Learning algorithms allow for improved reliability and accuracy of global mean surface temperature projections

Abstract

Climate predictions are only meaningful if the associated uncertainty is reliably estimated 1-3 .A standard practice for providing climate projections is to use an ensemble of projections.The ensemble mean serves as the projection while the ensemble spread is used to estimate the associated uncertainty 4, 5 . The main drawbacks of this approach are the fact that there is no guarantee that the ensemble projections adequately sample the possible future climate conditions and that the quantification of the ensemble spread relies on assumptions that are not always justified 4, 6 . The relation between the true uncertainties associated with projections and ensemble spreads is not fully understood 5 . Here, we suggest using simulations and measurements of past conditions in order to study both the performance of the ensemble 1 arXiv:1904.06773v1 [physics.ao-ph]

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.…”

BoostTree and BoostForest for Ensemble Learning

“…Finally, we will also apply BoostForest to challenging problems in biology, engineering, healthcare, etc., in which ensemble learning has found many successful applications [3][4][5][6][7][8][9][10][11][12][13] .…”

“…Ensemble learning 1,2 trains multiple base learners to explore the relationship between a set of covariates (features) and a response (label), and then combines them to produce a strong composite learner with better generalization performance. It has been successfully used in biology [3][4][5][6][7] , climate prediction 8 , healthcare 9,10 , materials design 11,12 , Moon exploration 13 , etc. For example, in biology, Wang et al 4 used an ensemble of neural networks to emulate mechanism-based biological models. They found that the ensemble is more accurate than an individual neural network, and the consistency among the individual models can indicate the error in prediction.…”

Trajectory Simulation and Prediction of COVID‐19 via Compound Natural Factor (CNF) Model in EDBF Algorithm

“…However, the CR received less attention in the COVID-19 modeling communities. One of the potential solutions is weighted ensemble method, which is popular in Meteorology (Yoo, et al, 2020), Socioeconomics (Boyce, et al, 2020), and Climatology (Strobach, et al, 2020). Therefore, the primary objective of this study was to quantify the influence of compound natural factor (CNF) on COVID-19 trajectory, and to quantify the contributions of their potential driving factors, including temperature, humidity, visibility, barometric pressure, wind speed, aerosol, and vegetation.…”