Abstract: A new outliers detection method for automatic soil moisture observation data based on characteristic curve is proposed. The main and basic idea of this method was feature extraction and the morphological matching between two soil moisture time series, and the detailed operation processes were as follows: firstly, the method took X as the expected checking time series and took Y as the corrected template time series, and also gave the range and elements of these two series. Secondly, the method decomposed series X and Y by empirical mode decomposition (EMD) method to obtain the recomposition series C and Q respectively. In this process, series C was the total accumulation of IMFs of series X and series Q was the total accumulation of IMFs of series Y. Thirdly, the method obtained series C' and Q' by using dynamic time warping (DTW) algorithm which was designed to align series C and Q. Fourthly, the method obtained the variation series D' whose elements were calculated by the variation coefficient between series C' and Q', and then, the method also traversed each element of series D' and marked the elements whose value was greater than threshold as overruns. The threshold was obtained by comprehensive calculating the standard deviation of series X and Y. Finally, the outliers in the checking series X could be found through the mapping relationships between series X and series D'. The example showed that: (1) the method did not need to introduce external factors such as soil physical constants and meteorological conditions, and avoided adding relevant parameters such as high and low boundary and slope in the calculation process. (2) The method used the continuous soil moisture data of the same depth from the same station instead of multi-station data comparison, and had no strict length consistency requirements for series X and series Y, so the calculation was more flexible and applicable. (3) The routine of the method was clear, and all of the input processes and output processes in this method were specific. The method was suitable for computer programming and business operation. The technical route of this method might be provided for other agrometeorological data quality control research.

Key words: Characteristic curve, Soil moisture, EMD, DTW, Outliers