import os import numpy as np import pandas as pd from scipy.interpolate import interp1d from scipy.optimize import curve_fit import netCDF4 as nc import matplotlib.pyplot as plt import seaborn as sns import matplotlib.font_manager as fm def process_single_file(base_folder_path, i, lat_range=(30, 40)): # 构建当前文件夹的路径 if i < 10: folder_name = f"atmPrf_repro2021_2008_00{i}" # 一位数，前面加两个0 elif i < 100: folder_name = f"atmPrf_repro2021_2008_0{i}" # 两位数，前面加一个0 else: folder_name = f"atmPrf_repro2021_2008_{i}" # 三位数，不加0 folder_path = os.path.join(base_folder_path, folder_name) # i should be day cache_path_nz = f"{base_folder_path}/{folder_name}_mean_ktemp_Nz.npy" cache_path_ptz = f"{base_folder_path}/{folder_name}_mean_ktemp_Ptz.npy" if os.path.exists(cache_path_nz) and os.path.exists(cache_path_ptz): mean_ktemp_Nz = np.load(cache_path_nz) mean_ktemp_Ptz = np.load(cache_path_ptz) print(f"Loaded cache for {folder_name}") return mean_ktemp_Nz, mean_ktemp_Ptz # 检查文件夹是否存在 if os.path.exists(folder_path): dfs = [] # 遍历文件夹中的文件 for file_name in os.listdir(folder_path): if file_name.endswith('.0390_nc'): finfo = os.path.join(folder_path, file_name) # print(f"正在处理文件: {finfo}") try: dataset = nc.Dataset(finfo, 'r') try: bad = int(dataset.getncattr('bad')) except AttributeError: bad = None if bad == 1: print(f"文件 {finfo} 被标记为坏文件，跳过。") continue # 提取变量数据 temp = dataset.variables['Temp'][:] altitude = dataset.variables['MSL_alt'][:] lat = dataset.variables['Lat'][:] lon = dataset.variables['Lon'][:] # 创建DataFrame df = pd.DataFrame({ 'Longitude': lon, 'Latitude': lat, 'Altitude': altitude, 'Temperature': temp }) dataset.close() # 剔除高度大于60的行 df = df[df['Altitude'] <= 60] df = df[(df["Temperature"] <= 49.85) & (df["Temperature" >= -120.15])] # 对每个文件的数据进行插值 alt_interp = np.linspace( df['Altitude'].min(), df['Altitude'].max(), 3000) f_alt = interp1d( df['Altitude'], df['Altitude'], kind='linear', fill_value="extrapolate") f_lon = interp1d( df['Altitude'], df['Longitude'], kind='linear', fill_value="extrapolate") f_lat = interp1d( df['Altitude'], df['Latitude'], kind='linear', fill_value="extrapolate") f_temp = interp1d( df['Altitude'], df['Temperature'], kind='linear', fill_value="extrapolate") # 计算插值结果 interpolated_alt = f_alt(alt_interp) interpolated_lon = f_lon(alt_interp) interpolated_lat = f_lat(alt_interp) interpolated_temp = f_temp(alt_interp) # 创建插值后的DataFrame interpolated_df = pd.DataFrame({ 'Altitude': interpolated_alt, 'Longitude': interpolated_lon, 'Latitude': interpolated_lat, 'Temperature': interpolated_temp }) # 将插值后的DataFrame添加到列表中 dfs.append(interpolated_df) except Exception as e: print(f"处理文件 {finfo} 时出错: {e}") # 按行拼接所有插值后的DataFrame final_df = pd.concat(dfs, axis=0, ignore_index=True) # 获取 DataFrame 的长度 num_rows = len(final_df) # 生成一个每3000个数从0到2999的序列并重复 altitude_values = np.tile( np.arange(3000), num_rows // 3000 + 1)[:num_rows] # 将生成的值赋给 DataFrame 的 'Altitude' 列 final_df['Altitude'] = altitude_values # 摄氏度换算开尔文 final_df['Temperature'] = final_df['Temperature'] + 273.15 lat_begin, lat_end = lat_range # 筛选出纬度在30到40度之间的数据 latitude_filtered_df = final_df[( final_df['Latitude'] >= lat_begin) & (final_df['Latitude'] <= lat_end)] # 划分经度网格，20°的网格 lon_min, lon_max = latitude_filtered_df['Longitude'].min( ), latitude_filtered_df['Longitude'].max() lon_bins = np.arange(lon_min, lon_max + 20, 20) # 创建经度网格边界 # 将数据分配到网格中 # latitude_filtered_df['Longitude_Grid'] = np.digitize( # latitude_filtered_df['Longitude'], lon_bins) - 1 latitude_filtered_df.loc[:, 'Longitude_Grid'] = np.digitize( latitude_filtered_df['Longitude'], lon_bins) - 1 # 对相同高度的温度取均值，忽略NaN altitude_temperature_mean = latitude_filtered_df.groupby( 'Altitude')['Temperature'].mean().reset_index() # 重命名列，使其更具可读性 altitude_temperature_mean.columns = ['Altitude', 'Mean_Temperature'] # 定义高度的范围（这里从0到最短段） altitude_range = range(0, 3000) all_heights_mean_temperature = [] # 用于存储所有高度下的温度均值结果 for altitude in altitude_range: # 筛选出当前高度的所有数据 altitude_df = latitude_filtered_df[latitude_filtered_df['Altitude'] == altitude] # 对Longitude_Grid同一区间的温度取均值 temperature_mean_by_grid = altitude_df.groupby( 'Longitude_Grid')['Temperature'].mean().reset_index() # 重命名列，使其更具可读性 temperature_mean_by_grid.columns = [ 'Longitude_Grid', 'Mean_Temperature'] # 添加高度信息列，方便后续区分不同高度的结果 temperature_mean_by_grid['Altitude'] = altitude # 将当前高度的结果添加到列表中 all_heights_mean_temperature.append(temperature_mean_by_grid) # 将所有高度的结果合并为一个DataFrame combined_mean_temperature_df = pd.concat( all_heights_mean_temperature, ignore_index=True) # 基于Altitude列合并两个DataFrame，只保留能匹配上的行 merged_df = pd.merge(combined_mean_temperature_df, altitude_temperature_mean, on='Altitude', how='inner') # 计算差值（减去wn0的扰动） merged_df['Temperature_Difference'] = merged_df['Mean_Temperature_x'] - \ merged_df['Mean_Temperature_y'] # 按Altitude分组 grouped = merged_df.groupby('Altitude') def single_harmonic(x, A, phi): return A * np.sin(2 * np.pi / (18 / k) * x + phi) # 初始化存储每个高度的最佳拟合参数、拟合曲线、残差值以及背景温度的字典 fit_results = {} fitted_curves = {} residuals = {} background_temperatures = {} for altitude, group in grouped: y_data = group['Temperature_Difference'].values x_data = np.arange(len(y_data)) wn0_data = group['Mean_Temperature_y'].values # 获取同一高度下的wn0数据 # 检查Temperature_Difference列是否全部为NaN if np.all(np.isnan(y_data)): fit_results[altitude] = { 'A': [np.nan] * 5, 'phi': [np.nan] * 5} fitted_curves[altitude] = [np.nan * x_data] * 5 residuals[altitude] = np.nan * x_data background_temperatures[altitude] = np.nan * x_data else: # 替换NaN值为非NaN值的均值 y_data = np.where(np.isnan(y_data), np.nanmean(y_data), y_data) # 初始化存储WN参数和曲线的列表 wn_params = [] wn_curves = [] # 计算wn0（使用Mean_Temperature_y列数据） wn0 = wn0_data # 对WN1至WN5进行拟合 for k in range(1, 6): # 更新单谐波函数中的k值 def harmonic_func( x, A, phi): return single_harmonic(x, A, phi) # 使用curve_fit进行拟合 popt, pcov = curve_fit(harmonic_func, x_data, y_data, p0=[ np.nanmax(y_data) - np.nanmin(y_data), 0]) A_fit, phi_fit = popt # 存储拟合结果 wn_params.append({'A': A_fit, 'phi': phi_fit}) # 使用拟合参数生成拟合曲线 WN = harmonic_func(x_data, A_fit, phi_fit) wn_curves.append(WN) # 计算残差值 y_data = y_data - WN # 使用残差值作为下一次拟合的y_data # 存储结果 fit_results[altitude] = wn_params fitted_curves[altitude] = wn_curves residuals[altitude] = y_data # 计算同一高度下的背景温度（wn0 + wn1 + wn2 + wn3 + wn4 + wn5） wn_sum = np.sum([wn0] + wn_curves, axis=0) background_temperatures[altitude] = wn_sum # 将每个字典转换成一个 DataFrame df = pd.DataFrame(residuals) # 使用前向填充（用上一个有效值填充 NaN） df.ffill(axis=1, inplace=True) # 初始化一个新的字典来保存处理结果 result = {} # 定义滤波范围 lambda_low = 2 # 2 km lambda_high = 15 # 15 km f_low = 2 * np.pi / lambda_high f_high = 2 * np.pi / lambda_low # 循环处理df的每一行（每个高度） for idx, residuals_array in df.iterrows(): # 提取有效值 valid_values = np.ma.masked_array( residuals_array, np.isnan(residuals_array)) compressed_values = valid_values.compressed() # 去除NaN值后的数组 N = len(compressed_values) # 有效值的数量 # 如果有效值为空（即所有值都是NaN），则将结果设置为NaN if N == 0: result[idx] = np.full_like(residuals_array, np.nan) else: # 时间序列和频率 dt = 0.02 # 假设的时间间隔 n = np.arange(N) f = n / (N * dt) # 傅里叶变换 y = np.fft.fft(compressed_values) # 频率滤波 yy = y.copy() freq_filter = (f >= f_low) & (f <= f_high) yy[~freq_filter] = 0 # 逆傅里叶变换 perturbation_after = np.real(np.fft.ifft(yy)) # 将处理结果插回到result字典中 result[idx] = perturbation_after # 处理背景温度和扰动温度数据格式 heights = list(background_temperatures.keys()) data_length = len(next(iter(background_temperatures.values()))) background_matrix = np.zeros((data_length, len(heights))) for idx, height in enumerate(heights): background_matrix[:, idx] = background_temperatures[height] heights = list(result.keys()) data_length = len(next(iter(result.values()))) perturbation_matrix = np.zeros((data_length, len(heights))) for idx, height in enumerate(heights): perturbation_matrix[:, idx] = result[height] perturbation_matrix = perturbation_matrix.T # 计算 Brunt-Väisälä 频率和势能 heights_for_calc = np.linspace(0, 60, 3000) * 1000 def brunt_vaisala_frequency(g, BT_z, c_p, heights): # 计算位温随高度的变化率 dBT_z_dz = np.gradient(BT_z, heights) # 计算 Brunt-Väisälä 频率，根号内取绝对值 frequency_squared = (g / BT_z) * ((g / c_p) + dBT_z_dz) frequency = np.sqrt(np.abs(frequency_squared)) return frequency # 计算势能 def calculate_gravitational_potential_energy(g, BT_z, N_z, PT_z): # 计算势能 return 0.5 * ((g / N_z) ** 2) * ((PT_z / BT_z) ** 2) g = 9.81 # 重力加速度 c_p = 1004.5 # 比热容 N_z_matrix = [] PT_z_matrix = [] for i in range(background_matrix.shape[0]): BT_z = np.array(background_matrix[i]) PT_z = np.array(perturbation_matrix[i]) N_z = brunt_vaisala_frequency(g, BT_z, c_p, heights_for_calc) PW = calculate_gravitational_potential_energy(g, BT_z, N_z, PT_z) N_z_matrix.append(N_z) PT_z_matrix.append(PW) ktemp_Nz = np.vstack(N_z_matrix) ktemp_Ptz = np.vstack(PT_z_matrix) mean_ktemp_Nz = np.mean(ktemp_Nz, axis=0) mean_ktemp_Ptz = np.mean(ktemp_Ptz, axis=0) # save to cache np.save(cache_path_nz, mean_ktemp_Nz) np.save(cache_path_ptz, mean_ktemp_Ptz) return mean_ktemp_Nz, mean_ktemp_Ptz else: print(f"文件夹 {folder_path} 不存在。") return None, None