Towards an unbiased filter routine to determine precipitation and evapotranspiration from high precision lysimeter measurements

摘要

Weighing lysimeters are considered to be the best means for a precise measurement of water fluxes at the interface between the soil-plant system and the atmosphere. Any decrease of the net mass of the lysimeter can be interpreted as evapotranspiration (ET), any increase as precipitation (P). However, the measured raw data need to be filtered to separate real mass changes from noise. Such filter routines typically apply two steps: (i) a low pass filter, like moving average, which smooths noisy data, and (ii) a threshold filter that separates significant from insignificant mass changes. Recent developments of these filters have identified and solved some problems regarding bias in the data processing. A remaining problem is that each change in flow direction is accompanied with a systematic flow underestimation due to the threshold scheme. In this contribution, we analyze this systematic effect and show that the absolute underestimation is independent of the magnitude of a flux event. Thus, for small events, like dew or rime formation, the relative error is high and can reach the same magnitude as the flux itself. We develop a heuristic solution to the problem by introducing a so-called "snap routine". The routine is calibrated and tested with synthetic flux data and applied to real measurements obtained with a precision lysimeter for a 10-month period. The heuristic snap routine effectively overcomes these problems and yields an almost unbiased representation of the real signal. (C) 2017 Elsevier B.V. All rights reserved.