16:02 〜 16:17
[MZZ47-04] Examination of the statistical forecast method of sunshine duration in Hokkaido
Photovoltaics (PV) is the one of the useful energy productions for mitigating global warming. This energy is depending on solar radiation and sunshine duration, so these electric productions have large variations. Therefore, PV is called as the unstable power generation. Electric power companies have to keep electric power balance between supply and demand. The large forecasting error makes energy power shortage or surplus. Therefore, high accuracy forecasting method is needed for stable energy supply by renewable energy. Hokkaido is attention PV penetration area in recent years. However, this area is considered PV unsuitable power plant because of heavy snowfall area. Although PV in Hokkaido enough electric power can be expected, due to do not reache baiu/meiyu front. But There aren’t almost studies to forecast solar radiation or sunshine duration in Hokkaido. Therefore, in this study, we used only surface observation data in Hokkaido and examined the statistical forecast method for sunshine duration.
This study used hourly data of see level pressure (SLP) in station of Japan meteorological Agency (JMA) at all 22 stations in Hokkaido. We accumulated the sunshine duration data between 9 to 13 hour at 6 stations ( Wakkanai, Rumoi, Abasiri, Obihiro, Nemuro, Suttu). The forecast method is used principal component analysis (PCA) of SLP in Hokkaido. Result of PCA, first principal component (Z1) indicated East-West SLP pattern, second principal component (Z2) indicated North-South SLP pattern and cumulative contribution rate is 90.0%. We forecasted accumulated sunshine duration using the statisical relationship between the principal scores (Z scores) of SLP at 9:00 and accumulated sunshine duration.
As the results, the Mean Bias Errors (MBE) of accumulated sunshine duration were -0.1 to 0.1 hour, the Root Mean Square Errors (RMSE) on almost stations were about 1.7 hour. The largest value of RMSE was shown at Nemuro (1.8 hour). It is considered that the advection fog caused the forecast error in Nemuro. This fog come from Pacific Ocean and occurs almost every day in summer. This events were difficult to prediction using only SLP patterns. In order to find out the main factor of forecast error in Nemuro, we forecasted accumulated sunshine duration using calculated Z score in each season. The MBEs in summer were overestimated only in Nemuro, and the RMSE shown highest in Nemuro. Of cause, the other factors (eg., sea ice, topography) also considered to affect forecast accuracy. The drift sea ice are changed surface wind fields, topography are blocked low level cloud, and so on. It is difficult to consider the effect of several factors with only SLP patterns.
This study used hourly data of see level pressure (SLP) in station of Japan meteorological Agency (JMA) at all 22 stations in Hokkaido. We accumulated the sunshine duration data between 9 to 13 hour at 6 stations ( Wakkanai, Rumoi, Abasiri, Obihiro, Nemuro, Suttu). The forecast method is used principal component analysis (PCA) of SLP in Hokkaido. Result of PCA, first principal component (Z1) indicated East-West SLP pattern, second principal component (Z2) indicated North-South SLP pattern and cumulative contribution rate is 90.0%. We forecasted accumulated sunshine duration using the statisical relationship between the principal scores (Z scores) of SLP at 9:00 and accumulated sunshine duration.
As the results, the Mean Bias Errors (MBE) of accumulated sunshine duration were -0.1 to 0.1 hour, the Root Mean Square Errors (RMSE) on almost stations were about 1.7 hour. The largest value of RMSE was shown at Nemuro (1.8 hour). It is considered that the advection fog caused the forecast error in Nemuro. This fog come from Pacific Ocean and occurs almost every day in summer. This events were difficult to prediction using only SLP patterns. In order to find out the main factor of forecast error in Nemuro, we forecasted accumulated sunshine duration using calculated Z score in each season. The MBEs in summer were overestimated only in Nemuro, and the RMSE shown highest in Nemuro. Of cause, the other factors (eg., sea ice, topography) also considered to affect forecast accuracy. The drift sea ice are changed surface wind fields, topography are blocked low level cloud, and so on. It is difficult to consider the effect of several factors with only SLP patterns.