In the future, monsoon rainfall over densely populated South Asia is expected to increase, even as monsoon circulation weakens1-3. By contrast, past warm intervals were marked by both increased rainfall and a strengthening of monsoon circulation4-6, posing a challenge to understanding the response of the South Asian summer monsoon to warming. Here we show consistent South Asian summer monsoon changes in the mid-Pliocene, Last Interglacial, mid-Holocene and future scenarios, characterized by an overall increase in monsoon rainfall, a weakening of the monsoon trough-like circulation over the Bay of Bengal and a strengthening of the monsoon circulation over the northern Arabian Sea, as revealed by a compilation of proxy records and climate simulations. Increased monsoon rainfall is thermodynamically dominated by atmospheric moisture following the rich-get-richer paradigm, and dynamically dominated by the monsoon circulation driven by the enhanced land warming in subtropical western Eurasia and northern Africa. The coherent response of monsoon dynamics across warm climates reconciles past strengthening with future weakening, reinforcing confidence in future projections. Further prediction of South Asian summer monsoon circulation and rainfall by physics-based regression models using past information agrees well with climate model projections, with spatial correlation coefficients of approximately 0.8 and 0.7 under the high-emissions scenario. These findings underscore the promising potential of past analogues, bolstered by palaeoclimate reconstruction, in improving future South Asian summer monsoon projections.

Hilbert space dimension is a key resource for quantum information processing1,2. Not only is a large overall Hilbert space an essential requirement for quantum error correction, but a large local Hilbert space can also be advantageous for realizing gates and algorithms more efficiently3-7. As a result, there has been considerable experimental effort in recent years to develop quantum computing platforms using qudits (d-dimensional quantum systems with d > 2) as the fundamental unit of quantum information8-19. Just as with qubits, quantum error correction of these qudits will be necessary in the long run, but so far, error correction of logical qudits has not been demonstrated experimentally. Here we report the experimental realization of an error-corrected logical qutrit (d = 3) and ququart (d = 4), which was achieved with the Gottesman-Kitaev-Preskill bosonic code20. Using a reinforcement learning agent21,22, we optimized the Gottesman-Kitaev-Preskill qutrit (ququart) as a ternary (quaternary) quantum memory and achieved beyond break-even error correction with a gain of 1.82 ± 0.03 (1.87 ± 0.03). This work represents a novel way of leveraging the large Hilbert space of a harmonic oscillator to realize hardware-efficient quantum error correction.