"""TrendeX expected-greenness model: CPU inference. Imports only numpy, pandas and xgboost. Files expected next to this module (the HF model repo TrendeX90/expected-greenness): model_a.ubj XGBoost model (predicts the NDVI anomaly) feature_schema.json feature order climatology.parquet cell_id, month, clim_ndvi (2001-2025 mean NDVI of that cell and month) weather_climatology.parquet cell_id, month, _mean (raw weather value minus this = model anomaly) conformal.json month -> [q10, q90] of residuals (P10/P90 band) from predict import load m = load("path/to/repo") out = m.predict(features_df) # features in feature_schema order out = m.what_if(cells_df, cell_ids, month=7, precip_pct=-30, lst_day_c=+1.5) Output: pred_anom, expected (= clim_ndvi + pred_anom), p10, p90. NDVI units. Meaning: greenness expected from weather and terrain alone. The gap to observed NDVI is change NOT explained by precipitation, temperature, snow and soil moisture. It is not a cause and not forage quality. """ from __future__ import annotations import json from pathlib import Path import numpy as np import pandas as pd import xgboost as xgb STATIC = ["elev_mean", "elev_std", "slope_mean", "steep_share", "aspect_sin", "aspect_cos", "tpi", "grass_frac", "barren_frac", "lc_type1"] WEATHER = ["precip_m", "precip_3m", "precip_winter", "lst_day_m", "lst_night_m", "lst_day_prev", "snow_spring", "snow_m", "soil_m", "swe_apr"] class ExpectedGreenness: def __init__(self, root: str | Path): root = Path(root) self.schema = json.loads((root / "feature_schema.json").read_text()) self.features = self.schema["features"] self.booster = xgb.Booster() self.booster.load_model(str(root / "model_a.ubj")) self.booster.set_param({"device": "cpu"}) self.clim = pd.read_parquet(root / "climatology.parquet").set_index(["cell_id", "month"]) self.wclim = pd.read_parquet(root / "weather_climatology.parquet").set_index(["cell_id", "month"]) self.conf = {int(k): v for k, v in json.loads((root / "conformal.json").read_text()).items()} # ------------------------------------------------------------------ core def predict(self, df: pd.DataFrame) -> pd.DataFrame: """df must contain every feature in feature_schema.json (NaN allowed for weather).""" missing = [c for c in self.features if c not in df.columns] if missing: raise ValueError(f"missing features: {missing}") X = df[self.features].astype("float32") pred = self.booster.predict(xgb.DMatrix(X, feature_names=self.features)) months = df["month"].astype(int).values lo = np.array([self.conf[m][0] for m in months]) hi = np.array([self.conf[m][1] for m in months]) exp = df["clim_ndvi"].values + pred return pd.DataFrame({"pred_anom": pred, "expected": exp, "p10": exp + lo, "p90": exp + hi}, index=df.index) # ------------------------------------------------------------------ helpers def features_from_raw(self, cells: pd.DataFrame, raw: pd.DataFrame) -> pd.DataFrame: """raw: cell_id, month and raw weather values named like WEATHER (mm, deg C, fractions, kg m-2). precip_3m = this + 2 previous months; precip_winter = Oct..Apr sum; snow_spring = Mar-May mean; swe_apr = April SWE. cells: rows of the cube's cells.parquet (for static features). """ key = pd.MultiIndex.from_arrays([raw["cell_id"].values, raw["month"].astype(int).values]) out = raw[["cell_id", "month"]].copy() for w in WEATHER: out[w] = raw[w].values - self.wclim[f"{w}_mean"].reindex(key).values if w in raw else np.nan out["clim_ndvi"] = self.clim["clim_ndvi"].reindex(key).values out = out.merge(cells[["cell_id"] + STATIC], on="cell_id", how="left") return out def what_if(self, cells: pd.DataFrame, cell_ids, month: int, precip_pct: float = 0.0, lst_day_c: float = 0.0, lst_night_c: float = 0.0, snow_spring_delta: float = 0.0, soil_pct: float = 0.0) -> pd.DataFrame: """Scenario vs typical weather: e.g. precip_pct=-30 means every precipitation input 30% below normal.""" ids = np.asarray(cell_ids) key = pd.MultiIndex.from_arrays([ids, np.full(len(ids), int(month))]) w = self.wclim.reindex(key) def frame(pp, ld, ln, ss, sp): f = pd.DataFrame({"cell_id": ids, "month": float(month)}) for k in ("precip_m", "precip_3m", "precip_winter"): f[k] = w[f"{k}_mean"].values * pp / 100.0 f["lst_day_m"] = f["lst_day_prev"] = ld f["lst_night_m"] = ln f["snow_spring"] = ss f["snow_m"] = 0.0 f["soil_m"] = w["soil_m_mean"].values * sp / 100.0 f["swe_apr"] = 0.0 f["clim_ndvi"] = self.clim["clim_ndvi"].reindex(key).values return f.merge(cells[["cell_id"] + STATIC], on="cell_id", how="left") res = self.predict(frame(precip_pct, lst_day_c, lst_night_c, snow_spring_delta, soil_pct)) base = self.predict(frame(0.0, 0.0, 0.0, 0.0, 0.0)) res.insert(0, "cell_id", ids) res["expected_typical_weather"] = base["expected"].values res["change_vs_typical"] = res["expected"].values - base["expected"].values # the scenario effect return res def load(root: str | Path = Path(__file__).parent) -> ExpectedGreenness: return ExpectedGreenness(root)