Improving Seasonal Forecast Using Probabilistic Deep Learning

Abstract

The path toward realizing the potential of seasonal forecasting and its socioeconomic benefits relies on improving general circulation model (GCM) based dynamical forecast systems. To improve dynamical seasonal forecasts, it is crucial to set up forecast benchmarks, and clarify forecast limitations posed by model initialization errors, formulation deficiencies, and internal climate variability. With huge costs in generating large forecast ensembles, and limited observations for forecast verification, the seasonal forecast benchmarking and diagnosing task proves challenging. Here, we develop a probabilistic deep learning‐based statistical forecast methodology, drawing on a wealth of climate simulations to enhance seasonal forecast capability and forecast diagnosis. By explicitly modeling the internal climate variability and GCM formulation differences, the proposed Conditional Generative Forecasting (CGF) methodology enables bypassing crucial barriers in dynamical forecast, and offers a top‐down viewpoint to examine how complicated GCMs encode the seasonal predictability information. We apply the CGF methodology for global seasonal forecast of precipitation and 2 m air temperature, based on a unique data set consisting 52,201 years of climate simulation. Results show that the CGF methodology can faithfully represent the seasonal predictability information encoded in GCMs. We successfully apply this learned relationship in real‐world seasonal forecast, achieving competitive performance compared to dynamical forecasts. Using this CGF as benchmark, we reveal the impact of insufficient forecast spread sampling that limits the skill of the considered dynamical forecast system. Finally, we introduce different strategies for composing ensembles using the CGF methodology, highlighting the potential for leveraging the strengths of multiple GCMs to achieve advantgeous seasonal forecast.

Hybrid forecasting: using statistics and machine learning to integrate predictions from dynamical models

“…Data-driven models are frequently used for downscaling low-resolution climate model simulations to reduce precipitation bias and make the outputs more skillful at the catchment scale. For instance, Generative Adversarial Networks (GANs) have been used to spatially downscale precipitation forecasts (Harris et al, 2022;Pan et al, 2022) to capture complex joint distributions between precipitation and initial climate conditions from climate simulations. Linear and kernel regression can be used to enhance the skill of decadal CMIP5 precipitation predictions (Salvi et al, 2017a).…”

A Neural Network‐Based Scale‐Adaptive Cloud‐Fraction Scheme for GCMs

“…All yield promising outcomes. For the same reasons, the networks with complex architectures exhibit superiorities in building tools for specific predictions such as El Niño (Ham et al., 2019; Nooteboom et al., 2018), precipitation (G. Chen & Wang, 2022; Ravuri et al., 2021; Shi et al., 2015) and clouds (J. Zhang et al., 2018), and for data processing (Kim et al., 2021; Leinonen et al., 2021; Pan et al., 2019, 2021, 2022; Rasp & Lerch, 2018; Y. Zhang et al., 2021).…”

Climate model-driven seasonal forecasting approach with deep learning

“…Seasonal forecast is defined as a variety of potential climate changes that are likely to occur in the coming months and seasons (Pan et al, 2022). This is crucial for governments and decision makers to better manage natural resources such as water, energy, and agriculture, as well as protect human health (Yuan et al, 2019; Talukder et al, 2021).…”