I would like to verify the most recent C3S multi-system monthly forecasts (t2m) against station observations in the Netherlands. For this purpuse I developed a methodology, but I am not entirely sure whether the methodology is correct. Here is the approach:
- Get hindcast data from all institutes contributing to the C3S-multi-system. This includes the most recent version of each model, with data between 1993 and 2016 (the hindcast period). The dataset is seasonal forecast: monthly statistics on single levels.
- Scale variance per model to mean variance of models per month (should this be done per month per leadtime? In other words, how is this done for the C3S-multi system?
- Compute average scaled ensemble mean anomaly per month and leadtime
- Detrend computed anomalies and station data to account for global warming.
- Compare model to station data covering the same time period.
Could someone comment on my approach; i.e. whether I have missed something?
Thanks!