Does textural heterogeneity matter? Quantifying transformation of hydrological signals in soils

- We applied a principal component analysis on simulated soil moisture time series.
- Mean temporal dynamics and effect of soil depth explained 86.7% of total variance.
- The nature of signal damping was equal for all textures but the intensity differed.
- Textural heterogeneity did not cause functional heterogeneity among time series.
- Valuable process information could be distinguished from uncorrelated noise.

SummaryTextural heterogeneity causes complex water flow patterns and soil moisture dynamics in soils that hamper monitoring and modeling soil hydrological processes. These patterns can be generated by process based models considering soil texture heterogeneities. However, there is urgent need for tools for the inverse approach, that is, to analyze observed dynamics in a quantitative way independent from any model approach in order to identify effects of soil texture heterogeneity. Here, studying the transformation of hydrological input signals (e.g., rainfall, snow melt) propagating through the vadose zone is a promising supplement to the common perspective of mass flux considerations. In this study we applied a recently developed new approach for quantitative analysis of hydrological time series (i) to investigate the effect of soil texture on the signal transformation behavior and (ii) to analyze to what degree soil moisture dynamics from a heterogeneous profile can be reproduced by a corresponding homogenous substrate. We used simulation models to generate three data sets of soil moisture time series considering homogeneous substrates (HOM), homogeneous substrates with noise added (NOISE), and heterogeneous substrates (HET). The soil texture classes sand, loamy sand, clay loam and silt were considered. We applied a principal component analysis (also called empirical orthogonal functions) to identify predominant functional patterns and to measure the degree of signal transformation of single time series. For the HOM case 86.7% of the soil moisture dynamics were reproduced by the first two principal components. Based on these results a quantitative measure for the degree of transformation of the input signal was derived. The general nature of signal transformation was nearly identical in all textures, but the intensity of signal damping per depth interval decreased from fine to coarse textures. The same functional patterns occurred in the HET data set. However, here the signal damping of time series did not increase monotonically with soil depth. The analysis succeeded in extracting the same signal transformation behavior from the NOISE data set compared to that of the HOM case in spite of being blurred by random noise. Thus, principal component analysis proved to be a very robust tool to disentangle between independent effects and to measure the degree of transformation of the input signal. The suggested approach can be used for (i) data processing, including subtracting measurement noise (ii) identification of factors controlling soil water dynamics, (iii) assessing the mean signal transformation in heterogeneous soils based on observed soil moisture time series, and (iv) model building, calibration and evaluation.