Improvement of climate predictions and reduction of their uncertainties using learning algorithms

Abstract

Abstract. Simulated climate dynamics, initialized with observed conditions, is expected to be synchronized, for several years, with the actual dynamics. However, the predictions of climate models are not sufficiently accurate. Moreover, there is a large variance between simulations initialized at different times and between different models. One way to improve climate predictions and to reduce the associated uncertainties is to use an ensemble of climate model predictions, weighted according to their past performances. Here, we show that skillful predictions, for a decadal time scale, of the 2 m temperature can be achieved by applying a sequential learning algorithm to an ensemble of decadal climate model simulations. The predictions generated by the learning algorithm are shown to be better than those of each of the models in the ensemble, the better performing simple average and a reference climatology. In addition, the uncertainties associated with the predictions are shown to be reduced relative to those derived from an equally weighted ensemble of bias-corrected predictions. The results show that learning algorithms can help to better assess future climate dynamics.

A new approach to extended‐range multimodel forecasting: Sequential learning algorithms

“…The skill of the sequential learning algorithms (SLAs) is first examined in the context of deterministic or point‐based skill, as in prior applications of the algorithms for climate prediction (e.g., Strobach and Bel 2015; 2016; 2020). As a subsequent step, the skill improvements of the methods are assessed in a probabilistic context, considering all the quantiles in “Qgrid”.…”

“…Each mixture was applied using all predictors (denoted as BOA, EGA) and the NWP‐based predictors only (denoted as BOA_NWP and EGA_NWP) in order to assess the presence of any added value from including reanalysis‐based information in the forecasts. The EGA method was implemented to benchmark the results against prior uses of sequential learning algorithms in climate prediction (e.g., Strobach and Bel 2015; 2016; 2020), but it has been considered here with several improvements with respect to its prior uses: EGA is trained for each in “Qgrid” using a 2‐year training period and optimizing a fixed learning rate across quantiles over this period.…”

Global Wildfire Outlook Forecast with Neural Networks

“…The uncertainties from both the prediction model and long-range meteorological forecasts need to be quantified. The uncertainty from long-range meteorological forecasts is likely larger [85][86][87]. It is conceivable that the global fire outlook forecast becomes an important part of a long-range meteorological forecast service as the quality of long-range meteorological forecast products improves, the forecast period becomes longer, and the nonlinear forecast model improves.…”

Quantifying the Uncertainties in an Ensemble of Decadal Climate Predictions

“…In order to demonstrate the generality of the methods presented here, we used both an equally weighted ensemble and a weighted ensemble for which the weights were generated by a learning algorithm. The learning algorithm that we used is the Exponentiated Gradient Average (EGA) (Cesa‐Bianchi & Lugosi, ; Kivinen & Warmuth, ) that was shown to outperform the equally weighted ensemble in decadal climate predictions (Strobach & Bel, , ). The learning algorithm used the first 10 years of the simulations to assign weights to the different models, and those weights were then used to generate the predictions for the following 20 years of the simulations.…”