location

Home>Multi-Model Ensemble>Introduction

Introduction

Seasonal Forecasts

  • Scientifically feasible due to improvement in numerical modeling
  • Higher uncertainty compared to short-range forecasts

Ensemble Forecasting

  • Main approach to improve forecast skill
  • Single model ensembles reduce the errors due to initial conditions
  • Systematic errors are more important for longer-term forecasts
    → Multi-model ensemble prediction is one of the solution to reduce these errors!

Multi-Model Ensemble (MME) Forecasting

  • An empirical statistical post-processing to produce an optimal forecast from a set of ensemble predictions
  • Computing the weights during training phase for each grid point of each model
  • Linear regression methods: simple ensemble mean, multiple linear regression , EOF-based multiple regression, Singular Value Decomposition
  • Non-linear regression methods: Artificial Neural Network, Genetic Algorithm, Fuzzy Theory
  • Notion of general MME scheme (See Fig.1)
    - Training phase: observed fields provide statistical relationships
Figure 1. The vertical dotted line denotes time t = 0; the area to the left denotes the training phase where a large number of forecast experiments are carried out. During this period, the available observed fields provide statistical relationships, which are then passed on to the area t > 0 (on the right).