1 下载安装netCDF4包

使用清华大学的镜像：
pip install -i https://pypi.tuna.tsinghua.edu.cn/simple netCDF4

2 编写程序

2.1 获取netCDF4文件的格式¹

#一、加载包
from netCDF4 import Dataset
import numpy as np
import os


path = "F:\\OCO2.SIF.all.daily.2001.nc"
csv_path = "F:\\test.csv"
dst = Dataset(path, mode='r', format="netCDF4")

# 二、查看nc文件中包含了什么
print(dst)

运行结果：
<class 'netCDF4._netCDF4.Dataset'>
root group (NETCDF4 data model, file format HDF5):
    dimensions(sizes): lat(360), lon(720), doy(92)
    variables(dimensions): float32 lat(lat), float32 lon(lon), int16 doy(doy), float32 all_daily_sif(doy, lat, lon)
    groups:

由上可见CSIF数据all_daily_sif是三维的，具体size为：
$doy(92)\times lat(360)\times lon(720)$
由于CSIF时间分辨率是4天，所以doy维是92；空间分辨率是0.5°，所以lat是360，lon是720。

#三、使用variables.keys()获取每个变量名，使用variables[i].ncattrs()获取每个变量的属性名
print(dst.variables.keys())
for i in dst.variables.keys():
    print('%s: %s' % (i, dst.variables[i].ncattrs()))

运行结果是：
dict_keys(['lat', 'lon', 'doy', 'all_daily_sif'])
lat: ['units', 'standard_name', 'axis', 'long_name']
lon: ['units', 'standard_name', 'axis', 'long_name']
doy: []
all_daily_sif: ['missing_value']

#四、输出每个variables[i]的信息：
print(dst.variables.keys())
for i in dst.variables.keys():
    print(dst.variables[i])
    print('\n')

运行结果：
dict_keys(['lat', 'lon', 'doy', 'all_daily_sif'])
<class 'netCDF4._netCDF4.Variable'>
float32 lat(lat)
    units: degrees_north
    standard_name: latitude
    axis: Y
    long_name: latitude
unlimited dimensions: 
current shape = (360,)
filling on, default _FillValue of 9.969209968386869e+36 used

<class 'netCDF4._netCDF4.Variable'>
float32 lon(lon)
    units: degrees_east
    standard_name: longitude
    axis: X
    long_name: longitude
unlimited dimensions: 
current shape = (720,)
filling on, default _FillValue of 9.969209968386869e+36 used

<class 'netCDF4._netCDF4.Variable'>
int16 doy(doy)
unlimited dimensions: 
current shape = (92,)
filling on, default _FillValue of -32767 used

<class 'netCDF4._netCDF4.Variable'>
float32 all_daily_sif(doy, lat, lon)
    missing_value: -999.9
unlimited dimensions: 
current shape = (92, 360, 720)
filling on, default _FillValue of 9.969209968386869e+36 used

#五、获取所有维度的size大小并输出
for i in dst.dimensions.keys():
    print('%s_sizes: %s' % (i, dst.dimensions[i].size))

运行结果：
lat_sizes: 360
lon_sizes: 720
doy_sizes: 92

2.2 提取nc数据转存为csv²³

# 从cn文件中获取所有的sif、lat、lon值
    IniSif = dst.variables['all_daily_sif'][:].data    
    latdata = dst.variables['lat'][:].data
    londata = dst.variables['lon'][:].data
    
    # 确定行索引名和列索引名，与xy轴方向一致，lat为行索引
    latname = []    
    lonname = []
    for lat in range(0, 360):
        latname.append(str(latdata[lat]))
    for lon in range(0, 720):
        lonname.append(str(londata[lon]))
    
    # 将sif值转换为表格dataframe格式
    sif = IniSif[0][:]
    sifframe = pd.DataFrame(sif, index=latname, columns=lonname)
    for doy in range(1,92):
        # 下面获得第doy天的所有sif
        sif = IniSif[doy][:]
        # 下面将第doy天的sif加上行列索引信息变成表格
        siflist = pd.DataFrame(sif,index = latname,columns =lonname)
        # 下面将已经生成表格合并
        sifframe = pd.concat([sifframe, siflist])

    sifframe.to_csv(csv_path,sep=',')
    dst.close()