AI 中文总结
本文通过对达尔瓦克拉通剖面的地震数据重新分析，采用自举法等技术校准广角地壳模型的不确定性，明确莫霍面深度约束及速度模型范围，指出采样与模型类别不确定性需分别报告。
AI 中文摘要
广角地壳模型的深度不确定性通常通过传播假设的拾取误差获得，但同一炮集的拾取结果相互关联，且速度-深度权衡作用于模型类别层面。本文重新研究了横穿达尔瓦克拉通的210公里佩鲁尔-奇克马加卢尔剖面，使用永久识别的2000个初级P1-P2初至波记录，以及来自7炮的9919个反射波，同时保留852个后续转弯的P4观测值用于条件测试。在固定生产速度下，基于互易性的快速行进表可负担地重新拟合嵌套完整炮集验证内的全部6个反射层，以及接收器坐标块自举法。预留的反射波预测的均方根（RMS）为68毫秒，标准化均方残差为2.07，且85%的覆盖率处于标称95%的限值内；校准具有强相位依赖性，其中莫霍面反射的RMS为73毫秒，标准化均方残差为2.12。初至波的RMS从全数据的17毫秒降至未触及炮的103毫秒，这是由孔径外推和对风化近地表的参数化盲区导致的。精确窗口均值确定莫霍面位于西部块体下方46.9公里，中部块体下方39.4公里，原始高差为7.47公里。在炮内重采样接收器得到±0.10公里的条件波动；删除整炮得到±0.40公里，且两项独立实验均表明原始高差被低估了0.36-0.48公里。可接受的速度模型范围为6.46-8.64公里/秒。1260个接收器处的成对低速层（LVL）边界反射在所有位置均超出不确定性范围，尽管内部速度形态仍未解决。必须分别报告采样不确定性和模型类别不确定性。
英文摘要
Depth uncertainties for wide-angle crustal models are usually obtained by propagating an assumed picking error, yet picks from one shot gather are mutually dependent and velocity-depth trade-offs act at the level of the model class. We revisit the 210-km Perur-Chikmagalur profile across the Dharwar Craton using a permanently identified ledger of 2,000 primary P1-P2 first arrivals and 9,919 reflections from seven shots, with 852 later-turning P4 observations held aside for a conditional test. At fixed production velocity, a reciprocity-based fast-marching table makes it affordable to refit all six reflectors inside nested complete-shot validation and a receiver-coordinate block bootstrap. Withheld reflections are predicted to 68 ms RMS with standardized mean-squared residual 2.07 and 85 per cent coverage inside nominal 95 per cent limits; calibration is strongly phase-dependent, the Moho reflection specifically giving 73 ms and 2.12. First arrivals degrade from a 17 ms all-data RMS to 103 ms on untouched shots, driven by aperture extrapolation and a parameterization blind to the weathered near-surface. Exact window means place the Moho at 46.9 km beneath the western block and 39.4 km beneath the central block, a raw contrast of 7.47 km. Resampling receivers within shots gives a +/-0.10 km conditional spread; deleting whole shots gives +/-0.40 km, and two independent experiments agree the raw contrast is biased low by 0.36-0.48 km. Admissible velocity models span 6.46-8.64 km. Paired LVL-boundary reflections at 1,260 receivers separate beyond uncertainty everywhere, though the internal velocity shape stays unresolved. Sampling uncertainty and model-class uncertainty must be reported apart.
Comments56 pages, 13 figures, 14 supplementary figures
