Publikationsdatum:
2017-08-25
Beschreibung:
In his comment, G. Bürger criticizes the conclusion that inflation of trends by quantile mapping is an adverse effect.He assumes that the argument would be ‘‘based on the belief that long-term trends and along with them future climate signals are to be large scale.’’ His line of argument reverts to the so-called inflated regression. Here it is shown, by referring to previous critiques of inflation and standard literature in statistical modeling as well as weather forecasting, that inflation is built upon a wrong understanding of explained
versus unexplained variability and prediction versus simulation. It is argued that a sound regressionbased downscaling can in principle introduce systematic local variability in long-term trends, but inflation systematically deteriorates the representation of trends. Furthermore, it is demonstrated that inflation by construction deteriorates weather forecasts and is not able to correctly simulate small-scale spatiotemporal structure.
Materialart:
Article
,
PeerReviewed
Format:
text