Elsevier

Advances in Water Resources

Volume 86, Part A, December 2015, Pages 170-183
Advances in Water Resources

Improving root-zone soil moisture estimations using dynamic root growth and crop phenology

https://doi.org/10.1016/j.advwatres.2015.10.001Get rights and content

Highlights

  • A WEB-SVAT model is extended to a multi-layer model with dynamic root distribution.

  • Impacts of plant root variations and phenological cycle are considered in new model.

  • Measurements collected from two sites are used to validate the new model.

  • New model provides improved soil moisture, transpiration and evaporation predictions.

Abstract

Water Energy Balance (WEB) Soil Vegetation Atmosphere Transfer (SVAT) modelling can be used to estimate soil moisture by forcing the model with observed data such as precipitation and solar radiation. Recently, an innovative approach that assimilates remotely sensed thermal infrared (TIR) observations into WEB-SVAT to improve the results has been proposed. However, the efficacy of the model-observation integration relies on the model's realistic representation of soil water processes. Here, we explore methods to improve the soil water processes of a simple WEB-SVAT model by adopting and incorporating an exponential root water uptake model with water stress compensation and establishing a more appropriate soil-biophysical linkage between root-zone moisture content, above-ground states and biophysical indices. The existing WEB-SVAT model is extended to a new Multi-layer WEB-SVAT with Dynamic Root distribution (MWSDR) that has five soil layers. Impacts of plant root depth variations, growth stages and phenological cycle of the vegetation on transpiration are considered in developing stages. Hydrometeorological and biogeophysical measurements collected from two experimental sites, one in Dookie, Victoria, Australia and the other in Ponca, Oklahoma, USA, are used to validate the new model. Results demonstrate that MWSDR provides improved soil moisture, transpiration and evaporation predictions which, in turn, can provide an improved physical basis for assimilating remotely sensed data into the model. Results also show the importance of having an adequate representation of vegetation-related transpiration process for an appropriate simulation of water transfer in a complicated system of soil, plants and atmosphere.

Introduction

Soil moisture is a key requirement in environmental monitoring and hydrological prediction. Within the past several decades, various modelling techniques have been developed to simulate soil water movement coupled with biophysical and surface energy processes [1], [2], [3]. Soil-vegetation-atmosphere transfer (SVAT) schemes represent one of the most advanced modelling approaches, which can be applied to monitor water and energy exchanges [4]. The broad category of SVAT schemes include a wide variety of approaches that partition surface net radiation into sensible and latent heat fluxes [5]. There are some structural differences between these methods in their treatment of surface temperature, which divides the SVAT approaches into two main categories: RS (Remote Sensing) and WEB (Water and Energy Balance) -SVAT [5]. In the RS-SVAT category, thermal-infrared (TIR) satellite observations are used as an input to diagnostic models for estimating instantaneous surface energy flux components of net radiation. Since RS-SVAT models do not contain prognostic equations for soil temperature or soil moisture states, they can only generate estimates for non-continuous instances in which remotely-sensed surface temperature retrievals are available. The Two-Source Energy Balance (TSEB) model [6] and the Surface Energy Balance (SEBAL) model [7] are examples of RS-SVAT models.

In the second category, prognostic water and energy balance (WEB) models are forced by meteorological input to assess the temporal variations of soil moisture and surface states. The main difference of this approach from the RS-SVAT models is that surface radiometric temperature is solved for via an internal surface energy balance which is coupled with a prognostic water balance calculation. As a result, WEB-SVAT models do not require remotely-sensed surface temperature observations and can be run continuously using meteorological observations. This enables the integration of TIR data or some of the products of RS-SVAT methods (e.g. predicted soil moisture or evapotranspiration) via data assimilation to improve the soil moisture estimations. Typically, WEB-SVAT models are more complicated than RS-SVAT schemes and often contain a larger number of parameters not only for partitioning the surface energy budget but also as a multi-objective model for predicting surface water states and fluxes as well.

WEB-SVAT modelling approaches obtain energy flux predictions by parameterizing components of the surface energy balance, i.e. net radiation (RN), sensible heat (H), latent heat (LE), and ground heat flux (G) as a function of surface aerodynamic temperature (Taero), soil and vegetation properties, forcing variables such as precipitation and incoming solar radiation, thereby numerically solving the following energy balance equation: RN=H+LE+G.

The model combines the energy flux predictions with vertical subsurface soil water modelling in order to predict the soil moisture content continuously [8].

Although the WEB-SVAT models share a common conceptual basis, the parameterization of equations can vary for different models resulting in predictions with a large variation [5]. The Common Land Model (CLM) [1] and Noah [3] are some examples of WEB-SVAT models. They have been developed as complex models with multiple soil layers in order to suit a wide range of applications.

The estimates from WEB-SVAT models contain some uncertainty due to model parameters and forcing errors. In addition, model deficiency in representing the land surface processes makes the results more uncertain. Data assimilation methods can be used to reduce random soil moisture prediction errors by constraining the model with ground or remotely sensed observations of surface states. The connection between surface and root zone states of the model plays an important role in data assimilation techniques designed to produce superior profile soil moisture estimations by propagating surface information into deeper layers of soil [9]. In this study we evaluate how a more realistic representation of the subsurface processes improves the surface-root zone linkage and affects model performance in estimating surface and root-zone states. Subsequently, the improvements can potentially lead to more appropriate results when surface observations are assimilated into the model.

The surface and root-zone coupling of WEB-SVAT models can be improved by adopting dynamic vegetation components. In general, WEB-SVAT methods parameterize plant transpiration using a stomatal resistance formulation that relates root zone water availability to canopy conductance based on atmospheric demand. Root depth variation plays an important role in estimating soil water loss by transpiration because it determines the soil depth from which the roots can extract water. Plants cannot extract water beyond this depth and water available for transpiration depends on water content within this depth. The importance of dynamic nature of the plant root system is not limited to growth stage-based varying rooting depth; rather it includes root density and distribution. Considering uniform distribution and constant density – which is the case in most of WEB-SVAT models – assumes that the amount of water transpired by the plant is uniformly distributed in the whole profile of the soil while in reality plant behaves differently due to non-homogeneous nature of root. As a result, dynamic rooting depth and distribution should be considered as a part of the model to simulate the soil water movement more realistic [10]. On the other hand, soil textural properties, vegetation type and characteristics of various species as well as growth stages and phenological cycle of the plant have an influence on transpiration activity and these effects may change the estimation of root zone soil moisture [11]. Incorporating biophysical and seasonal plant behaviour using observable indices make it possible to improve the vegetation parameterisation in evapotranspiration (ET) estimation and consequently root-zone soil moisture estimation of the model. Integrating WEB-SVAT models with other models simulating vegetation dynamics can make the aboveground-underground connection stronger; however, WEB-SVAT models will be more complex and some unexpected outcomes may be encountered.

Some crop models have plant growth and development linked dynamically to soil properties and moisture state driven by meteorological forcing. In these models, rooting depth and density is dynamically simulated as a function of various parameters such as soil moisture and temperature, crop and soil characteristics, climate and management variables such as irrigation and fertilizer application [12]. Although representation of vegetation components is typically more sophisticated in these models, their hydrological parameterisation for soil moisture and energy balance estimation is less complex than that in WEB-SVAT models. However, incorporating a relatively simple root growth model to a WEB-SVAT model can integrate the advantages of the crop models – considering the interaction of root depth and distribution with soil water content – with WEB-SVAT models.

A relatively simple WEB-SVAT model has been developed for some studies to make it more efficient to be integrated with remotely sensed observation systems and its state variables more compatible with observations (e.g., [8], [13], [14]). In order to make a distinction between the name of this specific WEB-SVAT model and more generic WEB-SVAT, the former model is called SWEB-SVAT hereafter.

The SWEB-SVAT model uses a two-layer soil-vegetation equations based on a force-restore model proposed by Noilhan and Planton [4] and adapted by Montaldo et al. [15] for the soil water balance with a very similar aerodynamic resistance structure and radiation parameterization to the parallel version of two source energy balance (TSEB) [6]. This model adopts the sophisticated scheme of TSEB model which provides separate estimation of transpiration and evaporation. Apart from the relative simplicity that enables easy modification of the soil water movement scheme and the vegetation parameterisation, using the SWEB-SVAT model has other advantages. Independent observations of surface temperature and soil moisture can be assimilated into the model to update the model predictions. In addition, not only the remotely sensed ET from various one-source algorithms, but also the separate estimate of transpiration derived from TSEB or other two-source models can be integrated in to the model. Structural similarity between SWEB-SVAT and TSEB makes it more robust in the choice of observations that can be assimilated into the model.

The SWEB-SVAT model has been used to propose the concept that root-zone soil moisture can be improved by assimilating thermal remote sensing observations into the model (e.g. [8], [14]). Although these attempts demonstrated some improvements, it appears that the over-simplifying the root biophysical processes or the transpiration parameterisation of the model may produce errors which hamper the ability of data assimilation to constrain root-zone soil moisture [9]. As is for many land surface models, adding more layers with assigning proportion of root distribution to each layer helps to simulate water percolation between the layers and extracting water from each layer based on the root contribution [16], [17].

In this paper the SWEB-SVAT model is extended to incorporate an exponential root water uptake model with water stress compensation proposed by Li et al. [17]. The model is modified to have more layers in order to provide a realistic dynamic root distribution. The impacts of plant root depth variations, growing stages and phenological cycle of plant on transpiration are considered in developing stages and the new developed model is validated using different data sets.

Section snippets

SWEB-SVAT model

The SWEB-SVAT adopts two soil layers, surface- and root-zone layers where the surface layer is a part of the root-zone layer. Surface and root-zone soil moisture variations are calculated by the water balance equation represented as: dθszdt=C1dsz[PgLES(ρλ)1]C2fd(θszθeq)dθrzdt=1drz[PgLEC(ρλ)1]LES(ρλ)1Qwhere θsz and θrz are surface- and root-zone soil moisture contents, dsz and drz surface- and root-zone depth, Pg throughfall, LES and LEC soil and canopy latent heat flux, ρ density of

Impacts of adding vegetation components

In order to evaluate the impact of each proposed change to the model, the three cases detailed above are compared with the baseline using the Ponca dataset. Soil moisture and latent heat estimates at 2 pm for each case are compared with the original SWEB-SVAT and new MWSDR models – including all of vegetation-based changes. The discrepancy between measured and simulated values is used to calculate a coefficient of variation (Cv) which is defined as the ratio of the root mean square error (RMSE)

Discussion and conclusions

Results of the experiments show that separate application of a NDVI-based canopy resistance (case 1) or adding four separate root layers (case 2) does not improve soil moisture and latent heat estimations significantly. Comparing error statistics of case 2 and 3 with SWEB-SVAT model indicates that having multiple root layers does not affect the final simulation of soil moisture unless the assumption of uniform root – and water uptake – distribution changes. Considering dynamic root density in

Acknowledgements

The authors would like to thank Rodger I. Young, Akuraju Venkata Radha and Robert Pipunic for their logistical support in operating and maintaining the Dookie site of the University of Melbourne as well as data collection and archiving. The valuable help from Fuqin Li with the initial programming of SWEB-SVAT model is also gratefully acknowledged. USDA is an equal opportunity provider and employer.

References (42)

  • NijssenB. et al.

    Global retrospective estimation of soil moisture using the variable infiltration capacity land surface model

    J Clim

    (2001)
  • HogueT.S. et al.

    Evaluation and transferability of the Noah land surface model in semiarid environments

    J Hydrometeorol

    (2005)
  • NoilhanJ. et al.

    A simple parameterization of land surface processes for meteorological models

    Mon Weather Rev

    (1989)
  • CrowW. et al.

    Intercomparison of spatially distributed models for predicting surface energy flux patterns during SMACEX

    J Hydrometeorol Sect

    (2005)
  • KumarS.V. et al.

    Role of subsurface physics in the assimilation of surface soil moisture observations

    J Hydrometeorol

    (2009)
  • GaylerS. et al.

    Assessing the relevance of sub surface processes for the simulation of evapotranspiration and soil moisture dynamics with CLM3.5: comparison with field data and crop model simulations

    Environ Earth Sci

    (2013)
  • ReichP.B. et al.

    The evolution of plant functional variation: traits, spectra, and strategies

    Int J Plant Sci

    (2003)
  • BrunF. et al.

    Brun models: evaluation, analysis, parameterization, and applications

    (2006)
  • CrowW. et al.

    Comparison of adaptive filtering techniques for land surface data assimilation

    Water Resour Res

    (2008)
  • MontaldoN. et al.

    Robust simulation of root zone soil moisture with assimilation of surface soil moisture data

    Water Resour Res

    (2001)
  • GaylerS. et al.

    Incorporating dynamic root growth enhances the performance of Noah-MP at two contrasting winter wheat field sites

    Water Resour Res

    (2014)
  • Cited by (9)

    • Data assimilation of high-resolution thermal and radar remote sensing retrievals for soil moisture monitoring in a drip-irrigated vineyard

      2020, Remote Sensing of Environment
      Citation Excerpt :

      The fine-scale WEB-SVAT model is applied to continuously track soil moisture evolution via a water balance approach. To facilitate a multiple soil layer configuration, the original force-restore model employed in Crow et al. (2008) is adapted to a simple conceptual representation of soil water source-sink processes (Hashemian et al., 2015). One surface layer at 0–0.05 m depth, and four subsurface layers are defined at 0.05–0.3, 0.3–0.6, 0.6–0.9, and 0.9–1.2 m, respectively.

    • Understanding long-term (1982–2013) patterns and trends in winter wheat spring green-up date over the North China Plain

      2017, International Journal of Applied Earth Observation and Geoinformation
      Citation Excerpt :

      Rather than only considering the impact of precipitation on phenology (Cong et al., 2013; Yu et al., 2003), we used remote sensing to directly measure soil moisture. In addition to the role of winter snow (approximately 10% of annual precipitation) in crop green-up, irrigation in early spring also plays an important role in the soil moisture conditions (Hashemian et al., 2015; Seghieri et al., 2009). The correlation analysis also showed that precipitation had non-significant impacts on soil moisture at each corresponding period with r ranging from 0.08 to 0.34 (p > 0.05).

    • Estimation of root zone storage capacity at the catchment scale using improved Mass Curve Technique

      2016, Journal of Hydrology
      Citation Excerpt :

      This variable directly determines the available water in the root zone that can be used by vegetation and thus controls the extent and type of vegetation (Tang et al., 2015). Besides, it affects evapotranspiration by controlling the maximum amount of water in root zone which is potentially available for vegetation transpiration (Hashemian et al., 2015). Further, it greatly influences the discharge and soil water storage, and thus the long-term water balance of a catchment (Zhang et al., 2001).

    View all citing articles on Scopus
    View full text