As unmanned aerial vehicles (UAVs) become increasingly prominent in modern naval reconnaissance, their operational effectiveness evaluation has transcended mere equipment performance testing, bearing strategic significance for revealing operational patterns, guiding tactical innovation, and supporting mission decision-making. However, traditional evaluation methods often focus only on correlation, failing to adequately control confounding bias caused by variables such as aircraft type and mission environment, thereby undermining decision credibility. To address this, we propose a Continuous Dose-Response Causal Effect Estimation Network, which decouples covariates into instrumental, confounding, and adjustment factors via an expert attention mechanism and employs a varying-coefficient network with integral probability metric constraints to achieve robust causal effect estimation for continuous operational decision variables. Validated on semi synthetic datasets generated from declassified data, the proposed model outperforms multiple mainstream baselines in prediction performance. Further case analysis reveals a significant nonlinear relationship between reconnaissance altitude and mission performance, with a global optimal interval? the causal effect exhibits substantial heterogeneity across different aircraft types, payloads, and mission types. This study provides a causal analysis perspective for data-driven UAV operational effectiveness evaluation.