Assessment of climate change impact on crop yield and irrigation water requirement of two major cereal crops (rice and wheat) in Bhaktapur district, Nepal

It is increasingly becoming clear that climate change is a real phenomenon and human activities such as burning of fossil fuel, deforestation, and so on are primarily to blame. As such, the recent anthropogenic emission of greenhouse gases is at its highest level (IPCC 2014). The impacts of increased temperature and elevated CO₂ level, intense or no rainfall, are widespread on natural systems and on humans (IPCC 2000, 2014). Water resources are affected, and hence the agricultural sector which could have long-term effects on food security (Malla 2008; IFPRI 2009). It is evident that increase in temperature and carbon dioxide (CO₂) have a positive impact on some crops, but the nutrient levels, soil moisture conditions, water availability for irrigation, and other crop-related conditions should be favorable in order to have better crop yield (Maximay 2014). Moreover, changes in precipitation pattern such as intense rainfall during a particular month are becoming more frequent and such events could have a devastating effect on crop production especially if they occurred in a sensitive phase of the crop, e.g., the flowering stage (Joshi et al. 2011).

All these events associated with climate change would pose further stress on farmers to produce more and more food for an increasingly growing and wealthier population (FAO 2011). The case is even more severe in peri-urban areas like Bhaktapur district, Nepal, which has experienced rapid urbanization of late (Shrestha et al. 2013b). To cope with such adverse effects (of climate change), different adaptive management practices especially applying proper fertilizer dose and water application need to be adopted. Furthermore, adaptation practices such as shifting of crop plantation date, adjusting cropping area and intensity might need to be considered (Iizumi & Ramankutty 2015).

There have been some studies which have quantified the impact of climate change on cereal crop yields at regional and national scale. For instance, Lal (2007) reported a decrease of 4–10% in cereal crop yield by 21st century in South Asia. Similarly, Shrestha et al. (2014) reported, based on a study in Myanmar, that future climate condition would increase the paddy rice yield and would thus increase the food security in the region. Acharya & Bhatta (2013) used a quantitative modelling approach to calculate the impact of climate change on agricultural gross domestic product of Nepal, and found positive impact due to increased precipitation in future. Karn (2014) reported a decrease of 4.2% in rice yield based on analysis made on 20 major rice growing district of Nepal. Palazzoli et al. (2015) carried out a study based on a physically based model in Indrawoti river basin of central Nepal, and found different estimates (−36% to +18% for wheat, and −17% to +12% for rice) of crop yield changes while using different future climate projection data. It is thus evident that very few studies have focused on peri-urban regions like Bhaktapur district.

In this context, this paper analyzes the impact of climate change on yield response of main cereal crops – rice and wheat in Bhaktapur district, Nepal. It also aims to find ways to stabilize the yield with plausible water and fertilizer application scenarios. An understanding of the impacts of recent climate trends on major cereal crops would help to anticipate impacts of future climate on the agriculture sector. We believe that the outcome of the study will facilitate in formulating suitable adaptation strategies to cope up with the adverse effects of climate change, thereby increasing food security of the district.

Several different techniques were applied for the data collection and will be discussed in the next sections.

Altogether 30 soil samples (see Figure 1) from different locations of the district were taken based on snowball sampling technique. Moreover, farmers of the sampled land were also supplied with questionnaires in order to collect information regarding farming practices, main factors affecting the plantation of crops, crop yield, variety of crops, crop phonological stages, and period, time and irrigation practices. To determine the soil texture class, collected soil samples were submitted to the Agricultural Technology Center (ATC), Nepal. Although the soil samples were taken from 30 cm depth, uniform soil profile is considered as suggested by Shrestha (2014).

The ETo calculator, developed by the Land and Water Division of the Food and Agriculture Organization (FAO), is used to calculate the potential evapotranspiration (ETo). The tool is based on a theoretical method proposed by Penman and Monteith (Allen et al. 1998) to calculate the ETo. The tool requires several climatic data such as temperature (maximum and minimum), relative humidity, wind speed, solar radiation etc., at user defined time steps. In this study, the tool was run for a daily time scale. Besides calculating the ETo, temperature data are also produced in a format suitable for the AquaCrop model (see below for details on the model).

Soil-plant-air-water (SPAW) is a model which is generally applied to simulate daily hydrologic water budgets of agricultural landscapes. The embedded hydrologic analysis involves the evaluation of soil water infiltration, conductivity, storage, and plant-water relationships (Saxton et al. 2006). Soil characteristics such as textural composition, organic matter content, as obtained from the laboratory tests, were supplied to the model which in turn simulated permanent wilting point, field capacity (FC), total available water (TAW), and saturated hydraulic conductivity (SAT) as outputs. These outputs are actually required by the AquaCrop model (see next section for details on the model).

AquaCrop is a crop-water productivity model which relates the soil, crop and atmospheric components. The soil component requires soil horizons of different texture composition, and for each textural class, hydraulic characteristics (generally the results of SPAW model) are required. The atmospheric component requires rainfall (generally observed at a meteorological station), temperature (generally the result of ETo calculator), evapotranspiration (generally the result of ETo calculator), and carbon dioxide concentration (generally taken as default value of the year 2000 measured at Mauna Loa Observatory, Hawaii). The crop component requires information about the crop such as phonology, crop cover, root depth, biomass production and harvestable yield, and field management conditions such as irrigation fertility, and field agronomic practices (Raes et al. 2012). The crop files for Nepal's general crops were adapted from the study of Shrestha (2014). The model was built up using the subsequent model outputs of the ETo calculator, SPAW model, and using information collected in the questionnaire survey. The model is then calibrated based on the actual yield, also obtained from the questionnaire survey.

Future climate projection data were fetched from the Nepal Climate Data Portal (NCDP 2014) with spatial resolution of 25 × 25 km. The data were based on Providing Regional Climates for Impacts Studies (PRECIS), one of the widely used dynamical downscaling tools developed at Met Office and Hadley Center, United Kingdom. The tool uses the atmospheric component of the HadCM3 Global Climate Model – GCM (Gordon et al. 2000; Jones et al. 2004, as cited in NCDP 2014). Data of a Regional Climate Model (RCM) run in PRECIS with imposed Lateral Boundary Condition (LBC) as HadCM3Q0 and ECHAM5, both with A1B scenario (IPCC 2000), were fetched in the NetCDF format (NCDP 2014). The future climate data (2030–2060) would then be used for simulating field management strategies for future climate change scenarios. As such, average data of all pixels covering the study area were estimated using ArcGIS.

Different water and fertilizer scenarios (Table 1) were formulated as plausible adaptation measures to cope with adverse climate change impact. The calibrated AquaCrop model was run to simulate the crop yield using these scenarios. Under water management scenario, rainfed (RF) and full irrigation (FI) conditions were formulated. These water management scenarios are currently practiced in the study area. Due to high household demand in the study area because of ever increasing population, and due to the need for allowing minimum discharge to maintain the ecological status of surface water, farmers are likely to face water scarcity and might not be able to use all the surface water sources for irrigation purposes in near future. Hence, another scenario (deficit irrigation) in which limited water is applied at the most sensitive growing phase of the crop (e.g. before and after flowering), is tested too. Deficit irrigation scheme is a rather promising and tested irrigation technique, especially in rain deficient conditions (Geerts et al. 2008, 2010; Geerts & Raes 2009; Shrestha et al. 2013c). As such, we tested two deficit irrigation scenarios – D1 and D2 (see description in Table 1). While RF and FI conditions are applicable for monsoon rice, the D1 and D2 are only applicable for winter wheat. Under the fertilizer management scenarios, different fractions of fertilizer as per National Recommended Fertilizer Dose (NRFD) were formulated (Table 1). Finally, all possible permutations of water and fertilizer application scenarios were tested.

It was found that 100% of the farmers grow rice during the monsoon season (June to September), and the overwhelming majority (>80%) grow wheat during the winter season. During winter, the rest (20%) grow maize. The statistics indeed justified our selection of rice and wheat being two major cereal crops of the study area. Besides, 80% of the respondents reported that they waited for rainfall in order to sow the crops, and the rest (20%) first examined moisture content in the soil and then fixed a sowing date.

Reported crop yield from the farmers (Table 2) of monsoon rice (4.9 t/ha) is comparable to the findings of MoAC (2010) which stand at 4.5 t/ha. This is also the case for the winter wheat in which responded reported yield (2.5 t/ha) nearly matches with the MoAC (2010) value of 2.69 t/ha. Finally, the phenological periods for monsoon rice and winter wheat are obtained from the questionnaire survey and it is reported that plantation dates for rice and wheat were July 1 and December 15, respectively.

ETo calculator revealed that the daily ETo have a decreasing trend for the base period (1979–2013) which is contrary to expectation as it is perceived that there would be rise in temperature due to climate change. However, as can be seen in Figure 2, the maximum and minimum temperatures are rather stable in the last decade or so. The decreasing ETo trend could then be due to the higher humidity levels in the atmosphere and decreased amount of solar radiation reaching the Earth's surface. It is well perceived that a small change in solar radiation can bring large amount of change in evapotranspiration (Gad & Gyar 2010).

Based on the soil physical characteristics as determined using Pedo-transfer functions from soil texture using the SPAW model (Saxton & Rawls 2006), we classified the soil samples into three classes namely S1, S2 and S3. The classification was based on the range of TAW values (refer to Table 3). Other characteristics of the soil are presented in the Table 3. Spatial distribution of soil samples is presented in Figure 1. It is clear that S3 was the most dominant type in the study area.

It is clear that monsoon rice yield does not seem to be too sensitive to the total rainfall during crop season. While large variation (600–1,395 mm) in the rainfall is evident, the variation in the monsoon rice yield (2.9–4.2 t/ha) is suppressed (Figure 5). The coefficient of correlation (r²) between them is thus very low (0.04). However, winter wheat yield is very sensitive to total rainfall occurring during crop season (Figure 6). Large variation in rainfall (25–210 mm) is also reflected in large variation in the yield (0.1–4.4 t/ha) with higher r² value of 0.37 between them. It therefore implies that RF irrigation can be practiced for monsoon rice while winter wheat needs irrigation infrastructure to ensure timely irrigation and better yield.

The monsoon rice yield in different fertilizer and water management scenarios is presented in Table 4. As can be seen, in the RF case, increment in fertilizer application from 0% to 150% of the NRFD resulted in a significant increment in the yield (up to 65%), and the result for the irrigated (FI) case is even higher (up to 74%). The net contribution of irrigation in the yield increment is below 4%. This implies that increment in fertilizer application should be practiced while the provision of even full irrigation would barely be beneficial. It might also be due to the fact that rainfall is enough during the crop growing period, and the development of an irrigation system might not be economically viable for rice anyway. These findings are consistent with the findings of Shrestha et al. (2013c) for the southern plain region (Terai) of Nepal.

The same for the winter wheat (Table 5) illustrated a rather different picture. Winter wheat yield could substantially be increased (up to 110%) by providing optimal fertilizer dose, and the contribution of irrigation is also substantial (up to 71%), unlike that observed in the case of monsoon rice (only up to 4%). Yield response indicated that the increment from the increased fertilizer dose would further be enhanced in the case of FI rather than RF and deficit irrigation (D1, D2). While it is evident that the yield was highest with full irrigation, the yield at deficit irrigation schemes (D1 and D2) would not be too low, especially for the D2 case. Hence, proper deficit irrigation schemes could be the option for the winter wheat along with optimal fertilizer application in order to increase the yield.

For all the fertilizer and water management scenarios, model simulation showed two contrasting results when different future (2030–2060) climatic data (HadCM3Q0 and ECHAM5) were forced. For instance, in calibrated conditions (fertilizer: 100% of the NRFD, and RF), the HadCM3Q0 based simulation showed a marked drop (−63.78%, see Table 6) in monsoon rice yield in future period (2030–2060) as compared to the base period (1979–2013) while the ECHAM5 based simulation showed +20.5% increment (see Table 6). Such significant differences in the yield when using two different climate projection data would surely have implications for policy makers. It should however be noted that different results when using different climatic data sets have been widely reported in the literature. McSweeney & Jones (2010) related such uncertainties to differences in the climate model formulation and the adopted downscaling techniques. Lately, researchers have used Mean Model Ensemble (MME) as future climate data in simulation models (McSweeney & Jones 2010) which can be perceived as balancing the extremes of the climate models. Some researchers (e.g. Prudhomme 2006) thus have warned of ‘misleading conclusions’ derived from different climate change projection models.

In contrast, the ECHAM5 based simulation showed that the crop yield would increase in future. This finding is in line with the results of Shrestha et al. (2014) in Myanmar; Bhatt et al. (2014) in Koshi river basin, Nepal; and Joshi et al. (2011) across Nepal. For all cases (RF and FI, and all fertilizer application conditions), the ECHMA5 based simulation showed higher yield than current yield. Furthermore, there exists less variation, as indicated by narrower box plots, in the yield which indicates that there would be sufficient rainfall in the monsoon rice growing season, should the ECHAM5 projection prevail in future.

While it is not clear if the monsoon rice yield would increase or decrease in future as both future climate data set indicated contrasting result, it is rather clear that the yield of winter wheat can easily be stabilized or even increased adopting proper water management scenarios (FI or D1). Such significant uncertainty in future yield of monsoon rice is indeed a dilemma for policy makes, hence, an effort was made in investigating what caused such a drastic decrease in monsoon rice yield when forcing the HadCM3Q0 projection.

Furthermore, there would be rather unfavorable rainfall occurrence and distribution in monsoon rice crop growing season (see Figure 10, right). The cumulative rainfall (mean) for the season would be less than 500 mm if HadCM3Q0 projection prevails in future while the ECHAM5 projection indicates comparable cumulative rainfall with that occurring in the base period.

The previous simulation results, in the case of monsoon rice, indicated a marked decrease in yield in all possible management scenarios, should HadCM3Q0 projection prevail in future. Moreover, rainfall and temperature seem to favor an earlier plantation date (mid-March, see Figure 11). As an adaptation measure for the climate change impact, and in order to stabilize the monsoon rice yield, crop plantation months were arbitrarily shifted and simulations with both water management scenarios, RF and FI, are carried out. It has to be noted that the worst future climatic scenario (HadCM3Q0 projection) has been considered here, as simulation based on ECHAM5 projection showed increment in monsoon rice yield.

On the other hand, the main reasons for a rather stable winter wheat yield (based on HadCM3Q0 projection) are due to favorable rainfall distribution and lessened temperature stress (Figure 10, left). However, the wider range of the error bars of the box plot of I_net, meaning higher variability, is of concern to policy makers. Furthermore, the I_net based on ECHAM5 projection is higher than base period indicating that less rainfall is expected during winter wheat's growing season which is also evident in Figure 11 (bottom left).

An assessment of climate change impact on irrigation water requirement and crop yield of two widely used cereal crops in Bhaktapur district, Nepal, was made with the help of social and analytical (using various models) techniques. Questionnaire survey with 30 farmers, selected using snowball sampling technique, was carried out to gain insights on the crop, water and fertilizer management practices, and harvested yield. Moreover, soil samples from the croplands of the selected farmers were taken and later analyzed in a laboratory to determine texture composition and organic matter content. SPAW tool was used to determine physical characteristics of the samples, and the ETo calculator was used to estimate daily potential evapotranspiration series. To study the crop-yield response on forced climatic, crop, soil, and management data, a yield simulation model namely, the AquaCrop model was calibrated. To realize the possible impacts of climate change 30 years of future climate data (2030–2060), as simulated by Providing Regional Climates for Impacts Studies (PRECIS) based on regional circulation model simulation of HadCM3Q0 and ECHAM5 for A1B scenario, were forced to the calibrated AquaCrop model. Finally, some realistic water and fertility management scenarios for the selected crops were formulated and simulated so as to explore ways of stabilizing the crop yield.

Laboratory and SPAW analysis confirmed silt loam as being the most dominant soil type in the study area. The AquaCrop model was able to simulate crop yield with reasonable accuracy. Future (2030–2060) crop yield simulations indicated decreased yield (based on HadCM3Q0 projection) and increased yield (based on ECHAM5 projection) of monsoon rice and stable yield of winter wheat as compared to base period (1979–2014) yield. Simulation results for the management strategies indicated that the monsoon rice and winter wheat yields were mainly constrained by water scarcity and fertility stress, and that such a situation can be overcome with improved fertilizer application and proper irrigation management. Moreover, winter wheat yields in the dry season could easily be stabilized and could even be increased with proper deficit irrigation strategies. Results also indicated that an increase in fertilizer application was more effective in fully irrigated conditions than in the RF condition. Analysis relating to shifting of crop plantation date showed a marked increase in the crop yield of the monsoon rice in the future period. This study therefore recommends shifting of crop plantation date, temperature resilient crop genotype (to overcome temperature stress), and proper water and fertilizer management to stabilize the crop yield.

Ms L. Shrestha received financial support from Center of Research for Environment, Energy and Water (CREEW), Nepal, to conduct this study. The authors are very grateful for Dr Nirman Shrestha and Mr Pabitra Gurung for providing valuable suggestions.