python delta降尺度

import xarray as xr import numpy as np from sklearn.neighbors import KDTree from sklearn.metrics import r2_score, mean_squared_error xr.set_options(display_max_rows=50) xr.set_options(display_width=800) high = xr.open_dataset(r'F:\pr\pr_2000y_12geyue.nc') high_values = high.pr.values extent = high.isel(time=0).pr.values high_lon = high.lon.values high_lat = high.lat.values high_ilon, high_ilat = np.meshgrid(high_lon, high_lat) valid_index = np.where(np.isnan(extent)==False) high_multiyearmonthlymean = high.groupby(high.time.dt.month).mean() high_grids_lonlat = np.vstack((high_ilon[valid_index].ravel(), high_ilat[valid_index].ravel())).T low = xr.open_dataset(r'F:\pr\cru_ts4.07.1901.2022.pre.dat.nc') low_values = low.pre.values low_lon = low.lon.values low_lat = low.lat.values low_ilon, low_ilat = np.meshgrid(low_lon, low_lat) low_grids_lonlat = np.vstack((low_ilon.ravel(), low_ilat.ravel())).T tree = KDTree(low_grids_lonlat, leaf_size=2) low_id = [] for i in range(high_grids_lonlat.shape[0]): dist, ind = tree.query(high_grids_lonlat[i, :].reshape(-1, 2), k=1) low_id.append(ind) resample = np.zeros((low_values.shape[0], high_values.shape[1], high_values.shape[2]), dtype=np.float32) * np.nan for i in range(low_values.shape[0]): low_value = low_values[i, :, :].ravel() resample[i, valid_index[0], valid_index[1]] = low_value[low_id].ravel() resample = xr.Dataset({"pre": (('time', 'lat', 'lon'), resample)}, coords={"time": low.time, "lat": high.lat, "lon": high.lon}) # resample.to_netcdf(r'G:\111kkk\resample.nc') # 使用低分辨率和高分辨率重合时段的数据计算delta resample = xr.open_dataset(r'F:\pr\resample.nc') resample_section = resample.sel(time=resample.time.dt.year.isin(range(1961, 2000))) resample_multiyearmonthlymean = resample_section.groupby(resample_section.time.dt.month).mean() delta = high_multiyearmonthlymean.pre / resample_multiyearmonthlymean.pre # delta.to_netcdf(r'G:\111kkk\delta.nc') result = [] for i in range(1, 13): data = resample.sel(time=resample.time.dt.month==i) * delta.isel(month=i-1) data = xr.Dataset({"pre": (('time', 'lat', 'lon'), data.pre.values.astype("float32"))}, coords={"time": data.time.values, "lat": data.lat.values, "lon": data.lon.values}) result.append(data) bc_down_scale = xr.merge(result) bc_down_scale = xr.Dataset({"pre": (('time', 'lat', 'lon'), bc_down_scale.pre.values.astype("float32"))}, coords={"time": low.time, "lat": high.lat, "lon": high.lon}) bc_down_scale.to_netcdf(r'F:\pr\delta_down_scale.nc')