Elsevier

Field Crops Research

Volume 231, 1 February 2019, Pages 93-104
Field Crops Research

Explicit wheat production model adjusted for semi-arid environments

https://doi.org/10.1016/j.fcr.2018.11.011Get rights and content

Highlights

  • A 50-m spatial resolution wheat production model for semiarid regions is developed.

  • Model integrates 3 sub-models: crop growth, soil hydrology and runoff models on GIS.

  • Model was evaluated vs SWC, biomass (AGB) and grain yield (GY) in 22 fields in Israel.

  • Model shows sensitivity to spatial variations in soils (SWC) and rain (AGB and GY).

  • Significant AGB and GY decreases were predicted for projected warming and drought in the study area.

Abstract

Current literature suggests that wheat production models are limited either to wide-scale or plot-based predictions ignoring pattern of habitat conditions and surficial hydrological processes. We present here a high-spatial resolution (50 m) non-calibrated GIS-based wheat production model for predictions of aboveground wheat biomass (AGB) and grain yield (GY). The model is an integration of three sub-models, each simulating elemental processes relevant for wheat growth dynamics in water-limited environments: (1) HYDRUS-1D, a finite element model that simulates one-dimensional movement of water in the soil profile; (2) a two-dimensional GIS-based surface runoff model; and (3) a one-dimensional process-driven mechanistic wheat growth model. By integrating the three sub-models, we aimed to achieve a more accurate spatially continuous water balance simulation with a better representation of root zone soil water content (SWC) impacts on plant development. High-resolution grid-based rainfall data from a meteorological radar system were used as input to HYDRUS-1D. Twenty-two commercial wheat fields in Israel were used to validate the model in two seasons (2010/11 and 2011/12). Results show that root zone SWC was accurately simulated by HYDRUS-1D in both seasons, particularly at the top 10-cm soil layer. Observed vs simulated AGB and GY were highly correlated with R2 = 0.93 and 0.72 (RMSE = 171 g m−2 and 70 g m−2) having low biases of -41 g m−2 (8%) and 52 g m−2 (10%), respectively. Model sensitivity test showed that HYDRUS-1D was mainly driven by spatial variability in the input soil characteristics while the integrated wheat production model was mostly affected by rainfall spatial variability indicating the importance of using accurate high-resolution rainfall data as model input. Using the integrated model, we predict decreases in AGB and GY of c. 10.5% and c. 12%, respectively, for 1 °C of warming and c. 7.7% and c. 7.3% for 5% reduction in rainfall amount in our study sites. The suggested model could be used by scientists to better understand the causes of spatial and temporal variability in wheat production and the consequences of future scenarios such as climate change.

Introduction

Crop models are increasingly being used in agriculture to support decision-making and planning at a large spatial scale, particularly in relation to risks associated to climate change issues such as decreases in yield and/or soil fertility (Rosenzweig et al., 2014; Schauberger et al., 2017; Wu et al., 2016). Among these models, process-based crop growth simulation models are the most useful in predicting crop responses to climate because they attempt to explain the mechanism of the processes rather than simply provide parameter predictions such as statistical models and models based on empirical relationships do (Challinor et al., 2009; Chenu et al., 2017; Holzkämper, 2017). Crop growth models have been originally developed at the plot scale and more recently used and evaluated at the regional and global scale but with a rather coarse spatial resolution. These models have been used to study the impacts of projected climate change over large agricultural areas (Challinor et al., 2016; Deryng et al., 2014; Zhao et al., 2017) but need to be adjusted for smallholding farm applications.

Process-based crop models are used to simulate crop growth dynamics and yield predictions under diverse management practices, climate and environmental conditions (Chenu et al., 2017). These models are either driven by plot-based meteorological information acquired from weather stations providing estimates at the plot scale (e.g. Brisson et al., 2003; Stöckle et al., 2003), or by grid-based modelled climate data, providing estimates at a regional to global scale but at a very coarse spatial resolution (at least a few to hundreds of kilometers, e.g. Stöckle et al., 2014). Due to a lack of spatial continuity in input weather data and/or the too coarse spatial resolution of climate data, these models suffer from a poor representation of water dynamics at the root zone essential to feedback process-based mechanistic crop models. The lack of a proper surface physics simulation, such as SWC dynamics, may result in model uncertainties, particularly in crops growing under water-limited conditions (Helman et al., 2019; Rajala et al., 2009).

To provide a better application to local conditions model predictions are often downscaled using either statistical or dynamic downscaling approaches (Fowler et al., 2007). However, such downscaling efforts might introduce biases in terms of unaccounted specific temperature extremes or rainfall patterns that could bias the simulated growth dynamics and yield (Cammarano et al., 2013; Cammarano and Tian, 2018; Hansen and Jones, 2000). Moreover, downscaled results may involve a high degree of uncertainty related to the original model parameters and model structure (Palosuo et al., 2011).

To overcome these drawbacks, the use of a one-dimensional mechanistic crop model is suggested for local applications (Boogaard et al., 1998; Brisson et al., 2009). The one-dimensional mechanistic model uses meteorological information acquired from a near weather station as an input instead of using a grid-based modelled climate data, providing estimates at the field/farm level. These kinds of models have been successfully applied in the estimation of wheat, maize and rice yields providing a potential decision supporting tool for local farming (Attia et al., 2016; Corbeels et al., 2016; Lopez et al., 2017).

Though promising, it requires the use of many input variables, which are often unavailable or difficult to acquire. It also lacks the spatially continuous representation needed to cope with issues of spatial variability in local soils and other environmental characteristics, which have been shown to affect crop yield production (Jégo et al., 2015). A combination of approaches is therefore suggested for merging process-based models using GIS tools to simulate crop growth processes over large areas with a relatively high spatial resolution (Huffman et al., 2015; Liu, 2009; Thorp et al., 2008).

A Digital Elevation Model (DEM) layer, for example, may provide high-resolution grid-based information relevant for computing topographic models (Tarboton, 1997) and potential incident radiation load (Fu and Rich, 2002), necessary for a more accurate simulation of plant growth in the one-dimensional mechanistic crop model (Amir and Sinclair, 1991a, 1991b). Meteorological radars may provide a more accurate grid-based rainfall data than statistical interpolation between weather stations, particularly in areas where rainfall is highly variable in time and space (Marra and Morin, 2015; Morin and Gabella, 2007).

Integrating the grid-based information with a simple mechanistic plant growth model in a geographic information system (GIS) environment may provide a spatially continuous estimation of crop production and yield at a very high spatial resolution (of a few tens of meters). Moreover, the high-resolution spatial hydrologic information may be used as an input into a one-dimensional hydrological model to improve SWC dynamics simulations providing a more accurate information on the available SWC at the root zone, which makes it an important factor in crop development and productivity (Lawes et al., 2009), particularly in water-limited environments (Acevedo et al., 1999; Amir and Sinclair, 1991b; Helman et al., 2018).

Here, we combine a one-dimensional Richard’s equation-based hydrological model with a two-dimensional runoff model and a mechanistic wheat growth model (Amir and Sinclair, 1991b, 1991a) on a grid-based GIS platform to provide high resolution aboveground wheat biomass and grain yield estimates. The integrated model, linking different processes on a grid-based platform, may be important for estimating AGB and GY variations in our region that present variable topography and significant slopes. To assess the importance of our integrated approach we first tested the spatial and temporal accuracy of the integrated model in 22 commercial wheat fields in Israel, validating the model with in-situ SWC, AGB and GY measurements (Section 3.1). Then, we examined the spatial sensitivity of the models through fixing each input parameter per model run (Section 3.2). Finally, we used the model to simulate field-level AGB and GY variations expected from projected rainfall and temperature changes in our region (Section 3.3).

Section snippets

Study area

The model was used in the study area located at Ramot Yssakhar (32°35′38.76″N, 35°27′48.96″E; Fig. 1a), which is a relatively hilly region in Northern Israel. At Ramot Yssakhar, there are numerous crop fields with a total agricultural area of c. 30 km2. Elevation in Ramot Yssakhar ranges between -150 and c. 270 m above sea level (m.a.s.l.) with a more narrow range of 30 and 260 m.a.s.l. at the area of the 22 fields used in this study (Fig. 1c). Typical slopes are 1° to 30°, with an average

Hydrus -1D

Fig. 3a,b show the seasonal evolution of observed (sensors) and modelled (HYDRUS-1D) volumetric SWC in the top-soil layers at location A (see Fig. 1c). HYDRUS-1D showed high sensitivity to rainfall, which was in accordance with observed SWC at the shallower soil depth (5–10 cm, Fig. 3a), but less at the deeper 30-cm soil layer (Fig. 3b). The lower sensitivity to rain events of observed SWC at the deeper layer was likely due to low percolation rate, which was less successfully simulated by HYDRUS

Discussion

The integrated model presented in this study links between water infiltration into the soil, water runoff flow and crop growth modeling at a high spatial resolution (50 m). The linkage between surface water dynamics and crop growth is of a great importance, particularly at the start of the season prior to the increase in surface roughness following vegetation growth. The use of three well-established sub-model components allowed to better quantifying water redistribution, which plays an

Acknowledgments

David Helman is a Fulbright Fellow 2018/2019. This research was supported by a grant from the Chief Scientist of the Israeli Ministry of Agriculture and Rural Development (IMARD; Grant #857061910). The authors thank Dr Enli Wang and two anonymous reviewers for thorough review and helpful comments, Shacham Mekorot for providing the radar data, the Israel Meteorological Service for the rainfall data and IMARD for sharing the meteorological data. Special thanks to farmers Tal Ofek, Eitan Avivi and

References (62)

  • J.W. Hansen et al.

    Scaling-up crop models for climate variability applications

    Agric. Syst.

    (2000)
  • D. Helman et al.

    Crop RS-Met: A biophysical evapotranspiration and root-zone soil water content model for crops based on proximal sensing and meteorological data

    Agric. Water Manag.

    (2019)
  • L.A. Hunt et al.

    Agronomic data: advances in documentation and protocols for exchange and use

    Agric. Syst.

    (2001)
  • P.D. Jamieson et al.

    Modelling nitrogen uptake and redistribution in wheat

    Field Crop Res.

    (2000)
  • G. Jégo et al.

    Impact of the spatial resolution of climatic data and soil physical properties on regional corn yield predictions using the STICS crop model

    Int. J. Appl. Earth Obs. Geoinf.

    (2015)
  • R.A. Lawes et al.

    Integrating the effects of climate and plant available soil water holding capacity on wheat yield

    Field Crop Res.

    (2009)
  • J. Liu

    A GIS-based tool for modelling large-scale crop-water relations

    Environ. Model. Softw.

    (2009)
  • J.R. Lopez et al.

    Integrating growth stage deficit irrigation into a process based crop model

    Agric. For. Meteorol.

    (2017)
  • F. Marra et al.

    Use of radar QPE for the derivation of Intensity–Duration–Frequency curves in a range of climatic regimes

    J. Hydrol.

    (2015)
  • T. Palosuo et al.

    Simulation of winter wheat yield and its variability in different climates of Europe: a comparison of eight crop growth models

    Eur. J. Agron.

    (2011)
  • A.B. Potgieter et al.

    Yield trends under varying environmental conditions for sorghum and wheat across Australia

    Agric. For. Meteorol.

    (2016)
  • A. Rajala et al.

    Spring wheat response to timing of water deficit through sink and grain filling capacity

    Field Crop Res.

    (2009)
  • S. Saadi et al.

    Climate change and Mediterranean agriculture: impacts on winter wheat and tomato crop evapotranspiration, irrigation requirements and yield

    Agric. Water Manage

    (2015)
  • T.R. Sinclair et al.

    A model to assess nitrogen limitations on the growth and yield of spring wheat

    Field Crop Res.

    (1992)
  • C.O. Stöckle et al.

    CropSyst, a cropping systems simulation model

    Eur. J. Agron.

    (2003)
  • C.O. Stöckle et al.

    CropSyst model evolution: from field to regional to global scales and from research to decision support systems

    Environ. Model. Softw.

    (2014)
  • I. Takken et al.

    The prediction of runoff flow directions on tilled fields

    J. Hydrol.

    (2001)
  • K.R. Thorp et al.

    Methodology for the use of DSSAT models for precision agriculture decision support

    Comput. Electron. Agric.

    (2008)
  • X. Yin et al.

    Performance of process-based models for simulation of grain N in crop rotations across Europe

    Agric. Syst.

    (2017)
  • E.H. Acevedo et al.

    Wheat production in Mediterranean environments

    Wheat Ecol. Physiol. Yield Determ.

    (1999)
  • H.L. Boogaard et al.

    WOFOST 7.1; User’s Guide for the WOFOST 7.1 Crop Growth Simulation Model and WOFOST Control Center 1.5. SC-DLO

    (1998)
  • Cited by (19)

    • Quantifying water use and groundwater recharge under flood irrigation in an arid oasis of northwestern China

      2020, Agricultural Water Management
      Citation Excerpt :

      Tian et al. (2019) used the HYDRUS-1D program successively simulated the temporal pattern of soil moisture response to rainfall in the upstream of the Heihe River in northwestern China. Miller et al. (2019) showed that the HYDRUS-1D software accurately simulated the SWC in the root zone of wheat in Israel. Daily actual soil evaporation matched well with potential soil evaporation data, indicating the sufficient water supply for atmospheric demand in the ten years (Fig. 3b).

    View all citing articles on Scopus
    1

    Currently at the Department of Earth and Planetary Sciences, Johns Hopkins University, Baltimore, MD, USA.

    View full text