用Python处理OMI L3 Gridded.he5数据并输出为geotiff图像

以OMI传感器的L3级Gridded数据为例（全球尺度，0.25 x 0.25度分辨率，文件格式为he5(HDF-EOS5)），首先用gdalinfo命令打开一个文件查看信息，如下图：

这里可以看到OMI L3 Gridded数据有4个子数据集，这里将提取第四个数据（ColumnAmountNO2TropCloudScreened）集进行处理。在gdalinfo里查找到目标子数据集的相关信息，如下图：

其中可以看到MissingValue和FillValue的值，需要进行处理，设置为NA；还可以看到offset值和scalefactor分别为0和1，说明图像没有位移和缩放；最后可以看到单位为摩尔每平方厘米，可以换算成克每平方米，这样小数点位数不会太多。HDF格式的图像文件比如MODIS的都可以这样处理，有需要的小伙伴可以拿去用。
代码如下：

"""
部分代码引用自Python GDAL/OGR Cookbook 1.0 documentation
新建一个geotiff单波段图像，长宽为he5数据的列数和行数（columns & rows）
GeoTransform坐标是从左上角开始到右下角结束
设置新的图像的spatial reference为地理坐标系（EPSG4326）
具体信息可以在“https://www.spatialreference.org/ref/epsg/”查询
"""

import os
import gdal
import osr
import numpy as np


def array2raster(newRasterfn, rasterOrigin, pixelWidth, pixelHeight, array):    
    cols = array.shape[1]  # obtain cols
    rows = array.shape[0]  # obtain rows
    originX = rasterOrigin[0]  # upper left corner X
    originY = rasterOrigin[1]  # upper left corner Y

    format = 'GTiff'
    driver = gdal.GetDriverByName(format)
    
    # create a single band raster
    outRaster = driver.Create(newRasterfn, cols, rows, 1, gdal.GDT_Float32)
    # set GeoTransform parameters
    outRaster.SetGeoTransform((originX, pixelWidth, 0, originY, 0, pixelHeight))
    # read band 1
    outband = outRaster.GetRasterBand(1)
    outband.WriteArray(array)
    # EPSG4326
    outRasterSRS = osr.SpatialReference()
    outRasterSRS.ImportFromEPSG(4326)
    outRaster.SetProjection(outRasterSRS.ExportToWkt())
    outband.FlushCache()

def main(newRasterfn, rasterOrigin, pixelWidth, pixelHeight, array):
    reversed_arr = array[::-1]
    array2raster(newRasterfn, rasterOrigin, pixelWidth, pixelHeight, reversed_arr)

# find .he5 files and process
in_dir = r'G:\DATA\OMINO2_L3\he5' # input dir
out_dir = r'G:\DATA\OMINO2_L3\tif' # output dir
file_list = os.listdir(in_dir)
for file in file_list:
    if file.endswith('.he5'):
        print('Processing >>> ' + file)
        src_ds = gdal.Open(os.path.join(in_dir, file))
        # open sub dataset
        sub_ds = src_ds.GetSubDatasets()
        # # print some info
        # print('The number of sub-datasets is : {}'.format(len(sub_ds)))
        # for sd in sub_ds:
        # print('Name: {0}\nDescription:{1}\n'.format(*sd))
        no2_ds = gdal.Open(sub_ds[3][0]).ReadAsArray() # NO2 tropcloudscreened = 4th
        # so2_ds = gdal.Open(sub_ds[1][0]).ReadAsArray() # SO2 = 2nd
        # date cleaning
        # set Filling/Missing Value (-1.2676506e+30) to NaN
        data = no2_ds[:]        
        data[data > 2e+16] = np.nan
        data[data < 0] = np.nan
        # molecules/cm^2 to grams/m^2
        # data = data * (1 / (6022 * 10**20)) * 46 * 10**7

        if __name__ == '__main__':
            # keep date in output files
            fn = os.path.splitext(file)[0][19:28]
            fn = fn.replace('m', '_')
            newRasterfn = os.path.join(out_dir, fn + '.tif')
            # define upper left corner and pixel size
            rasterOrigin = (-180, 90)
            x_size = 0.25
            y_size = -0.25
            print('Writing ... ' + newRasterfn)
            main(newRasterfn, rasterOrigin, x_size, y_size, data)
    else:
        print("No '.he5' file found ...")
print('... ... ... ... COMPLETED ... ... ... ...')