Article Information

Luis Giménez
 

Department of Plant Production. Mario A. Cassinoni Experimental Station. Faculty of Agronomy, University of the Republic. Paysandú 60000, Uruguay.

^*Corresponding Author: Luis Giménez, Department of Plant Production. Mario A. Cassinoni Experimental Station. Faculty of Agronomy, University of the Republic. Paysandú 60000, Uruguay

Received: 10^th June-2019; Revised: 28^th June-2019 ; Accepted: 29^th June-2019;

Copyright: ©2019 Luis Giménez. This is an open-access article distributed under the terms of the Creative Commons Attribution License , which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited

Citation:

Luis Giménez. Aquacrop Model Evaluation in Maize Under Different Water Availabilities in the Western of Uruguay. International Journal of Plant, Animal and Environmental Sciences 9 (2019): 103-117.

Abstract

Keywords

Crops Simulation, Grain Production, Water Deficit, Irrigation

Article Details

ABBREVIATIONS

AQ-Aquacrop; B- biomass; b- Regression Coefficient; CC- canopy cover; CCx- maximum root depth; CDC- canopy declined coefficient; CDI- controlled deficit irrigation; CGC- canopy growth coefficient; CP- critical period; DI-deficit irrigation; Dr- root zone humidity depletion; Es- soil evaporation; ETcact- actual crop; evapotranspiration; ETo- reference evapotranspiration; fHi- adjusted factor; HIo- reference crop harvest index; IRR- irrigation; Kc- crop coefficient; KcTrx- maximum standard transpiration coefficient; Kex- water stress coefficient; Kr- evaporation reduction coefficient; Ks- stress crop coefficient; Ksat- saturated hydraulic conductivity; LAI- leaf area index; MAE- mean absolute error; NCP- noncritical period; NRMSE- normalized root mean square error; p- fraction; AW that the crop can extract without occurring water deficit; PP- precipitation; REW- easily evaporable water; RMSE- root mean square error; SWC- soil water content; Ta- actual transpiration of the crop; TAW- total available water; Tmax- maximum temperature; Tmin- minimum temperature; WP*- normalized water productivity; Y- yield
 

INTRODUCTION

Summer crops have the largest sown area in Uruguay and, thereby, a high relevance in the country's exports [26]. More than 1.3 Mha are planted [17] and the main environmental limitation that explain their Y inter annual variation is water availability [24,3,12]. However, most of summer crops planted area ??is rainfed. Though according to [9], long term annual rain average is 1300 mm, atmospheric demands in summer are high and soils water recharges are due exclusively by precipitation (PP) which present a high variability [6]. In addition, soil water storage capacity (SWC) of Uruguay’s agricultural soils is low in relation to crops potential consumption and it covers only between 20 and 30% of their requirements[12]. Because of its morphophysiological characteristics, maize is a summer crop that has high sensitivity to water deficiencies, mainly around flowering, critical period (CP), as severe water deficiencies cause significant Y decreases[4]. Grain filling is also a cycle stage in which hydric stress causes Y decreases as the grains weight is determined [11,8].
 

Currently, irrigation water availability does not present great limitations in the country [9] but supplementary irrigation is scarcely used for different reasons, among which high energy costs stand out [5]. This is one of the main arguments that prevent a greater adoption of irrigation, particularly in corn. Traditionally, irrigation has been used to obtain potential Y, covering the total crops water needs. However, in such energy costs situations, deficit irrigation (DI) and controlled deficit irrigation (CDI) strategies are alternatives that allow improving water use efficiency [10] and to reduce irrigation direct costs.
 

On the other hand, crops models that adequately simulate productivity with different water managements are tools that help to improve irrigation planning. It is possible to evaluate the cost decreases by using an alternative water management other than to satisfy crops complete demand throughout the cycle, and thereby to provide knowledge to improve the adoption of irrigation. In any case, it is essential to evaluate models locally and to study their adaptation in different situations of water availability.
 

FAO’s AQ model [25] was proposed to simulate B production and Y in diverse crops, under different irrigation strategies, including the DI [17] and rainfed conditions [23]. Model focuses on water, it has the advantage of using a scarce number of parameters and, at the same time, it is of low sophistication which allows applicability in a wide users’ spectrum [23]. This work’s main objective was to evaluate AQ performance in maize in the western littoral of Uruguay, in different situations of water availability: full irrigation, CDI and rainfed crops.
 

MATERIALS AND METHODS

Model was parameterized in experiments carried out to study the effect of water deficiencies on corn Y, developed in an experimental irrigation field located at Paysandú (32º 22 'S and 58º 03' W), during 2009, 2010, 2011 and 2012 seasons. The experimental field soil is a Pachic Argiudoll. Tillage carried out in both experiments was conventional. Soil main hydrological characteristics are shown in Table 1.
 

Table 1. Experimental site soil physical and hydraulic properties.

For each studied season, Table 2 shows monthly values of average daily maximum and minimum temperatures (Tmax.and Tmin., °C), mean daily solar radiation (MJm^-2d^-1), mean reference evapotranspiration (ETo, mm), mean PP (mm), and all crop growing season parameters totals.
 

Trials were planted on October 27, 22, 29 and 27, respectively in 2009, 2010, 2011 and 2012, and harvested on March 2, February 28, March 13 and February 28 of 2010, 2011, 2012 and 2013, in that order. DK 692, an intermediate cycle hybrid characterized by a high Y potential in the cultivars’ national evaluation was used.
 

Plots were 5 m long and 3.5 m wide and were composed of 5 rows at a distance of 0.7 m. Target population was 100000 plants ha^-1 which was achieved in the four seasons. Weeds chemical control was carried out with an atrazine and metolachlor herbicides mixture, in doses of 1.5 l ha^-1 of commercial product of each one, pre-sowing applied.
 

Treatments and crop stages in which AQ model calibration and validation was performed were: T1 = full irrigation, T2 = water deficiencies in CP, T3 = water deficiencies during grain filling, T4 = deficiencies in the vegetative stage plus in CP and T5 = rainfed. Treatment T4 was evaluated in 2010, 2011 and 2012 since it was not possible to install it in 2009. In crop stages of T2, T3 and T4 with no caused water deficiencies, they were under hydric comfort conditions.
 

Treatments T1, T2, T3 and T4 were established through two types of interventions: a) supplementary irrigation and b) interception of the PP through rain shelters in water deficiencies treatments. Irrigation was drip type, with 1.49 l h^-1 flow 0.2 m spaced emitters tapes. It was applied according to the soil water content (SWC) variation, estimated through a simplified daily soil water balance, considering 0.7 m the maximum rooting depth.
 

Table 2. Monthly and all crop growing season values of daily average maximum and minimum temperatures, monthly daily mean and total growing season solar radiation, monthly average and total growing season reference evapotranspiration and precipitations.

SWC variations were estimated by the expression: ΔSWC=PP+IRR–ETc_act, where: ΔSWC=SWC variation (mm), IRR=irrigation (mm) and ETc_act= actual crop evapotranspiration (mm), this latter calculated as ETc_act=ETo x Kc x Ks, being ETo (mm)  the reference evapotranspiration estimated using the FAO-PM method [2], Kc=crop coefficient as those proposed by FAO, and Ks=crop stress coefficient, using for its calculation the formula proposed by FAO-PM method [2] where Ks=TAW-Dr/TAW(1-p), where Dr= root zone humidity depletion (mm), that is, the missing water in relation to soil field capacity and "p" is the fraction of the total available water (TAW) that the crop can extract without occurring water deficits. Used "p" were 0.4 TAW in the CP and 0.6 TAW in non-critical period (NCP).
 

SWC was measured by means of a calibrated neutron probe 503 DR HIDROPROBE (InstroTek Inc., Martinez, CA, USA).
 

Table 3 shows net irrigation and ETc_act for each treatment and season evaluated. Climatic parameters to estimate ETo were measured through an automatic weather station Vantage Pro 2^TM Model 6510 (Davis Instruments, Hayward, CA), located approximately 2000 m far from the experimental site. Irrigation was completed when SWC reached 90%. In treatments with water deficiencies, SWC was allowed to descend to 20% and then raised by irrigation up to 40% in NCP and up to 60% during CP.
 

To intercept the PP, 3.5m wide x 5m long x 2.5m maximum high steel frames with water proof canvas covers rain shelters were built. Rain shelters were placed immediately before each PP event and removed immediately after, trying not to substantially modify solar radiation and temperature conditions.
 

In each treatment, total B and Y were measured in 9 linear meters of the three central rows of each plot and, in addition, number m^-2 and weight of grains were determined. In T1 in 2009 and 2011, leaf area index (LAI) was measured through a Decagon AccuPar LP 80 ceptometer and transformed into CC through the empirical equation proposed by Hsiao et al. [16] where CC = 1,005[(1-exp (-0.6 x LAI)]^1,2. A completely randomized blocks experimental design was used. ANOVAS and means contrasts were performed to compare the evaluated treatments, and Tukey’s test at a significance level of 5% was used. Analysis were performed using the statistical package SAS v.9.2.
 

Table 3. Adjusted Evapotranspiration (ETc_act) and net irrigation, both in mm, in each treatment and season.

AQ is a crop growth model that calculates B and Y considering actual transpiration (Ta, mm) separated from soil evaporation (Es, mm). ETc_act(mm day^-1) is calculated as the sum of Ta and Es, where: Ta = Ks x CC x Kc_Trx x ETo (1) and Es = Kr x (1-CC) x Ke x ETo (2) where Kc_Trx(non-dimensional) is the maximum standard transpiration coefficient or maximum basal crop coefficient when CC = 100%; CC (%); Ks (0-1) is the water stress coefficient; Kex is the completely wet and unshaded soil surface evaporation coefficient (non-dimensional) and Kr (0-1) is the evaporation reduction coefficient [23]. Therefore, the calculation of the two ET components is mainly linked to CC simulation.
 

AQ model combines four sub-models: (1) soil water balance, (2) development, growth and Y, (3) atmosphere, rainfall, evaporative atmospheric demand through ETo and concentration of CO₂, (4) crop management that includes irrigation and fertilization [23].
 

B (kg ha^-1) is estimated by the model using the crop water transpired in the growing season and WP* (g m^-2). WP* represents the B produced per surface unit considering the accumulated transpiration, after adjusting for the atmospheric concentration of CO₂ and ETo [23]. Y (kg ha^-1) is calculated as: Y = fHI HIo x B (3). HIo is the reference crop harvest index, which indicates the proportion of B in the grains and fHI is an adjustment factor that integrates five water stress factors related to leaf growth, stomatal closure, B reduction due to stress before anthesis, and failure in pollination.
 

CC in AQ is a fundamental parameter equivalent to the fraction of soil covered by the canopy. Model does not allow the usage of data to construct the CC curve, but it allows to calibrate the CC curve. CC calculations are made through three phases [23]: the first uses an exponential function of time,which begins at crop emergence and ends when half of the maximum CC is reached, and CC growth rate defined by the CGC parameter. The second phase uses another exponential function until the maximum CC (CCx) is reached, being the shape of the curve given by the same CGC parameter. The last phase refers to the decrease in canopy coverage after the start of crop senescence and the curve shape is defined by the CDC parameter. To parameterize the CC curves, this is, finding the CCx, CGC and CDC with best fitting to the observed values ??of B and Y, observed LAI data can be used to calculate CC values ??through a function proposed by Hsiao et al., [17].

Model input data [23] include daily meteorological data of (1) max and min T (°C), PP and ETo. Atmospheric data are referred to the annual concentration of CO₂ (2). Crop data that refer to: i) sowing dates, date on which the maximum CC is reached, date on which the maximum root depth is reached, crop senescence starting date, and date of maturity, (ii) maximum Kc_Tr value, (iii) minimum and maximum root depth Zr (m) and root expansion shape factor, iv) initial coverage (CCo) and CCx, CGC and CDC, (v) WP*, (vi) HIo, (vii) Ks related to canopy expansion, stomatal closure, early canopy senescence and aeration stress (3). Soil data for five maximum horizons soil. For each horizon the data requires horizon depth (m), field capacity (m³m^-3), permanent wilting point (m³m^-3), water content at saturation (m³m^-3), and saturated hydraulic conductivity (Ksat,m.s^-1). In relation to the soil profile, the easily evaporable water (REW, mm) and the curve number (4). Irrigation data, dates and water depth applied (5). Field management practices related to salinity, soil fertility and runoff reduction.
 

Model calibration in corn was carried out in T1 of 2009, 2010, 2011 and 2012 seasons experiments. Calibration process was started running the model with the conservative parameters proposed by Hsiao et al. [17]. Then, based on CC observations, parameters that give the CC curve shape, i.e. CGC and CDC, were adjusted. Subsequently, the WP* was adjusted by means of a sharp trial and error process with the objective that the differences between observed and simulated B and Y values were the minimum possible.
 

In this study AQ was statistically evaluated with four indexes, used to measure the adjustment quality of simulation models, which are detailed below:
 

1)The root mean square error (RMSE) that expresses the variance of the residual errors, which values vary between 0 and +∞, and its formula is:

being Oi the observed and Si the simulated values.
 

2) The normalized RMSE (NRMSE) defined as the relation between the RMSE and the observations mean value, expressed in percentage, which formula is:

3) Willmott’s index (d)(1982), varying in a rank of -∞ and 1. The model has best adjustment when the index is close to l and it is considered a bad adjustment when “d” values are negative. 

4) Mean absolute error (MAE) that expresses the magnitude of the mean estimated errors.

5) Regression coefficient (b) which, if being close to 1 indicates that the simulated values are statistically close to those observed.

RESULTS AND DISCUSSION

Aquacrop calibration

Table 4 shows the conservative and non-conservative parameters used in AQ calibration   in corn, including adjustments made in CGC and CDC that shape the CC curve. Adjusted WP* value is also shown in treatments T1 of 2009, 2010, 2011 and 2012 seasons. Values of the coefficients ??obtained for the calibration of the CC curve were: CGC=13.8%d^-1, CDC=1.081%GD^-1 and WP*=34 gr m^-2. Calibrated parameters presented values ??close to those proposed by Hsiao et al.[17] in the initial calibration of AQ in maize. In case of WP* the same value as that achieved by Abedinpour et al. [1] was adjusted. On the other hand, values ??obtained for CGC and CDC are higher than those indicated by Paredes et al. [21] for maize. It is possible that this adjustment response of the CGC and CDC parameters has a certain dependency on local conditions of temperatures and radiation, since they are the factors that, in absence of water and nutrients deficiencies, control crop development and growth, respectively. In the remaining conservative parameters, values ??proposed in the initial calibration were used [17]. Table 4 shows the set of conservative and non-conservative parameters used in model calibration


Table 4. Conservative and non-conservative parameters used in Aquacrop model calibration for corn.

Figure 1 shows observed and simulated values of the CC curve in 2009 (a) and 2011 (b) seasons, with CGC and CDC parameters adjusted in the calibration process. CC curve is a key aspect in the operation of AQ, since model estimates crop transpiration through the CC and, in turn, Ta is used for total B calculation. Similarly, Abedinpour et al.[1], obtained good CC estimates under water availability conditions adjusted to corn requirements without nitrogen restrictions.
 

Figure 2 shows values observed and simulated by AQ of total aerial B (a) and Y (b). In both variables, a good fit was found between observed and simulated values.

Figure 1. Values simulated by Aquacrop (line) and observed (markers) of canopy cover (CC) in full irrigation treatments (T1) of 2009 (a) and 2011 (b) seasons.

Figure 2. Relationship between values observed and simulated by Aquacrop in biomass (a) and yield (b) in full irrigation treatments experiments (T1) of 2009, 2010, 2011 and 2012 seasons.

Table 5 shows statistical indicators used to evaluate model performance in the calibration. In 2009 and 2011 seasons in which they were measured, CC was well simulated by AQ, NRMSE presented lower than 7% values being considered excellent simulations when this indicator is less than 10%, the concordance indicator "d" presented a value of 0.99 while the best result for this indicator is 1. In case of B and Y, statistical indicators used to evaluate model showed that under conditions of crop full irrigation both variables were well simulated by AQ in the four seasons.
 

Table 5. Root mean square error (RMSE), normalized root mean square error (NRMSE), Willmott’s concordance index ("d"), mean absolute error (MAE) and regression coefficient (b) for observed and simulated biomass, yield and canopy coverage values in Aquacrop calibration.

Figure 3 shows SWC simulated by the model and observed in T1 in the four seasons. Congruence between observed and estimated values can be seen in most of the water measurements carried out.This justifies the good estimates that AQ made of total B and Y, since for the model, water availability is the key aspect for estimating both variables.
 

According to the results obtained, it can be concluded that adjustments made in crop parameters CGC, CDC and WP*, proposed by Hsiao et al. [17], allowed good simulations of corn B and Y in non-limiting water conditions in which the model was calibrated.
 

Figure 3. Full irrigation treatments in 2009, 2010, 2011 y 2012 soil water content evolution simulated by Aquacrop (blue line) and observed (red markers), soil water content at field capacity (black line) and soil water content at permanent wilting point (green line). Soil water content (vol%).
 

Similarly, in different regions it was found that under no water stress crop conditions, model simulated well CC, total Band Y [7]. Also, [1], added that model presented safe and adjusted simulations of B and Y in situations of full irrigation and when irrigation covered 75% of ETc.
 

Aquacrop validation

Model validation was made in treatments with water deficiencies caused in crop stages CP (2), grain filling (T3), vegetative stage plus in CP (T4) and rainfed crops (T5). Table 6 shows B and Y values observed and simulated by AQ and their differences in the four seasons.
 

Table 6. Total biomass and yield and differences (Diff.) between Aqua crop simulated  (Sim.) and observed (Obs.) values, in full irrigation (T1), water deficiencies around flowering (T2), water deficiencies in grain filling (T3), water deficiencies in vegetative stage plus around flowering (T4) and rainfed crop (T5) treatments, in 2009, 2010, 2011 and 2012 seasons.

In the majority of treatments with caused water deficiencies (T2, T3 and T4) it was found that AQ model simulated maize B and Y with errors. In treatment with water deficiencies in grain filling (T3), model over estimated B and Y in the four seasons. Hydric deficiencies during grain filling cause early crop senescence losses of B, diminution of the canopy coverage and, as a consequence of this, diminution of Y due to lower grains weight. [7,16] agree that AQ model does not simulate well severe water deficiencies in corn final stages. Hydric deficiencies in grain filling cause losses of B and Y that vary between 20 and 30% of the crop potential [11,8]. In the AQ simulations carried out, B and Y losses during grain filling due to severe water deficiencies varied between 1 and 12%. Simulations results show clearly that AQ does not have certainty to model water deficiencies effects in grain filling, and therefore does not simulate well the B and Y decreases that occur in that stage.
 

In case of water deficiencies caused in CP (T2), in two of the seasons AQ over estimated B and Y, and in the other two seasons it underestimated both variables. Additionally, it was verified that in seasons with relatively scarce PP, model underestimated them. In contrast, in 2009 season with abundant PP throughout the cycle, and in 2012 with abundant PP greater than 400 mm in crop initial stages alternating with deficiencies in other stages, AQ overestimated both variables. In reviewed literature there is no record of model performance evaluation in situations with severe water deficiencies solely around flowering. This is a maize key development stage and water deficiencies cause significant Y decreases [11,8,4] due to corn strong apical dominance that induces spikes to be relegated to the male panicle in crop water stress situations [4]. Water stress in CP affects viable grains number negatively, with Y decrease consequences [13,27,15].Water deficiencies located in this stage can lead to losses between 40 and 50% of the potential Y [11,8]. In no evaluated case with deficiencies around corn CP AQ simulations were correct.
 

In 2010, 2011 and 2012 seasons, effects of accumulated water deficiencies in the vegetative stage plus in CP (T4) were evaluated. In this case, AQ simulations of B and Y presented the biggest errors. As seen in Table 6, water deficiencies in T4 were the longest among those evaluated, model showed strong underestimations in 2010 and 2011, both in B and Y. In contrast, in 2012 errors in the simulation were in the opposite direction, that is, AQ simulated higher values ??of B and Y to those observed. As already mentioned, in 2010 and 2011 the PP were scarce while relatively abundant in 2012 and the trend in the simulations was the same as that found in T2, that is to say, in seasons with scarce PP simulations underestimated B and Y, and in years with elevated PP, AQ showed over estimates in both variables.
 

Results obtained in rainfed crops show different levels of adjustment according to the evaluated season. In season 2009, characterized by PP exceeding 1500 mm during the crop cycle, allowing a high water availability throughout it, the model showed a good fit between B and Y simulated and observed values, as happened with crop full irrigation treatments, confirming that when water requirements in maize are covered, either by irrigation or by PP, model simulates well B and Y.
 

In contrast, in 2010 and 2011 seasons, scarce PP during December (79 and 48 mm, respectively), and January (43 mm, until 25^th. In 2010 and 67 mm in 2011), caused significant water deficiencies in CP of rainfed crops. In this situation, AQ underestimated B and Y with respect to the observed values, as occurred with treatments with deficiencies caused in CP, and with deficiencies accumulated in the vegetative stage plus in CP, in both seasons. This AQ behaviour to simulate well under good water availability and to present errors in severe water stress situations was verified and discussed by [15,7,19]. In 2012 season, when rainfed crop experienced better water conditions than in 2010 and 2011, alternating good water availability and scarcity stages, the model overestimated total B and Y.
 

Table 7 shows the statistical indicators of fitting goodness used to evaluate AQ simulations of B and Y for treatments with caused crop water deficiencies and rainfed. It was found that RMSE and NRMSE presented high values. RMSE in all cases was far above 0, which is the perfect adjustment value, and NRMSE exceeded 20%, considered the upper limit for adjusted simulations. MAE values ??were high in all cases and Willmott’s "d" match index showed values ??far from the unit, and the regression coefficient b indicated overestimates for the T3 and underestimates for the T2, T4 and T5 treatments, in both variables.
 

Table 7. Root mean square error (RMSE), normalized root mean square error (NRMSE), mean absolute error (MAE), Willmott’s concordance index ("d") and linear regression coefficient (b) for water deficiencies around flowering (T2), in grain filling (T3), in vegetative plus around flowering (T4) and in rainfed crops (T5) treatments, in seasons 2009, 2010, 2011 and 2012.

Figure 4 shows root zone SWC simulations evolution carried out by AQ, in treatments and seasons evaluated. Water evolution presented the expected trends, both in treatments with crops full irrigation with high water contents throughout the cycle, and in those with water deficiencies caused in different stages. In rainfed crops the behaviour was different in each season due to PP irregularity.
 

Figure 4. AQ model simulations evolution of corn cycle root zone soil water content (SWC, vol%) in treatments: a) full irrigation (T1), b) deficiencies in CP (T2), c) deficiencies in grain filling (T3), d) deficiencies in vegetative stage plus in PC (T4), and e) rainfed (T5). Respective seasons and colour lines: 2009 light blue, 2010 red, 2011 black, and 2012 green.
 

Figure 5 presents SWC evolution of AQ simulations and measurements in 2011 season, in evaluated treatments. It is clear the good fit between SWC in AQ simulations and in observed values in crop full irrigation situation. On the other hand, discrepancies between soil water simulations and measurements were verified in different stages of the crop cycle, both in treatments with caused water deficiencies (T2, T3 and T4) and T5. In T2 and T4, model simulated a SWC below the measurements in most of the stages, even with low water levels close to permanent wilting point. In turn, as indicated, in both treatments of 2011 season model strongly underestimated B and Y (Table 6).
 

Figure 5. Root zone soil water content (SWC, vol%) evolution simulated by Aquacrop (blue line), measurements made with neutron probe (red markers), content at field capacity (light blue line) and content at permanent wilting point (green line), in 2011 season, in treatments with full irrigation (T1), water deficiencies around flowering (T2), water deficiencies in grain filling (T3), water deficiencies in the vegetative stage plus around flowering (T4) and rainfed crop (T5).
 

Similarly, [1] and [7] founded that in situations of severe water deficiencies during crop cycles in which only 50% of the ETc was covered, simulations of B and Y developed by AQ in maize were not adjusted to the observed data. Figure 5 shows root zone SWC simulations in T3 are mostly above the measurements and model overestimated B and Y. Again, in T5, model simulated SWC values ??lower than measurements and, in turn, underestimated B and Y, as indicated above.
 

Figure 6 shows the relationship between simulated and observed values ??of total B and Y for all the treatments with caused water deficiencies and rainfed crop.
 

Figure 6: Relationship between observed and AQ simulated values of total biomass (a) and yield (b) for the set of treatments: water deficiencies around flowering (T2), water deficiencies in grain filling (T3), water deficiencies in vegetative stage plus around flowering (T4) and rainfed crop (T5).
 

Observed and simulated values ??dispersion demonstrates AQ problems to model both variables in situations of corn crop water deficiencies. According to the information presented in figure 5, it is possible that errors are due to problems in root zone water content simulation when there are severe deficiencies. Based on the results obtained, we interpret that model does not simulate adequately the effects of water deficiencies in corn. It is possible that AQ does not present good adjustments of Ks for severe water deficiencies. On the other hand, from results obtained in rainfed crops, it can be inferred that model operation depended on PP characteristics in each season. In seasons in which PP were abundant and allowed good water availability for most of the crop cycle, AQ performed adjusted simulations of B and Y. On the contrary, when PP were scarce and there were severe water deficiencies, simulations showed errors in relation to values ??observed in both variables.
 

CONCLUSIONS

AQ model was evaluated during four years in corn experiments with different water availabilities. After adjustments made in the calibration process in CGC, CDC and WP* parameters, model showed good predictions of B and Y in situations of crops full irrigation.
 

On the opposite, in conditions of water deficiencies in CP, AQ simulated with errors. In seasons with scarce PP it underestimated B and Y, as in high water availability seasons it overestimated results in both variables. In situations with deficiencies in grain filling, model overestimated B and Y through the four seasons. In treatments with water deficiencies accumulated in vegetative stage plus in CP and in seasons with low PP, model strongly underestimated both variables while overestimated them in high PP seasons.
 

Results in rainfed crops confirmed AQ good functioning when water availability is not restrictive, in this case due to abundant PP. However, in seasons with water deficiencies, model underestimated the observed values ??of B and Y. Results obtained lead to infer that AQ does not present adjusted stress coefficients values ??and that this produces problems in soil water modelling that later are translated into B and Y simulations with errors in severe water deficiencies situations.
 

ACKNOWLEDGMENT

The author wishes to thank INIA’s Promotion of Agricultural Technology Fund for financing the experiments through the FPTA 231 research project. Many thanks also to Undergraduate and Master students who developed their thesis works in the mentioned project.
 

Conflicts of Interest

The author declares no conflict of interest.
 

 REFERENCES

1. Abedinpour, M., Sarangi, A., Rajput, T.B.S., Singh, Man., Pathak, H., Ahmad, T. 2012. Performance evaluation of AquaCrop model for maize crop in a semi-arid environment.  Agricultural Water Management, 110: 55-66. 
2. Allen, R., Pereira, L.S., Raes, D., Smith, M. 1998. Crop Evapotranspiration. Guidelines for computing crop water requirements. FAO Irrigation and Drainage Paper, 56, pp: 300.
3. Andersen, J., Alagarswamy, G., Rotz, C., Ritchie, J., LeBaron, A. 2001. Weather impacts on maize, soybean and alfalfa production in the great lakes region. Agronomy Journal, 93: 1059-1070.
4. Andrade, F., Cirilo, A., Uhart, S., Otegui, M.E. 1996. Ecophysiology of corn crop. (in Spanish) La Barrosa y Dekalb press. Argentina. pp: 292.
5. Bachino, F. 2012. Analysis of costs and profitability in sprinkler irrigation. In 2^nd. International Seminar On Irrigation in Crops and Pastures (in Spanish). Paysandú. Uruguay. pp. 83-97.
6. Baethgen, W.E., Terra, R.2010. Irrigation in a changing climate (in Spanish). Work presented in: International Seminar on Extensive Irrigation in Crops and Pastures Potential (1°, 2010, Paysandú, Uruguay). Boscana. Montevideo, Uruguay. pp. 7–18.
7. Bitri, M., Grazhdani, S.2015. Performance Evaluation of AquaCrop Model for Irrigated Field Maize in South-eastern Albania. Journal of International Environmental Application and Sciences 10 (3): 375-383.
8. Çakir, R. 2004. Effect of water stress at different development stages on vegetative and reproductive growth of corn. Field Crops Research 8 (1): 1-16.
9. FAO-AQUASTAT. 2016. Uruguay. (Online). Consulted December 20, 2016. Available in: http://www.fao.org/nr/water/aquastat/data/wrs/readPdf.html?f=URY-WRS_eng.pdf
10. Fereres, E., Soriano, M.A. 2007. Deficit irrigation for reducing agricultural water use. Journal of Experimental Botany, 58 (2): 147–159.
11. Giménez, L. 2012. Production of corn with water stress caused in different stages of development. (in Spanish). Agrociencia (Uruguay) 16 (2):  92-102.
12. Giménez, L., García Petillo, M. 2011. Evapotranspiration of summer crops for two climatically contrasting regions of Uruguay. (in Spanish). Agrociencia (Uruguay) 15 (2): 100-108.
13. Grant, R., Jackson, B., Kiniry, J., Arkin, G. 1989. Water deficit timing effects on yield components in maize. Agronomy Journal, 81: 61-65.
14. Greaves, G. E., Wang Y-M. 2016. Assessment of FAO AquaCrop model for simulating maize growth and productivity under deficit irrigation in a tropical environment. Water, 8 (12), 557.  doi.org/10.3390/w8120557
15. Hall, A.J., Lemcoff, J.H., Trápani, N. 1981. Water stress before and during flowering in maize and its effects on yield, its components, and their determinants. Maydica, 26: 19-38.
16. Heng, L.K., Hsiao, T.S., Evett, S. Howell, T., Steduto, P. 2009. Validating the FAO AquaCrop Model for Irrigated and Water Deficient Field Maize. Agronomy Journal, 101: 488–498.
17. Hsiao, T.C., Heng, L.K., 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.
18. MGAP-DIEA (Ministry of Livestock, Agriculture and Fisheries-Directorate of Agricultural Statistics). 2016. Winter agricultural survey series 337. (Online). Consulted in November 2016. Available in: http://www.mgap.gub.uy/sites/default/files/encuesta_agricola_invierno_2016.pdf.
19. Katerji, N., Campi, P., Mastrorilli, M. 2013. Productivity, evapotranspiration, and water use efficiency of corn and tomato crops simulated by AquaCrop under contrasting water stress conditions in the Mediterranean region. Agricultural Water Management, 130: 14-26.
20. Nash, J.E., Sutcliffe, J.V. 1970. River flow forecasting through conceptual models. I. A discussion of principles. Journal of Hydrology, 10: 282–290.
21. Paredes, P., de Melo-Abreu, J.P., Alves, I., Pereira, L.S. 2014. Assessing the performance of the FAO Aquacrop model to estimate maize yields and water use under full and deficit irrigation with focus on model parameterization. Agricultural Water Management, 144: 81-97.
22. Paredes, P., Rodrigues, G.C., Alves, I., Pereira, L.S. 2014. Partitioning evapotranspiration, yield prediction and economic returns of maize under various irrigation management strategies. Agricultural Water Management, 135: 27-39.
23. 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 (3): 438-447. 
24. Sawchik, J., Ceretta, S. 2005.Consumption of water by soybeans of different maturity groups in different production environments (in Spanish). In. Summer Crops Technical Day.  Diffusion Activities Series, 417. pp. 41-45.
25. Steduto, P., Raes, D., Hsiao, T.S. 2009. Concepts and applications of AquaCrop: the FAO crop water productivity model. Crop Modelling and Decision Support, pp. 175–191.
26. Uruguay, Siglo XXI. 2016. Foreign trade report. Exports and Imports of Uruguay (in Spanish). (Online). Consulted July 20, 2016. Available in:http://www.uruguayxxi.gub.uy/informacion/wp-content/uploads/sites/9/2016/07/Informe-mensual-de-comercio-exterior-Junio-2016.pdf
27. Westgate, M., Boyer, J. 1986. Reproduction at low silk and pollen water potentials in maize. Crop Science, 26: 951-956.
28. Willmott, C.J. 1982. Some comments on the evaluation of model performance. Bulletin American Meteorological Society, 63: 1309-1313.