Observationally-constrained projections of September ice-free Arctic by selecting optimal fingerprint strategy

The timing of an ice-free Arctic is a key focus of scientific research, yet it remains highly uncertain. Detection and Attribution (D&A) analyses have demonstrated that anthropogenic global warming is the dominant driver of current decline in Arctic sea ice coverage, employing the optimal fingerprinting method. Consequently, projections constrained by D&A results are expected to provide more reliable estimates of the timing of an ice-free Arctic. With the inclusion of anthropogenic aerosol-only simulations in the Detection and Attribution Model Intercomparison Project (DAMIP) of the Coupled Model Intercomparison Project Phase 6 (CMIP6), many studies adopted the fingerprint strategy combines greenhouse gas forcing only, anthropogenic aerosol forcing only, and natural forcing only. However, biases in sea ice responses to aerosol forcing can reduce the accuracy of this fingerprint strategy, thereby compromising the reliability of the constraint-based projection. This study addressed this limitation by comparing four different fingerprint strategies within the perfect model framework. It identifies the all-forcing, greenhouse gas forcing only, and natural forcing only as the optimal fingerprint strategy for attributing Arctic sea ice area changes since the late 1970s. Scaling factors derived from this strategy more effectively correct the model biases in simulating observed sea ice area response to external forcing compared to other strategies. Using this optimal attribution-based constraint projection, the Arctic could become ice-free as early as 2033 under high emission scenario, or by 2037 under a moderate emission scenario, in September.