Estimating crop water requirements for irrigation scheduling of tomato using Aquacrop 5.0 and Cropwat 8.0 models under scarcity and unreliability of rainfall in Harare

.


INTRODUCTION
Global food demand is rising and the uncertainty of food supply increasing due to the impact of climate change *Corresponding author. E-mail: gmuroyiwa2009@yahoo.com Author(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License and rising variability of rainfall. The global population is projected to increase up to 9.1 billion in 2050 and 2.4 billion for Africa by 2015 (FAO, 2015;Hall et al., 2017). At least 60% more food production is needed by 2050 to feed the world's population and hence there is a need for more sustainable production of sufficient, safe and nutritious food supply for future food security (FAO, 2015). Recent estimates from the Food and Agriculture Organisation of the United Nations suggest that one in four people in Africa lack adequate food to sustain an active and healthy life (Bremner, 2012;FAO, 2015) and the impact of climate change has overwhelmed food security systems on the African continent (FAO, 2016). About 70% of people in developing countries live in rural areas and they depend on subsistence farming which has been characterised by low productivity as a result of marginal and erratic rainfall (Nethononda and Odhiambo, 2011;Vermeulen et al., 2012). Climate change is expected to exacerbate the pressure on the planet's available water resources with a parallel increase in the irrigation water requirements by up to 70 to 90% through 2050 (Garrote et al., 2014;Kreins et al., 2015). As a result, the shortage of existing resources threatens the stability of agricultural crop production and will eventually overwhelm the planet's food security in the near future (Vote et al., 2015). One of the potential promising prospects for alleviating the increasing water scarcity is to exploit the available irrigation surface and groundwater in a more sustainable way (Vote et al., 2015). Using precision irrigation techniques, it is possible to apply less water while maintaining or even improving crop quality and yield (Munyaradzi et al., 2013a, b). Recent studies have indicated that the use of drip irrigation systems could result in better yields as compared to full irrigation, when the same amount of water is applied on arable crops and vegetables (Rashidi and Keshavarzpour, 2011;Al-Said et al., 2012;Vote et al., 2015;Tsakmakis et al., 2017) and a higher water use efficiency is obtained from precision irrigation (Raeth, 2020(Raeth, , 2021. Full irrigation at 100% crop water requirements, is a condition where crop growth is fully supported by applied water and stands as a control versus the 80, 60 and 50% ETc. This guarantees achievement of maximum crop production, as plants are supplied with the optimum water to counterbalance the evapotranspiration demand (Allen et al., 1998). Any irrigation scheduling strategy applying less water than that applied in the full irrigation is then considered as deficit irrigation (Muroyiwa et al., 2022). The effects of deficit irrigation on the final crop production and water consumption has been studied extensively for a variety of arable crops and vegetables (Qiu and Meng, 2013;Bakhsh et al., 2012;Jinxia et al., 2012;Igbadun et al., 2012;Tsakmakis et al., 2017). It is summarized that deficit irrigation has the potential to decrease the water consumption per unit of crop yield, compared to the full irrigation strategy (Geerts and Raes, 2009). Simulation models measuring the effects of water use on fruit yields at farm level have been used for irrigation and water management (Sam-Amoah et al., 2013). Testing the yield response to various water applications in the field or controlled experiments is time-consuming and expensive (Bitri et al., 2014) and due to such limitations, modeling has become an important tool for studying and developing promising deficit irrigation strategies (Blum, 2009;Geerts and Raes, 2009) which allow a better assessment of various environment and management factors, as well as crop water stress affecting crop yield (Bray, 1997). If this is combined with frequency analysis and historical climatic input data, the irrigation timing and dose irrigation application can be optimized for varying weather conditions (Pereira et al., 2002;Liu et al., 2007;Popova and Pereira, 2008). The models allow integrated testing of various yield factors such as biomass, canopy cover, water productivity and harvest index to determine the appropriate amount of irrigation for a variety of conditions (Liu et al., 2007).
The models were run for full irrigation (100% ETc) and deficit irrigation scenarios (50-95%) ETc, to examine the potential water use and yield obtained. Therefore, guidelines for irrigation applications were developed for small scale farmers in the area to allow for a sustainable use of the scarce resource. The guidelines can be used in another cycle of calibration and field-testing for possible future work. To assist the small scale farmers in the application of irrigation water to tomato crop in a particular area, the guidelines are summarized in irrigation charts which consider the local weather conditions, soil type, management conditions and water availability. Due to the declining water resources, there is an urgent need to increase crop water productivity  through a timely application of irrigation water Zwart and Bastiaanssen, 2004) thus avoiding over-and under-irrigation guaranteeing optimal growing conditions throughout the season .
In this paper, the objective of the research was to carry out a systematic model study for tomato (Lycopersicon esculentum) irrigation management in Harare using the Cropwat 8.0 and Aquacrop 5.0 models. This study aimed to compare estimation methods of crop water requirements and irrigation scheduling method using the two models to check the significance difference for adoption in different situations in Harare. The models were calibrated using data obtained from field experiments undertaken at University of Zimbabwe's Thornpark research station in Harare from 2014 to 2017 mainly for the case of crop cultivation during the hot dry season and also for supplementary irrigation during the wet season. The charts provide guidelines for the adjustment of drip-irrigation (DI) applications to the actual weather conditions.

Aquacrop and Cropwat models
Cropwat 8.0 is a decision-support tool developed by the FAO and designed for practical use by agronomists, agro-meteorologists and irrigation engineers (Antoine, 1998;Bernardi, 2004). It is an empirical process-based crop model that is used to calculate crop water and irrigation requirements from crop and climate data (Vote et al., 2015). It can be used to estimate crop performance under both rain-fed and irrigated conditions based on calculations of the daily soil water balance and can evaluate the farmer's irrigation practice and establish water supply schedules for different cropping patterns within an irrigation scheme, for a maximum of 20 different cultivars (FAO Water Development and Management Unit, 2013;FAO, 2014).
Aquacrop 5.0 is a dynamic water-driven crop-growth model developed by FAO Land and Water Division to simulate biomass and yield response of herbaceous crops under varying water use, management and environmental conditions Steduto et al., 2009a, b). It was developed as a tool for farmers, agronomists, engineers, water managers, economists and policy makers. Aquacrop is similar to Cropwat and incorporates the relatively novel concept of crop water productivity (WP) in order to transform the estimated crop evapotranspiration to final crop yield (Steduto et al., 2007). Detailed information is needed for soil, crop and field management files making Aquacrop a more sophisticated model (Table 1) and input data requirements (Table 2) are similar to those of Cropwat for the climate file. Consequently, the output results of Aquacrop (Table 3), in addition to the soil water balance data include information regarding the crop's final dry above-ground biomass, dry yield, water productivity, harvest index (HI) and yield. The model also generates crop development and production for daily, 10 day, monthly and yearly periods together with canopy development from transplant, root deepening to maturity, growth stages and the crop calendar (Tables 6 and 8).
This model is suitable for evaluating the effect of irrigation schedules on crop yield. It calculates the soil water balance, considering rainfall, irrigation, capillary rise, runoff, evaporation, transpiration, and deep percolation and to simulate plant growth and yield; it requires climate, soil, crop, and field management characteristics to be specified in the model.
Aquacrop and Cropwat models are FAO's agronomic models which assess the impact of irrigation technology and strategy for estimation of crop water requirements. The main differences are summarized in Table 1.

Aquacrop 5.0 model
AquaCrop interfaces with the main components of the soil-crop-atmosphere continuum and hence can generate an irrigation schedule with depth criteria (back to field capacity or fixed net application) and time criteria (fixed interval and allowable depletion in mm or % of readily available water RAW) and hence a calendar (Tables 9 and 10) can be generated with the addition of climate data (rainfall, temperature, CO2, ETo). The model can be used to develop a seasonal deficit irrigation schedule for a specific crop and can be used to optimise irrigation practices by comparing simulated model outputs with actual field data. The Aquacrop model determines an irrigation program that ensures that soil water content within the crop root zone is fully depleted at the time of harvest. It assesses the impact of soil properties (e.g. soil fertility) and management practices on yield and determines the optimal planting date based on probability analysis of historical rainfall and ETo data. At a larger scale, Aquacrop can be used to assess the effect of weather and climate on crop production and water use. For instance, the impact of rainfall variability on crop yields in rain-fed areas can be predicted using historical climate data. Through the use of data from an automatic weather station the model outputs can also be used to map the yield potential of deficit irrigation as well as rainfed system. Possible implications of a changing climate (air temperatures, humidity, solar radiation and atmospheric CO 2 concentrations) on future crop production and water use can be simulated using the model.
The Aquacrop model version 5.0 used in the current study to evaluate irrigation water use and crop yield is more ideal for applications requiring iterative simulations as in this study.

Cropwat 8.0 model
Cropwat estimates reference evapotranspiration (ET o ), crop water requirement (CWR), irrigation scheduling, and irrigation water requirement, using rainfall, soil, crop, and climate data ( Table 4). The program allows the development of irrigation schedules for different management conditions and the calculation of scheme water supply for varying crop patterns. Cropwat 8.0 can be used when local data is not available, and measures standard crop and soil data. When local data is available, the data files can be modified or new files can be created. The development of irrigation schedules in Cropwat 8.0 is based on a daily soil-water balance using various userdefined options for water supply and irrigation management conditions (Table 7).

METHODOLOGY
Aquacrop was calibrated and validated for the tomato crop method using field data obtained from a trial run between 2014 and 2017 at  Thornpark station, University of Zimbabwe. The calibration and validation exercise is quite extensive to be included in this manuscript and has been dealt with extensively by Muroyiwa et al. (2022). Simulations were run from a series of historical data which covered a period of 30 years from 1991 to 2020. To formulate guidelines for DI applications, the crop water requirements were considered and involved the following four steps as adapted from Geerts et al. (2010).
Step one: Crop development from transplant to harvest is simulated under rain-fed conditions where 100% crop water requirements is used (Table 11).
Step two: Simulations are run for the complete growing cycle from April to August of each year from 2014 to 2017 for the hot and dry period.
Step three: The simulation results were statistically analysed and indicative degrees of crop development for different years were obtained.
Step four: The generated schedules with varying irrigation intervals were simplified and translated into easy readable charts The following inputs were used for both models: Long-term historical climate data (1991 to 2020) including the weather data for the whole year: Daily rainfall and air temperature and daily reference evapotranspiration (ETo) and the default CO 2 . Local soil physical characteristics: Type of soil, restrictive soil layer, rooting depth. Crop characteristics of the tomato variety: Planting data, growth cycle, development, maximum effective root zone, plant density, initial canopy cover.

Effects of drought stress on biomass growth, transpiration and senescence were monitored.
Management practices: (1) Irrigation method, deficit.   (2) Irrigation events: Aquacrop 5.0: (i) Time criteria-allowable depletion (% of RAW) (ii) Depth criteria-fixed net application (mm) Cropwat 8.0: (1) Fixing the interval (2) Adjusting the depth to a constant value for no yield reduction and minimum water loss and the 100% readily available soil moisture depletion: (i) Maximum number of iterations selected.
(ii) Complete climate and irrigation files for four cropping seasons created and put into the models.
Eleven simulations for each season run to obtain an average for each crop season.
(1) Models run for the dry season to evaluate crop yield and total water use for each of the simulations.
(2) Specific day, 10-day and a monthly interval chosen for the irrigation scheduling.
The irrigation application depth was considered as fixed. Fixed application depths in combination with a variable irrigation interval resulted in an efficient use of the irrigation water. During field work there were four treatments for each trial, that is, 100, 80, 60, and 50% ETc but for the models, eleven simulations were done at 5% intervals with a range from 50 to 100% ETc. From field work data and model simulations, the calendar was developed using the input parameters for 60% ETc which had the highest water use efficiency.

Case study of Harare, Thornpark station
The research was confined to the University of Zimbabwe's Thornpark farm (-17.42 1981 -2017), showed that the mean monthly maximum temperature of 28°C were recorded in the month of November (Figure 1). Precipitation falls mainly in November to March. The other months are generally dry. On average the area receives a mean annual rainfall total of about 800 mm and has a mean monthly minimum temperature of 7C and a mean annual average temperature of 25C in summer (Figure 1). The land is relatively flat with slopes of 2% or less and has deep to moderately deep well drained red clay-loam soils (Mhizha, 2010). The land has typical water holding capacity within 0.80 m of the soil of about 175 mm, with an available water content of 12%v/v (Mhizha, 2010) ( Table 1). The applicability of the Aquacrop model to simulate water use of tomato in Harare has been evaluated by Muroyiwa et al. (2022) who conducted field experiments at Thornpark research site, University of Zimbabwe (Table 5). He concluded that the model could be used to evaluate the water use of tomato in Zimbabwe, Harare and indicated that with minor calibration and validation of the model parameters, the Aquacrop can simulate growth, water use and yield with acceptable accuracy. Parameter values established during field work for the tomato crop were used to calibrate the Aquacrop model used in the current study. The cropping season considered for simulations was from April to end of August. The planting date was 25 April and 27 August for maturity date and the irrigation depletion was at 60% ETc. The length of the irrigation season was 124 days. The decision variable was a vector of ten real numbers representing maximum allowable soil water depletion levels as percentages of readily available water (RAW). Irrigation was applied when soil water depletion levels reached these values. Irrigation application depth was back to field capacity. These parameters were also applied to the Cropwat 8 model

RESULTS AND DISCUSSION
The precision for Aquacrop model was good with a Pearson correlation coefficient of 0.9 for biomass showing that the models' simulation performance was good since the correlation coefficient was close to zero. The Pearson correlation coefficient for the simulated Cropwat and Aquacrop models showed a value of 0.06 and a p-Value of 0.81 which is expected due to the difference in calibration of the models. The study, showed some negative aspects of the Aquacrop models' performance under different irrigation schedules for simulation of tomato while showing that Cropwat can also be used effectively to estimate crop water requirements. The maximum irrigation requirements per cropping season for Aquacrop were at 555.3 mm and for Cropwat it was 675.0 mm. The yield obtained from the Cropwat model was at 2.93 t h -1 versus a yield of 3.40 t h -1 for the Aquacrop. The water productivity for Cropwat stood at 0.94 kg m -3 and for Aquacrop it was 0.97 kg m -3 . Under the scenarios presented and other conditions of the     ; Biomass production, 10.831 t ha-1; Expected yield, 3.40 t ha-1; Actual water use, 471.6 mm. Interval in days: Initial 23; Development 42; Mid 27; Late 20. Record rainfall before irrigation; Subtract rainfall from water requirements to obtain the irrigation amount. Source: Authors research has shown that yield potential can vary greatly between crops, crop varieties and within single cultivars are based on micro-climates, soil environments and nutrient availability (Lovelli et al., 2017;Singh et al., 2010). Although the Cropwat model is relatively good, Hess (2010) highlighted some shortfalls that included its inability to carry soil moisture over calendar years due to the fact that simulations are programmed to run for discrete, individual years despite the facility to use daily values of rainfall and ETo. Raeth (2020Raeth ( , 2021 noted that CropWat appears to be using incoming solar radiation instead of net radiation in its calculations, with the result that one would expect higher estimations of irrigation requirements.

Conclusion
Cropwat 8.0 and Aquacrop 5.0 can be used to identify water constraints for the tomato crop in environments where technical capacity and knowledge is limited. Based on crop, soil, meteorological data, CO 2 , groundwater and field management, crop water requirements and irrigation scheduling of tomato in Harare could be estimated effectively using both models. The study revealed that there was a moderate correlation between simulated crop water requirements for Cropwat model and the simulated crop water requirement of the Aquacrop model and hence model performance indicators showed that the models can be used effectively for irrigation scheduling using different calibration parameters leading to the development of an irrigation calendar and this can entail another cycle of calibration and field-testing for possible future work.
Aquacrop model simulated well the study area under different climatic conditions and hence, the model is highly recommended for use, due to its merit of being user friendly, simulates well observed canopy cover,  Source: Authors biomass, water use efficiency, irrigation water, and yield for all areas of water application, addressing the conditions where water is a key limiting factor for crop production. It is thus advisable for farmers and end-users to adopt the Aquacrop model in the development of irrigation guidelines for the irrigation of tomato as it compares attainable and actual yields in a field, and simulates water productivity, crop water requirement and irrigation application depth. The developed irrigation schedule for both Aquacrop and Cropwat should be validated and calibrated in each area of application. Cropwat has been observed to be good for general design, planning and operation of irrigation systems and provides a rapid assessment of crop performance under water-limiting conditions whereas Aquacrop gives the finer details.
Cropwat and Aquacrop have similar functions and can be used to predict water availability and crop response to current and future agro-climatic conditions. However, in this respect, the Aquacrop model is considered more superior in that it can account for the rising atmospheric concentrations of CO 2 as well as increasing surface temperatures whereas the Cropwat model can account only for increasing temperatures. The Aquacrop model unlike the Cropwat model, normalises water productivity for atmospheric evaporative demand and CO 2 concentrations and is relatively insensitive to variation in soil nutrient status, this enables the quantitative assessment of water-limited productivity between different agro-ecological zones, crops and seasons. As an evolution of Cropwat, Aquacrop reproduces the crop environment more accurately through more advanced crop routines including the partitioning of ETc into nonproductive soil evaporation and productive crop transpiration and final yield into biomass and harvest index. Aquacrop allows the user to better define the soil profile by incorporating up to five horizons of variable textural composition and depth within the root zone, whereas the Cropwat model allows the user to specify only one (that is, maximum rooting depth) or two layers (that is, top-soil and subsoil), respectively. Finally, both the Cropwat and Aquacrop are relatively easy to manipulate and have been widely adopted within the global scientific and other user communities (Table 6).