The influence of aggregation and statistical post-processing on the sub-seasonal predictability of European temperatures

The succession of European surface weather patterns has limited predictability because disturbances quickly transfer to the large scale flow. Some aggregated statistic however, like the average temperature exceeding a threshold, can have extended predictability when adequate spatial scales, temporal scales and thresholds are chosen. This study benchmarks how the forecast skill horizon of probabilistic 2-meter temperature forecasts from the ECMWF sub-seasonal forecast system evolves with varying scales and thresholds. We apply temporal aggregation by rolling window averaging and spatial aggregation by hierarchical clustering. We verify 20 years of re-forecasts against the E-OBS data set and find that European predictability extends at maximum up to week 4. Simple aggregation and standard statistical post-processing extend the forecast skill horizon with two and three skillful days on average, respectively.
The intuitive notion that higher levels of aggregation capture the larger scale and lower frequency variability and therefore tap into an extended predictability, holds in many cases. However, we show that the effect can saturate and that regional optimums exist, beyond which extra aggregation reduces the forecast skill horizon. We expect that such windows of predictability result from specific physical mechanisms that only modulate and extend predictability locally. To optimize sub-seasonal forecasts for Europe, aggregation should in certain cases thus be limited.