Fractional Integral Statistical Model: A new way for climate prediction and projection from the perspective of scaling

It is well recognized that climate memory is one the origins for climate predictability, but how to include the concept of climate memory into the climate prediction, is still an open question. Here in this work, we suggest the Fractional Integral Statistical Model (FISM), a generalized stochastic climate model, as a new way for this purpose. With FISM, one can extract the “forcing-induced direct component ε(t)” and the “memory-induced indirect component M(t)” from a given variable x(t). By predicting ε(t), one can further obtain the predicted x(t) using FISM. Different from traditional prediction approaches which normally focus on x(t), here this new strategy based on FISM clarifies the climate memory impacts. From this new perspective, we have quantified the climate memory induced predictability, and developed a temperature response model that can project the future warming trend. Compared to CMIP6 simulations, our approach projects lower global warming levels over the next few decades. A further examination indicates that many CMIP6 models overestimated the climate memory, which might contribute to the overestimated future warming trend.