#;+ #; :Author: Dr. Jing Wei (Email: weijing_rs@163.com) #;- import os from time import sleep # import gdal import netCDF4 as nc import numpy as np from glob import glob import requests import json import pandas as pd import concurrent.futures # from osgeo import osr #Define work and output paths WorkPath = r'/root/r_base/O3' OutPath = WorkPath #Define air pollutant type #e.g., PM1, PM2.5, PM10, O3, NO2, SO2, and CO, et al. AP = 'O3' #Define spatial resolution #e.g., 1 km ≈ 0.01 Degree SP = 0.01 #Degrees if not os.path.exists(OutPath): os.makedirs(OutPath) path = glob(os.path.join(WorkPath, '*.nc')) #线程运行函数 def thread_work(vll_work, start, end): result_work_array = [] for value_work, lat_ind_work, lon_ind_work in vll_work: params = { "lng": "{:.6f}".format(lon[lon_ind_work]), "lat": "{:.6f}".format(lat[lat_ind_work]) } flag = True while(flag): try: response_work = requests.get(url="http://103.116.120.27:9527/queryPoint",params = params) flag = False except Exception as e: print(f"请求错误一次：{e}") sleep(10) pass res_json_work = json.loads(response_work.text) #坐标在国内 list_local_work = res_json_work['v']['list'] if len(list_local_work) > 0: try: if len(list_local_work) == 1: province_city_work = [local_work['ext_path'] for local_work in list_local_work if local_work['deep'] == '0'] tmp_result_work = [province_city_work[0], province_city_work[0], value_work] result_work_array.append(tmp_result_work) else: province_city_work = [local_work['ext_path'] for local_work in list_local_work if local_work['deep'] == '1'] tmp_result_work = [province_city_work[0].split(" ")[0], province_city_work[0].split(" ")[1], value_work] result_work_array.append(tmp_result_work) except Exception as e: print("发生成异常"+json.dumps(list_local_work)) else: print(f"这是一个空的坐标：{lon[lon_ind_work]}，{lat[lat_ind_work]}\n") # if len(result_work_array) % 100 == 0 : # print(f"当前线程处理开始{start},结束{end}, 已经处理的个数为{len(result_work_array)}\n") return result_work_array for file in path: #全部年份的 file_path = "result.csv" #提取出来年份 year = file.split("_")[4] print(f"当前处理年份{year}") f = nc.Dataset(file) #Read SDS data data = np.array(f[AP][:]) #Define missing value: NaN or -999 data[data==65535] = np.nan #-999 #Read longitude and latitude information lon = np.array(f['lon'][:]) lat = np.array(f['lat'][:]) #获取非空索引 indices = np.where(~np.isnan(data)) #获取非空值 values = data[indices] #拼接（value, lat, lon） vll = list(zip(values, indices[0], indices[1])) #继续索引记录文件地址 index_path = "index_"+year+".txt" # 尝试以读取模式打开文件 try: with open(index_path, 'r') as file: # 如果文件存在，读取文件内容 index = file.readline() # 如果文件不存在，则新建文件并写入内容 except FileNotFoundError: index = 0 vll = vll[int(index):] #将一年的数据拆分成多大一块 batch_size = 100000 total_len = len(vll) if total_len == 0: continue batch_start = 0 # 多少个线程 max_workers = 10 with concurrent.futures.ThreadPoolExecutor(max_workers) as executor: for i in range(total_len // batch_size + 1): # 尝试以读取模式打开文件 try: # 如果文件存在，读取文件内容 result_all = pd.read_csv(file_path) # 如果文件不存在，则新建文件并写入内容 except FileNotFoundError: result_all = [] batch_end = min(batch_start + batch_size, total_len) vll_one = vll[batch_start:batch_end] batch_start = batch_end result_array = [] #并行调用接口获取坐标对应城市 start = 0 avg = len(vll_one)//max_workers remainder = len(vll_one) % max_workers all_task = [] for i in range(max_workers): if i < remainder: end = start + avg + 1 else: end = start + avg all_task.append(executor.submit(thread_work, vll_one[start:end], start, end)) start = end for future in concurrent.futures.as_completed(all_task): data = future.result() result_array = result_array + data #相同地区求平均 columns = ['province', 'city', year] result_df = pd.DataFrame(result_array, columns=columns) if len(result_all) == 0 : result_all = result_df.groupby(['province', 'city']).mean().reset_index() else: result_one = result_df.groupby(['province', 'city']).mean().reset_index() #合并 if year in result_all.columns: print("============新加的数据================") print(result_one) concatenated_df = pd.concat([result_all[['province', 'city', year]], result_one]) # 使用 groupby 进行聚合 grouped_df = concatenated_df.groupby(['province', 'city']).mean().reset_index() result_all = pd.merge(result_all, grouped_df, on=['province', 'city'], how='outer', suffixes=('', '_total')) #替换掉 result_all[year] = result_all[year+"_total"] result_all = result_all.drop([year+"_total"], axis=1) else: result_all = pd.merge(result_all, result_one, on=['province', 'city'], how="outer") print("============合并后的数据================") print(result_all.head()) result_all.to_csv("result.csv",index=False) with open(index_path, 'w') as file: # 如果文件存在，读取文件内容 file.write(str(batch_end+int(index))) # LonMin,LatMax,LonMax,LatMin = lon.min(),lat.max(),lon.max(),lat.min() # N_Lat = len(lat) # N_Lon = len(lon) # Lon_Res = SP #round((LonMax-LonMin)/(float(N_Lon)-1),2) # Lat_Res = SP #round((LatMax-LatMin)/(float(N_Lat)-1),2) # #Define Define output file # fname = os.path.basename(file).split('.nc')[0] # outfile = OutPath + '/{}.tif' .format(fname) #Write GeoTIFF # driver = gdal.GetDriverByName('GTiff') # outRaster = driver.Create(outfile,N_Lon,N_Lat,1,gdal.GDT_Float32) # outRaster.SetGeoTransform([LonMin-Lon_Res/2,Lon_Res,0,LatMax+Lat_Res/2,0,-Lat_Res]) # sr = osr.SpatialReference() # sr.SetWellKnownGeogCS('WGS84') # outRaster.SetProjection(sr.ExportToWkt()) # outRaster.GetRasterBand(1).WriteArray(data) # print(fname+'.tif',' Finished') # #release memory # del outRaster f.close() # result_all.reset_index(inplace=True) # print(result_all.head()) # result_all.to_csv("result.csv",index=False)