Drought is a principal determinant of yield variability in rain-fed wheat systems, with climate change expected to exacerbate both the frequency and severity of water deficits. However, knowledge gaps remain in quantifying (i) yield loss probability across different drought indices and (ii) the dynamic thresholds at which drought induces yield losses under divergent climate scenarios. A systematic quantification of these relationships is essential to improve the empirical foundation for risk assessment and adaptive strategies in water-limited agricultural systems. This study analyses future wheat yield loss probability and dynamic drought thresholds in southeastern Australia using the APSIM model and copula functions, comparing a soil water index (SPAWI) against a precipitation index (SPI). We found a higher future wheat yield loss probability for SPAWI-based drought (5-20% greater than for SPI), underscoring the limitation of rainfall-only indices by neglecting soil buffer effects during drought. Drought thresholds were higher for SPAWI than SPI, due to soil moisture buffering, and lower in wetter regions. Including CO2 fertilization increases yields and partially offsets drought impacts, lowering both loss probabilities and thresholds, while climate-model choice remains the dominant source of projected threshold shifts. Our analysis demonstrates that drought index selection influences yield-loss risk projections, and the quantified shifts in drought yield thresholds under climate change reveal key soil moisture buffering effects and CO2 mitigation potential. These findings provide evidence-based drought thresholds to guide adaptive management in dryland wheat cropping systems under climate change.