The WFD tool was tested in farmers’ fields to improve farm-level irrigation decision-making and water use efficiency (Schmitter et al., 2016), which, in turn, could improve crop yields and therefore farm income. Farmers using the WFD irrigated depending on the response of the WFD indicators as these popped up when soils reached their field capacity at the depth of installation. The farmers under the FTP group used water based on their knowledge. To evaluate the effects of the irrigation scheduling tool (WFD) on crop and water productivity, farmers who did not have access to the tool were assigned to the FTP treatment. This study followed a complete randomized block design with replications for each crop and treatment. In 2016, 17 farmers (7 farmers under WFD treatment and 10 farmers under FTP) grew cabbage and carrots, whereas in 2017, only 12 farmers decided to continue (3 under WFD and 9 farmers under FTP) growing cabbage and carrot.

Cabbage was transplanted after a nursery period, 3 months earlier than the plantation time. The land size per crop varied between 50 and 100 m². The fertilizer type used was diammonium phosphate (DAP) and urea at the time of planting and in a split application, respectively. The fertilizer amount and time of application are shown in Table 1.

The data collected during the field experiment included planting dates, irrigation, fertilizer application dates and amounts, soil moisture content, crop height, and yields. In addition, detailed field information (area of the field, planting and/or transplanting date, and plant density) was collected. Soil samples were collected and analyzed to determine soil physical and chemical properties such as soil texture, field capacity, available organic matter, pH, total nitrogen, available phosphorus, and electric conductivity.

The soil texture ranged from sandy loam to clay loam (Table 2). The estimated field capacity and wilting point followed the variability in organic matter and textural differences between the fields. Field capacity (FC) ranged between 23 and 36%, whereas permanent wilting point (PWP) ranged between 11 and 23%.

Instruments such as WFD for managing the on-farm water and soil moisture access tubes to monitor the soil moisture content were installed in the experimental plots. The WFD was used to indicate when the wetting front had reached the end of the root zone, so farmers could improve their irrigation decision-making.

In both seasons, access tubes were installed in six farmer fields (three access tubes in each treatment) to measure the soil moisture content up to the depth of 1 m (i.e., 10, 20, 30, 40, 60, and 100 cm) for each crop using the PR2/4 profile probe, manufactured by Delta-T Devices Ltd. A portable/handheld HH2 display device/reader was used to obtain soil moisture readings at regular intervals throughout the production period. The probe is known for its high measurement accuracy and user-friendly design (Teferi et al., 2023). The soil moisture measurements were categorized into two groups: the effective root zone (0–60-cm depth) to estimate the change in soil moisture content within the effective root zone and the deeper soil (60–100 cm) to estimate the deep percolation portion. These measured values are used for the soil water balance equation (see below).

The soil water balance is computed based on the principle of conservation of mass and entails estimating the different components of the soil water balance. The general soil water balance equation is shown in Equation 1. The temporal scale of the SWB was based on the crop growth stage (initial, development, mid, and late season stages).

where Peff = effective rainfall (mm), Irr = irrigation (mm), ETc = crop evapotranspiration during the study period (mm), Q = surface runoff (mm), D = drainage (percolation, mm), Δ S = variation in the soil water storage before and after in each crop stage (mm), Θ1 - Θ2 = volumetric water content from a soil horizon in each stage, in two consecutive crop stage measurements (cm³ cm⁻³), and Dzi = thickness of each depth in each stage (mm).

The different soil water balance components were measured at the experimental plots, including rainfall, irrigation, runoff, deep percolation, and change in soil moisture. After rewriting the equation, the crop evapotranspiration of the two vegetable crops (cabbage and carrot) was estimated using Equation 3.

The crop coefficient (Kc) was estimated as the ratio of crop evapotranspiration (ETc) and reference evapotranspiration (ETo) for each cropping stage and treatment (Equation 6) (Allen et al., 1998). For each crop stage, ETc was calculated using the soil water balance equation (Equation 3). ETo refers to the potential evaporation (Allen et al., 1998). The FAO Penman–Monteith method was used to estimate ETo using daily climate data (sunshine hour, solar radiation, minimum and maximum temperatures, relative humidity, and wind speed).

The main reason for estimating Kc for each treatment was to evaluate whether improved management practices affected the crop coefficient. Finally, the Kc value estimated with the soil water balance method was compared with the FAO regional Kc values. This helps to check whether the FAO Kc regional values are representative of Ethiopian climate conditions.

A digital elevation model (DEM) from the United States Geological Survey (USGS) Earth Explorer with a 30-m spatial resolution was used to set up the APEX model. The climate data for the APEX model (precipitation, temperature, humidity, wind speed, and solar radiation) were taken from the automatic weather station installed on the site.

The APEX model is a field or watershed scale distributed hydrological model (Arnold et al., 1994; Gassman et al., 1998; Williams et al., 1998). In the APEX model, the watershed is divided into multiple sub-watersheds of homogeneous land use, soil, and slope. The model operates on a daily time step, driven by climate data and agricultural management schedules. The model is capable of simulating detailed field conditions, including crop management and growth, as well as the hydrology of various agricultural management practices (Saleh and Gallego, 2007; Wang et al., 2011; Worqlul et al., 2018). In this study, the APEX model was used to estimate the actual and potential evaporation of the practices using the agricultural management practices as input to the model.

The APEX model provides five options to estimate the potential evaporation, namely, Penman–Monteith, Penman, Priestley–Taylor, Hargreaves, and Baier–Robertson. In this study, the Penman–Monteith method was adopted. The model estimates the actual evaporation from the soil and transpiration from the plant separately. Actual evaporation is estimated as a function of potential evaporation and leaf area index, while actual soil water evaporation is estimated as an exponential function of soil depth and water content (Williams et al., 2006).

In general, the crop coefficient is estimated as the ratio of actual evapotranspiration and reference evapotranspiration (Allen et al., 1998). In the study by Allen et al. (1998), the reference evapotranspiration is represented with a reference surface of hypothetical grass of a uniform height of 12 cm. There are some crops such as rice, which can have Kc values greater than one during maturity. However, in the APEX model, the crop coefficient was estimated as a fraction of actual and potential evaporation for both practices and vegetables after calibrating the model for observed cabbage and carrot yield. The performance of the APEX-simulated yield was compared with the observed yield, and the model performance was evaluated using root mean square error (RMSE) and percent bias (PBIAS).

The performance of the APEX-simulated yield was compared with the observed yield, and the model performance was evaluated using root mean square error (RMSE) and percent bias (PBIAS). Both model performance indicators measured the average difference between the APEX model predicted yield and the actual observed vegetable yield. In contrast, the APEX and soil water balance estimated crop coefficient was evaluated through the coefficient of determination, providing a quantitative measure of the goodness of fit for both the APEX and soil water balance estimated crop coefficient.

Crop water demand and crop coefficient depend on the climate variables, mainly on rainfall and temperature. In the future, the presumed climate will have an immense effect on both the crop coefficient and the crop water requirements (CWR). To evaluate the effect of future climate on the CWR and Kc, predicted climate data were collected from the Coordinated Regional Downscaling Experiment (CORDEX) archive (Christensen et al., 2014). The CORDEX archive improves and reframes regional-scale climate projections, which help evaluate the impacts of climate change in the future. For all major inhabited areas of the world, the CORDEX-projected climate data are available at 25-km spatial resolution (Teichmann et al., 2020). For this study, the rainfall and minimum and maximum temperature data from the medium representative pathways 4.5 (RCP4.5) were used for the period 2006–2,100. The future projected climate data were evaluated for three time horizons representing a 30-year time window [2011–2040 (2025), 2041–2070 (2055), and 2071–2,100 (2085)]. The historical CORDEX data matching the observed climate data for the period 2008–2019 were used to evaluate the performance of the RCP4.5. Finally, the projected climate data were used as input to the calibrated APEX model to assess the effect of climate change on the crop coefficient and crop water requirement in the projected time horizons. The APEX model was also used to estimate climate change impact on the actual evapotranspiration and crop coefficient during the future time horizons using the projected climate data as input to the calibrated model (2025, 2055, and 2085).

For both seasons and crops, the FTP treatment showed a higher irrigation water variability compared to the WFD treatments (p < 0.05) (Figure 3). The tool helped farmers to better manage how much water to apply as the tool indicated when the water moved below the root zone, whereas farmers who did not have the tool applied the water to the crops based on their experiences, which might have led to under- or over-irrigation.

In 2016, throughout the growing seasons, the deep percolation of cabbage in the WFD and FTP was 31 mm and 20 mm, while the deep percolation of carrot in the WFD and FTP was 22 mm and 30 mm, respectively. In 2017, the corresponding values for cabbage were 29 and 23 mm, while for carrot, they were 21 and 27 mm, respectively. In 2016, the change in soil moisture for WFD and FTP treatment plots of cabbage accounted for 56 and 28 mm, and for carrot, it accounted for −6 and 11 mm, respectively. In 2017, the change in soil moisture for WFD and FTP treatment plots of cabbage accounted for 102 and 39, and for carrot, it accounted for −2 and −2 mm, respectively (Figure 4).

The actual crop evapotranspiration of cabbage and carrot was estimated for the period 2016 and 2017 using the daily climate data. The daily crop evapotranspiration was aggregated over the initial, development, mid, and late growing seasons (Figure 4). In 2016, the growing period for cabbage and carrot was 119 and 129 days, respectively, whereas, in 2017, the growing period for cabbage and carrot was 118 and 127 days, respectively. The water management practices influenced the duration of different cropping stages.

In 2016, the average amount of ETc for cabbage in the WFD and FTP treatments was 439 and 393 mm, and for carrot, it was 535 and 550 mm, respectively. In 2017, cabbage under the WFD and FTP treatment groups used 448 and 514 mm and carrot under the WFD and FTP treatment groups used 522 and 556 mm, respectively. According to Allen et al. (1998), the crop water requirement of cabbage and carrot ranges from 350 to 500 mm. For both seasons, ETc estimation indicated a similar pattern for both crop and treatment groups.

The 2-year average ETc from both crops showed a different amount of water applied along the different growing stages. The ETc was almost the same at the initial, development, and harvesting stages for both treatments and crops. However, during the middle stage, ETc in the FTP treatment was significantly lower for cabbage and significantly higher for carrot than that in the WFD treatment. This is because in the carrot plots, the soil surface is fully covered and the canopy cover is high compared with cabbage plots.

The Kc of cabbage in the WFD treatments at the initial, middle, and harvest seasons ranged from 0.66 to 0.79, 1.20 to 1.27, and 0.83 to 0.86, and in the FTP treatments, Kc ranged from 0.62 to 0.75, 1.1 to 1.26, and 0.55 to 0.86, respectively. Whereas for carrot, Kc in the WFD treatments at initial, middle, and late stages ranged from 0.69 to 0.71, 1.19 to 1.27, and 0.82 to 0.86, and Kc in the FTP treatments ranged from 0.28 to 1.05, 1.25 to 1.36, and 0.86 to 0.89, respectively (Figure 5). All the Kc values in each crop stage between the two treatments showed no significant difference (p < 0.05).

The FAO crop coefficients fell within the estimated range for some stages, crops, and treatments, especially the Kc at the initial stage for carrot, which was significantly lower in both treatments than in FAO. On the other hand, during the mid-stage of cabbage, the Kc values in the field were significantly higher than in the FAO. This reflects the need for local validation of Kc values based on climate conditions and varietal differences. The Kc values of cabbage and carrot in FAO estimations are not estimated by considering the Ethiopian local condition. FAO-estimated Kc values of cabbage and carrot reported under Allen et al. (1998) are for desert climate conditions of California, United States. Hence, according to the study by Allen et al. (1998), the Kc value of cabbage and carrot in each of their stages was the same, and the values at the initial, mid, and late seasons are 0.7, 1.05, and 0.95, respectively.

The APEX model simulated yield captured the observed cabbage and carrot yield with an average root mean square error of 1.04 t/ha and 3.3 t/ha and PBIAS of 2.0 and 5.5%, respectively. The daily crop coefficient estimated as a ratio of the simulated actual and potential evaporation of the respective vegetables is plotted with the FAO-estimated regional values for both crops and practices (Figure 6). The APEX-simulated crop coefficient has captured the FAO crop coefficient pattern very well, with R-square values ranging between 0.73 and 0.90 for cabbage and between 0.77 and 0.98 for carrots (Appendix A). The APEX-simulated crop coefficient aggregated over the growing period compared with the soil water balance estimated crop coefficient captured the pattern very well with R-square ranging between 0.5 and 0.75 for cabbage and 0.66 and 0.96 for carrot. The crop coefficient estimated with the APEX model indicated a maximum value of one, and it is observed when there is enough soil moisture after irrigation or rainfall to meet the potential evaporation (i.e., when actual evaporation equals potential evaporation of the respective vegetables). However, the Kc reported under FAO is a fraction of a reference crop’s potential evaporation (i.e., potential evaporation of a hypothetical grass reference crop with an assumed crop height of 0.12 m, a fixed surface resistance of 70 s/m, and an albedo of 0.23) and crop’s actual evapotranspiration. The APEX-simulated crop coefficient was sensitive for irrigation applications and indicated a larger difference, especially at the initial stage of crop growth due to differences in canopy cover. The simulated crop coefficient for carrot during the late stage indicated a significant difference from the FAO generic crop coefficient curve (Appendix A). This is due to senescence processes, or deterioration of carrot leaves with age, where the leaf area decreases significantly after mid-maturity, leading to a divergence from a typical FAO crop coefficient curve.

The projected climate for the three time horizons was compared with the observed temperature (baseline period; 2008 to 2019) (Figures 7A,B). In all the future time horizons, the projected minimum temperature showed an increasing trend during the dry season and the main rainfall season (June to October) compared to the baseline period. The average annual minimum temperature is expected to increase by 23, 31, and 34% from the baseline in 2025, 2055, and 2085, respectively. The maximum temperature will increase during the dry season and decrease during the rainy season (Figure 7B). The climate data analyzed during the cabbage and carrot growing period indicated that maximum and minimum temperatures are expected to increase during the future time horizon. The minimum temperature is expected to increase by 1.5% in 2025, 6.5% in 2055s, and 7.8% in 2085. The increasing minimum temperature coupled with the increasing maximum temperature during the dry season will significantly affect the evaporation and crop water requirement.

The average annual projected rainfall is expected to increase by 9.1, 0.7, and 1.2% from the baseline period in 2025, 2055, and 2085, respectively (Figure 8). The monthly average rainfall is expected to increase in January, March, June, and November for the future time horizons and is expected to decrease in February, April, July, October, and December from the baseline period. Overall, the dry season rainfall (November to May) is expected to decrease across all time horizons. The projected rainfall of the dry season is expected to decrease by 20, 1.2, and 3.8% compared to the baseline period in 2025, 2055, and 2085, respectively.

Temperature increase together with the rainfall reduction in the future time horizons will likely increase the crop water demand and therefore the need for supplementary irrigation during the dry season.

The effect of climate change on evapotranspiration and crop coefficient was evaluated using the calibrated APEX model for the baseline period. The result indicated that under climate change scenarios, the growing seasons’ potential evapotranspiration is expected to increase by 2.5, 5.1, and 6.0% in 2025, 2055, and 2085 from the baseline period, respectively (Figure 8). As the potential evaporation increases during the future time horizon, the water demand in each of the cropping stages (at initial, development, mid, and late seasons) will increase.

The crop coefficient for cabbage and carrot estimated as a fraction of simulated potential and actual evaporation is shown in Figures 9, 10, respectively. The project time horizon’s simulated crop coefficient indicated a similar trend with the simulated baseline crop coefficient with R-square ranging between 0.51 and 0.64 for cabbage and between 0.51 and 0.77 for carrot. The simulated crop coefficient indicated a higher variation in 2085 with 19% for cabbage and 24% for carrot, while the 2025 period simulated Kc indicated the least coefficient of variation (16 and 21% for cabbage and carrot, respectively).

The projected climate parameters (minimum and maximum temperatures and rainfall) will have an immense impact on the future ETc and Kc (Figure 11). Hence, wise utilization and conservation of water resources will become increasingly important to avoid the risk of water scarcity for food production.