In Germany, 31% of the largest error in photovoltaic day-ahead power forecasts for the years 2013-2014 were associated with fog and low stratus events. A detection algorithm for low stratus risk (LSR) is developed and applied as post-processing to the NWP model forecasts. The aim of the detection algorithm is to supply day-ahead warnings and support the decision-making process of the TSOs.
The LSR algorithm considers the lowest model levels where low cloud can be produced by the model (~800hPa). Thresholds for the inversion strength and the saturation deficit determine the behaviour of the LSR. With strict criteria, the LSR gives conservative estimations of the low cloud cover (comparable to the operational NWP output) and with less constrictive criteria, broader areas are marked to be potentially affected by low stratus. Even though the LSR inherits the errors of the underlying NWP model, the resulting LSR serves as additional information to the likelihood of low stratus occurrence (fig 1.).
Validation is performed against the Nowcasting Satellite Application Facility (NWCSAF) cloud type classification product. The combination of thresholds for the LSR is dependent on the users different sensitivity to risk as validation produces a large range of results in critical success index (CSI) and false alarm ratio (FAR).
Application in our research:
- Generate a climatology of fog in Ireland.
- Explore the use of NWCSAF for my research.
- Apply similar methods to high resolution ensemble forecasts in Ireland.
- Is this method of probability assessment better than using an ensemble forecast system?