Prediction and forecasting

Prediction and forecasting of hydrologicalvariability is a growing area of research within Euro-FRIEND Project 3. The method of Bayesian merging developed to correct biases in climate model seasonal forecasts was adapted to evaluate the capability of Global Climate Model (GCM) to forecast Europe’s climate several month ahead. Precipitation hindcasts with 1- to 3-month lead time of seven ensemble runs developed for the European DEMETER project were compared with the ENSEMBLES gridded observational dataset and biases of each model assessed. Monthly hindcasts where merged with rainfall climatology according to the performances of their corresponding GCM to produce probabilistic ensemble forecasts. These forecasts where disaggregated to a daily time-step using an analogue-based technique, generating 20-member ensembles time series for each grid, then input into the G2G gridded hydrological model to generate daily river flow hindcasts

(a) observed rainfall, (b) historical climatology, (c) Bayesian merged DEMETER hindcasts ensemble mean and (d) river flow hindcasts (in mm) for February 1995 (28 Feb for river flow).

Initial results show that the merged forecasts perform better than climatology for 1-month lead time; but, at a 3-month lead, biases are generally too large for forecasts to significantly outperform the climatology (Lavers; Prudhomme; Hannah).

© 2016 EURO-FRIEND

Top