Interpolating Method for Missing Data of Daily Air Temperature and Its Error Analysis
It is the foundation of building up a continuous meteorological datasets for int erpolating the missing meteorological record. A method of estimating missing dai ly air temperature (daily maximum, minimum and average air temperature) for several continuous months is proposed, whichis improved from the linear regression model that are usually used to interpolate the missing data for a single day or several days. A series of improvements are taken on, such as: (1)The number of n eighboring referenced meteorological stations and size of time window about sampl e data are determined by optimized method; (2)The linear regression model is set up between the station of missing data and neighboring referenced stations, and the least absolute deviation(LAD) instead of the least squareroot estimate (LS E)is treated as the object function when the model parameter is solved for higher computing efficiency and parameter stability; (3)In order to reduce extreme err or, the averaged value between LAD method and DeGaetano's standardized method is used as the final result. A lot of interpolating experiments are done with the data of Caidian Meteorology Station, Hubei Province from 1961 to 2006 and the results indicate: (1)The leas t interpolating error can be obtained when 4 neighboring referenced stations, tim e window with 8 years and 15 days is used; (2)The frequency of mean absolute err or of daily maximum, minimum and average air temperature within ±0.8℃ is 94.1% 、84.8%、96.1% respectively. The monthly average and correlation coefficient bet ween actual value and interpolated value are all over 0.886 with the significant level of 0.05 and 0.01 by ttest respectively.