Technical note: Improving the AWAT filter with interpolation schemes for advanced processing of high resolution data

Abstract

Abstract. Weighing lysimeters with appropriate data filtering yield the most precise and unbiased information for precipitation (P ) and evapotranspiration (ET). A recently introduced filter scheme for such data is the AWAT (Adaptive Window and Adaptive Threshold) filter (Peters et al., 2014). The filter applies an adaptive threshold to separate significant from insignificant mass changes, guaranteeing that P and ET are not overestimated, and uses a step interpolation between the significant mass changes. In this contribution we show that the step interpolation scheme, which reflects the resolution of the measuring system, can lead to unrealistic prediction of P and ET, especially if they are required in high temporal resolution. We introduce linear and spline interpolation schemes to overcome these problems. To guarantee that medium to strong precipitation events abruptly following low or zero fluxes are not smoothed in an unfavourable way, a simple heuristic selection criterion is used, which attributes such precipitations to the step interpolation. The three interpolation schemes (step, linear and spline) are tested and compared using a data set from a grass-reference lysimeter with 1 min resolution, ranging from 1 January to 5 August 2014. The selected output resolutions for P and ET prediction are 1 day, 1 h and 10 min. As expected, the step scheme yielded reasonable flux rates only for a resolution of 1 day, whereas the other two schemes are well able to yield reasonable results for any resolution. The spline scheme returned slightly better results than the linear scheme concerning the differences between filtered values and raw data. Moreover, this scheme allows continuous differentiability of filtered data so that any output resolution for the fluxes is sound. Since computational burden is not problematic for any of the interpolation schemes, we suggest always using the spline scheme.

Prediction of soil evaporation measured with weighable lysimeters using the FAO Penman–Monteith method in combination with Richards’ equation

“…This step function is subsequently interpolated using spline functions to produce a time series with 1-min temporal resolution of noise-removed and interpolated lysimeter weights. Further information and an extensive validation of this method can be found in Hannes et al (2015) and Peters et al (2014Peters et al ( , 2016Peters et al ( , 2017.…”

“…To smooth the time series of mass measurements obtained with 1-min resolution and avoid smoothing errors on the estimation of precipitation and evapotranspiration, the adaptive window and adaptive threshold filter method (AWAT filter) developed by Peters, Nehls, Schonsky, and Wessolek (2014); Peters, Nehls, and Wessolek (2016); and Peters et al (2017) was used. To manage time-variable noise levels, the filter uses an adaptive threshold (δ) and adaptive averaging window width (ω) (Peters et al, 2016). For our calculations, d varied between 0.02 (0.02 kg) and 0.2 mm (0.2 kg), and w between 3 and 31 min.…”

Lysimeter based evaporation and condensation dynamics in a Mediterranean ecosystem

“…b) Time-series smoothing: is necessary to remove noise from the time series before the partitioning and data analysis based on ∆W (Schrader et al, 2013). Since noise in this type of data is not constant in time due to wind for example (Nolz et al, 2013), we apply a routine with adaptive averaging window widths (ω) and adaptive ∆W thresholds (δ) (AWAT) from Peters et al (2014Peters et al ( , 2016Peters et al ( , 2017. As an intermediate result, the AWAT algorithm produces a step function of lysimeter weight.…”

Terrestrial water loss at night: global relevance from observations and climate models

“…A minimum of 5 min and a maximum of 45 min are assumed for the moving-average window, and a minimum of 0.01 mm and a maximum of 0.25 mm are assumed for the threshold values to distinguish signal from noise. A piecewise cubic Hermitian spline is used to interpolate between points of significant mass change (Peters et al, 2016), after applying an 85th percentile "snap routine" at inflection points (Peters et al, 2017). We estimate dew formation from hourly weight increases in the lysimeter when a co-located rain gauge does not record precipitation in that hour or the next.…”