MERRA2/GFS Case studies for WRF forecasts

We have looked at shortwave (SW) radiation errors in the MERRA2 and ERA-Interim reanalysis datasets compared with seven stations around Ireland: Belmullet, Birr, Clones, Dublin, Kilkenny, Malin Head and Valentia.

We want to run WRF forecasts for certain case study dates. We will use archived GFS global forecast data to drive WRF forecasts for these dates. GFS archive data are available to download for dates from 2005 (see this earlier post), while some obs data stop in 2008, so we must choose within those years. Continue reading “MERRA2/GFS Case studies for WRF forecasts”

Download archived GFS data

I’d like to run some WRF forecasts for dates on which there were large errors in the shortwave (SW) radiation from reanalysis data. To do this, I need to get forecast data from a global model.

The GFS global model forecast data are archived here: https://www.ncdc.noaa.gov/data-access/model-data/model-datasets/global-forcast-system-gfs

To request archived data, click on the HAS link for the data you want – I’ll use the 1º GFS data, as it goes back further.

Continue reading “Download archived GFS data”

Journal Club: 13-06-2017

Article title: Long-term patterns of European PV output using 30 years of validated hourly reanalysis and satellite data

This article looked at the validation of hourly PV simulations across Europe, using MERRA and MERRA-2 reanalysis and the Meteosat-based CM-SAF SARAH (Surface Solar Radiation Data Set) satellite dataset using 30 years of simulated data.

Simulated and measured hourly power production for a site in Czech Republic

MERRA and MERRA-2 generally overestimate the PV output, while SARAH generally underestimates. This is expected as the satellite derived SARAH is better able to resolve irradiance-relevant weather events that are not properly modelled in MERRA. This is possibly due to the low spatial resolution of the reanalysis and therefore it does not consider the local topography. Other disadvantages include inaccurate cloud modelling and an overestimation of atmospheric transparency during clear-sky conditions (due to insufficient consideration of aerosols).

Uncorrected MERRA has similar accuracy to SARAH therefore both datasets should use some correction. Corrections include correcting a systematic bias by matching the simulations to the mean bias in modelling individual sites. Both datasets exhibit systematic bias but SARAH models the shape of the power output more accurately (see figure above).

Future research:

  • Look at the possibility of using the SARAH dataset.
  • Look at the renewables.ninja website.
  • Potential to advance the post-processing techniques used in this paper.