POWER V10 hourly GHI, DNI, DHI, GTI, etc. based on CERES SYN1deg(Ed4.1) since 2001¶

Hourly GHI, DHI and DNI and the correction¶

The CERES SYN1deg (Ed4.1) provides all-sky hourly GHI, DHI, and DirHI along with hourly solar zenith angle (SZA) on a 1° latitude by 1° longitude grid system, and the data cover March 2000 to near present.

The hourly GHI agree well with the Baseline Surface Radiation Network (BSRN) data, but the hourly DHI is positively biased against the BSRN, and the DNI, derived by dividing the DirHI by the cosine of SZA, is negatively biased, which mean the errors of DHI and DirHI largely cancel each other when they are summed up to get GHI.

In spite of being biased against the BSRN, the uncertainties of DHI and DNI as represented by the standard deviations of their errors are smaller than the ones from the DirIndex model results. Therefore, we decided to make a bias correction of the hourly DHI and DNI. Since the SZA and cloud fraction (CLFR) are two dominant factors in the variability of DHI and DNI and on an hourly basis, the latitude (ϕ) also plays a role, the bias correction is performed in the ϕ-cos (SZA)-CLFR 3D phase space.

To prepare for the hourly correction, the latitude, ϕ from pole to pole is first divided into 5 latitudinal zones. In each zone, the ranges of both cos (SZA) and CLFR are divided into 100 equal parts, and this forms a 3D matrix in the ϕcos(SZA)-CLFR phase space with 5x100x100 3D bins. The DNI bias against the BSRN data is then calculated in each of the bins, and the relative bias, R_bias, is calculated as the ratio of the bias to that of the mean DNI. The resulting R_bias forms a 3D lookup table in the ϕcos (SZA)-CLFR space.

The correction of the hourly DNI is then performed as follows: For each original hourly DNI, first look for its R_bias according to its corresponding latitudinal zone, cos (SZA) and CLFR, then multiply it by the factor (1 - R_bias).

The correction of the hourly DHI is done in the same fashion.

Table 1 shows the comparison statistics before the correction.

The daily and monthly means of DNI and DHI in local time are then calculated from the corrected GMT hourly mean DNI and DHI.

Table 1 shows the hourly comparison statistics before the correction. Tables 2 and 3 show the hourly comparison statistics of the DNI and DHI, respectively, along with that of their daily and monthly means.

Table 1. CERES SYN1deg (Ed4.1)-BSRN hourly GHI, DHI, DirHI, and DNI comparison statistics over the period from 2000-03 to 2020-12. The Greek letter ρ stands for correlation coefficient; σ for standard deviation of the differences from BSRN; μ_DATA for the mean of the parameter; N is the number of data points being compared. Note that the DirIndex model results over the period 2000-04 to 2016-12 are included for intercomparison. The units of bias, RMS, σ, and μ_DATA are W/m2.

Table 2. CERES SYN1deg (ED4.1)-BSRN direct normal irradiance (DNI) comparison from 2000-04 to 2020-12 after the bias correction has been performed. The definitions of symbols are the same as in Table 1.

Table 3. CERES SYN1deg(ED4.1)-BSRN diffuse horizontal irradiance (DHI) comparison from 2000-04 to 2020-12 after the bias correction has been performed. The definitions of symbols are the same as in Table 1.