NNFO trainning dataset

BP_inc(7,2580480)为神经网络输入数据,BP_out(2580480)为输出数据(天定反射率)。BP_inc与BP_out与预设条件的对应关系通过以下python程序计算得到:

ratio = [0, 0.2, 0.4, 0.6, 0.8, 1.0]  ## 云水路径/(云水路径+云冰路径)
CWP = [0.001, 0.01, 0.0316, 0.1, 0.3162, 1, 3.1623, 10, 31.6228, 100, 316.2278, 1000]  ## 云水路径+云冰路径,单位为kg m-2
RE_CLOUD = [2, 6, 10, 14, 18, 22, 26, 30] ## 云滴有效粒子半径,单位为um
BRDF = [0.001, 0.05, 0.10, 0.15, 0.20, 0.25, 0.30]  ## 下垫面反照率
SZA = [0, 10, 20, 30, 40, 50, 60, 70]  ## 太阳天顶角,单位为°
VZA = [0, 10, 20, 30, 40, 50, 60, 70]  ## 卫星天顶角,单位为°
SAZ = [0, 20, 40, 60, 80, 100, 120, 140, 160, 180]  ## 太阳与卫星的相对方位角,单位为°

### LUT -- 上述条件下对应的天顶发射率

N0, N1, N2, N3, N4, N5, N6 = len(ratio), len(CWP), len(RE_CLOUD), len(BRDF), len(SZA), len(VZA), len(SAZ)

n_dimy = 7
BP_inc = np.zeros((n_dimy,N0*N1*N2*N3*N4*N5*N6))
BP_out = np.zeros(N0*N1*N2*N3*N4*N5*N6)

ii = 0
for i0 in range(6):
    for i1 in range(N1):
      for i2 in range(N2):
         for i3 in range(N3):
            for i4 in range(N4):
               for i5 in range(N5):
                  for i6 in range(N6):
                      
                      BP_inc[0,ii] = ratio[i0]
                      BP_inc[1,ii] = np.log(CWP[i1])
                      BP_inc[2,ii] = RE_CLOUD[i2]
                      BP_inc[3,ii] = BRDF[i3]
                      BP_inc[4,ii] = SZA[i4]
                      BP_inc[5,ii] = VZA[i5]
                      BP_inc[6,ii] = SAZ[i6]
                      
                      BP_out[ii] =  LUT[i0, i1, i2, i3, i4, i5, i6]
                      
                      ii += 1