When predicting the surface subsidence in mining areas, the observed data at monitoring
points inevitably exist gross error, leading to the deviation from predicting elevation of
observed points. The principle of robust estimation is used to solve the subsidence
prediction of mining area surface when there exists gross error in observed data. The results
show that when the observed data are mixed with gross error, the least square fitting
prediction data significantly deviate from the actual situation. The subsidence prediction
result by the robust estimation method is consistent with the actual subsidence, which can
truly reflect the subsidence at monitoring points, which provides effective data for ensuring
safe operation in mining areas.