Ground-based networks have been developed to determine the spatiotemporal distribution of aerosols using radiometers. In this study, the accuracy of the calibration constant (V0) for the sky radiometer (POM-02) which is used by SKYNET was investigated. The temperature dependence of the sensor output was also investigated, and the dependence in the 340, 380, and 2200 nm channels was found to be larger than for other channels, and varied with the instrument. In the summer, the sensor output had to be corrected by a factor of 1.5 to 2 % in the 340 and 380 nm channels and by 4 % in the 2200 nm channel in the measurements at Tsukuba. In the other channels, the correction factors were less than 0.5 %. The accuracy of V0 from the normal Langley method is between 0.2 and 1.3 %, except in the 940 nm channel. The effect of gas absorption was less than 1 % in the 1225, 1627, and 2200 nm channels. The degradation of V0 for shorter wavelengths was larger than that for longer wavelengths. The accuracy of V0 estimated from the side-by-side measurements was 0.1 to 0.5 %. The V0 determined by the improved Langley (IML) method had a seasonal variation of 1 to 3 %. The RMS error from the IML method was about 0.6 to 2.5 %, and in some cases, the maximum difference reached 5 %. The trend in V0 after removing the seasonal variation was almost the same as for the normal Langley method. The calibration method for water vapor in the 940 nm channel was developed using an empirical formula for transmittance. The accuracy of V0 was better than 1 % on relatively stable and fine days. A calibration method for the near-infrared channels, 1225, 1627, and 2200 nm, was also developed. The logarithm of the ratio of the sensor output can be written as a linear function of the airmass, by assuming that the ratio of the optical thicknesses between the two channels is constant. The accuracy of V0 was better than 1 % on days with good conditions.