Data filtration methods of field weighing lysimeters with high temporal resolution for soil water balance components determination

Faisal Zeidalkilani

Master's thesis

Abstract:

Weighing lysimeters are state of the art equipment that allow for direct measurement of the terrestrial water balance components. Their temporal and weighing resolution makes them prone to an array of errors, both systematic and random. Random errors can significantly affect lysimetric measurements, where external mechanical disturbances result in stochastic distortions that mask the magnitude of the true measurement signal. Such errors manifest as noise that requires proper filtration prior to utilizing the measured data.This study aims to adopt a suitable filtering scheme and to optimize its parameters by tailoring them to the site’s soil properties, meteorological conditions, and lysimeter specifications. An investigation period of 164 days has been chosen to ensure the process’s representativeness. The lysimeter utilized for the purposes of this study is the SFL-300 (METER Group AG), located at the experimental field of CULS. The comprehensive filtering routine by Hannes et al. (2015) has been employed to filter the lysimeter measurements. The comprehensive filtering routine consists of 5 consecutive filtration steps. The manual, threshold, and median filters address discrete measurement errors. While the subsequent smoothing and oscillation threshold filters tackle the remaining diffuse noise. A prominent feature of the selected filtering scheme is the oscillation threshold filter. Its calculation yielded a value of 12.8 g, compared to the software’s default value of 30 g. When the filtered summarized masses for both the default and calculated parameters are compared, the calculated parameters show a better fit to the raw data than the default ones.Furthermore, in order to validate both parameters results, their cumulative precipitation values were compared to that of a reference rain gauge. The calculated parameters display slightly better correlation with the rain gauge, with a coefficient of determination R2= 0.9993 compared to R2=0.9990 for the default parameters. Additionally, the normalized root mean square error for the calculated parameters was 1.86%, compared to 1.98% for the default ones. While the standard deviation for the calculated parameters was 118.8 mm compared to 116.5 mm for the default parameters.

Keywords

Noise Filtration Methods, Oscillation Threshold Filter, Comprehensive Filtering Routine, Water Balance Components, Weighing Lysimeters

Language used: English

Date on which the thesis was submitted / produced: 24. 7. 2020

Thesis defence

• Supervisor: Ing. Kamila Báťková, Ph.D.

Czech University of Life Sciences Prague

Faculty of Agrobiology, Food and Natural Resources

Master programme / field:
Natural Resources and Environment / Natural Resources and Environment