Abstract: Fishery-dependent and independent data each have strengths and limitations for estimating abundance indices. Commercial catch-per-unit-effort (CPUE) offers broad spatio-temporal coverage but suffers from gear selectivity and preferential sampling, while scientific surveys provide standardized sampling but limited coverage. Integrating these data sources is particularly challenging in mixed fisheries where multiple gear types with different selectivity patterns operate concurrently. Small yellow croaker (Larimichthys polyactis) in the East China Sea represents such a complex fishery, supporting important commercial fisheries while exhibiting strong spatio-temporal dynamics influenced by environmental conditions and gear-specific catchability. This study aimed to develop a robust CPUE standardization approach for small yellow croaker by integrating multi-gear commercial fishery data and scientific survey data within a Bayesian spatio-temporal modeling framework, evaluating alternative spatial structures and data integration strategies to obtain more reliable abundance indices for stock assessment. We analyzed 39 434 commercial fishing records from 158 vessels operating in September during 2010-2023, covering three major gear types: trawl, gillnet, and stow net, complemented by scientific survey data from 90~120 stations annually. An INLA-based Bayesian spatio-temporal generalized linear mixed model with gamma distribution and log-link was developed, incorporating year effects, gear effects, environmental covariates (depth, distance to coast, bottom temperature, bottom salinity), and their interactions. Models with independent spatial fields substantially outperformed shared spatial field models for both commercial and survey data, with the optimal model (M1) including independent spatial fields, linear environmental effects, and gear-environment interactions achieving the lowest DIC (7 786) and WAIC (7 838) values. Gamma distribution provided superior predictive performance (R²=0.76, RMSE=616) compared to lognormal distribution (R²=0.65, RMSE=784). Gear-environment interactions significantly improved model fit, revealing differential environmental responses: salinity positively affected all gears but most strongly influenced trawl catch rates (effect size 0.262), while distance to coast showed negative effects on trawl (-0.259) and stow net (-0.129) but negligible effects on gillnet. Spatial random effects revealed persistent positive anomalies in the northern East China Sea (30~33°N), indicating this region as core habitat not fully explained by environmental covariates. Annual abundance indices from integrated modeling showed pronounced interannual variability, with peaks in 2015 and notable declines during 2016—2020, followed by recovery in 2022-2023. The INLA-GLMM framework with independent spatio-temporal fields effectively disentangles gear-specific catchability, environmental effects, and true abundance variation, providing a robust foundation for stock assessment and fisheries management of this important species.