Regional and seasonal characteristics of forecast errors of downward shortwave radiation by the Japan Meteorological Agency mesoscale model for the forecast of a photovoltaic power production

Forecasts of the downward shortwave radiation (DSW), which have large variability in both the temporal and spatial scales, are necessary for the forecast of a photovoltaic (PV) power production and stabilization of an electric power system with other power generation systems. For forecasting the PV power production, it is also necessary to investigate forecast errors of the DSW by a mesoscale model (MSM) at the Japan Meteorological Agency (JMA) compared with surface observed irradiance dataset. In this study, regional and seasonal characteristics of forecast errors of the DSW values by the MSM in Japan are investigated based on a statistical evaluation during three years from 2008 to 2011. For the surface observation of solar irradiance, the DSW values are measured with a pyranometer at JMA meteorological observations at about 50 stations in Japan. Since the surface DSW datasets measured and calibrated by the JMA are of the highest quality in Japan, surface-observed DSW datasets of the JMA are used in this analysis to validate the DSW. Numerical weather predictions of the DSW are performed with an MSM developed by the JMA. The model domain of MSM is a region surrounding the Japan Islands with a domain size of 3600 km × 2900 km × 21.8 km in the x, y, and z directions, respectively. The horizontal grid spacing is 5 km. The horizontal grid size of the radiative transfer processes is 10 km, since radiative transfer processes are calculated by thinning out every two model grids in the x and y directions to reduce calculation costs. The MSM calculates the radiative transfer processes at 15-min intervals. The forecast outputs of 33-h integration times with four different initialization times (03, 09, 15 and 21 UTC) are used in this analysis. For forecasts of the PV power production for one-day ahead, particularly, the MSM outputs of 03 UTC initialization times are used. Evaluations of forecast errors of the DWS show that the mean bias error (MBE) values of the DSW range from -50 to 50 W/m2 in a year. The root mean square error (RMSE) values in winter are about 90-100 W/m2, while the RMSE values in summer approaches up to 150 W/m2. For the seasonal characteristics of forecast errors of the DSW, evaluations of forecast errors of the DSW indicate that the DSW values are generally underestimated (overestimated) in summer (winter) compared with the observations. The dependence on the different four initialization times is also found, suggesting that the RMSE values for the initial time of 21 UTC are generally smaller than the other (03, 09 and 15 UTC) initial times. As expected, the short lead time of 21 UTC tends to cause generally better DSW forecasts compared with those for the rest of other initialization times. Regarding regional characteristics of forecast errors, the mapping of the MBE and RMSE values in Japan indicate that the annual MBE (RMSE) values in the mid-latitude region from 30N to 42N range from -50 W/m2 to 50 W/m2 (from 100 W/m2 to 140 W/m2), while significant underestimations (under -50 W/m2) and forecast errors (the RMSE values range from 140 W/m2 to 200 W/m2) are seen over the Pacific Ocean in the southern Japan. Forecasts of the DSW for the relatively lower latitude (sub-tropical) region (< 30N) tend to be difficult than those in other regions in Japan. Detailed investigations for forecast errors of the DSW show that the forecasted DSW has regional and seasonal characteristics. Examples of time series of daily and hourly DSW forecasts on each month shows that the DSW forecast in the clear-sky conditions is in line with the observations, while those in the cloudy sky conditions produce often relatively large forecast errors. For cases of relatively large errors in the DSW forecasts in the central part of Japan (84 cases during the three years near Tokyo), in which forecast errors is normalized by the extraterrestrial solar irradiance > 0.2, appearance frequency of cloud types based on visual monitoring indicate that specific cloud types; cirrus (Ci). Altocumulus (Ac), stratiform (St) or cumulus (Cu) clouds, are often observed. Verifications in the DSW forecasts of the MSM at JMA indicated that improvements in cloud radiation processes in MSM; microphysical and radiative transfer processes, are necessary for accurate forecasting the DSW and/or the PV power production.