Solar Energy at Forecast Applications Branch

The LAPS system (laps.noaa.gov) is being used to produce rapid update, high resolution analyses and forecasts of solar radiation, also known as global horizontal irradiance (GHI). LAPS is highly portable and can be run onsite, particularly when high-resolution and rapid updating is needed. This allows the user to assimilate their own observational data merged with centrally available observations and to set up the analysis/forecast configuration to their liking. The cloud analysis uses satellite (including IR and 1-km resolution visible imagery, updated every 15-min), METARs, radar, aircraft and model first guess information to produce an hourly 3-D field of cloud fraction, cloud liquid, and cloud ice. The cloud analysis and satellite data together are used to produce a gridded analysis of total solar radiation. The analysis is verified against solar radiation measurements that are independent (not used in the analysis). Several domains are being run and verified presently at NOAA/ESRL/GSD/FAB.

The total solar radiation forecast is being run on most domains, where the WRF model is being initialized by LAPS, using the same cloud analysis package that drives the analysis fields mentioned above. The Colorado domain produces hourly analyses and hourly LAPS/WRF 5-km forecasts initialized every 6 hours. It is verified with about 20 Oklahoma mesonet stations. The HWT domain is running 1-km analyses every 15 minutes. LAPS/WRF 3-km forecasts are initialized every hour with 15 minute forecast output. We have the ability to run forecasts down to around 1-km resolution.

Real-time verification of the analyses and forecasts (including images) can be seen at this link, and help illustrate some metrics that can be used to validate Global Horizontal Irradiance. Red lines are the mean analyzed or forecast total (direct plus indirect) solar radiation for all the stations in the domain. Green is the mean of the observed. Blue is the RMS error comparing the analyzed/forecast to the observed. For analysis plots, purple dots gauge the hypothetical clear sky radiation values. Generally the red curves are anticipated to agree with the green curves. Blue curves near zero indicate a perfect agreement. When the red/green curves are near the purple dots in the analysis, we have a clear sky regime. When they are below the purple dots it is a cloudy regime. Thus we can evaluate the analysis performance on both clear and cloudy days.

The domain with the most available GHI observations is the "HWT" domain, covering all of the Oklahoma and West Texas mesonets with about 150 pyranometers, available via MADIS from the Oklahoma and West-Texas mesonets. These can be considered of medium quality, and are most useful for Bias and RMSE verification on cloudier days when the errors of the analysis/forecast can be larger than the observations. On clearer days, the large number of stations present can help in refining estimates of bias, assuming the observation errors follow a gaussian distribution. While it is true that some residual systematic bias errors are possible with the MADIS observations, we are finding this type of verification does provide useful guidance in making improvements to the LAPS system. We are hoping to acquire the more accurate (though less numerous) SURFRAD observations as an additional dataset to verify against, once they have closer to real-time availability.

A second link is a powerpoint explaining more about how the system works. We are also doing some related work with all-sky imagery.

Real-Time Verification and Imagery | Powerpoint Summary | AGU Poster (2012) | All-Sky Imagery

Contact: Steve.Albers@noaa.gov