Hostname: page-component-7b7567c8c-wxvwp Total loading time: 0 Render date: 2026-11-14T12:17:55.224Z Has data issue: false hasContentIssue false
  • English
  • Français

Simulation of crop production in weed-infested fields for data-scarce regions

Published online by Cambridge University Press:  11 February 2016

H. VAN GAELEN,
N. DELBECQUE,
B. ABRHA,
A. TSEGAY and
D. RAES
H. VAN GAELEN*
Affiliation:
Department of Earth and Environmental Sciences, KU Leuven – University of Leuven, Celestijnenlaan 200 E, 3001 Leuven, Belgium
N. DELBECQUE
Affiliation:
Department of Earth and Environmental Sciences, KU Leuven – University of Leuven, Celestijnenlaan 200 E, 3001 Leuven, Belgium
B. ABRHA
Affiliation:
Department of Dryland Crop and Horticultural Sciences, Mekelle University, P.O. Box 231, Mekelle, Ethiopia
A. TSEGAY
Affiliation:
Department of Dryland Crop and Horticultural Sciences, Mekelle University, P.O. Box 231, Mekelle, Ethiopia
D. RAES
Affiliation:
Department of Earth and Environmental Sciences, KU Leuven – University of Leuven, Celestijnenlaan 200 E, 3001 Leuven, Belgium
*
* To whom all correspondence should be addressed. Email: hanne.vangaelen@ees.kuleuven.be
Rights & Permissions [Opens in a new window]

Summary

Weed infestation is a major yield-reducing factor that also decreases crop water productivity. Yet weeds are often neglected in crop productivity simulation studies, because existing empirical equations and mechanistic models are not widely applicable or have a high demand for input data and calibration. For that reason, AquaCrop, a widely applicable crop water productivity model, was expanded with a weed management module which requires only two easily obtainable input variables: (i) relative leaf cover of weeds, and (ii) weed-induced increase of total canopy cover. Using these inputs, AquaCrop directly simulates soil water content, crop canopy development and production as it is observed in weed-infested fields. Despite this simple approach, AquaCrop performed well to simulate soil water content in the root zone (relative root-mean-square error (RRMSE) of 5–13%), canopy cover (RRMSE of 15–22%), dry above-ground crop biomass during the season (RRMSE of 21–39%) and at maturity (RRMSE of 5–6%) and yield (RRMSE of 11–25%) of barley and wheat grown under different weed infestation levels and environments. The current study illustrates that the AquaCrop model can be used to assess the effect of weed infestation on crop growth and production, using a simple approach that is applicable to diverse environmental and agronomic conditions, even in data-scarce regions.

Information

Type
Crops and Soils Research Papers
Information
The Journal of Agricultural Science , Volume 154 , Issue 6 , August 2016, pp. 1026 - 1039
Check for updates
Copyright
Copyright © Cambridge University Press 2016 

INTRODUCTION

Nowadays, studies on agricultural productivity rely increasingly on simulation models. Being more cost- and time-efficient, crop models offer an attractive supplement to field experiments. Crop models are not only used to estimate potential yield levels and investigate causes of yield gaps, but also to evaluate management strategies that can boost crop productivity and water use efficiency. Initially, most crop models were developed to simulate crop production at field scale for historical time series. However, model application has evolved towards large-scale studies as well as prediction of future crop productivity.

While most crop models consider yield-limiting factors such as water stress and nutrient deficiencies, biotic factors such as weeds are often neglected. Notwithstanding, considerable yield losses due to weeds are not only faced by smallholder farmers in developing countries (FAO 2009), but also occur in large-scale intensive cropping systems in developed countries (Swanton et al. Reference Swanton, Harker and Anderson1993; Pimentel et al. Reference Pimentel, Lach, Zuniga and Morrison2000; Milberg & Hallgren Reference Milberg and Hallgren2004). In addition, weeds transpire water and thereby reduce water availability to the crop. This unproductive water consumption is critical in drought-prone regions, where optimizing crop water productivity is a prerequisite for sustainable crop production.

To include the effect of weed infestation in simulation studies, one could apply empirical equations that predict crop yield losses based on variables such as weed density (Cousens Reference Cousens1985), relative time of weed emergence (Cousens et al. Reference Cousens, Brain, O'Donovan and O'Sullivan1987), relative leaf area (Kropff et al. Reference Kropff, Lotz, Weaver, Bos, Wallinga and Migo1995) and relative leaf cover of the weeds (Lotz et al. Reference Lotz, Kropff, Wallinga, Bos and Groeneveld1994). However, these empirical equations entirely rely on locally calibrated parameters, which hinder extrapolation to other weed species, crop species, locations, environmental conditions and management practices (Kropff et al. Reference Kropff, Weaver and Smits1992; Murphy et al. Reference Murphy, Lemerle, Jones and Harden2002). Also, process-oriented, mechanistic simulation models such as ALMANAC (Kiniry et al. Reference Kiniry, Williams, Gassman and Debaeke1992), APSIM (Keating et al. Reference Keating, Carberry, Hammer, Probert, Robertson, Holzworth, Huth, Hargreaves, Meinke, Hochman, McLean, Verburg, Snow, Dimes, Silburn, Wang, Brown, Bristow, Asseng, Chapman, McCown, Freebairn and Smith2003), CROPSIM (Chikoye et al. Reference Chikoye, Hunt and Swanton1996) and INTERCOM (Kropff & van Laar Reference Kropff and van Laar1993) have been developed or extended to study crop–weed interactions. However, high requirements for input data, parameter calibration and validation, impede efficient application of these models for a wide range of environmental conditions and cropping systems (Weaver Reference Weaver1996). Particularly in data-scarce regions, application of existing mechanistic simulation models is impractical.

The AquaCrop crop water productivity model (Hsiao et al. Reference Hsiao, Heng, Steduto, Rojas-Lara, Raes and Fereres2009; Raes et al. Reference Raes, Steduto, Hsiao and Fereres2009; Steduto et al. Reference Steduto, Hsiao, Raes and Fereres2009) was developed by the Food and Agriculture Organization of the United Nations (FAO) to estimate yield for herbaceous crops cultivated under various environmental conditions and management practices. Although the model is mechanistic by nature, simulated processes in the crop–soil system are largely simplified. Notwithstanding these simplifications, AquaCrop provides accurate productivity estimates based on a limited number of easily obtainable input variables and parameters. This makes the model practical to apply in data-scarce regions, as well as for studies on a regional scale (Lorite et al. Reference Lorite, García-Vila, Santos, Ruiz-Ramos and Fereres2013; Kim & Kaluarachchi Reference Kim and Kaluarachchi2015). AquaCrop has been applied to assess irrigation and soil fertility management strategies (Geerts et al. Reference Geerts, Raes, Garcia, Taboada, Miranda, Cusicanqui, Mhizha and Vacher2009, Reference Geerts, Raes and Garcia2010; Shrestha et al. Reference Shrestha, Raes and Sah2013a , Reference Shrestha, Raes, Vanuytrecht and Sah b ; Tsegay et al. Reference Tsegay, Vanuytrecht, Abrha, Deckers, Gebrehiwot and Raes2015), but only recently a weed management module was developed for implementation in AquaCrop version 5 (expected to be released in 2015). Like the AquaCrop model, this module was developed pursuing an optimal balance between model accuracy, robustness, transparency and input requirements.

The current study discusses the new AquaCrop calculation algorithms for crop yield simulation in weed-infested fields. Moreover, the performance of the AquaCrop model to simulate soil water content, canopy cover (CC) development and crop production in weed-infested fields is evaluated for two different grain crops grown in various environmental and agronomic conditions.

MATERIALS AND METHODS

The AquaCrop model for weed-infested conditions

AquaCrop simulates crop productivity using a four-step process. In a first step, the green crop CC is simulated. The expansion of the crop CC from its initial value (CC0) to the maximum canopy cover (CC x ) is described by a logistic function determined by the canopy growth coefficient (CGC). At the end of the growing season, canopy senescence is described by means of the canopy decline coefficient (CDC). In a second step, crop transpiration (Tr) is simulated considering reference evapotranspiration (ET0) and a crop transpiration coefficient (KcTr) that is proportional to the simulated CC (Eqn 1). Next, crop transpiration is converted into dry above-ground biomass production (B) (Eqn 2). In a final step, crop biomass is converted to crop yield (Y) by means of the harvest index (HI) (Eqn 3). Crop yield per unit of water evapotranspired (ET) is given by the ET crop water productivity (WPET) (Eqn 4).

(1) $${\rm T}{{\rm r}_i} = {\rm K}{{\rm s}_i} \times {\rm K}{{\rm c}_{{\rm T}{{\rm r}_i}}} \times {\rm E}{{\rm T}_{{0_i}}}$$
(2) $${\rm B} = {\rm K}{{\rm s}_{{b_i}}} \times {\rm W}{{\rm P}^{^\ast}} \times \mathop \sum \limits_{i = 1}^n \displaystyle{{{\rm T}{{\rm r}_i}} \over {{\rm E}{{\rm T}_0}_i}} $$
(3) $${\rm Y} = {\rm HI} \times {\rm B}$$
(4) $${\rm W}{{\rm P}_{{\rm ET}}} = \displaystyle{{\rm Y} \over {\mathop \sum \nolimits_{i = 1}^n {\rm E}{{\rm T}_i}}}$$

where Tr i is crop transpiration (mm/day) on day i, ET0i is reference evapotranspiration (mm/day), KcTri is the crop transpiration coefficient (no units), Ks i is the soil water stress coefficient (no units), Ks bi is the cold stress coefficient for biomass production (no units), B is cumulative dry above-ground biomass production (g/m2), WP* is the normalized crop water productivity (g/m2), n is the number of sequential days spanning the growing period (day), Y is dry mass of yield production (g/m2), HI is the harvest index (g/g), WPET is ET crop water productivity (g/m3) and ET i is evapotranspiration (mm/day).

During this four-step simulation process, the model accounts for the effect of various abiotic stresses. To account for water stress, AquaCrop determines the soil water content in the root zone (Wr) using a soil water balance that tracks all incoming (rainfall, irrigation and capillary rise) and outgoing (runoff, deep percolation, evaporation and transpiration) daily water fluxes. When the soil water content in the root zone drops below process- and crop-specific thresholds, soil water stress affects processes, including root zone expansion, crop canopy expansion and early senescence, transpiration and yield formation. Besides water stress, AquaCrop considers the effect of soil fertility (Van Gaelen et al. Reference Van Gaelen, Tsegay, Delbecque, Shrestha, Garcia, Fajardo, Miranda, Vanuytrecht, Abrha, Diels and Raes2015), soil salinity and temperature stress. A more detailed description of the AquaCrop model calculation procedure and algorithms can be consulted in Raes et al. (Reference Raes, Steduto, Hsiao and Fereres2012).

When weed stress is considered, AquaCrop directly simulates crop canopy development as it is observed in a weed-infested field (CCW, Fig. 1). In addition, AquaCrop simulates canopy development of the crop–weed mixture (CCTOT, Fig. 1), which will be referred to hereafter as ‘total vegetation’. Simulation of total vegetation CC is crucial since the denser canopy in weed-infested fields affects soil evaporation, transpiration and consequently water availability in the root zone and crop water productivity. In AquaCrop, total vegetation is represented by a theoretical crop with higher CC, but otherwise identical characteristics to the crop growing in a weed-free field (CCWF). Hence, crop characteristics such as phenology, rooting depth, growing cycle length and sensitivity to abiotic stresses are also applicable to the total vegetation. The denser canopy is reflected by both a higher initial and maximum CC. Due to this denser canopy and assumption of equal growing cycle length, the canopy decline rate of the total vegetation is also higher compared to the decline rate of the crop.

Fig. 1.

Weed infestation affects the simulated (1) initial canopy cover (CC), (2) maximum CC and (3) canopy decline rate. As weeds take up empty space or suppress the crop, the crop CC in weed-infested conditions (CCW) is lower compared to weed-free conditions (CCWF). Moreover, the CC of the total vegetation (CCTOT) is higher and has a faster decline compared to the crop CC in weed-free conditions (CCWF).

Simulation of both the total vegetation and crop CC in weed-infested fields is completely determined by two user-specified inputs: (i) the weed infestation level or amount of weeds and (ii) the weed-induced increase of total CC.

The weed infestation level is quantified by means of the relative leaf cover of weeds (RC, Eqn 5) as defined by Lotz et al. (Reference Lotz, Kropff, Wallinga, Bos and Groeneveld1994).

(5) $${\rm RC} = \displaystyle{{{\rm WC}} \over {{\rm C}{{\rm C}_{{\rm TOT}}}}} = \displaystyle{{{\rm WC}} \over {{\rm WC} + {\rm C}{{\rm C}_{\rm W}}}}$$

where RC is the relative leaf cover of weeds (m2/m2), WC (m2/m2) is the area covered by weeds per unit ground area, CCW (m2/m2) is the area covered by the crop per unit ground area in a weed-infested field, and CCTOT (m2/m2) is the area covered by the total vegetation (crop–weed mixture) per unit ground area.

Relative weed cover is a multi-species canopy characteristic that varies during the growing season. Since the weed's share in leaf area at time of canopy closure is regarded as a good indicator of crop–weed competition (Kropff & Spitters Reference Kropff and Spitters1991), AquaCrop requires input of RC observed at the time maximum crop CC is reached.

The weed-induced increase of total CC (f weed) is defined by Eqn (6).

(6) $${\,f_{{\rm weed}}} = \displaystyle{{{\rm C}{{\rm C}_{x,{\rm TOT}}}} \over {{\rm C}{{\rm C}_{x,{\rm WF}}}}}$$

where f weed is the weed-induced increase of total CC (no units), CC x,TOT is the maximum total vegetation CC (m2/m2) and CC x,WF is the maximum crop CC in weed-free conditions (m2/m2).

A large f weed value indicates that weeds predominantly fill any empty gaps in the crop CC, which results in a strong increase of the total CC. A small f weed value, on the other hand, indicates that weeds predominantly suppress crop growth by ‘stealing’ light. This results in a small increase in total CC. Hence, for a certain weed infestation level (RC), differences in weed competitive abilities can result in different total vegetation CCs and consequently different f weed values. The AquaCrop user defines f weed either directly or by specifying CC x,TOT for the selected weed infestation level and optimal growing conditions (no fertility, water or salinity stress). In addition, AquaCrop can automatically determine f weed based on a user-specified canopy expansion factor (f shape). This f shape factor fixes the relation between the observed weed infestation level (RC) and the maximum total vegetation CC (CC x,TOT) for optimal growing conditions (Fig. 2), and consequently represents the competitive ability of weeds to compete with the crop for light.

Fig. 2.

Relationship between maximum total vegetation canopy cover (CC x,TOT) and relative cover of weeds (RC) is determined by the canopy expansion factor (f shape). This example presents a crop with a maximum CC (CC x,WF) of 0·8 in weed-free conditions.

An AquaCrop simulation for weed-infested conditions starts by simulating total vegetation CC (CCTOT) by multiplying the crop CC under weed-free conditions (CC x,WF) with f weed. Next, CCTOT is corrected for water, soil fertility or soil salinity stress using the stress thresholds calibrated for the crop. Thereby it is assumed that weeds are equally as sensitive to those stresses as the crop. Subsequently, the crop CC for weed-infested conditions (CCW) is derived from CCTOT based on RC (Eqn 5). Finally, simulation of CCW enables simulation of crop transpiration, crop biomass, crop yield and crop water productivity in weed-infested fields using Eqns (1–4).

Since weeds take up water and affect the soil water balance, AquaCrop calculates the soil water content in the root zone of a weed-infested field considering the total vegetation CC. Evaporation and transpiration of the total vegetation are simulated considering the denser canopy of weed-infested v. weed-free fields (CCTOT v. CCWF). The simulated soil water content is used to correct the simulated crop and total vegetation CC, crop transpiration and yield formation in a weed-infested field for water stress. It should be noted that increased water stress is the only mechanism through which weeds affect the HI in AquaCrop. An additional adjustment to the HI for weeds is not considered. Furthermore, weeds increase nutrient stress by ‘stealing’ crop nutrients. For that reason, the soil fertility level is adapted during simulation. In addition, a weed-induced increase of total CC is no longer considered (f weed is 1), since low soil fertility restricts total CC development to the level that can be reached in weed-free conditions (CC x,TOT equals CC x,WF).

Field experiments

The AquaCrop weed management algorithms were tested against two sets of experimental data (Table 1). The first set comprised data of four experiments conducted with barley (Hordeum vulgare L.) in the drought-prone, degraded highlands of Tigray in northern Ethiopia. The second set consisted of data from a field experiment with wheat (Triticum aestivum L.) conducted in semi-arid Wagga Wagga, Australia.

Table 1.

Experimental sites, setup and environmental conditions of the five experiments. Water treatments consist of absence (S0) or presence (S1) of water stress. Seasonal aridity indices represent the ratio of total rainfall to reference evapotranspiration (ET0 ) during the growing season

* Seasonal rainfall was affected by the presence of a rain shelter in the S1 treatment.

All experiments were set up with a split-plot design in which water treatment was the main factor and weed the treatment sub-factor (Table 1). In S1 water treatments, crop development or production were affected by water stress, while crops did not suffer water stress in S0 treatments. Occurrence of water stress was caused by insufficient rainfall, inadequate irrigation or the presence of a rain shelter. Since observed differences between water treatments were not significant for barley in all four experiments (Abrha Reference Abrha2013), only one water treatment was retained for the current study (Table 1). In Dejen and Maiquiha, naturally occurring weeds were retained and weed treatments consisted of different hand weeding frequencies: no weeding, 1 time weeding at 21 days after emergence (only in 2009) and frequent weeding (at least three times). The majority of weed species were broad-leaved (e.g. Scorpiurus muricatus), but grasses (e.g. Avena sp., Digitaria sp.) and sedges (e.g. Cyperus sp.) were also present. In Mekelle, wild oat (Avena fatua L.) was sown together with barley at proportions of 0·00, 0·05, 0·20 and 0·50 of the total amount of seeds. In Wagga Wagga, ryegrass (Lolium rigidum Gaud.) was sown aiming at a weed density of 250 plants/m2. At all experimental sites, the plots were kept free from pests, diseases and undesired weeds throughout the growing season. Moreover, all plots were kept at optimal fertility to ensure that neither crops nor weeds would suffer from nutrient stress. During the growing season observations of local daily weather, soil characteristics, irrigation and soil fertility management, soil water content, crop phenology, crop and weed canopy (leaf area index (LAI) or green CC), dry above-ground crop biomass and crop yield, were recorded. Observations of the second data set were retrieved from the paper by Deen et al. (Reference Deen, Cousens, Warringa, Bastiaans, Carberry, Rebel, Riha, Murphy, Benjamin, Cloughley, Cussans, Forcella, Hunt, Jamieson, Lindquist and Wang2003) or shared by those authors. Unfortunately, not all data could be retrieved; some observations of the weed-infested plots were missing (e.g. yield and soil water content) or observations were limited to average values without records of the deviation between replications. More detailed information on the experimental setup and data collection is described by Deen et al. (Reference Deen, Cousens, Warringa, Bastiaans, Carberry, Rebel, Riha, Murphy, Benjamin, Cloughley, Cussans, Forcella, Hunt, Jamieson, Lindquist and Wang2003) for wheat, and Abrha et al. (Reference Abrha, Delbecque, Raes, Tsegay, Todorovic, Heng, Vanutrecht, Geerts, Garcia-Vila and Deckers2012) and Abrha (Reference Abrha2013) for barley.

Model input

Observations of local climatic data, irrigation practices, soil characteristics and initial soil water content were used as inputs in AquaCrop (version 5). Simulations were conducted assuming non-limiting soil fertility for all plots. Default crop parameters were used as a starting point; thereafter non-conservative crop parameters were adjusted to match the characteristics of the local cultivar and environment (Table 2). Crop development stages were specified in growing degree days (GDD) to enable temperature-dependent crop canopy development. Conservative crop parameters, which by definition are independent of cultivar, management and geographical location, were kept as default, except for wheat. Since AquaCrop does not consider typical processes for winter crops such as vernalization, dormancy and cold acclimation, simulation of winter wheat development required adaptation of some conservative crop parameters (Table 2) following the example of Vanuytrecht (Reference Vanuytrecht2013).

Table 2.

Key crop parameters for barley and wheat grown at the experimental sites. Values that were changed from the default are presented in bold

* Values for Mekelle.

Weed management inputs (RC and f weed), listed in Table 3, were determined applying Eqns (5) and (6) to available CCWF, CCW and WC observations of well-watered plots at time of maximum crop CC. For wheat experiments, LAI values were first converted to CC values using Eqn (7) (Kropff & van Laar Reference Kropff and van Laar1993) with an extinction coefficient (k) of 0·6 and 0·5 for wheat and ryegrass, respectively (Lantinga et al. Reference Lantinga, Nassiri and Kropff1999; Acevedo et al. Reference Acevedo, Silva, Silva, Curtis, Rajaram and Gómez Macpherson2002).

(7) $${\rm CC} = 1 - \exp \left( { - \mathop \sum \limits_{i = 1}^n {k_i} \times {\rm LA}{{\rm I}_i}} \right)$$

where CC is the green canopy cover (m2/m2), k i is the light extinction coefficient (no units), and LAI i the LAI (m2/m2) of species i as observed in a field with n species.

Table 3.

Weed treatments with selected AquaCrop weed management input: relative weed cover (RC) and weed-induced total canopy cover increase (fweed ) with corresponding canopy expansion factor (fshape ). Water treatments consist of absence (S0) or presence (S1) of water stress

Since weed species were similar for all weed treatments of the same experiment, a single canopy expansion factor (f shape) was selected for all treatments. The f shape value was selected so that the resulting f weed values, automatically determined by AquaCrop, approached the observed f weed values well. Moreover, water availability in wheat plots affected the amount of weeds at time of canopy closure so that different RC values were selected for both water treatments.

Model performance evaluation

The AquaCrop model was evaluated for its performance to simulate soil water content in the root zone, crop CC, total vegetation CC, crop biomass and crop yield both for weed-free and weed-infested conditions. Thereby observations of all experimental sites and water treatment, listed in Table 1, were included. Performance was assessed using graphical displays (plots of simulated v. observed values) and statistical indicators: relative root-mean-square error (RMSE) (Loague & Green Reference Loague and Green1991), Nash-Sutcliffe model efficiency (Nash & Sutcliffe Reference Nash and Sutcliffe1970) and coefficient of determination (R 2; Eqn 10). When the number of observations was limited, performance was assessed based on the deviation (Mitchell & Sheehy Reference Mitchell, Sheehy, Kropff, Teng, Aggarwal, Bouma, Bouman, Jones and van Laar1997).

(8) $${\rm RRMSE} = \sqrt {\displaystyle{{\mathop \sum \nolimits_{i = 1}^n {{({O_i} - {P_i})}^2}} \over n}} \times \displaystyle{{100} \over {\bar O}}$$
(9) $${\rm EF} = 1 - \displaystyle{{\mathop \sum \nolimits_{i = 1}^n {{({P_i} - {O_i})}^2}} \over {\mathop \sum \nolimits_{i = 1}^n {{({O_i} - \bar O)}^2}}}$$
(10) $$R^2 = {\left( {\displaystyle{{\mathop \sum \nolimits_{i = 1}^n ({O_i} - \bar O)({P_i} - \bar P)} \over {\sqrt {\mathop \sum \nolimits_{i = 1}^n {{({O_i} - \bar O)}^2}} \; \sqrt {\mathop \sum \nolimits_{i = 1}^n {{({P_i} - \bar P)}^2}}}}} \right)^2}$$
(11) $${\rm deviation} = \displaystyle{{{P_i} - {O_i}} \over {{O_i}}} \times 100$$

where RRMSE (%) is the relative root-mean-square error, O i are the observed values, P i are the predicted values, $\bar Q$ is the mean of the observed values, $\bar P$ is the mean of the predicted values, n is the number of observations, R 2 (no units) is the coefficient of determination and EF (no units) is the Nash–Sutcliffe model efficiency. Model performance was classified based on RRMSE values as excellent (RRMSE < 10%), good (10% < RRMSE < 20%), fair (20% < RRMSE < 30%) and poor (RRMSE > 30%), according to Jamieson et al. (Reference Jamieson, Porter and Wilson1991).

RESULTS

Two examples, presented in Fig. 3, illustrate how AquaCrop simulates the effect of weed infestation on barley canopy and biomass development. The total vegetation CC increased due to presence of weeds, while crop CC reduced. As a consequence, less biomass was produced. Even though both experiments had a similar weed infestation level of about 0·50 (Table 3), the simulated weed-induced increase of total CC was strong in the study of Dejen (2009), while it was negligible in Mekelle (2010). This was the result of selecting a different f weed value (Table 3). Since the effect of weeds on crop CC differed between the two experiments, the weed-induced reduction in biomass also differed. The simulated decrease of crop final biomass due to 0·50 RC was about 43% for Mekelle, but only 39% for Dejen. Also, field observations indicated that due to the high crop sowing density in Mekelle there was almost no empty space in the canopy for weeds to occupy. While in Dejen only some of the weeds competed with the crop for light, all the weeds in Mekelle suppressed the crop by outcompeting it for light, thereby causing a larger reduction in biomass and water productivity.

Fig. 3.

Simulated (lines) and observed (symbols) CC (top) and dry above-ground biomass (bottom) in well-watered barley plots in Dejen 2009 (left) and Mekelle 2010 (right). Crop CC and biomass are affected by the weed infestation level: weed-free (black) and weed-infested with a relative weed cover of about 0·50 (grey). The dashed line and open symbol present the total vegetation CC in weed-infested conditions. Error bars indicate ± s.d. for three replications.

The goodness-of-fit statistics (Table 4) present the performance of AquaCrop to simulate different crop variables of both weed-free and weed-infested barley and wheat plots including all experimental sites and water treatments.

Table 4.

Relative root-mean-square error (RRMSE), Nash-Sutcliffe model efficiency (EF) and coefficient of determination (R2 ) for n number of observations of average soil water content in the root zone, crop canopy cover in weed-free conditions (CCWF ) or weed-infested conditions (CCW ), total vegetation canopy cover (CCTOT ), dry above-ground crop biomass (B) during the growing season and at phenological maturity, and crop yield in weed-free and weed-infested barley and wheat plots (all experimental sites and water treatments included). Dashes indicate that observations for performance assessment were unavailable

AquaCrop made fair predictions of barley CC development and biomass build up during the growing season, as shown by the RRMSE values of between 21 and 24% (Table 4). Regardless of the fair biomass predictions during the growing season, biomass at maturity was the most accurately predicted crop variable for both weed treatments (RRMSE values of 5–6%). Moreover, AquaCrop made very good predictions of soil water content in the root zone (RRMSE of about 12–13%), despite only fair predictions of CC and consequently transpiration. Furthermore, Table 4 illustrates that simulations were slightly less accurate for weed-infested barley plots compared to weed-free conditions. The RRMSE values of weed-infested treatments were only 1·5–2·5% higher than weed-free treatments for soil water content and biomass. Predictions of barley biomass at maturity were even better than for weed-free conditions. Although final biomass was predicted as excellent for all weed treatments, model performance decreased for grain yield depending on the weed treatment; barley yield predictions were very good for weed-free conditions (RRMSE of 12%), but only fair for weed-infested conditions (RRMSE of 25%). Figure 4 shows that yield was over-estimated for all weed treatments, while biomass was predicted accurately. This indicates over-estimation of the HI.

Fig. 4.

Observed v. simulated dry above-ground biomass at maturity (left) and grain yield (right) of barley grown in weed-free (black) and weed-infested (grey) plots with water stress (circle) or without water stress (diamond) at the three experimental sites in 2009 & 2010. Error bars indicate ± s.d. for three replications.

Simulation of the soil water content in the root zone of weed-free wheat plots was excellent, as indicated by the small RRMSE value of 4·5% (Table 4). However, because soil water content simulations were less good for the S0 treatment, EF and R 2 values were rather low. Unfortunately, observations of soil water content were not available for weed-infested fields. Furthermore, AquaCrop performed well for simulations of crop and total vegetation CC, with RRMSE values between 15 and 21%. In contrast, performance was poor for simulation of crop biomass during the season. Figure 5 illustrates that AquaCrop was not able to capture the slow biomass build-up during winter, even though conservative crop parameters were altered to match the characteristics of winter wheat (Table 2). Introduction of weeds further decreased model performance. In weed-infested conditions, RRMSE values were 3–9% higher than in weed-free conditions. In spite of AquaCrop's limitations to accurately simulate biomass development, biomass at maturity was predicted very accurately with deviations as small as −1 to 2%. For both water treatments, final biomass in weed-free plots was under-estimated, but over-estimated for weed-infested plots.

Fig. 5.

Simulated (lines) and observed (symbols) CC (top) and dry above-ground biomass (bottom) in well-watered (left) and water-stressed (right) wheat plots in Wagga Wagga, 1998. Crop CC and biomass are affected by the weed infestation level: weed-free (black) and weed-infested (grey) with a relative weed cover of about 0·20 for well-watered and 0·15 for water-stressed plots. The dashed line and open symbol present the total vegetation CC in weed-infested conditions. Data points represent an average value over five replications. Error bars indicating the s.d. were not available for the wheat dataset.

DISCUSSION

Model approach

Data availability is a major bottleneck for application of existing crop–weed competition models. In contrast, AquaCrop requires very few input variables and parameters that can be easily obtained. Simulation for weed-infested conditions requires only two additional inputs (RC and f weed), which can be easily determined by analysing digital photographs or from remote sensing images (Lotz et al. Reference Lotz, Kropff, Wallinga, Bos and Groeneveld1994; Burgos-Artizzu et al. Reference Burgos-Artizzu, Ribeiro, Tellaeche, Pajares and Fernández-Quintanilla2009). If such digital images are not available, one can also rely on visual estimates done in the field (Andújar et al. Reference Andújar, Ribeiro, Carmona, Fernández-Quintanilla and Dorado2010). Since the two weed management inputs need to be specified for the weed mixture, knowledge of the exact weed species is not required. Moreover, this multiple-species approach is an important advantage, because competition by a single weed species is rarely encountered in farmers’ fields.

AquaCrop relies completely on the input of relative weed cover to simulate partitioning of resources between crop and weeds. Light partitioning is by definition represented by RC. Water partitioning requires the total transpiration be divided into crop and weed transpiration based on total vegetation CC and crop CC. Since the latter are related through RC, partitioning of water is clearly determined by RC. Also, nutrient partitioning requires the soil fertility stress level to be adapted based on RC. One could argue that such an RC-based approach neglects the competitive ability of weeds v. crop to obtain light, water and nutrients. Also, differences in sensitivity to scarcity of these resources seem to be neglected. However, this is not true since f weed incorporates the competitive ability of crop and weed to obtain light. In addition, RC serves as a proxy for competitive ability and sensitivity to water and soil fertility stress. Also, Aldrich (Reference Aldrich1987) found light competition to be representative for total competition, since canopy size and structure is the result of competition for light, water and nutrients, as well as allelopathic interactions. Moreover, photosynthesis is not just involved in light competition, but also provides the energy for uptake of nutrients and water.

Furthermore, AquaCrop uses a static approach regarding weed infestation, as it uses a constant value of RC at time of maximum CC to define the effect of weeds on crop CC during the whole growing season. Such a static approach neglects the dynamics of weed cover, which can increase or decrease during the growing season depending on the competitive ability of crop v. weed. However, the absolute error made with this static approach is negligible at the beginning of the growing season, because CC is still very small and little biomass is produced during the canopy expansion phase. In contrast, the error could be larger in mid-season, when crop CC is large and most biomass is produced. Although the current study proved that good model results can be obtained with the static approach, future research should indicate whether RC dynamics in mid-season have a large impact on model results, and should be incorporated in the model.

Finally, it should be noted that AquaCrop's representation of the crop–weed vegetation as a single theoretical crop with the same growing cycle as the weed-free crop neglects possible differences between the crops’ and weeds’ growing cycles. Nevertheless, this simplification is valid both for late-emerging weeds, that cannot outgrow the crop, and for early-emerging weeds, which are removed during land preparation. Moreover, weeds with a life cycle similar to that of the crop will usually be the most successful competitors (Zimdahl Reference Zimdahl2013) and consequently lead to major yield losses.

Model performance

AquaCrop performed well for simulated barley growth and production. This indicates that the selected crop parameters and inputs were a good representation of the local environment and cropping system. As the default crop parameters had been calibrated and validated by Abrha et al. (Reference Abrha, Delbecque, Raes, Tsegay, Todorovic, Heng, Vanutrecht, Geerts, Garcia-Vila and Deckers2012) for the same local barley variety grown in weed-free conditions, this is no surprise. Overestimation of yield indicated incorrect simulation of the HI, especially in weed-infested plots. This could be due to inaccurate settings of the barley water stress thresholds, or because AquaCrop disregards any direct effect of weeds on grain formation. Field experiments show that weeds can reduce barley grain yield by lowering the number of ear bearing tillers, number of grains per ear and 1000-kernel weight (Wilson & Peters Reference Wilson and Peters1982; Morishita & Thill Reference Morishita and Thill1988). Introducing an RC-dependent stress coefficient that affects the HI could improve model simulations. However, this would also lead to extra parameter uncertainty in the model.

Model performance to simulate wheat production was less good compared to barley, in particular for simulation of biomass during the growing season. It is expected that introduction of an extra stress coefficient to represent slow crop development because of low winter temperatures, as previously proposed by Vanuytrecht (Reference Vanuytrecht2013), would significantly improve AquaCrop's performance for both weed-free and weed-infested winter wheat simulations. Despite the poor in-season biomass predictions, AquaCrop made excellent predictions of wheat biomass at maturity. In practice, final biomass, and not intermediate biomass, is most crucial to assess weed-induced yield losses.

Moreover, AquaCrop's performance for the wheat dataset was well within the performance range of four mechanistic crop–weed competition models (ALMANAC, APSIM, CROPSIM and INTERCOM) that were tested with the same dataset by Deen et al. (Reference Deen, Cousens, Warringa, Bastiaans, Carberry, Rebel, Riha, Murphy, Benjamin, Cloughley, Cussans, Forcella, Hunt, Jamieson, Lindquist and Wang2003). Final biomass deviation ranged between −27 and 25% for the four other models, but was a maximum of 2% for AquaCrop. However, this model performance comparison should be interpreted with care, as the AquaCrop approach is very different from mechanistic crop–weed competition models. The latter predict both crop and weed growth, partitioning of resources between crop and weed, and the resulting crop yield loss due to weeds. AquaCrop, on the other hand, uses input of the observed effect of weeds on crop CC at time of canopy closure to simulate the effect of weeds on the soil water balance and crop production.

Finally, it should be noted that because of the experimental setup, the current study focused mainly on simulation of the outcome of crop–weed competition for light. Model performance for water competition was assessed based on the weed-infested, water-stressed treatments, which were only available for one of the barley experiments (at Maiquiha in 2009) and the wheat experiment. Although model performance proved very promising for the current experiments, additional research including different water stress levels, ranging between low and very severe water stress, could further reveal to what extent AquaCrop can accurately capture the effect of weeds on soil water content and crop water availability. Furthermore, competition for nutrients was not considered in the current study, because the soil fertility level was optimal in all experiments. Therefore, further evaluation of AquaCrop for weed-infested, soil-fertility-stressed fields remains necessary.

Model application

Once properly calibrated to the local environment and cropping system, AquaCrop can be used to simulate crop growth, production and water productivity in weed-infested fields. Notwithstanding its simple approach, the new weed module is widely applicable. This was demonstrated by model evaluation for various experimental setups including different locations, soil types, climatic conditions, water treatments, crops, weed species and weed infestation levels. Furthermore, limited input and calibration requirements make AquaCrop applicable in data-scarce regions. If data are sparse, weed management inputs can even be specified in qualitative terms. A user can select one of the predefined RC classes, ranging between ‘very poor’ and ‘perfect’ weed management, or one of the predefined f weed classes ranging between ‘very weak’ and ‘very strong’ weed-induced increase of total CC. In addition, the transparent simulation procedure and user-friendly interface enable application by non-specialists.

Addition of weed infestation as a production- and water-limiting factor improves accuracy of yield predictions and yield gap analysis in weed-infested areas. Other potential model applications include investigation of tolerable weed infestation levels, as well as assessment of yield loss mitigation strategies such as changing sowing density or crop type. AquaCrop has the important advantage that weed infestation scenarios can be evaluated based on both simulated crop yield and crop water productivity. Assessing both of these productivity indicators is vital in water-scarce regions, where management decisions should be made in view of limited water availability.

Even though potential applications are numerous, the AquaCrop approach has limitations. The model can only support strategic management decisions and cannot be used for assessment of tactical decisions. For example, model simulations can provide information on tolerable weed infestation levels, but can optimize neither pesticide dose nor timing of weed control operations. Moreover, the model cannot provide additional insights into the competition mechanisms. This would require a more detailed, process-based approach as adopted by other mechanistic models. Finally, it is clear that AquaCrop can support weed management decisions only from an agronomic point of view. However, by linking AquaCrop to an economic model, weed management can be optimized accounting for factors including labour and time requirement of weed control operations, prices of herbicides and prices of crop products. An example was set by Dunan et al. (Reference Dunan, Moore and Westra1994, Reference Dunan, Westra and Moore1999), who optimized weed management based on both agronomic and economic aspects. Moreover, the linkage of AquaCrop to economic models has been demonstrated by García-Vila et al. (Reference García-Vila, Fereres, Mateos, Orgaz and Steduto2009), García-Vila & Fereres (Reference García-Vila and Fereres2012) and Cusicanqui et al. (Reference Cusicanqui, Dillen, Garcia, Geerts, Raes and Mathijs2013).

CONCLUSIONS

The AquaCrop approach to simulate crop production in weed-infested fields proves to be a very intuitive approach that requires just two easily obtainable input variables: the relative leaf cover of weeds and weed-induced increase of total CC. This makes the model applicable to all herbaceous crops grown in competition with any type or number of weeds. Despite its simple approach, AquaCrop performs good to simulate soil water content, crop development and crop production in weed-infested barley and wheat fields over a wide range of environmental and agronomic conditions. Further testing of the model remains necessary to assess model performance for weed-infested conditions combined with nutrient limitations or severe water stress. Because of its simple but accurate simulation procedure, wide applicability and low data requirements, AquaCrop is a practical tool to investigate the effect of weed infestation on crop production in data-scarce regions.

We would like to thank the Flemish Interuniversity Council (VLIR) for funding the field experiments with barley in Ethiopia, and the Research Foundation – Flanders (FWO) for funding development and evaluation of the simulation approach. Moreover, we are grateful to Dr Bill Deen and Prof. Roger Cousens for sharing the wheat dataset. Finally, we thank all anonymous reviewers for their valuable suggestions.

References

REFERENCES

Abrha, B. (2013). Barley (Hordeum vulgare L.) yield prediction and its gap analysis in Geba catchment, northern highlands of Ethiopia. Ph.D. Thesis, KU Leuven University, Leuven, Belgium.Google Scholar
Abrha, B., Delbecque, N., Raes, D., Tsegay, A., Todorovic, M., Heng, L., Vanutrecht, E., Geerts, S., Garcia-Vila, M. & Deckers, S. (2012). Sowing strategies for barley (Hordeum vulgare L.) based on modelled yield response to water with AquaCrop. Experimental Agriculture 48, 252–271.CrossRefGoogle Scholar
Acevedo, E., Silva, P. & Silva, H. (2002). Wheat growth and physiology. In Bread Wheat – Improvement and Production. FAO Plant Production and Protection Series No. 30 (Eds Curtis, B. C., Rajaram, S. & Gómez Macpherson, H.), pp. 53–89. Rome: FAO. Available from: http://www.fao.org/docrep/006/y4011e/y4011e06.htm (accessed 26 October 2015).Google Scholar
Aldrich, R. J. (1987). Predicting crop yield reductions from weeds. Weed Technology 1, 199–206.CrossRefGoogle Scholar
Andújar, D., Ribeiro, A., Carmona, R., Fernández-Quintanilla, C. & Dorado, J. (2010). An assessment of the accuracy and consistency of human perception of weed cover. Weed Research 50, 638–647.CrossRefGoogle Scholar
Burgos-Artizzu, X. P., Ribeiro, A., Tellaeche, A., Pajares, G. & Fernández-Quintanilla, C. (2009). Improving weed pressure assessment using digital images from an experience-based reasoning approach. Computers and Electronics in Agriculture 65, 176–185.CrossRefGoogle Scholar
Chikoye, D., Hunt, L. A. & Swanton, C. J. (1996). Simulation of competition for photosynthetically active radiation between common ragweed (Ambrosia artemisiifolia) and Dry Bean (Phaseolus vulgaris). Weed Science 44, 545–554.CrossRefGoogle Scholar
Cousens, R. (1985). A simple model relating yield loss to weed density. Annals of Applied Biology 107, 239–252.CrossRefGoogle Scholar
Cousens, R., Brain, P., O'Donovan, J. T. & O'Sullivan, P. A. (1987). The use of biologically realistic equations to describe the effects of weed density and relative time of emergence on crop yield. Weed Science 35, 720–725.CrossRefGoogle Scholar
Cusicanqui, J., Dillen, K., Garcia, M., Geerts, S., Raes, D. & Mathijs, E. (2013). Economic assessment at farm level of the implementation of deficit irrigation for quinoa production in the Southern Bolivian Altiplano. Spanish Journal of Agricultural Research 11, 894–907.CrossRefGoogle Scholar
Deen, W., Cousens, R., Warringa, J., Bastiaans, L., Carberry, P., Rebel, K., Riha, S., Murphy, C., Benjamin, L. R., Cloughley, C., Cussans, J., Forcella, F., Hunt, T., Jamieson, P., Lindquist, J. & Wang, E. (2003). An evaluation of four crop: weed competition models using a common data set. Weed Research 43, 116–129.CrossRefGoogle Scholar
Dunan, C. M., Moore, F. D. & Westra, P. (1994). A plant process-economic model for wild oats management decisions in irrigated barley. Agricultural Systems 45, 355–368.CrossRefGoogle Scholar
Dunan, C. M., Westra, P. & Moore, F. D. (1999). A plant process economic model for weed management decisions in irrigated onion. Journal of the American Society for Horticultural Science 124, 99–105.CrossRefGoogle Scholar
FAO (2009). The lurking menace of weeds – Farmers’ enemy No. 1. FAO News Article, 11 August 2009. Rome: FAO. Available from: http://www.fao.org/news/story/en/item/29402/icode/ (verified 17 September 2015).Google Scholar
García-Vila, M. & Fereres, E. (2012). Combining the simulation crop model AquaCrop with an economic model for the optimization of irrigation management at farm level. European Journal of Agronomy 36, 21–31.CrossRefGoogle Scholar
García-Vila, M., Fereres, E., Mateos, L., Orgaz, F. & Steduto, P. (2009). Deficit irrigation optimization of cotton with AquaCrop. Agronomy Journal 101, 477–487.CrossRefGoogle Scholar
Geerts, S., Raes, D., Garcia, M., Taboada, C., Miranda, R., Cusicanqui, J., Mhizha, T. & Vacher, J. (2009). Modeling the potential for closing quinoa yield gaps under varying water availability in the Bolivian Altiplano. Agricultural Water Management 96, 1652–1658.CrossRefGoogle Scholar
Geerts, S., Raes, D. & Garcia, M. (2010). Using AquaCrop to derive deficit irrigation schedules. Agricultural Water Management 98, 213–216.CrossRefGoogle Scholar
Hsiao, T. C., Heng, L., Steduto, P., Rojas-Lara, B., Raes, D. & Fereres, E. (2009). AquaCrop – The FAO crop model to simulate yield response to water: III. Parameterization and testing for maize. Agronomy Journal 101, 448–459.CrossRefGoogle Scholar
Jamieson, P. D., Porter, J. R. & Wilson, D. R. (1991). A test of the computer simulation model ARCWHEAT1 on wheat crops grown in New Zealand. Field Crops Research 27, 337–350.CrossRefGoogle Scholar
Keating, B. A., Carberry, P. S., Hammer, G. L., Probert, M. E., Robertson, M. J., Holzworth, D., Huth, N. I., Hargreaves, J. N. G., Meinke, H., Hochman, Z., McLean, G., Verburg, K., Snow, V., Dimes, J. P., Silburn, M., Wang, E., Brown, S., Bristow, K. L., Asseng, S., Chapman, S., McCown, R. L., Freebairn, D. M. & Smith, C. J. (2003). An overview of APSIM, a model designed for farming systems simulation. European Journal of Agronomy 18, 267–288.CrossRefGoogle Scholar
Kim, D. & Kaluarachchi, J. (2015). Validating FAO AquaCrop using Landsat images and regional crop information. Agricultural Water Management 149, 143–155.CrossRefGoogle Scholar
Kiniry, J. R., Williams, J. R., Gassman, P. W. & Debaeke, P. (1992). A general, process-oriented model for two competing plant species. Transactions of the ASAE (USA) 35, 801–810.CrossRefGoogle Scholar
Kropff, M. J. & Spitters, C. J. T. (1991). A simple model of crop loss by weed competition from early observations on relative leaf area of the weeds. Weed Research 31, 97–105.CrossRefGoogle Scholar
Kropff, M. J. & van Laar, H. H. (1993). Modelling Crop–Weed Interactions. Wallingford, UK & Los Banos, Philippines: CAB International & IRRI.Google Scholar
Kropff, M. J., Weaver, S. E. & Smits, M. A. (1992). Use of ecophysiological models for crop–weed interference: relations amongst weed density, relative time of weed emergence, relative leaf area and yield loss. Weed Science 40, 296–301.CrossRefGoogle Scholar
Kropff, M. J., Lotz, L. A. P., Weaver, S. E., Bos, H. J., Wallinga, J. & Migo, T. (1995). A two parameter model for prediction of crop loss by weed competition from early observations of relative leaf area of the weeds. Annals of Applied Biology 126, 329–346.CrossRefGoogle Scholar
Lantinga, E. A., Nassiri, M. & Kropff, M. J. (1999). Modelling and measuring vertical light absorption within grass-clover mixtures. Agricultural and Forest Meteorology 96, 71–83.CrossRefGoogle Scholar
Loague, K. & Green, R. E. (1991). Statistical and graphical methods for evaluating solute transport models: overview and application. Journal of Contaminant Hydrology 7, 51–73.CrossRefGoogle Scholar
Lorite, I. J., García-Vila, M., Santos, C., Ruiz-Ramos, M. & Fereres, E. (2013). AquaData and AquaGIS: two computer utilities for temporal and spatial simulations of water-limited yield with AquaCrop. Computers and Electronics in Agriculture 96, 227–237.CrossRefGoogle Scholar
Lotz, L. A. P., Kropff, M. J., Wallinga, J., Bos, H. J. & Groeneveld, R. M. W. (1994). Techniques to estimate relative leaf area and cover of weeds in crops for yield loss prediction. Weed Research 34, 167–175.CrossRefGoogle Scholar
Milberg, P. & Hallgren, E. (2004). Yield loss due to weeds in cereals and its large-scale variability in Sweden. Field Crops Research 86, 199–209.CrossRefGoogle Scholar
Mitchell, P. L. & Sheehy, J. E. (1997). Comparison of predictions and observations to assess model performance: a method of empirical validation. In Applications of Systems Approaches at the Field Level (Eds Kropff, M. J., Teng, P. S., Aggarwal, P. K., Bouma, J., Bouman, B. A. M., Jones, J. W. & van Laar, H. H.), pp. 437–451. Systems Approaches for Sustainable Agricultural Development vol. 6. Dordrecht, The Netherlands: Springer.CrossRefGoogle Scholar
Morishita, D. W. & Thill, D. C. (1988). Factors of wild oat (Avena fatua) interference on spring barley (Hordeum vulgare) growth and yield. Weed Science 36, 37–42.CrossRefGoogle Scholar
Murphy, C., Lemerle, D., Jones, R. & Harden, S. (2002). Use of density to predict crop yield loss between variable seasons. Weed Research 42, 377–384.CrossRefGoogle Scholar
Nash, J. E. & Sutcliffe, J. V. (1970). River flow forecasting through conceptual models part I – A discussion of principles. Journal of Hydrology 10, 282–290.CrossRefGoogle Scholar
Pimentel, D., Lach, L., Zuniga, R. & Morrison, D. (2000). Environmental and economic costs of nonindigenous species in the United States. BioScience 50, 53–65.CrossRefGoogle Scholar
Raes, D., Steduto, P., Hsiao, T. C. & Fereres, E. (2009). AquaCrop – The FAO crop model to simulate yield response to water: II. Main algorithms and software description. Agronomy Journal 101, 438–447.CrossRefGoogle Scholar
Raes, D., Steduto, P., Hsiao, T. C. & Fereres, E. (2012). AquaCrop Reference Manual, AquaCrop Version 4.0. Rome: FAO.Google Scholar
Shrestha, N., Raes, D. & Sah, S. K. (2013 a). Strategies to improve cereal production in the Terai Region (Nepal) during dry season: simulations with Aquacrop. Procedia Environmental Sciences 19, 767–775.CrossRefGoogle Scholar
Shrestha, N., Raes, D., Vanuytrecht, E. & Sah, S. K. (2013 b). Cereal yield stabilization in Terai (Nepal) by water and soil fertility management modeling. Agricultural Water Management 122, 53–62.CrossRefGoogle Scholar
Steduto, P., Hsiao, T. C., Raes, D. & Fereres, E. (2009). AquaCrop – The FAO crop model to simulate yield response to water: I. Concepts and underlying principles. Agronomy Journal 101, 426–437.CrossRefGoogle Scholar
Swanton, C. J., Harker, K. N. & Anderson, R. L. (1993). Crop losses due to weeds in Canada. Weed Technology 7, 537–542.CrossRefGoogle Scholar
Tsegay, A., Vanuytrecht, E., Abrha, B., Deckers, J., Gebrehiwot, K. & Raes, D. (2015). Sowing and irrigation strategies for improving rainfed tef (Eragrostis tef (Zucc.) Trotter) production in the water scarce Tigray region, Ethiopia. Agricultural Water Management 150, 81–91.CrossRefGoogle Scholar
Van Gaelen, H., Tsegay, A., Delbecque, N., Shrestha, N., Garcia, M., Fajardo, H., Miranda, R., Vanuytrecht, E., Abrha, B., Diels, J. & Raes, D. (2015). A semi-quantitative approach for modelling crop response to soil fertility: evaluation of the AquaCrop procedure. Journal of Agricultural Science, Cambridge 153, 1218–1233.CrossRefGoogle Scholar
Vanuytrecht, E. (2013). Crop response to climate change. Impact on agricultural production and the soil water balance in the Flemish region of Belgium. Ph.D. Thesis, KU Leuven University, Leuven, Belgium.Google Scholar
Weaver, S. E. (1996). Simulation of crop–weed competition: models and their applications. Phytoprotection 77, 3–12.CrossRefGoogle Scholar
Wilson, B. J. & Peters, N. C. B. (1982). Some studies of competition between Avena fatua L. and spring barley I. The influence of A. fatua on yield of barley. Weed Research 22, 143–148.CrossRefGoogle Scholar
Zimdahl, R. L. (2013). Fundamentals of Weed Science, 4th edn. Amsterdam: Elsevier.Google Scholar
Figure 0

Fig. 1.Weed infestation affects the simulated (1) initial canopy cover (CC), (2) maximum CC and (3) canopy decline rate. As weeds take up empty space or suppress the crop, the crop CC in weed-infested conditions (CCW) is lower compared to weed-free conditions (CCWF). Moreover, the CC of the total vegetation (CCTOT) is higher and has a faster decline compared to the crop CC in weed-free conditions (CCWF).

Figure 1

Fig. 2.Relationship between maximum total vegetation canopy cover (CCx,TOT) and relative cover of weeds (RC) is determined by the canopy expansion factor (fshape). This example presents a crop with a maximum CC (CCx,WF) of 0·8 in weed-free conditions.

Figure 2

Table 1.Experimental sites, setup and environmental conditions of the five experiments. Water treatments consist of absence (S0) or presence (S1) of water stress. Seasonal aridity indices represent the ratio of total rainfall to reference evapotranspiration (ET0) during the growing season

Figure 3

Table 2.Key crop parameters for barley and wheat grown at the experimental sites. Values that were changed from the default are presented in bold

Figure 4

Table 3.Weed treatments with selected AquaCrop weed management input: relative weed cover (RC) and weed-induced total canopy cover increase (fweed) with corresponding canopy expansion factor (fshape). Water treatments consist of absence (S0) or presence (S1) of water stress

Figure 5

Fig. 3.Simulated (lines) and observed (symbols) CC (top) and dry above-ground biomass (bottom) in well-watered barley plots in Dejen 2009 (left) and Mekelle 2010 (right). Crop CC and biomass are affected by the weed infestation level: weed-free (black) and weed-infested with a relative weed cover of about 0·50 (grey). The dashed line and open symbol present the total vegetation CC in weed-infested conditions. Error bars indicate ± s.d. for three replications.

Figure 6

Table 4.Relative root-mean-square error (RRMSE), Nash-Sutcliffe model efficiency (EF) and coefficient of determination (R2) for n number of observations of average soil water content in the root zone, crop canopy cover in weed-free conditions (CCWF) or weed-infested conditions (CCW), total vegetation canopy cover (CCTOT), dry above-ground crop biomass (B) during the growing season and at phenological maturity, and crop yield in weed-free and weed-infested barley and wheat plots (all experimental sites and water treatments included). Dashes indicate that observations for performance assessment were unavailable

Figure 7

Fig. 4.Observed v. simulated dry above-ground biomass at maturity (left) and grain yield (right) of barley grown in weed-free (black) and weed-infested (grey) plots with water stress (circle) or without water stress (diamond) at the three experimental sites in 2009 & 2010. Error bars indicate ± s.d. for three replications.

Figure 8

Fig. 5.Simulated (lines) and observed (symbols) CC (top) and dry above-ground biomass (bottom) in well-watered (left) and water-stressed (right) wheat plots in Wagga Wagga, 1998. Crop CC and biomass are affected by the weed infestation level: weed-free (black) and weed-infested (grey) with a relative weed cover of about 0·20 for well-watered and 0·15 for water-stressed plots. The dashed line and open symbol present the total vegetation CC in weed-infested conditions. Data points represent an average value over five replications. Error bars indicating the s.d. were not available for the wheat dataset.