Abstract: Rapid and accurate forest fire detection is critical for disaster emergency response, yet the applicability of individual remote sensing indices is highly scenario-dependent. Using multi-temporal Landsat 8 OLI and Sentinel-2 MSI data from 98 fire events in southern China (2019–2021), this study employed a monthly historical statistical baseline and introduced Cohen's d effect size to systematically evaluate the sensitivity of 10 remote sensing indices (NDVI, EVI, NBR, NBR2, NBRT, BAI, NDWI, NDMI, MNDWI, and BSI) to fires. Multi-index combination strategies were then constructed, including the improved cascaded detection method, Two-Stage version 2 (TSv2). Results show that NBR2 exhibited the highest fire sensitivity, with a median |d| of 6.74 and a statistically significant proportion of 91.8%, followed by NDVI (median |d| = 6.08) and NDWI (median |d| = 5.76). Case-scale analysis revealed that burn indices (NBR2, NBRT) were the most sensitive in approximately 68.4% of cases, whereas vegetation and water indices were the most sensitive in about 27.4% of cases, indicating that a single index could not adequately cover all fire scenarios. Multi-index combination evaluation showed that TSv2 performed the best under moderate and high threshold conditions. At the threshold condition of T = 15.0, TSv2 achieved a true positive rate (TPR) of 79.3%, a 32.2 percentage-point improvement over the single-index OR method (47.1%). TSv2 achieved fire detection with a two-stage cascading mechanism of "OR-based, then trimmed mean deviation weighted", with Stage 1 (OR method on any of the three optimal indices) rapidly capturing pronounced fires and Stage 2 (trimmed mean deviation-weighted method) refining borderline cases, thereby establishing a detection framework that is permissive in entry yet rigorous in screening. This study provides methodological support for multi-index synergistic fire remote sensing monitoring, and the proposed cascaded detection framework has broad application value for remote sensing monitoring of weak-signal fires and other disaster types.