Applying transfer function-noise modelling to characterize soil moisture dynamics: a data-driven approach using remote sensing data

Michiel Pezij,Denie C.M Augustijn,Dimmie M.D Hendriks,Suzanne J.M.H Hulscher

doi:10.1016/j.envsoft.2020.104756

Michiel Pezij, Denie C.M Augustijn + Show 2 more

Open Access

https://doi.org/10.1016/j.envsoft.2020.104756

Copy DOI

Abstract

The increasing availability of remotely sensed soil moisture data offers new opportunities for data-driven modelling approaches as alternatives for process-based modelling. This study presents the applicability of transfer function-noise (TFN) modelling for predicting unsaturated zone conditions. The TFN models are calibrated using SMAP L3 Enhanced surface soil moisture data. We found that soil moisture conditions are accurately represented by TFN models when exponential functions are used to define impulse-response functions. A sensitivity analysis showed the importance of using a calibrated period which is representative of the hydrological conditions for which the TFN model will be applied. The IR function parameters provide valuable information on water system characteristics, such as the total response and the response times of soil moisture to precipitation and evapotranspiration. Finally, we encourage exploring the possibilities of TFN soil moisture modelling, as predicting soil moisture conditions is promising for operational settings.

Highlights

Soil moisture is a key component of the hydrological cycle, linking surface and subsurface hydrological processes (Entekhabi and Rodriguez-Iturbe, 1994; Vereecken et al, 2016)
The results show that transfer function-noise (TFN) modelling using the Predefined Impulse Response Function In Continuous Time (PIRFICT) method can be applied to predict surface soil moisture conditions in the Twente region using SMAP surface soil moisture remote sensing data as cali bration set
We studied the applicability of transfer function-noise modelling (TFN) for describing and predicting soil moisture dynamics

Summary

Introduction

Soil moisture is a key component of the hydrological cycle, linking surface and subsurface hydrological processes (Entekhabi and Rodriguez-Iturbe, 1994; Vereecken et al, 2016). Three methods exist for estimating soil moisture at various spatiotemporal scales: in situ (Dobriyal et al, 2012; Susha Lekshmi et al, 2014), remote sensing (Fang and Lakshmi, 2014; Petropoulos et al, 2015; Zhuo and Han, 2016) and hydrological modelling (Vischel et al, 2008; Zhuo and Han, 2016). Remote sensing and hydrological modelling are alternative sources for providing spatially distributed soil moisture information at larger scales. Sensed soil moisture information is often retrieved using active and passive microwave sensors (Petropoulos et al, 2015). Vegetation dynamics, surface roughness, and satellite sensor uncertainty significantly affect remote sensing retrievals (Petropoulos et al, 2015; Benninga et al, 2019)

Methods

Results

Discussion

Conclusion