Abstract: Uneven land subsidence may endanger urban development and people's safety. The land subsidence in Tianjin city is particularly serious.Aiming at this problem, this paper collects the leveling observation data of Tianjin city from 2005 to 2012 and from 2016 to 2017. The Kalman filter model is constructed based on the state vector of the settlement amount, settlement rate, and settlement acceleration rate at a fixed level to filter the level observation data in Tianjin city over the years. According to the results of filtering, this paper uses the method of polynomial weighted interpolation to determine the weights by the information of distance, settlement rate and settlement acceleration rate, and to interpolate the land subsidence. The accuracy of various interpolation methods is compared by using the root mean square error as the precision evaluation parameter. Based on the experimental analysis of the interpolation results, the cumulative average ground settlement in Tianjin city from 2005 to 2017 is 394.477 5 mm, and the maximum settlement is 1 143.5 mm. The main settlement areas are Beichen, Dagang, Tanggu and other areas.And with the increase of time, these areas present funnel-like subsidence. Experiments prove that this method combined with Kalman filter and polynomial weighted interpolation can better reflect the spatial and temporal distribution of land subsidence, and predict the settlement in the future, which has a certain reference significance for the urban development and construction of Tianjin.

Key words: leveling data, polynomial weighted interpolation, Kalman filter, land subsidence, spatial-temporal distribution