiny range reads and
        # effectively hangs.  h5py opens lazily (only the superblock + root group)
        # and reads each dataset's header on first access, so we pay that cost
        # spread across the batch loop (with visible progress) instead of upfront.
        import h5py

        print(f"Opening ERA5 NetCDF over S3 (lazy, h5py): {nc_path}", flush=True)
        raw_file = fs.open(nc_path, "rb", block_size=8 * 1024 * 1024, cache_type="readahead")
        h5 = h5py.File(raw_file, "r")
        raw_feats = [
            v.decode() if isinstance(v, bytes) else str(v)
            for v in h5["dynamic_features"][:]
        ]
        # nc time axis is a contiguous daily series 1950-01-01..2023-12-31, which
        # equals default_t_range's date_range -> align positionally (no decoding).
        if h5["0"].shape[0] != nt:
            raise RuntimeError(
                f"nc time length {h5['0'].shape[0]} != expected {nt}; "
                "positional time alignment invalid"
            )
        # map each raw feature column index -> cleaned specific name
        feat_targets = {}  # col_index -> cleaned name (only mapped ones)
        for idx, feat in enumerate(raw_feats):
            cleaned = self._clean_feature_names([self._DYN_RENAME.get(feat, feat)])[0]
            if cleaned in cleaned_var_lst:
                feat_targets[idx] = cleaned
        print(f"  opened. {n} stations x {nt} timesteps x {len(feat_targets)} vars", flush=True)

        n_batches = (n + batch_size - 1) // batch_size
        for start in range(0, n, batch_size):
            end = min(start + batch_size, n)
            bnum = start // batch_size + 1
            if all(progress[start:end]):
                print(f"Batch {bnum}/{n_batches}: already done, skipping", flush=True)
                continue
            print(f"Batch {bnum}/{n_batches}: {end-start} stations ...", flush=True)
            buffers = {vn: np.full((end - start, nt), np.nan) for vn in cleaned_var_lst}
            for j, stn in enumerate(tqdm(stations[start:end], desc=f"batch {bnum}")):
                var = str(int(stn) - 1)  # nc data_vars are 0-based watershed idx
                try:
                    arr = h5[var][:]  # (nt, n_features), positional time
                    for idx, vn in feat_targets.items():
                        buffers[vn][j] = arr[:, idx].astype("float64")
                except Exception as e:
                    print(f"  WARN {stn}: {e}", flush=True)
            for vn in cleaned_var_lst:
                root[vn][start:end, :] = buffers[vn]
            progress[start:end] = 1
            print(f"Batch {bnum}/{n_batches}: done", flush=True)
        h5.close()
        raw_file.close()
        self._write_zarr_units(root, "dynamic")
        print(f"Timeseries zarr written to: {out}", flush=True)

    @property
    def _attributes_cache_filename(self):
        return "hysets_attributes.nc"

    @property
    def _timeseries_cache_filename(self):
        return "hysets_timeseries.nc"

    @property
    def default_t_range(self):
        return ["1950-01-01", "2023-12-31"]

    def cache_attributes_xrdataset(self):
        """Override base method to add calculated p_mean from precipitation timeseries.

        This method:
        1. Calls parent method to create base attribute cache
        2. Reads precipitation timeseries data
        3. Calculates mean precipitation (p_mean) for each basin
        4. Adds p_mean to the attribute dataset
        5. Saves the updated cache
        """
        # Step 1: Create base attribute cache using parent method
        print("Creating base attribute cache...")
        super().cache_attributes_xrdataset()

        # Step 2: Load the base cache file
        cache_file = self.cache_dir.joinpath(self._attributes_cache_filename)
        with xr.open_dataset(cache_file) as ds_attr:
            ds_attr = ds_attr.load()  # Load into memory

        print("Calculating p_mean from precipitation timeseries...")

        # Step 3: Read precipitation timeseries for all basins
        basin_ids = self.read_object_ids().tolist()

        try:
            # Read full precipitation timeseries
            prcp_ts = self.read_ts_xrdataset(
                gage_id_lst=basin_ids,
                t_range=self.default_t_range,
                var_lst=["precipitation"],
            )

            # Step 4: Calculate temporal mean for each basin
            # The result is a DataArray with dimension (basin,)
            p_mean_values = prcp_ts["precipitation"].mean(dim="time")

            # Add units attribute
            p_mean_values.attrs["units"] = "mm/day"
            p_mean_values.attrs["description"] = (
                "Mean daily precipitation (calculated from timeseries)"
            )

            # Step 5: Add p_mean to the attribute dataset
            ds_attr["p_mean"] = p_mean_values

            print(f"Successfully calculated p_mean for {len(basin_ids)} basins")

        except Exception as e:
            print(f"Warning: Could not calculate p_mean from precipitation data: {e}")
            print("Creating p_mean with NaN values as placeholder")
            # Create p_mean with NaN values if calculation fails
            p_mean_nan = xr.DataArray(
                np.full(len(basin_ids), np.nan),
                coords={"basin": basin_ids},
                dims=["basin"],
                attrs={
                    "units": "mm/day",
                    "description": "Mean daily precipitation (not available)",
                },
            )
            ds_attr["p_mean"] = p_mean_nan

        # Step 6: Save the updated cache file
        print(f"Saving updated attribute cache with p_mean to: {cache_file}")
        ds_attr.to_netcdf(cache_file, mode="w")
        print("Successfully saved attribute cache with p_mean")

    _subclass_static_definitions = {
        "area": {"specific_name": "area_km2", "unit": "km^2"},
        "p_mean": {"specific_name": "p_mean", "unit": "mm/day"},
    }

    _dynamic_variable_mapping = {
        StandardVariable.STREAMFLOW: {
            "default_source": "observations_cms",
            "sources": {
                "observations_cms": {"specific_name": "q_cms_obs", "unit": "m^3/s"},
                "observations_mm": {"specific_name": "q_mm_obs", "unit": "mm/day"},
            },
        },
        StandardVariable.PRECIPITATION: {
            "default_source": "observations",
            "sources": {"observations": {"specific_name": "pcp_mm", "unit": "mm/day"}},
        },
        StandardVariable.TEMPERATURE_MAX: {
            "default_source": "observations",
            "sources": {
                "observations": {"specific_name": "airtemp_c_2m_max", "unit": "°C"}
            },
        },
        StandardVariable.TEMPERATURE_MIN: {
            "default_source": "observations",
            "sources": {
                "observations": {"specific_name": "airtemp_c_2m_min", "unit": "°C"}
            },
        },
        StandardVariable.TEMPERATURE_MEAN: {
            "default_source": "observations",
            "sources": {
                "observations": {"specific_name": "dptemp_c_mean_2m", "unit": "°C"}
            },
        },
        StandardVariable.SOLAR_RADIATION: {
            "default_source": "observations",
            "sources": {
                "observations": {"specific_name": "solrad_wm2", "unit": "W/m^2"},
                "net": {"specific_name": "solradnet_wm2", "unit": "W/m^2"},
            },
        },
        StandardVariable.EVAPORATION: {
            "default_source": "observations",
            "sources": {
                "observations": {"specific_name": "evap_mm", "unit": "mm/day"},
                "snow": {"specific_name": "evap_mm_snow", "unit": "mm/day"},
            },
        },
        StandardVariable.SNOW_WATER_EQUIVALENT: {
            "default_source": "observations",
            "sources": {"observations": {"specific_name": "swe_mm", "unit": "mm"}},
        },
        StandardVariable.SURFACE_PRESSURE: {
            "default_source": "observations",
            "sources": {
                "observations": {"specific_name": "airpres_hpa", "unit": "hPa"}
            },
        },
        StandardVariable.U_WIND_SPEED: {
            "default_source": "observations",
            "sources": {
                "observations": {"specific_name": "windspeedu_mps", "unit": "m/s"}
            },
        },
        StandardVariable.V_WIND_SPEED: {
            "default_source": "observations",
            "sources": {
                "observations": {"specific_name": "windspeedv_mps", "unit": "m/s"}
            },
        },
        StandardVariable.LONGWAVE_SOLAR_RADIATION: {
            "default_source": "downward",
            "sources": {
                "downward": {"specific_name": "lwdownrad_wm2", "unit": "W/m^2"},
                "net": {"specific_name": "lwnetrad_wm2", "unit": "W/m^2"},
            },
        },
        StandardVariable.SNOW_DENSITY: {
            "default_source": "observations",
            "sources": {
                "observations": {"specific_name": "snowdensity_kgm3", "unit": "kg/m^3"}
            },
        },
    }

    def cache_timeseries_xrdataset(self, batch_size=1000):
        """Cache timeseries to NetCDF in batches (14425 stations × hourly data)."""
        if not hasattr(self, "aqua_fetch"):
            raise NotImplementedError("aqua_fetch attribute is required")

        unit_lookup = {}
        if hasattr(self, "_dynamic_variable_mapping"):
            for std_name, mapping_info in self._dynamic_variable_mapping.items():
                for source, source_info in mapping_info["sources"].items():
                    unit_lookup[source_info["specific_name"]] = source_info["unit"]

        gage_id_lst = self.read_object_ids().tolist()
        total_stations = len(gage_id_lst)
        original_var_lst = self.aqua_fetch.dynamic_features
        cleaned_var_lst = self._clean_feature_names(original_var_lst)
        var_name_mapping = dict(zip(original_var_lst, cleaned_var_lst))

        n_batches = (total_stations + batch_size - 1) // batch_size
        print(
            f"Start batch processing {total_stations} stations, "
            f"{batch_size} stations per batch ({n_batches} batches)"
        )
        self.cache_dir.mkdir(parents=True, exist_ok=True)

        batch_num = 1
        for batch_idx in range(0, total_stations, batch_size):
            batch_end = min(batch_idx + batch_size, total_stations)
            batch_stations = gage_id_lst[batch_idx:batch_end]
            print(
                f"\nProcessing batch {batch_num}/{n_batches} "
                f"(stations {batch_idx}-{batch_end - 1})"
            )
            try:
                batch_data = self.aqua_fetch.fetch_stations_features(
                    stations=batch_stations,
                    dynamic_features=original_var_lst,
                    static_features=None,
                    st=self.default_t_range[0],
                    en=self.default_t_range[1],
                    as_dataframe=False,
                )
                dynamic_data = (
                    batch_data[1] if isinstance(batch_data, tuple) else batch_data
                )

                new_data_vars = {}
                time_coord = dynamic_data.coords["time"]
                for original_var in tqdm(
                    original_var_lst,
                    desc=f"Variables (batch {batch_num})",
                    total=len(original_var_lst),
                ):
                    cleaned_var = var_name_mapping[original_var]
                    var_data = []
                    for station in batch_stations:
                        if station in dynamic_data.data_vars:
                            station_data = dynamic_data[station].sel(
                                dynamic_features=original_var
                            )
                            if "dynamic_features" in station_data.coords:
                                station_data = station_data.drop("dynamic_features")
                            var_data.append(station_data)
                    if var_data:
                        combined = xr.concat(var_data, dim="basin")
                        combined["basin"] = batch_stations
                        combined.attrs["units"] = unit_lookup.get(
                            cleaned_var, "unknown"
                        )
                        new_data_vars[cleaned_var] = combined

                batch_ds = xr.Dataset(
                    data_vars=new_data_vars,
                    coords={"basin": batch_stations, "time": time_coord},
                )
                batch_filepath = self.cache_dir.joinpath(
                    f"batch{batch_num:03d}_hysets_timeseries.nc"
                )
                batch_ds.to_netcdf(batch_filepath)
                print(f"Saved batch {batch_num} -> {batch_filepath}")

            except Exception as e:
                print(f"Batch {batch_num} processing failed: {e}")
                import traceback
                traceback.print_exc()
                continue

            batch_num += 1

        print(f"\nAll batches processed! Total {batch_num - 1} batch files saved")

    def read_ts_xrdataset(
        self,
        gage_id_lst: list = None,
        t_range: list = None,
        var_lst: list = None,
        sources: dict = None,
        **kwargs,
    ) -> xr.Dataset:
        """Read timeseries from the batch-saved cache (standard names + sources)."""
        if self._is_cloud():
            # cloud: base class opens the zarr and handles selection/renaming
            return super().read_ts_xrdataset(
                gage_id_lst=gage_id_lst, t_range=t_range,
                var_lst=var_lst, sources=sources, **kwargs,
            )
        if (
            not hasattr(self, "_dynamic_variable_mapping")
            or not self._dynamic_variable_mapping
        ):
            raise NotImplementedError(
                "This dataset does not support the standardized variable mapping."
            )
        if var_lst is None:
            var_lst = list(self._dynamic_variable_mapping.keys())
        if t_range is None:
            t_range = self.default_t_range

        target_vars_to_fetch = []
        rename_map = {}
        for std_name in var_lst:
            if std_name not in self._dynamic_variable_mapping:
                raise ValueError(
                    f"'{std_name}' is not a recognized standard variable for this dataset."
                )
            mapping_info = self._dynamic_variable_mapping[std_name]
            is_explicit_source = sources and std_name in sources
            sources_to_use = []
            if is_explicit_source:
                provided_sources = sources[std_name]
                if isinstance(provided_sources, list):
                    sources_to_use.extend(provided_sources)
                else:
                    sources_to_use.append(provided_sources)
            else:
                sources_to_use.append(mapping_info["default_source"])

            needs_suffix = is_explicit_source and len(sources_to_use) > 1
            for source in sources_to_use:
                if source not in mapping_info["sources"]:
                    raise ValueError(
                        f"Source '{source}' is not available for variable '{std_name}'."
                    )
                actual_var_name = mapping_info["sources"][source]["specific_name"]
                target_vars_to_fetch.append(actual_var_name)
                output_name = f"{std_name}_{source}" if needs_suffix else std_name
                rename_map[actual_var_name] = output_name

        import glob

        batch_pattern = str(self.cache_dir / "batch*_hysets_timeseries.nc")
        batch_files = sorted(glob.glob(batch_pattern))
        if not batch_files:
            print("No batch cache files found, starting cache creation...")
            self.cache_timeseries_xrdataset()
            batch_files = sorted(glob.glob(batch_pattern))
            if not batch_files:
                raise FileNotFoundError("Cache creation failed, no batch files found")

        if gage_id_lst is None:
            gage_id_lst = self.read_object_ids().tolist()
        gage_id_lst = [str(gid) for gid in gage_id_lst]

        relevant_datasets = []
        for batch_file in batch_files:
            try:
                ds_batch = xr.open_dataset(batch_file)
                batch_basins = [str(b) for b in ds_batch.basin.values]
                common_basins = list(set(gage_id_lst) & set(batch_basins))
                if common_basins:
                    missing_vars = [
                        v for v in target_vars_to_fetch if v not in ds_batch.data_vars
                    ]
                    if missing_vars:
                        ds_batch.close()
                        raise ValueError(
                            f"Batch {os.path.basename(batch_file)} missing: {missing_vars}"
                        )
                    ds_selected = ds_batch[target_vars_to_fetch].sel(
                        basin=common_basins, time=slice(t_range[0], t_range[1])
                    )
                    relevant_datasets.append(ds_selected)
                ds_batch.close()
            except Exception as e:
                print(f"Failed to read batch file {batch_file}: {e}")
                continue

        if not relevant_datasets:
            raise ValueError(
                f"Specified stations not found in any batch files: {gage_id_lst}"
            )
        if len(relevant_datasets) == 1:
            final_ds = relevant_datasets[0]
        else:
            final_ds = xr.concat(relevant_datasets, dim="basin")

        final_ds = final_ds.rename(rename_map)
        existing_basins = [b for b in gage_id_lst if b in final_ds.basin.values]
        if existing_basins:
            final_ds = final_ds.sel(basin=existing_basins)
        return final_ds

default_t_range property

__init__(uri, region=None, download=False)

Initialize HYsets dataset.

Parameters:

Name	Type	Description	Default
uri	str	Path to the data directory	required
region	Optional[str]	Geographic region identifier (optional)	None
download	bool	Whether to download data automatically (default: False)	False

Source code in hydrodataset/hysets.py

def __init__(
    self, uri: str, region: Optional[str] = None, download: bool = False
) -> None:
    """Initialize HYsets dataset.

    Args:
        uri: Path to the data directory
        region: Geographic region identifier (optional)
        download: Whether to download data automatically (default: False)
    """
    super().__init__(uri)
    self.region = region
    self.download = download
    # cloud path: aqua_fetch cannot read S3, use cache_*_to_zarr instead
    if str(uri).startswith("s3://"):
        return
    self.aqua_fetch = HYSETS(uri)

read_object_ids()

Source code in hydrodataset/hysets.py

def read_object_ids(self) -> np.ndarray:
    if self._is_cloud():
        fs = self._make_s3fs()
        uri = str(self.data_source_dir).rstrip("/")
        with fs.open(f"{uri}/{self._STATIC_REL}".removeprefix("s3://")) as fh:
            idx = pd.read_csv(fh, index_col="Watershed_ID", sep=",",
                              usecols=["Watershed_ID"]).index
        return np.array(sorted((str(i) for i in idx), key=lambda x: int(x)))
    return super().read_object_ids()

read_ts_xrdataset(gage_id_lst=None, t_range=None, var_lst=None, sources=None, **kwargs)

Read timeseries from the batch-saved cache (standard names + sources).

Source code in hydrodataset/hysets.py

def read_ts_xrdataset(
    self,
    gage_id_lst: list = None,
    t_range: list = None,
    var_lst: list = None,
    sources: dict = None,
    **kwargs,
) -> xr.Dataset:
    """Read timeseries from the batch-saved cache (standard names + sources)."""
    if self._is_cloud():
        # cloud: base class opens the zarr and handles selection/renaming
        return super().read_ts_xrdataset(
            gage_id_lst=gage_id_lst, t_range=t_range,
            var_lst=var_lst, sources=sources, **kwargs,
        )
    if (
        not hasattr(self, "_dynamic_variable_mapping")
        or not self._dynamic_variable_mapping
    ):
        raise NotImplementedError(
            "This dataset does not support the standardized variable mapping."
        )
    if var_lst is None:
        var_lst = list(self._dynamic_variable_mapping.keys())
    if t_range is None:
        t_range = self.default_t_range

    target_vars_to_fetch = []
    rename_map = {}
    for std_name in var_lst:
        if std_name not in self._dynamic_variable_mapping:
            raise ValueError(
                f"'{std_name}' is not a recognized standard variable for this dataset."
            )
        mapping_info = self._dynamic_variable_mapping[std_name]
        is_explicit_source = sources and std_name in sources
        sources_to_use = []
        if is_explicit_source:
            provided_sources = sources[std_name]
            if isinstance(provided_sources, list):
                sources_to_use.extend(provided_sources)
            else:
                sources_to_use.append(provided_sources)
        else:
            sources_to_use.append(mapping_info["default_source"])

        needs_suffix = is_explicit_source and len(sources_to_use) > 1
        for source in sources_to_use:
            if source not in mapping_info["sources"]:
                raise ValueError(
                    f"Source '{source}' is not available for variable '{std_name}'."
                )
            actual_var_name = mapping_info["sources"][source]["specific_name"]
            target_vars_to_fetch.append(actual_var_name)
            output_name = f"{std_name}_{source}" if needs_suffix else std_name
            rename_map[actual_var_name] = output_name

    import glob

    batch_pattern = str(self.cache_dir / "batch*_hysets_timeseries.nc")
    batch_files = sorted(glob.glob(batch_pattern))
    if not batch_files:
        print("No batch cache files found, starting cache creation...")
        self.cache_timeseries_xrdataset()
        batch_files = sorted(glob.glob(batch_pattern))
        if not batch_files:
            raise FileNotFoundError("Cache creation failed, no batch files found")

    if gage_id_lst is None:
        gage_id_lst = self.read_object_ids().tolist()
    gage_id_lst = [str(gid) for gid in gage_id_lst]

    relevant_datasets = []
    for batch_file in batch_files:
        try:
            ds_batch = xr.open_dataset(batch_file)
            batch_basins = [str(b) for b in ds_batch.basin.values]
            common_basins = list(set(gage_id_lst) & set(batch_basins))
            if common_basins:
                missing_vars = [
                    v for v in target_vars_to_fetch if v not in ds_batch.data_vars
                ]
                if missing_vars:
                    ds_batch.close()
                    raise ValueError(
                        f"Batch {os.path.basename(batch_file)} missing: {missing_vars}"
                    )
                ds_selected = ds_batch[target_vars_to_fetch].sel(
                    basin=common_basins, time=slice(t_range[0], t_range[1])
                )
                relevant_datasets.append(ds_selected)
            ds_batch.close()
        except Exception as e:
            print(f"Failed to read batch file {batch_file}: {e}")
            continue

    if not relevant_datasets:
        raise ValueError(
            f"Specified stations not found in any batch files: {gage_id_lst}"
        )
    if len(relevant_datasets) == 1:
        final_ds = relevant_datasets[0]
    else:
        final_ds = xr.concat(relevant_datasets, dim="basin")

    final_ds = final_ds.rename(rename_map)
    existing_basins = [b for b in gage_id_lst if b in final_ds.basin.values]
    if existing_basins:
        final_ds = final_ds.sel(basin=existing_basins)
    return final_ds