Optimized GNSS tropospheric delay model for high-precision dam deformation monitoring in large height difference environments

Tropospheric delay is a critical error in Global Navigation Satellite System (GNSS) high-precision positioning, particularly in dam deformation monitoring requiring millimeter-level accuracy. In small-area relative positioning, tropospheric delays between stations are often mitigated through inter-station differencing. However, in large height difference environments common in dam monitoring scenarios (e.g., stations spanning reservoir banks and mountain valleys), residual tropospheric delays persist after differencing, severely degrading vertical (Up direction) positioning accuracy. While analytical models approximate these residuals, their performance remains limited in complex terrains due to insufficient error characterization. This study proposes a Gradient Boosting Decision Tree (GBDT) machine learning model that integrates practical meteorological data to optimize the relative tropospheric delay correction for dam deformation monitoring under extreme height variations. Observation data of GNSS stations and meteorological parameters with a height difference of up to near 700 m were used to verify the validity of the model. Results show that the model effectively captures nonlinear tropospheric delay variations in height-varying dam environments. When baseline height differences exceed 400 m, the GBDT method achieves an average vertical positioning accuracy of 6.5 mm, outperforming the ERA5 model and stochastic process estimation by 22% and 28%, respectively. This enhancement directly benefits dam safety monitoring by improving the reliability of vertical deformation measurements. It should be noted that this is a preliminary study and presents a modeling approach rather than a finalized model. The proposed method provides a robust solution for mitigating tropospheric errors in GNSS-based dam monitoring operating in topographically challenging regions, ensuring higher data fidelity for early warning and risk assessment. • Proposes LS and GBDT model to correct tropospheric delay in GNSS dam monitoring with large height differences. • Validation shows a substantial performance gain (22-28%) in vertical positioning over established methods. • Presents a preliminary modeling approach that enhances data fidelity for dam safety early warning.