Modeling the impacts of climate-smart practices on soil–water interaction and wheat yield under climate change in central Ethiopia

Abstract

This study aimed to model the impact of climate-smart agricultural (CSA) practices on soil‒water balance, water use efficiency (WUE_ET), and wheat yield in the face of climate change. The AquaCrop version 7.1 model was used to estimate the water balance and yield under the baseline (1981–2010) and future (2050s, RCP4.5) climate scenarios. We evaluated five CSA practices, varying in tillage, residue management, and water management, based on experiments conducted in 2020 and 2021. Observed data on wheat (Triticum aestivum L.) grain yield and surface runoff were used for model calibration (2020) and evaluation (2021). The model was evaluated using four performance indicators and found to be robust. The treatments included farmers’ conventional practices (CPs), soil bunds (SBs), crop residues (CRs), integrated conservation practices (ICPs), and berken plows (BPs). The results show that climate change is likely to reduce grain yield and WUE_ET under CP by 1% and 16.3%, respectively, by 2050 compared to the current 2021 period. All CSA practices studied increased grain yield and WUE_ET over the CP in both periods. Under future climates, ICP showed a greater relative grain yield (Y = 4.51 t/ha), water use efficiency (WUE_ET = 1.32 kg m³), and other soil water balances, followed by CR, BP, and SB over CP. Overall, ICP has shown tremendous potential for climate change adaptation among the other CSA practices tested. Therefore, adaptation to future climate conditions must integrate different practices, and the novel ICP can be promoted as a climate-smart practice in similar farming systems and agro-ecological settings.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-39954-x.

Introduction

In Sub-Saharan African (SSA) nations, approximately 417 million rural inhabitants depend on agriculture, which is predominantly small-scale, rain-fed, and highly sensitive to climatic variability¹. Contemporary notions and emerging concerns about agriculture have focused on increasing agricultural production sustainably under changing climate conditions without harming the environment. This is because agriculture may not satisfy the food demand of ever-growing populations², given the stagnation and/or declining productivity related to severe land degradation^3,4, including soil fertility depletion, green water losses through evaporation^1,5, blue water losses through runoff ^1,6, and climate change impacts^7,8. This implies that efforts toward sustainable agricultural intensification are paramount for enhancing crop production per unit area to meet food demand and ensure food security without further jeopardizing ecosystem stability.

Climate variability, exceptionally prolonged dry spells, has led to recurrent yield reductions and sometimes total crop failures, resulting in food insecurity in East African countries such as Ethiopia⁹. This problem is exacerbated in areas where 70–85% of the rainfall is nonproductive and is lost through evaporation, runoff, and deep percolation from crop systems¹. Consequently, only 15–30% of rainfall is productive and lost through transpiration, which is considered green water flow. Green water flow is primarily determined by changes in water storage, as the difference between total input and output is offset by changes in storage. The total amount of water received (rainfall) in cultivated fields is partitioned into surface runoff (Qs), deep percolation (R), transpiration (T), soil evaporation (E), and soil moisture storage variation (ΔS). These partitions can be categorized as blue or green water ^1,10. Blue water is the portion lost from the soil surface as Qs and R and eventually joins lakes, rivers, seas, and dams¹¹. Green water is the water stored in the soil and the portion that returns to the atmosphere via vapor fluxes of transpiration (T) and soil evaporation (E)¹¹. These vapor fluxes could have occurred from the plant body, wetted soil surface, and water bodies. Transpiration (T) is considered productive green water, whereas soil evaporation (E) is regarded as non-productive green water^12,13. This distinction is because yield and biomass production are functions of transpiration^10,14. This was because the amount of water transpired by crops is used as a measure of water-use efficiency in a given system.

A study conducted in the drought-prone area of northern Ethiopia showed that the impact of dry spells reduced yields by 42–73% in non-conserved fields15. Furthermore, conventional practices, which involve repeated tillage at shallow depths (13–16 cm) and crop residue removal, exacerbate crop failure and drought vulnerability^6,16,17. As a result, the yield gap between farmers’ yields and achievable yields for major grains is estimated to be up to 70% partition. Narrowing the yield gap while non-productive water loss through evaporation, surface runoff, and deep percolation in the soil‒water balance partition remains a major challenge^18,19. Therefore, closing the yield gap while increasing and ensuring water productivity using different ‘climate-smart’ practices in the face of climate variability remains a concern.

However, a lack of sufficient information and knowledge on the benefits of sole and/or combined application of existing CSA practices at different spatial and temporal scales has been the central constraint for climate change adaptation, sustainable intensification, and adoption of sustainable CSA practices in the agricultural landscapes of Ethiopia^20,21. There is inadequate research on CSA practices such as soil bunds, mulching, conservation tillage, and their integration to optimize water use efficiency and wheat yield productivity in the changing climate without harming the environment. Additionally, adequate research has not been conducted to evaluate the overall performance of CSA practices on hydrological functions under a range of climatic conditions and different agroecological settings²¹. According to Adgo et al.²²and Stillhardt et al.²³, the socioeconomic and ecological benefits of some CSA practices are gradual and long-term. This conclusion is supported by Adimassu et al.²⁴and Biratu et al.²⁰, who reported that the implementation of physical soil and water conservation structures may improve multiple ecosystem services and functions but may not improve grain yield in the first 1 to 3 years. This is mainly because a portion of the cultivated area is occupied by physical structures, thereby reducing the crop area. In addition to area loss factors related to time for impact, inadequate technical quality and implementation design also affect the adoption of physical structures by resource-poor farmers in most parts of the country 24,25. Despite the availability of individual best practices, there is a lack of integrated research that systematically identifies, evaluates, and combines multipurpose agricultural practices to address existing production challenges and future changing climates. In particular, empirical evidence remains limited on how the strategic integration of such practices can simultaneously enhance crop yields while delivering short-term cobenefits. Therefore, it is indispensable to evaluate and identify the most effective CSA practices that increase the water-use efficiency of crops and complementary services in the short term. This study aimed to (1) evaluate the impacts of CSA practices on wheat yield and biomass, (2) assess the impacts of CSA practices on water use efficiency, and (3) simulate yields, water use efficiency, and soil water balance under different CSA practices for the current (2020–2021) and future (2050) climates of Kulumssa. In doing so, this study investigated effective ‘climate-smart’ practices for conserving and harnessing rainwater, thereby improving agricultural production and other environmental benefits.

Materials and methods

Study site

The study was conducted at the Kulumssa Agricultural Research Center (KARC) in the Central Rift Valley Basin (CRVB), Oromia National Regional State, approximately 170 km south of Addis Ababa (Fig. 1).

Fig. 1.

Map of the study area of the KARC in the Central Rift Valley Basin (CRVB) in Ethiopia.

The area experiences a unimodal rainfall pattern that extends from March to September. The KARC farmland has a gently undulating topography with a slope gradient of 0–30%. The mean annual maximum temperature is 23 °C, and the mean annual rainfall is 788 mm. The total rainfall during the experimental growing season was approximately 862 mm. The dominant soil type in the KARC is Vertic Luvisols. The soil is moderately drained, and the texture is clay with a strong medium angular blocky structure.

Method

Treatments

Following previous meta-data synthesis and review studies on LM practices in Ethiopia²⁰, five distinct treatments were selected for experiment-based field evaluation and simulation studies. These five treatments (i.e., SB, CR, ICP, BP, and CP) were arranged in a Randomized Complete Block Design (RCBD) replicated three times for the 2020 and 2021 main cropping seasons. Descriptions of these CSA treatments are presented below.

Conventional practice (CP) Plots were cross-plowed via an ox-driven Maresha plow (Fig. 2a). This conventional tillage using the Maresha plow results in V-shaped furrows with unplowed strips of land in between, requiring multiple rounds of plowing and cross-plowing^26,27. This conventional tillage method required repeated cross-plowing, which was performed up to five times. We used conventional tillage practices as the control treatment.

Fig. 2.

Tillage practices, soil bunds, and crop-residue applications studied. where (a) local Maresha plow, (b) Crop residue-treated plot, (c) Berken plow used for the first till, (d) Wing-Berken plow used at planting time to make wide furrows, and (e) Soil bund constructed across the slope.

Crop residue (CR) Plots were treated by vertical mulching with 2 t ha^{− 1} of wheat straw (Fig. 2b) (approximately equivalent to 66% of the biomass left on the soil surface)²⁸. In doing so, the upper soil surface was covered with crop residue (Fig. 2b). This operation was performed 40 days before sowing. The CR shields the soil from intense sunlight and raindrop impact, helping to reduce soil crusting, erosion, and runoff and restore and maintain soil fertility²⁹.

Berken plow (BP) Plots contour plowed (i.e., perpendicular to the slope of the plots) via ox-driven Berken (Fig. 2c, d). The soil was tilled twice before sowing: once after the first rain and once after the onset of rain. A Berken plow (Fig. 2c) was used for the first tillage operation to ripen the soil to a depth of 28 cm along the contour lines, resulting in a deep center cut³⁰. This deep central furrow, approximately 28 cm, was created by the Berken plow, which is significantly deeper than the 10–15 cm furrow produced by the standard conventional oxen plow used in Ethiopia²⁷. The second winged plow (Fig. 2d) was used to till the plot before sowing. The last tillage was performed on the sowing date. The Berken plow reduces the frequency of plowing and omits cross-plowing, thereby helping retain water, carbon, and other essential nutrients in the soil^27,31. The Berken plow creates U-shaped furrows and a deep central cut that promotes water infiltration and root development, leading to reduced runoff, decreased erosion, and reduced loss of essential nutrients such as organic carbon^27,31.

Soil bund (SB) Plots were treated with a soil bund structure (Fig. 2e). To construct the SB, the soils were excavated and placed downhill to form soil embankments across the slope, with a basin at its upper side, and the layout was based on a 1.2 m vertical interval between two bunds³². The SB slows down or halts the movement of overland flow or surface runoff, effectively minimizing soil erosion³³. Therefore, SB was constructed as barriers to capture and hold all runoff within the space between two bunds and accumulate eroded soil in the basin located behind the lower bund²⁹.

Runoff plot establishment and field management

A runoff plot experiment was established at the KARC during the 2020 (for calibration) and 2021 (evaluation) main rainy seasons at an 8–11% slope range. The standard hydrologically isolated runoff plots were established with 35 m long and 6 m wide plot sizes and arranged across the slope (Fig. 3a). The spacing between plots was 0.5 m. Galvanized iron sheet metal fencing surrounded three sides of each plot (upper, left, and right) to create hydrologically isolated conditions. Iron sheets were inserted 30 cm into the ground, leaving 30 cm above the surface to prevent lateral flow from adjacent land (Fig. 3b).

Fig. 3.

A full schematic representation (top) of the actual experimental layout of the five treatments embodied by (a) a Google Earth image taken in 2020 and (b) a photograph taken 100 days after planting; (c) trough with 60 slot flow divisors. where SB soil bund; CR crop residue; CP conventional practice; ICP integrated conservation practice; BP berken plowing.

Fifteen rectangular boxes/troughs, each made from a 4 mm-thick metal sheet, were used to collect runoff. Each trough had sixty slot dividers made from square metal pipes (3 cm × 3 cm). The slot divisors were assembled 20 cm from the bottom and arranged 7 cm apart horizontally. The back side of the trough was bent approximately 10 cm below the soil at the bottom edge of the plot, extending over the trough lip to prevent runoff from seeping beneath the lip. Barrels were installed at the lower tips of the plots to collect overflow. One of the sixty trough slots was connected via a pipe to the next barrel, which had a capacity of 110 L. Barrels were used to sample the overflow volume from the troughs. Fifty-nine slots of the trough were used to dispose of the excess runoff when full overflow occurred (Fig. 3c). Calibration was performed before and during the first measurement to check that each slot had an equal chance of disposing of runoff from the plots.

All crop management techniques were implemented in accordance with regional recommendations. Wheat (Triticum aestivum L.) was used as the test crop, and the recommended seed rate was 130 kg ha^{− 1}. Locally recommended nitrogen and phosphorus were supplied at a rate of 138 kg N per hectare and 69 kg P₂O₅ per hectare, respectively, equally for all treatments. Split application of N fertilizer was supplied half at planting, while the other half was supplied at the tillering stage of wheat. The fertilizer sources of N and P were urea and DAP, respectively³⁴.

Data collection

Measurement of rainfall and runoff

The KARC meteorological station utilized an automatic rain gauge to collect daily rainfall data. The volume of runoff was measured under five distinct treatments during the two main seasons (June–September) in 2020 and 2021. Every morning at 9:00 AM, the water levels in all boxes, troughs, tanks, and barrels were measured. The amount of water collected in the barrels was adjusted to account for direct and trough-collected rainfall and sediment volume. This method, adopted from McHugh et al.¹⁵, was used to adjust and calculate the runoff volume, taking into account the sediment volume and water content.

Soil physicochemical properties

Our study investigated various soil physical properties to gain insights into soil health and to inform model inputs. Soil samples were taken at the beginning and end of the experiment, postharvest, for physico-chemical property analysis. At the end of the experiment, five soil samples were collected within a 2 m radius of the central point to make composite samples from each plot at 0–20 and 20–40 cm soil depths based on a crisscross sampling procedure using an auger. Additionally, we collected samples from the upper and lower sides of the bund where the soil bund was treated.

The moisture content at the field capacity (FC) and wilting point (WP) at depths of 0–20 cm and 20–40 cm was determined. Standard procedures were followed, including the use of a pressure plate apparatus to determine the FC (at a suction of 0.33 bars) and WP (at a suction of 15 bars). The difference between FC and WP allowed us to calculate the plant-available water capacity (PAWC).

Soil bulk density (g/cm³) was determined based on undisturbed soil samples collected using a sharp-edged steel cylinder at 0–20 cm and 20–40 cm depths. Sets of undisturbed soil samples were taken from two pits at each sampling depth for bulk density determination. The pits were dug to a depth of 60 cm, and soil samples were taken from the 0–20 and 20–40 cm soil depths in each pit using a hand trowel. Similarly, core samples from the pits on the upper and lower sides of the bund were also considered. The bulk density (g cm⁻³) was determined after drying the core samples in an oven at 105 °C for 24 h³⁵. The bulk density of each sample was calculated by dividing the total oven-dry mass of the sample by the volume of the soil core used in the sample collection. The bulk density was corrected for coarse fragment content after sieving the crushed soil samples through a 2 mm sieve and weighing the soil fractions < 2 mm (mineral soil fraction) and > 2 mm (rock and ravel) coarse fragments. The proportion of coarse fragments was calculated by dividing the coarse fragment weight (g) by the total weight of the sample (g) multiplied by 100 (corrected ρb is equal to ρb, the proportion of coarse fragments). Additionally, we estimated the soil organic carbon pools at depths of 0–20 cm and 20–40 cm. The soil parameters used in this study included saturated hydraulic conductivity (Ksat), which was estimated using SPAW software based on measured soil properties.

Soil moisture was measured using automatic soil moisture sensors (ML3–HH2 Soil Moisture Sensor) and gravimetrically at depths of 0–20 cm and 20–40 cm. The volumetric soil water content (g/cm³) was calculated by converting the gravimetric water content to a volumetric basis. This conversion involved multiplying the gravimetric soil water content by the soil bulk density (ρb) to obtain the specific gravity of the soil water (γ). Soil moisture was also measured across different plant growth stages.

Crop data collection and analysis

The number of days after planting at emergence, maximum canopy cover, onset of senescence, and physiological maturity were recorded and then cross-checked against a reference book of wheat varieties released by the EIAR in 2016. The leaf area index (LAI) was calculated based on the maximum width and length of leaves directly measured for five randomly selected plants using a pocket meter in an ‘X’ pattern at the emergence and maturity stages from each treatment plot. Then, the Ritchie model was used to calculate the canopy ground cover (CC), which was calculated using the Ritchie model based on the LAI and the following equation³⁶:

1

where CC denotes canopy cover, LAI denotes the leaf area index, and K denotes the extinction coefficient (assigned as 0.65 for wheat).

Agronomic data on the grain yield and aboveground biomass of wheat were collected from a 2 m × 2 m sample area in each plot, with three replicates. The combined average values were used for analysis. The grain yield was measured as the weight of the harvested grain and adjusted to 12.5% of the moisture content of the wheat grain.

Climate data collection and analysis

Climate datasets, including observed daily rainfall, maximum temperature (Tmax), minimum temperature (Tmin), wind speed, solar radiation, and relative humidity, for the baseline and observed periods (1981–2021) were obtained from the KARC. After thoroughly checking for missing and outlier values, homogeneity testing of the climate datasets (ETo) was performed using the Penman‒Monteith method with the AquaCrop model.

Projected changes in rainfall and temperature for the study were analyzed using eight combinations of four global climate models (GCMs) (CNRM-CM5, EC-EARTH, HadGEM2-ES, and MPI-ESM-LR) and two regional climate models (RCMs) (RCA4 and CLMcom). The data were obtained from the Coordinated Regional Downscaling Experiment (CORDEX) for Africa dataset (http://www.cordex.org).

Using the delta change method, GCM-RCM outputs were downscaled to a station level, and daily values were generated for the study area³⁷. Daily rainfall and temperature for the 2050 climate were derived from the observed 2021 baseline by multiplying daily rainfall by the rainfall change factor (based on Eq. 2) and adding the temperature change factor to daily temperature values (based on Eq. 3) under the RCP 4.5 climate scenario. Only a single RCP4.5 emission scenario was considered for analysis, as the selected RCP showed limited divergence from other RCP scenarios for the period 2050 in the study area. The reason that we use the RCP4.5 scenario is up to 2050; the differences among these scenarios are relatively small for the following two reasons. First, emission pathways only begin to diverge strongly after mid-century. Second, climate system inertia (lagged response of oceans and atmosphere) means that near-term projections are similar across scenarios. Thus, in the study area, the temperature and precipitation projections under RCP4.5 were not substantially different from those under RCP2.6 or RCP8.5 for 2050. Therefore, projected changes in daily rainfall and mean daily temperature from the eight GCM-RCM pairings and single RCP (RCP4.5) for the year 2050 were generated and used for analysis in the study.

2

3

where P_Δ and T_Δ are the delta change perturbed daily climate change variables, P_obs and T_obs are the observed climate variables in the reference period, Δ_P and Δ_T are the changes in climate as simulated by the RCMs, and are the daily climate means on an annual basis, and the indices ‘ref’ and ‘fut’ indicate the reference and future periods, respectively.

Model description

A standard AquaCrop model with a graphical user interface (GUI) and database version 7.0 was used in this study. AquaCrop has several important modules that are used to simulate transpiration, evaporation, surface runoff, deep percolation, grain yield, biomass production, yield gaps, soil‒water balance, and water productivity³⁸.

Model inputs

The parameters required by the AquaCrop model were obtained from field experiments and other primary and secondary data. AquaCrop uses both conservative and nonconservative parameters. While conservative parameters are inherent to the model, the relevant nonconservative parameters used in this study are presented in the supplementary materials, along with their determination methods and sources.

Model calibration and evaluation

In this study, the model is primarily designed to capture the effects of water availability, soil characteristics, and climate variables on crop performance related to CSA practices, but it assumes that nutrient supply is nonlimiting. Thus, yield reductions or growth constraints arising from deficiencies in nitrogen, phosphorus, or other essential nutrients and constraints related to diseases, weeds, and insect pests are outside the scope of AquaCrop’s simulation. Inorganic fertilizer was applied at the recommended rate for the experimental site, and mechanical weeding and chemical (i.e., herbicides, fungicides, and insecticides) application measures were taken to control diseases, weeds, and insect pests. Additionally, the modeling framework did not incorporate certain aspects of heat stress, salinity stress, or capillary rise. Capillary rise was determined based on experimental observations in the field. To validate and calibrate the model, we used grain-yield and surface-runoff data from field experiments conducted during the 2020 and 2021 cropping seasons. The first year’s data were used to calibrate the model, while the second year’s data were used to evaluate the model. For yield, the calibrated coefficients are maximum canopy cover (CCx), rooting depth, and the length of building up HI values to be 85–96% and 46 days, respectively. Similarly, the curve number (CN) was sensitive to runoff generated by the model and was calibrated to estimate runoff.

Subsequently, the simulation’s performance was assessed using various indicators with selected observed parameter values. Following calibration and validation, model simulations were conducted based on current (2021) and future (2050s) climate data to quantify the climate change impacts on the different water balance components, yield, and WUE of wheat under different management scenarios.

The model outputs were evaluated against the observed parameters using selected performance indicators. Accordingly, the Nash–Sutcliffe efficiency (NSE), percentage bias (Pbias), coefficient of determination (R²), and root mean square error (RMSE) were used to evaluate the model outputs.

The RMSE quantifies the deviation between the observed and simulated results across all treatments. It measures model uncertainty, which indicates the range over which the model overestimates or underestimates the observation, and is calculated based on the following equation³⁹:

4

where S_i and O_i denote the simulated and measured values, respectively, and n is the number of observations.

The R² indicates the amount of variance explained by the model compared to the observed data and was calculated using the following equation⁴⁰:

5

where S_i and O_i are the simulated and measured values, respectively; S ®S is the mean of S; and O is the mean of O.

The PBIAS values of larger and smaller simulated and observed values were calculated using the following equation⁴¹:

6

where Oi and Si are the measured and simulated data.

The NSE determines the relative change in residual variance between the observed and simulated values and is calculated using the following equation⁴²:

7

where Oi and Si are the measured and simulated data, respectively, and Oi and Si are the means for both the measured and simulated numbers of observations.

Water use efficiency (WUE) determination

Water use efficiency (WUE) was computed based on the ratio of grain yield to total rainfall evapotranspiration and transpiration using the following formula¹⁹.

8

9

10

where WUE_P, WUE_ET, and WUE_T are the water productivity in kg m^− 3 for total rainfall, ET (mm) is the evaporation of water from the soil surface plus transpiration from the crop, Y is the grain yield in kg m² per season, T is transpiration (mm) per season, and P is the rainfall (mm) per season.

Climate change adaptation analysis

The analysis of the adaptation potential of CSA practices was conducted for both present and future climate scenarios. Therefore, following⁴³, the impacts of CSA practices were calculated for the current climate and for adaptation purposes as follows.

11

12

where A is the yield of a CSA practice under the baseline climate, B is the yield of the conventional practice under the baseline climate, C is the yield of the CSA practice under the future climate, and D is the yield of the conventional practice under the future climate.

The framework was applied to simulate both actual and potential yields under CSA practices for future periods. AquaCrop was used to simulate yield gaps between actual and potential yields in the 2050s across the different CSA practices tested.

Statistical analysis

The normality of the residuals and homoscedasticity of the data were assessed before analysis. Analysis of variance (ANOVA) was used to analyze the effects of treatments on the observed variables using the general linear model. It was used to analyze wheat biomass, grain yield, soil water partitions, and water use efficiency in the studied CSA. HSD multiple comparison tests were used to separate means when there were statistically significant differences (p < 0.05) among treatments. The relative mean difference (%) for the response variables was calculated based on the following formula:

13

Results

Model robustness

The simulated grain yield and surface runoff closely matched the observed values. Table 1 presents the model robustness indicators, namely, PBIAS, NSE, RMSE, and R². Overall, the model fit was strong for all treatments except CP. Specifically, R² values for both observed and simulated yield and runoff were consistently high, exceeding 0.95 and 0.89, respectively, across all treatments.

Table 1.

AquaCrop model simulation performance evaluation based on yield and runoff.

Variables	CSA practices	Performance indicators
		RMSE	R2	PBIAS	NSE
Yield	SB	0.009	0.977	0.102	0.962
	CR	0.025	0.971	0.238	0.958
	CP	0.037	0.958	-0.409	0.906
	ICP	0.027	0.947	0.259	0.905
	BP	0.003	0.964	0.027	0.961
Runoff	SB	0.977	0.969	-0.556	0.919
	CR	1.139	0.969	0.290	0.989
	CP	6.312	0.969	1.549	0.810
	ICP	0.987	0.894	-0.587	0.850
	BP	1.991	0.998	-0.741	0.985

SB soil bund, CR crop residue, ICP integrated conservation practice, BP berken plow and CP conventional practices. RMSE root mean square error, PBIAS percentage of bias, R²  coefficient of determination, and E Nash-Sutcliffe efficiency.

Projected rainfall and temperature

Future climate change conditions were projected using a single representative pathway, the RCP4.5 scenario. Accordingly, rainfall in the study area is projected to vary by 4.2–16% in 2050 relative to the baseline period of 1986–2010. The temperature is expected to increase by 0.7–1.25 °C, with an average increase of 0.98 °C. Additionally, ET is projected to increase by 12.5–18.4% across all CSA practices in 2050 relative to 2021. Conversely, in 2050, ET is projected to decrease across all CSA practices by 0.63% to 14.7% relative to the control period.

Effect of CSA practices on soil water balance

The partitions of rainfall into various components, including transpiration (T), evaporation (E), surface runoff (Qs), soil water storage (S), and deep percolation (R), across different CSA practices under both current (2020–2021) and future (2050) climate conditions. Based on observed and projected planting dates, rainfall amounts decreased in the sequence 2020 > 2050 > 2021. Conversely, ETo increased in the sequence: 2050 > 2021 > 2020. The study findings highlighted significant variations in how rainfall is partitioned into different components of the soil‒water balance across the tested CSA practices under current and future climate scenarios. All components of the water balance showed statistically significant differences (p < 0.05) among the various CSA practices under both current and projected climate conditions.

Transpiration (T) is a crucial component of the soil water balance. The maximum transpiration (T = 254.3 mm) was recorded in 2021, whereas under current and future climates, the highest transpiration (T = 279.1 mm) occurred with the ICP practice. In contrast, the minimum transpiration (T = 220.1 mm) was observed under the SB practice in the current 2020 scenario, and T = 257.4 mm was observed under the CP practice in the future climate of 2050 under the RCP4.5 scenario. Across current and future climates, transpiration (T) followed the order: ICP > CR > BP > SB > CP. Additionally, ICP, CR, and BP exhibited the highest transpiration-to-rainfall ratios, ranging from 38 to 47%, while CP and SB showed relatively lower ratios, ranging from 35 to 45%. The increased transpiration in CR, ICP, and BP practices underscores their effectiveness in enhancing water availability in the soil root zone, making them favorable strategies for crop production.

Apart from transpiration, the portion of rainfall that returned to the atmosphere through evaporation (E) is presented in the supplementary materials. Maximum soil evaporation (E = 97.4 mm and E = 137.6 mm) was observed under CP in 2021 and the future climate, respectively. Conversely, minimum evaporation (E = 44.5 mm) occurred under ICP in the current 2021 climate and (E = 62.3 mm) under CR in the future climate. The CP, SB, and BP practices showed the highest evaporation-to-precipitation ratios.

In contrast, CR and ICP exhibited the lowest evaporation-to-rainfall ratios due to the application of crop-residue organic material in the plots. Surface runoff (Qs) represents a significant loss of rainwater. The maximum surface runoff (Qs = 185.8 mm) was observed under CR in the current climate of 2020 and (Qs = 134.9 mm) under CP in the future climate. Conversely, the minimum surface runoff (Qs = 49.6 mm and Qs = 57.5 mm) occurred under ICP in the current and future climates, respectively. The maximum proportion of surface runoff relative to total rainfall ranged from 23 to 29% under CP, whereas the minimum was 8% to 13% under ICP. Surface runoff-to-precipitation ratios were lowest under the SB, ICP, and BP practices and highest under the CP and CR practices. Under current and future climates, the least surface runoff occurred in the order of ICP < SB < BP < CR < CP.

The portions of rainwater were stored in the soil root zone and percolated for ground discharge. However, AquaCrop outputs for soil water balance components do not separately report soil water storage; instead, they provide data on water infiltration into the soil and deep percolation. Therefore, we computed soil water storage by subtracting water lost as surface runoff and deep percolation from total rainfall. This portion of water includes soil-stored water used by crops, transpired water, and water lost through soil evaporation. The deep percolation simulated was higher under ICP and SB in both the current and projected climates. The maximum deep percolation (R = 358.9 mm) occurred in the current climate of 2020 and (R = 269.1 mm) in the projected climate under ICP. In contrast, the minimum deep percolation (R = 133 mm) occurred in the current climate of 2021 and (R = 134.3 mm) in the future climate under CP. Proportionally, deep percolation ranged from 45 to 57% under ICP at its maximum and from 23 to 35% under CP at its minimum. Overall, the simulation results indicated that deep percolation increased in the order of ICP > CR > BP > CP under the current climate. However, under future climate conditions, greater deep percolation was observed in the order ICP > BP> CR > CP.

Impact of CSA practices on the LAI, CC, grain yield, and biomass of wheat

Figure 4 shows that the LAI and CC differed across the tested treatments. On average, LAI and CC differed significantly (P < 0.05) among treatments and CP.

Fig. 4.

Leaf area index (LAI) (m2 m2) and canopy cover (CC) (%) under different CSA practices measured in 2020–2021, where SB soil bund, CR crop residue, ICP integrated conservation practice, BP berken plow; and CP conventional practices.

The average LAI increased by 30.6, 30.0, and 26.8% under CR, ICP, and BP, respectively, compared to that under CP. Similarly, CC varied significantly among CSA practices, with increases averaging 2.9%, 6.9%, 6.9%, and 6.3% under SB, CR, ICP, and BP, respectively, compared to CP. Wheat grain yield and biomass were simulated for the current (2020 and 2021) and projected (2050) climate scenarios. Figure 5 illustrates the impact of CSA practices on wheat grain yield and biomass under observed (2020–2021) and projected (2050) climates.

Fig. 5.

Wheat grain yield and biomass under different land management (CSA) practices in the current and future climates. where SB soil bund, CR crop residue, ICP integrated conservation practice, BP berken plow, and CP conventional practices. Note Treatments differ significantly (p < 0.05).

Simulated wheat grain yield and biomass differed significantly (p < 0.05) across the tested CSA practices in both current and projected climates. SB, CR, ICP, and BP significantly (p < 0.05) increased wheat grain yield by an average of 4.2%, 16.4%, 13.0%, and 15.6%, respectively, compared to the control (CP) in the projected future (2050) climate. Similarly, wheat biomass increased by 0.2%, 6.7%, 7.6%, and 6.2% under SB, ICP, CR, and BP, respectively, compared to CP in the projected future (2050) climate. The increase in biomass was not statistically significant compared to the increase in grain yield under all CSA practices. Compared with 2021, grain yield decreased slightly under CP in the projected 2050 climate. However, compared to the average yield of the observed years (2020–2021), higher yields were simulated under all CSA practices in the projected 2050 climate. Across all CSA practices, grain yield increased in the order 2050 > 2021 > 2020 relative to CP.

Effect of CSA practices on water use efficiency (WUE)

The water use efficiency for total rainfall (WUE_P), transpiration (WUE_T), and evapotranspiration (WUE_ET) under various treatments in both current (2020–2021) and projected (2050) climate conditions is depicted in Fig. 6. Significant differences in simulated WUE_ET among the tested treatments were observed in both the current (2020 and 2021) and projected 2050 climates. Figure 6 illustrates trends in water-use efficiency for evapotranspiration (WUE_ET) under observed and projected climates. In 2020, WUE_ET increased by an average of 14.1% under both ICP and CR but decreased by 1.3% and 11.3% under BP and SB, respectively, compared to CP. However, in 2021, WUE_ET showed an average increase of 4.4%, 28.9%, 23.7%, and 12.3% under SB, CR, ICP, and BP, respectively, compared to CP. Similarly, in 2050, WUE_ET increased by 6.0%, 36.8%, 31.6%, and 14.7% under SB, CR, ICP, and BP, respectively, compared with CP.

Fig. 6.

Water use efficiency based on (a) total rainfall (WUE_P), (b) transpiration (WUE_T), and (c) evapotranspiration (WUE_ET) under different CSA practices. where SB soil bund, CR crop residue, ICP integrated conservation practice, BP berken plow; and CP conventional practices, Obs observation, Sim simulation, and Proj projected period. Note Treatments differ significantly (p < 0.05).

The highest WUE for transpiration (WUE_T) occurred in the order of 2020 > 2021 > 2050 across all treatments. However, on average, the highest WUE_T was observed in the following descending order: ICP > CR > BP > SB > CP during the observed years (2020–2021). Consequently, WUE_T increased by 11.3%, 10.1%, 9.7%, and 6.8% under ICP, CR, BP, and SB, respectively, compared with CP in the observed years (2020–2021). Under projected 2050 climate conditions, the highest WUET occurred in the order ICP > CR > SB > BP > CP. Accordingly, under ICP, CR, SB, and BP, WUE_T increased by 11.9%, 11.4%, 9.9%, and 8%, respectively, compared to CP in the projected 2050 climate. This increase is attributed to the use of crop residue and physical structures that reduce significant water loss from evaporation and surface runoff. Therefore, under both observed and projected climate conditions, practices that mitigate evaporation and runoff are highly effective in increasing transpiration and WUE_T.

The trend in water use efficiency for total rainfall (WUE_P) appeared similar to that of WUE_ET and WUE_T across all treatments in both observed and projected climates (Fig. 6). These indicators of water use efficiency (WUE_ET, WUE_T, and WUE_P) have diverse implications and applications¹⁹. For example, the interest in determining factors affecting WUE varies among breeders working with crops, agronomists focusing on water use and management, and conservationists in agriculture, spanning from field to landscape scales.

According to Rockström and Barron¹², WUE_ET provides a more comprehensive assessment of WUE compared to transpiration alone. This is because of the shift in water vapor due to reduced soil evaporation from increased biomass production under moisture-conservation practices, thereby reducing runoff. The increased WUE_ET resulting from vapor shift leads to more efficient utilization of evaporated water under optimal practices. This improvement positively impacts water productivity at watershed and basin scales.

Effectiveness of CSA practices in climate change adaptation

The study investigated the adaptive capacity of CSA practices under projected 2050 climate conditions. CSA practices have varying effects on grain yield and WUEET relative to conventional practices (CP) or control treatments in both observed (2020–2021) and projected 2050 climates. Among the tested CSA practices, ICP showed the greatest potential to enhance grain yield and WUE in both observed and projected 2050 climates. Relative to the CP treatment, CSA practices effectively increased grain yield by 5.4 to 19.6% in the current observed climate (2020–2021) and by 9.9 to 21.2% in the future climate. Similarly, compared with CP, WUEET in CSA treatments increased by 0.3–1.22% under the current climate and by 11.7–40.2% under the future climate. These findings underscore the superior effectiveness of CSA practices over conventional methods.

Conversely, CSA practices exhibited positive adaptive effects under projected future climates, significantly influencing yield and WUE_ET. The results indicate that CSA practices enhance future climate adaptation by 5.7–35.1% relative to control treatments. Among the tested CSA practices, ICP exerted the greatest influence on grain yield and WUEET, whereas SB had the least impact. ICP emerges as a robust practice for climate adaptation under both current and future climates, followed by CR and BP, which outperformed SB. Overall, in future climates, ICP, CR, BP, and SB increased yield by approximately 20.3%, 17.5%, 16.3%, and 5.7%, respectively, relative to CP. Moreover, compared with conventional practices, ICP, CR, BP, and SB are estimated to increase the climate change impacts on WUE_ET by approximately 35.1%, 32.6%, 17.2%, and 7.81%, respectively. Therefore, compared with observed conditions, CSA practices are projected to substantially increase wheat grain yield and WUEET under future 2050 climates. Notably, WUE_ET is more effective than yield as a metric for evaluating the benefits of these practices.

Under the baseline current climate conditions, CSA resulted in an approximately 0.21 t ha (5.25%) higher yield than conventional practices. Similarly, under future climate conditions, CSA yields 0.46 t ha (12.1%) higher than those under conventional practices. Accordingly, the adaptation effect that was observed through yield change over time indicates that CSA increases yield slightly by 1.2% under the future climate, while conventional practice decreases yield by 5% under the future climate. Conventional farming shows a yield decline under climate change, while CSA maintains stability and even slightly improves.

Discussion

The AquaCrop model demonstrates strong robustness in simulating both yield and runoff under different climate-smart agricultural (CSA) practices, as evidenced by the performance indicators in Table 1. For yield, the model achieves consistently high coefficients of determination R2 values ranging from 0.947 to 0.977 and Nash–Sutcliffe efficiency (NSE) values above 0.90, indicating excellent agreement between simulated and observed data. Practices such as SB and BP show strong performance, with very low RMSE values (0.009 and 0.003, respectively) and minimal bias (PBIAS close to zero), reflecting the model’s ability to capture yield dynamics with high precision. Even in cases where bias is slightly higher (e.g., CP with − 0.409), the NSE and R2 values remain high, suggesting that the model maintains reliability despite localized deviations. For runoff, the model also shows robust performance, although with more variability across practices. High R2 values (0.894–0.998) and NSE values (0.810–0.989) confirm that the model effectively reproduces runoff patterns. The SB, CR, and BP practices stand out with strong predictive accuracy, while CP exhibits higher RMSE (6.312) and PBIAS (1.549), indicating reduced reliability under that practice. Nonetheless, the overall consistency of high NSE values across most practices highlights the model’s resilience in capturing hydrological responses. Overall, these results suggest that AquaCrop is a reliable tool for evaluating CSA interventions, with strong robustness in yield simulations and generally solid performance in runoff predictions, albeit with some sensitivity to specific practices. Others studies^{43– 48}Additionally, this also confirms the precision of the AquaCrop model in simulating grain yields and soil and water content under different environmental conditions.

The study explored the comparative performance of CSA practices on water inflow and outflow across all soil‒water balance components under current and future climate conditions. Compared with CP, CSA practices such as ICP, CR, BP, and SB significantly enhanced the productivity of soil water balance components. However, the ability of CSA practices to increase the productivity of different water balance components varied. Among the evaluated practices, ICP, which integrates residue management, tillage, and structural improvements, notably increased blue and green water productivity by reducing losses from surface runoff and evaporation while enhancing transpiration and deep percolation. ICP effectively minimized surface runoff and evaporation losses while maximizing transpiration and deep percolation. This comprehensive approach addresses multiple ecological, agricultural, and economic benefits, enhancing the resilience of rainfed agriculture to climate change.

Improved drainage resulting from reduced surface runoff is particularly advantageous. In addition to conserving soil and nutrients that would otherwise be lost through runoff, evening out stream discharge by decreasing flows in the rainy season and increasing them in the dry season benefits downstream users. Deep percolation and soil water storage in the root zone are influenced differently by various CSA practices under changing climate conditions, where increased evaporation and transpiration due to atmospheric demand (linked to temperature and other factors) play significant roles. BP, which enhances rainwater storage in the soil root zone but increases evaporation, may result in reduced deep percolation. In contrast, integrated practices such as ICP, which combine various strategies, offer substantial and diverse benefits.

The effects of CSA practices on soil water balance mirror those on water-use efficiency, as improvements in water-balance components directly enhance efficiency. Specifically, within the soil water balance, ICP significantly improved water-use efficiency and grain yield compared with other practices. CR and BP showed effects similar to CP. The study evaluated CSA practices that integrated multiple components (tillage, residue management, and water management), making it difficult to quantify the individual benefits of each component within a treatment. While each component contributes incrementally to the system, specific components may exert greater influence than others⁴⁹.

CR enhances deep percolation and water storage while reducing soil evaporation losses. Retaining surface residue conserves soil moisture, suppresses weeds, and enhances soil organic matter and structure⁵⁰. BP improves deep percolation and water storage while reducing surface runoff. SBs enhance deep percolation and minimize surface runoff⁵. Integrating BP with CR enhances soil physical properties, such as infiltration rate and aggregate stability, through residue retention^5,51. SB significantly contributes to excess water storage, reduces surface runoff, mitigates soil erosion and downstream flooding, and enhances groundwater discharge^5,24. Despite potential yield penalties from land-use change, SB is strongly recommended in areas experiencing declining groundwater levels, severe water shortages, and soil erosion²⁴. SB offers significant environmental and ecological benefits compared to immediate crop production²², while CR and BP provide short-term improvements in crop production.

Therefore, to optimize wheat yield, water use efficiency, and water balance components within agricultural landscapes, an ICP approach integrating CR, BP, and SB practices is crucial. Compared with the baseline (1981–2010), ET increases exceed projected rainfall under 2050 climate conditions. In such scenarios, CSA practices that enhance the productivity of farming systems and water-use efficiency are imperative. Technologies aimed at reducing ET and increasing crop yield per unit of water lost as ET offer potential solutions to water scarcity issues⁴⁹. Similarly, previous studies in Ethiopia have demonstrated that implementing various conservation tillage practices improves water balance components by reducing evaporation and runoff while increasing transpiration, soil water storage, and deep percolation^5,6,19. This study provides new insights into water balance, yield gaps, and the potential for water-use efficiency across different CSA practices under current and future scenarios.

CSA practices, therefore, have significant practical implications for farmers, extension systems, and markets. The adoption of CSA strategies with wide sole and combined application of conservation tillage and in situ water harvesting has been shown to improve yields, incomes, and household resilience to climate shocks. However, practical challenges include limited evidence and information, and knowledge gaps among smallholder farmers. Addressing these issues requires capacity building, participatory research, and strong institutional support. At the farm level, CSA practices reduce greenhouse gas emissions and enhance soil health, while at the community level, they foster ecosystem services such as water regulation, soil conservation, and landscape restoration. Ultimately, the practical success of CSA depends on enabling policies, available research-proven CSA technologies, farmers’ genuine participation, and inclusive governance structures that bridge local realities with national and global climate goals.

Conclusion

This study indicates that, as projected by multiple GCMs under a single representative pathway for the RCP4.5 scenario, climate change is expected to affect wheat yield and productivity. The sole and combined application of CSA strategies helps reduce some of these impacts by minimizing water loss, primarily through evaporation and runoff. ICP practices, which integrate crop residues, physical structures, and improved tillage, have demonstrated improvements in yield and WUEEP under both current and future climate conditions. During the 2020–2021 climate period, ICP achieved the highest relative grain yield, water-use efficiency, and deep percolation while minimizing surface runoff, transpiration, and evaporation compared with other practices. The AquaCrop model demonstrated robust performance in simulating the projected impacts of CSA practices on crop yield, WUEEP, and soil water balance components under future climates that mirror those observed in current conditions. Accordingly, the AquaCrop simulation result confirms that CSA practices enhance the resilience of rainfed wheat production to climate change, thereby improving productivity. CSA practices outperform conventional farming in both current and future climates. Conventional practice shows a yield decline under climate change, while CSA maintains stability and even a slight improvement. Although the modeling approach demonstrated strong performance, uncertainties persist due to inherent assumptions embedded in the model structure, input parameterization, and management representations. Further limitations arise from the reliance on global climate model projections and downscaling techniques, which may affect both the magnitude and direction of projected climate change impacts on wheat productivity. Moreover, the evaluation of CSA strategies in this study was constrained to a 2-year field experiment conducted at a single site. Accordingly, the following recommendations are presented in light of these study-specific limitations.

• Conduct additional site-specific studies to evaluate the performance of CSA practices across diverse agro-ecological zones and multiple representative concentration pathways (RCPs).

• Assess CSA strategies across multiple locations and contrasting agro-climatic conditions to enhance the generalizability of results.

• Promote continued research and field-level testing of emerging CSA technologies to strengthen climate change adaptation strategies.

• Integrate long-term empirical field experiments with comprehensive economic and financial profitability analyses in diverse environments.

Supplementary Information

Below is the link to the electronic supplementary material.

Author contributions

AAB was involved in the conceptualization, data collection, analysis, preparation of the original draft, and searching for resources and funding. BB; SGG; FA; TH; DKA; WA; ZA; SFD; LT, and AMM were involved in the conceptualization, analysis, review, and editing.

Funding

This research was funded by the International Foundation for Science (IFS) [grant number 1I1_W_043323]; UK Research & Innovation (UKRI) through the Global Challenges Research Fund (GCRF) programme, under the project Social and Environmental Trade-offs in African Agriculture (SENTINEL), led by IIED and in part implemented by the Regional Universities Forum for Capacity Building in Agriculture (RU-FORUM) [grant number ES/P011306]; and the British Ecological Society (BES), financing this research through the Ecologists in Africa 2021 [grant number EA21/1204].

Data availability

The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors declare no competing interests.