File size: 10,486 Bytes
76d61a0 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 | from __future__ import annotations
from pathlib import Path
from typing import Any
import numpy as np
import xarray as xr
from scipy.ndimage import binary_dilation
from .input_data import ConcatInput, L2AIIInput
from .utils import future_offsets, inv_zscore, load_stats, open_memmap, parse_time, zscore
class CILabel:
source_name = "ci"
def __init__(self, config: dict[str, Any]):
self.config = dict(config)
self.root = Path(self.config["root"])
self.expected_hw = self.config.get("shape_hw")
if self.expected_hw is not None:
self.expected_hw = tuple(int(v) for v in self.expected_hw)
self._xr_open_kwargs = dict(decode_cf=False, mask_and_scale=False, decode_times=False)
@property
def row_shape(self) -> tuple[int, int]:
if not self.expected_hw:
raise ValueError("ci.shape_hw must be configured or inferred before memmap use")
return tuple(self.expected_hw)
def path_for_time(self, timestamp: str):
ts = parse_time(timestamp).strftime("%Y%m%d%H%M")
day = ts[:8]
nc = self.root / day / f"{ts}_label.nc"
if nc.exists():
return nc
npz = self.root / day / f"{ts}_layers.npz"
if npz.exists():
return npz
return None
def load_label(self, timestamp: str) -> np.ndarray:
path = self.path_for_time(timestamp)
if path is None:
raise FileNotFoundError(f"CI label not found for {timestamp}")
if path.suffix == ".npz":
with np.load(path, allow_pickle=False) as data:
if "masks" not in data:
raise KeyError(f"'masks' key not found in {path}")
masks = data["masks"]
if masks.ndim != 3:
raise ValueError(f"npz masks must be (N,H,W), got {masks.shape}: {path}")
label = (masks > 0).any(axis=0).astype(np.uint8)
else:
with xr.open_dataset(path, **self._xr_open_kwargs) as ds:
var = list(ds.data_vars)[0]
arr = ds[var].values
if arr.ndim != 2:
raise ValueError(f"CI label must be 2D, got {arr.shape}: {path}")
label = (~np.isnan(arr)).astype(np.uint8)
if self.expected_hw is None:
self.expected_hw = tuple(int(v) for v in label.shape)
if tuple(label.shape) != tuple(self.expected_hw):
raise ValueError(f"CI shape mismatch: {label.shape} != {self.expected_hw}")
return label
@staticmethod
def smooth(label: np.ndarray, base: float = 0.5, radius: int = 3) -> np.ndarray:
base = float(base)
radius = int(radius)
if not (0.0 < base < 1.0):
raise ValueError(f"base must be between 0 and 1, got {base}")
if radius < 0:
raise ValueError(f"radius must be >= 0, got {radius}")
positive = np.asarray(label) > 0.5
smoothed = np.zeros(positive.shape, dtype=np.float32)
smoothed[positive] = 1.0
if radius == 0 or not positive.any():
return smoothed
structure = np.ones((3, 3), dtype=bool)
prev = positive
for distance in range(1, radius + 1):
dilated = binary_dilation(positive, structure=structure, iterations=distance)
ring = dilated & ~prev
smoothed[ring] = base ** distance
prev = dilated
return smoothed
def open_memmap(self, dat_path: str | Path, n_rows: int, dtype: str = "uint8", mode: str = "r") -> np.memmap:
return open_memmap(dat_path, dtype, (int(n_rows), *self.row_shape), mode=mode)
def load_memmap_row(self, dat_path: str | Path, row_idx: int, n_rows: int, dtype: str = "uint8") -> np.ndarray:
mm = self.open_memmap(dat_path, n_rows=n_rows, dtype=dtype, mode="r")
return np.asarray(mm[int(row_idx)])
class BTLabel:
source_name = "bt"
file_dtype = "float16"
DEFAULT_CONDITIONS = {
"KI": {"op": ">=", "value": 30.0},
"LI": {"op": "<=", "value": -2.0},
"SI": {"op": "<=", "value": 2.0},
"CAPE": {"op": ">=", "value": 500.0},
"TTI": {"op": ">=", "value": 42.0},
}
def __init__(self, config: dict[str, Any], concat_input: ConcatInput, l2_input: L2AIIInput):
self.config = dict(config)
self.concat_input = concat_input
self.l2_input = l2_input
self.future_minutes = int(self.config.get("future_minutes", 60))
self.interval_minutes = int(self.config.get("interval_minutes", 10))
self.lead_minutes = list(self.config.get("lead_minutes") or future_offsets(self.future_minutes, self.interval_minutes))
self.var_name = str(self.config.get("var_name", "ir105"))
self.mask_var_name = str(self.config.get("mask_var_name", self.var_name))
self.bt_threshold_k = float(self.config.get("bt_threshold_k", 233.0))
self.expansion_km = float(self.config.get("expansion_km", 50.0))
self.pixel_size_km = float(self.config.get("pixel_size_km", 2.0))
self.conditions = dict(self.DEFAULT_CONDITIONS)
self.conditions.update(self.config.get("aii_conditions", {}))
self.stats_path = self.config.get("stats_path") or self.concat_input.stats_path
self.stats = load_stats(self.stats_path)
self.normalization = str(self.config.get("normalization", "zscore")).lower()
self.eps = float(self.config.get("eps", 1e-6))
self.expected_hw = self.config.get("shape_hw") or self.concat_input.expected_hw
if self.expected_hw is not None:
self.expected_hw = tuple(int(v) for v in self.expected_hw)
@property
def row_shape(self) -> tuple[int, int, int]:
if not self.expected_hw:
raise ValueError("BT shape_hw must be configured or inferred before memmap use")
h, w = self.expected_hw
return (1, h, w)
@property
def mask_row_shape(self) -> tuple[int, int]:
if not self.expected_hw:
raise ValueError("BT shape_hw must be configured or inferred before memmap use")
return tuple(self.expected_hw)
def _condition_mask(self, arr: np.ndarray, op: str, value: float) -> np.ndarray:
arr = np.asarray(arr, dtype=np.float32)
finite = np.isfinite(arr)
if op == ">=":
return finite & (arr >= float(value))
if op == ">":
return finite & (arr > float(value))
if op == "<=":
return finite & (arr <= float(value))
if op == "<":
return finite & (arr < float(value))
raise ValueError(f"unsupported condition op: {op}")
def circular_footprint(self) -> np.ndarray:
radius_pixels = int(np.ceil(self.expansion_km / self.pixel_size_km))
yy, xx = np.ogrid[-radius_pixels : radius_pixels + 1, -radius_pixels : radius_pixels + 1]
distance_km = np.sqrt(xx**2 + yy**2) * self.pixel_size_km
return distance_km <= self.expansion_km
def build_mask(self, timestamp: str, shape_hw: tuple[int, int]) -> np.ndarray:
aii = self.l2_input.load_raw_dict(timestamp)
masks = []
for name, rule in self.conditions.items():
if name not in aii:
raise KeyError(f"BT mask requires L2 variable {name!r}")
arr = np.asarray(aii[name], dtype=np.float32)
if tuple(arr.shape) != tuple(shape_hw):
raise ValueError(f"AII/BT shape mismatch for {name}: {arr.shape} != {shape_hw}")
masks.append(self._condition_mask(arr, str(rule["op"]), float(rule["value"])))
aii_mask = np.logical_and.reduce(masks)
start_data = self.concat_input.load_raw_dict(timestamp)
start_bt = self.concat_input.get_variable(start_data, self.mask_var_name)
if tuple(start_bt.shape) != tuple(shape_hw):
raise ValueError(f"start mask-variable shape mismatch: {start_bt.shape} != {shape_hw}")
cold = np.isfinite(start_bt) & (start_bt <= self.bt_threshold_k)
expanded = binary_dilation(cold, structure=self.circular_footprint()) if self.expansion_km > 0 else cold
return aii_mask & ~expanded
def normalize_bt(self, arr: np.ndarray) -> np.ndarray:
arr = np.asarray(arr, dtype=np.float32)
if self.normalization in {"none", "raw", "false"}:
return arr
if self.normalization == "zscore":
stat = self.stats[self.var_name]
return zscore(arr, stat["mean"], stat["std"], self.eps)
raise ValueError(f"unsupported BT normalization: {self.normalization}")
def denormalize(self, arr: np.ndarray) -> np.ndarray:
arr = np.asarray(arr, dtype=np.float32)
if self.normalization in {"none", "raw", "false"}:
return arr
stat = self.stats[self.var_name]
return inv_zscore(arr, stat["mean"], stat["std"], self.eps)
def load_label(self, timestamp: str) -> np.ndarray:
data = self.concat_input.load_raw_dict(timestamp)
bt = self.concat_input.get_variable(data, self.var_name)
if self.expected_hw is None:
self.expected_hw = tuple(int(v) for v in bt.shape)
if tuple(bt.shape) != tuple(self.expected_hw):
raise ValueError(f"BT shape mismatch: {timestamp} {bt.shape} != {self.expected_hw}")
normalized = self.normalize_bt(bt)
return normalized[np.newaxis, ...].astype(np.float32, copy=False)
def load_mask(self, timestamp: str) -> np.ndarray:
data = self.concat_input.load_raw_dict(timestamp)
bt = self.concat_input.get_variable(data, self.var_name)
if self.expected_hw is None:
self.expected_hw = tuple(int(v) for v in bt.shape)
if tuple(bt.shape) != tuple(self.expected_hw):
raise ValueError(f"BT mask shape mismatch: {timestamp} {bt.shape} != {self.expected_hw}")
return self.build_mask(timestamp, tuple(self.expected_hw)).astype(np.uint8, copy=False)
def open_memmap(self, dat_path: str | Path, n_rows: int, dtype: str | None = None, mode: str = "r") -> np.memmap:
return open_memmap(dat_path, dtype or self.file_dtype, (int(n_rows), *self.row_shape), mode=mode)
def load_memmap_row(self, dat_path: str | Path, row_idx: int, n_rows: int, dtype: str | None = None) -> np.ndarray:
mm = self.open_memmap(dat_path, n_rows=n_rows, dtype=dtype or self.file_dtype, mode="r")
return np.asarray(mm[int(row_idx)], dtype=np.float32)
|