Continuous optical-electrical field diagnostics separate environmental modulation from irreversible perovskite...

Outdoor photovoltaic records confound reversible weather response with irreversible degradation. Here, we establish a continuous optical-electrical field-diagnostic framework that separates shared environmental modulation from persistent cell-specific failure and tests optical timing, transferability and predictive value. We monitored encapsulated perovskite cells in Alpine and Mediterranean climates using synchronised electrical, spectral, thermal, humidity and passive RGB measurements. A fixed C1/C3 peer reference removed the dominant cohort-wide response and exposed four persistent Alpine terminal failures; seasonal recovery of both reference cells showed that the deep winter minimum was predominantly reversible. An analysis-locked retrospective colour replay matched three of four retrospective terminal markers (optical-first offsets of 1, 4 and 6 days), and a locked marker-definition audit then resolved the timing. With an earlier daily-persistent marker, C2 became coincident, C8 electrical-first, and the only optical-first change, at C5/C6, was shared with non-failing C3/C4. The Alpine record therefore establishes day-scale optical-electrical co-evolution around failure, and the same-date C3/C4 control shows that this optical-first change is not device-specific. At Paterna, optical changes occurred 18 and 13 days after electrical onset and did not reproduce the Alpine timing. A date-disjoint hidden test established reliable passive measurement of wetness and thaw, which did not improve three-day health prediction. Endpoint diffraction distinguished PbI₂-rich photocurrent-loss states from the lower-ratio mixed voltage-current failure. Over the common winter window, Alpine survivors recovered while Paterna C3/C4 declined despite overlapping per-cell integrated yields. The framework separates environmental response, irreversible failure, optical timing and cross-climate transfer within a single field-diagnostic workflow.