Temporal trends of climatic variables and water footprint of rice and wheat production in Punjab, India from 1986 to 2017

The agriculture sector is vulnerable to climate change and related changes in the hydrological cycle. In order to understand the changes in climatic variables and their implications for agricultural water consumption, the present study aims to analyse the temporal variability of climatic factors and water footprint (WF) of rice and wheat during the period 1986–2017 in Ludhiana, Punjab. Further, it aims to identify the dominant climatic factors that cause variation in reference evapotranspiration (ETo) and WF of rice and wheat. WF was estimated using CROPWAT, and Path analysis was used to determine the dominant climate variables. Temporal trends of climate variables were analysed using the Mann– Kendall test. The total WF of both rice and wheat shows a significant declining trend over the past 32 years. Sunshine duration and wind speed were the dominant factors influencing the variability of total WF of rice and wheat, respectively, whereas rainfall strongly influenced the green and blue WF of rice and wheat. Rainfall had a high variability, and consequently, irrigation water requirement was highly fluctuating. This indicates the significant impact of present and projected erratic pattern of precipitation on agriculture due to climate change and reiterates the importance of adaptive measures like rainwater harvesting and capacity building.


GRAPHICAL ABSTRACT INTRODUCTION
The agriculture sector is one of the most vulnerable sectors to the risks of climate change and related changes in the hydrological cycle (Smit & Skinner ). Large-scale changes in the hydrological cycle like increase in atmospheric water vapour, changes in precipitation, soil moisture and run-off have been linked to global warming (Bates et al. ).
Climatic factors along with non-climatic drivers like 'population growth, economic development, urbanization, land use changes and water management responses' competing for water resources can have profound impacts on water availability for both rainfed and irrigated agriculture (Cisneros et al. ). Irrigation accounts for 70% of global water withdrawals and more than 90% of consumptive water use (IPCC ).
With the projected expansion in irrigated area and cropping intensity, it is estimated that future irrigation water demand would surpass water availability in various regions under climate change scenario (Wada et al. ).
Water footprint (WF) is, in general, an indicator of direct and indirect freshwater appropriation, measured in terms of water volumes consumed (evaporated or incorporated into a product) and polluted per unit of time. The volumetric WF comprises three components: green WF refers to the consumption of rainwater; the blue WF refers to water consumed from surface and groundwater sources; while grey WF is an indicator of volume of water polluted, i.e. the freshwater volume required to assimilate the pollutant load to bring it to natural condition/ambient standards (Hoekstra et al. ). Temporal trends in WF reflect changes in crop water use over time for a given place (Lu et al. ). The WF green and WF blue are computed based on reference evapotranspiration (ET o ) and precipitation and therefore directly associated with water availability in a given region. ET o is a measure of the 'evaporative demand of the atmosphere' which solely depends on climatic parameters (Allen et al. ). ET o is a key variable in the hydrological process and determines the availability of water for plant growth (Gao et al. ). Among crops, rice and wheat have the largest blue water footprints, together accounting for 45% of the global blue WF (Mekonnen & Hoekstra ). India is a major food producer, where the agriculture sector accounts for 90% fresh water use (Dhawan ). It is also a water-stressed country that is expected to face severe water constraints by 2050 (OECD ). High water stress has been found to contribute to high virtual water content values (Fader et al. ). The study region Punjab also known as the 'bread basket of India' is one of the largest producers of rice and wheat in India (Department of Food Civil Supplies & Consumer Affairs Govt. of Punjab ). In Punjab, 85% of water consumption is accounted for by the agriculture sector (Gulati et al. ), of which groundwater accounts for 90-97% of the irrigation in the Central Zone of Punjab (Sarkar et al. ). Punjab is facing a massive depletion in its water

Study area
The state of Punjab is located in north-western India. It extends from 29 32 0 to 32 32 0 north latitude and 73 55 0 to 76 50 0 east longitude and comprises a geographical area of 5.03 million hectares (Mha), i.e. 1.54% of the total geographical area of India. Of this, 83.4% of the land (4.20 Mha) is cultivated, and rice and wheat are the major crops. The groundwater in 80% of the geographical area in Punjab is overexploited (Gulati et al. ). Punjab is divided into five agroclimatic zones (ACZ), of which the 'Central plain zone', comprising 36% of the total area of Punjab, is the largest (Rang et al. ). The Central plains also account for two-thirds of the total rice and wheat production in Punjab (Sarkar et al. ). The study district Ludhiana is located in the Central plain zone of Punjab and is therefore assumed to be representative of Punjab (Figure 1).

Data collection
Meteorological data for Ludhiana for the year 1986-2018 was acquired from the Department of Agrometeorology, Punjab Agriculture University, Ludhiana. The data included maximum temperature, minimum temperature, relative humidity (RH), wind speed, sunshine hours and rainfall.
Yield data for rice and wheat production of Ludhiana dis- 'the amount of water lost through evapotranspiration' is identical in value to CWR which is defined as 'the amount of water required to compensate the evapotranspiration loss from the cropped field' (Allen et al. ). ET c and CWR are identical in value, but since this study is focused on water resources, the term 'crop water requirement' has been used for further analysis.

Calculation of WF
WF of a product is expressed as water volume per unit of product (in m 3 /t). It generally has three components: blue water, green water and grey water. In this study, only WF green and WF blue , which are rainwater and irrigation water components, have been considered as these two depend on climate. WF green and WF blue have been computed as follows (Hoekstra et al. ): where CWU green and CWU blue are the green and blue water components, respectively, of crop water use that is equivalent to the summation of daily evapotranspiration (in mm/ day) over the length (in days) of the growing period (lgp); Y is the crop yield (Y, tons/hectare or t/ha), ET green represents green water evapotranspiration; ET blue , i.e. the blue water evapotranspiration or field-evapotranspiration of irrigation water, also denoted as irrigation water requirement (IWR), is the difference between the total crop evapotranspiration and effective precipitation (P eff ). ET blue is 0 when effective rainfall exceeds crop evapotranspiration. Crop evapotranspiration (ET c ) and effective rainfall (P eff ) were derived from CROPWAT output, which were further used to calculate green and blue WF. For the input in CROPWAT, Punjab-specific crop coefficients for wheat were derived from Kaur et al. (). The planting dates were kept constant for all years (25 June for rice transplantation and 5 November for wheat sowing). Using soil texture inputs where S statistic and VAR(S) were derived from MK test output in XLSTAT.

Impact of climatic factors on WF
Path analysis and correlation analysis were used to ascertain the dominant climatic factors affecting WF of rice and wheat crops. Since the data were not normally distributed,

Temporal variations of climatic factors
The descriptive statistics of climatic factors and results of the MK test for annual and seasonal trends are presented in Table 1. Temporal trends of annual climatic factors are presented in Figure 3 (Table 1). Since sunshine duration and radiation are strongly correlated (ρ ¼ 0.99), only sunshine duration was used in further analysis. The coefficient of variation (CV) of rainfall was found to be 34%, and the highest annual rainfall was

Temporal variations of ET o , CWR and IWR
ET o showed a statistically significant downward trend in the duration 1986-2017 (p < 0.05) (Figure 4). It was found to be decreasing at the rate of 0.012 mm/a ( Table 1). The average annual ET o for the study period was found to be 3.87 mm.
Thus, the trend of increasing air temperature and decreasing evapotranspiration confirms the existence of an 'evaporation paradox' in Ludhiana, Punjab. A significant downward trend was also found for seasonal ET o in the showed a non-significant declining trend at the rate of 2.05 and 0.93 mm/a for rice and wheat, respectively. As compared to CWR, IWR showed a greater fluctuation because of variability in rainfall. If rainfall is less, the same amount of water is compensated by irrigation; therefore, a high variability in rainfall consequently leads to high variability in IWR of crops. IWR (CV ¼ 294%) for rice was found to have a greater variation than wheat IWR (CV ¼ 29%) (Table A6), which could be because rice is grown in the monsoon season and rainfall was found to have CV of 41%. Further, correlation analysis was used to determine the relationship between CWR (ET c ) and climatic factors. In the case of rice, a significant positive relationship was found between CWR and sunshine hours (ρ ¼ 0.68) ( showed a significant increasing trend over the period 1986-2017. Rice and wheat yield were found to increase at the rate of 35 and 31 kg/ha/a, respectively (Figure 4).

Interannual variability in WF of rice and wheat
Interannual variability in green WF   (Table A6). Correlation analysis indicated a significant positive effect between WF green of rice and rainfall (ρ ¼ 0.71) followed by WF green of rice and RH (ρ ¼ 0.36). There was a negative correlation of temperature (ρ ¼ À0.37) with the WF green of rice (p < 0.05) ( Table 2).
This was consistent with the results of path analysis that showed that rainfall and RH significantly influenced WF green during the rice-growing season (p < 0.05) (Table 3). Additionally, sunshine duration was also found to significantly influence WF green . Similar to the WF green of rice, the correlation analysis of WF green of wheat revealed a significant positive effect of rainfall (ρ ¼ 0.91) followed by RH (ρ ¼ 0.49) and a negative effect of temperature (ρ ¼ À0.44) (p < 0.05). According to path analysis results, the WF green of wheat in the Rabi (winter) season was significantly influenced (p < 0.05) by rainfall only.

Interannual variability in blue WF
Similar to the trend of WF green , the blue WF of both rice and wheat did not show a significant trend over the duration 1986-2017 ( Figure 8). The average blue WF for rice and wheat, respectively, for 32 years was found to be 296 and 334 m 3 /t (Table A6). Similar to the IWR of rice, the blue WF of rice was found to be highly fluctuating varying from 0 m 3 /t (in several years) to as high as 1,448 m 3 /t in 1987 (Table A5). This is because of variability in rainfall. In the case of rice, correlation analysis indicated a significant negative effect of rainfall (ρ ¼ À0.72) and RH (ρ ¼ À0.49) on the WF blue of rice (  (Table 3).

Interannual variability in total WF
Annual variability of WF for rice and wheat is presented in Figure 9. The total WF of both rice and wheat showed a significant decrease over the past 32 years (p < 0.05) declining at the rate of 19 and 6 m 3 /t/a, respectively. The average total WF of rice and wheat for the period 1986-2017 was found to be 1,535 and 540 m 3 /t, respectively. If the 32-year period is divided into three periods, period I (1986-1995),  period II (1996II ( -2005 and period III (2006III ( -2017: the average WF of rice for each period was 1,721, 1,641 and 1,293 m 3 /t, respectively, with a percentage decrease of 5% between period I and II and 21% between period III and II. The average WF for wheat was 629, 507, 493 m 3 /t, respectively, for the corresponding periods. In terms of percentage, the wheat WF decreased by 19% between the periods I and II, and 3% between the periods II and III. The contribution of the WF green and WF blue , respectively, to the total WF was 80.7 and 19.3% for rice, and 38.2 and 61.8% for wheat. The annual total WF of rice was found to be positively correlated with sunshine duration (ρ ¼ 0.79) (p < 0.05). The WF total of wheat was found to be positively correlated with sunshine duration (ρ ¼ 0.73) followed by wind speed (ρ ¼ 0.61) (p < 0.05). Results of path analysis indicated that the total WF of rice grown in Kharif (monsoon/autumn) season in Ludhiana was mainly influenced by sunshine duration (p < 0.05), while wind speed influenced the WF of wheat. The details of regression weights for the relationship between WF of rice and wheat and climatic factors are presented in supplementary Table A9. Increase in the amount of aerosols and other air pollutants was found to be the major reason behind this phenomenon (Stanhill & Cohen ), while in South Asia, it was found to be primarily driven by cloud cover (Kambezidis et al. ).

DISCUSSION
Wind speed is another factor that has been shown to affect aerosol concentration ( The proportion of green WF (81%) in total water con-

ACKNOWLEDGEMENT
One of the authors acknowledges the fellowship received from the host institute.

DECLARATION OF INTEREST
The authors declare no conflict of interest.

DATA AVAILABILITY STATEMENT
All relevant data are included in the paper or its Supplementary Information.