Abstract
In the case study of Tangshan city, Hebei Province, China, this paper analyzes the temporal change of the blue agricultural water footprint (WF) during 1991–2016 and discusses the applicability of different climate change models during 2017–2050. Results show effective rainfall, wind speed and maximum temperature are leading factors influencing the blue agricultural WF. Relative error analysis indicates that the HadGEM2-ES model is the most applicable for climate change projections in the period of 2017–2050. Agricultural blue WF is about 1.8 billion m3 in RCP2.6, RCP4.5 and RCP8.5 emission scenarios, which is almost equal to the average value during 1991–2016.
INTRODUCTION
Due to the uneven distribution of water at spatial and temporal scales, more than two billion people live in water-stressed areas (Oki & Kanae 2006). Simulations have shown that 59% of the world's population will face water shortage by 2050 (Rockström et al. 2009). Agricultural water use currently accounts for ∼70% of freshwater withdrawals both globally and in China. The water footprint (WF) of a product was first defined in 2003 as the volume of freshwater used for production at the place where the product is actually produced (Hoekstra & Hung 2002) and it consists of green WF, blue WF, and grey WF. In this paper, grey WF is not considered in the WF calculation for crops. Green WF is defined as rainwater consumption that is stored in soil and evaporated during production during crop growth. Blue WF refers to surface and groundwater that is consumed by irrigation and evaporated, and the agricultural blue WF is considered as the theoretical irrigation water requirement.
Climate change is a global issue that may have dramatic impacts on ecosystems, social economics, and agriculture. Changes in the availability of water, particularly for agriculture, due to climate change have been observed and reported globally (Burn & Hesch 2007). Temperature increase and precipitation decrease in the Zayandeh–Rud River Basin is expected to cause increasing irrigation water demand from 2015 to 2055 (Gohari et al. 2013). Based on crop data and changes in the seasonal timing of water demand, irrigation water requirements in the Guadalquivir River Basin of Spain are projected to increase 15%–20% by the 2050s (Rodríguez-Díaz et al. 2007). The satellite weather application platform (SWAP) model, global change models (GCM), and regional climate models (RCM) have been used to predict local impacts of climate change on irrigation water demand in Turkey (Yano et al. 2007). The expected water requirements of global irrigation, without measures to alleviate climate change impacts, may increase ∼20% by 2080 (Fischer et al. 2007). The impacts of climate change on agricultural water use, especially on irrigation water demand, have been reported in numerous studies in China (Zhang & Cai 2013). Existing studies have shown that the influence of climate change on regional irrigation water demand varies greatly among different areas (Zang et al. 2013; Wang et al. 2016). Research on the effects of climate change on the agricultural blue WF is of great significance for guiding agricultural management to cope with climate change.
At present the climate change model is the main approach to predicting future climate and its associated human response. However, initial and boundary conditions, scenarios, observations, model parameters, and structure of climate simulations all cause some degree of uncertainty in simulation results, and uncertainty is more prominent at the regional scale (IPCC 2013; Knutti & Sedláček 2013). Therefore, evaluating the adaptability of climate change models to different regions is essential for climate change model use.
The main objectives of this paper were to: (1) identify key meteorological variables influencing the blue WF of major crops and calculate the relative error of these variables with different climate change models during 1991–2016; and (2) determine an applicable model for a specific region and use it to calculate the WF in 2017–2050. The results will help guide regional water use management.
MATERIALS AND METHODS
Site description and data
Located in the east of Hebei, a province in north China, Tangshan covers an area of ∼13,472 km2 (Figure 1). The city is characterized by a typical mainland monsoon climate with annual precipitation of 644.2 mm. The per capita water resources in Tangshan are 329 m3, which are less than 15% of that in China and water resources in the region are in short supply. Agricultural water consumption accounts for about 60% of the total regional water use and crop irrigation water takes up more than 90% in this region.
Daily meteorological data from 1991 to 2016 were acquired from the China Meteorological Data Service Center (http://data.cma.cn/). Parameters included average air temperature, average relative humidity, average wind speed, sunshine hours, and precipitation. Time series of crop yield were taken from annals of statistics for Tangshan. Data from the coupled model inter-comparison project phase 5 (CMIP5) released by the Earth System Grid Federation were downloaded from the internet with a grid distance of 25 km, and data used in calculations were interpolated to 0.5 degrees with a bilinear interpolation method.
Methods
Agricultural blue WF
During computation, the CWR is assumed to be equal to crop evapotranspiration (ETc, mm/day), which can be calculated by multiplying the potential reference evapotranspiration (ET0, mm/day) and the crop coefficient (Kc). In this study, ET0 was calculated using the Penman–Monteith equation recommended by the FAO in 1998. The parameter Kc is obtained based on current values in Tangshan City and Hebei Province combined with parameters from FAO-56.
Meteorological yield

RESULTS AND DISCUSSION
Applicability of climate change model
Agricultural WF in Tangshan
In Tangshan, the main cultivated crops include soybeans, sorghum, oil crops, cotton, vegetables, tubers, muskmelon, watermelon, winter wheat, tobacco, summer maize, and rice. Among the 12 kinds of crops, the WF of summer maize, winter wheat, oil crops, cotton, vegetables, and rice accounted for more than 94% of the entire agricultural WF of the city. Therefore, to simplify other analyses in this paper, these six crops were selected to represent the major crops of Tangshan.
Agricultural blue WF
The agricultural blue WF is closely related to CWR and effective precipitation, which reflect the combined effect of growth period, precipitation, topography, and meteorological conditions. In Tangshan city, the WF of summer maize is the highest, followed by winter wheat; however, the blue WF of winter wheat exceeds that of summer maize because the growth period of summer maize is consistent with the rainy season (June–September), which makes it possible for some of its water requirement to be satisfied by green water.
The total blue WF displayed an insignificant downward trend of −0.27 billion m3 per decade from 1991 to 2016, which is favorable to the current water crisis. The blue WF showed strong annual fluctuations because it was affected by meteorological variables, especially precipitation, which had strong annual variations. According to specific crops, the blue WF of winter wheat and rice (significantly) and of summer maize and oil crops (insignificantly) decreased, whereas that of vegetables and cotton increased significantly. The total volume of the blue WF was between 1.17 billion m3 and 2.82 billion m3, with an average value of 1.79 billion m3. The volume of the blue WF for grain crops including winter wheat, summer maize, and rice was 80% of that for the whole region. Therefore, reducing the irrigation water demand for these three types of crops is critical for water conservation.
Simulation accuracy of meteorological factors with different climate models
Relative error of different climate models
. | RE . | TE . | |||||
---|---|---|---|---|---|---|---|
Winter wheat . | Summer maize . | Rice . | Cotton . | Oil crops . | Vegetables . | ||
(a) GFDL-ESM2M model | |||||||
Effective precipitation | 6.21 | 3.95 | 3.51 | 4.36 | 3.95 | 2.93 | 4.59 |
Wind speed | 5.49 | 12.88 | 4.87 | 5.50 | 1.29 | −1.63 | 7.2 |
Maximum temperature | −1.20 | −0.87 | −1.11 | −1.19 | −0.87 | −1.29 | 1.07 |
(b) HadGEM2-ES model | |||||||
Effective precipitation | 5.01 | 2. 52 | 2.55 | 3.67 | 2.52 | 4.29 | 3.57 |
Wind speed | 3.72 | 6.44 | 7.14 | 6.32 | 16.44 | −3.17 | 5.95 |
Maximum temperature | −1.18 | −0.66 | −0.43 | −0.86 | −0.66 | −0.84 | 0.87 |
(c) IPSL-CM5A-LR model | |||||||
Effective precipitation | 9.33 | 1. 36 | 3.77 | 3.87 | 1.36 | 5.01 | 4.85 |
Wind speed | 7.79 | 7.71 | 3.21 | 3.74 | 7.71 | −1.07 | 6.33 |
Maximum temperature | −1.66 | 0 | −0.68 | −0.57 | 0 | −1.33 | 0.83 |
(d) MIROC-ESM-CHEM model | |||||||
Effective precipitation | 9.70 | 4.16 | 3.27 | 5.67 | 4.16 | 6.82 | 6.2 |
Wind speed | 29.44 | 6.22 | 2.80 | 37.49 | 6.22 | 43.83 | 18.17 |
Maximum temperature | −1.05 | −0.56 | −0.40 | −0.56 | −0.56 | −6.90 | 1.21 |
(e) NorESM1-M model | |||||||
Effective precipitation | 9.03 | 2.18 | 3.27 | 4.60 | 2.58 | 5.15 | 5.06 |
Wind speed | 6.40 | 10.12 | 3.01 | 4.20 | 10.12 | −2.64 | 6.89 |
Maximum temperature | −1.35 | −0.83 | −0.97 | −1.00 | −0.83 | −1.40 | 1.08 |
. | RE . | TE . | |||||
---|---|---|---|---|---|---|---|
Winter wheat . | Summer maize . | Rice . | Cotton . | Oil crops . | Vegetables . | ||
(a) GFDL-ESM2M model | |||||||
Effective precipitation | 6.21 | 3.95 | 3.51 | 4.36 | 3.95 | 2.93 | 4.59 |
Wind speed | 5.49 | 12.88 | 4.87 | 5.50 | 1.29 | −1.63 | 7.2 |
Maximum temperature | −1.20 | −0.87 | −1.11 | −1.19 | −0.87 | −1.29 | 1.07 |
(b) HadGEM2-ES model | |||||||
Effective precipitation | 5.01 | 2. 52 | 2.55 | 3.67 | 2.52 | 4.29 | 3.57 |
Wind speed | 3.72 | 6.44 | 7.14 | 6.32 | 16.44 | −3.17 | 5.95 |
Maximum temperature | −1.18 | −0.66 | −0.43 | −0.86 | −0.66 | −0.84 | 0.87 |
(c) IPSL-CM5A-LR model | |||||||
Effective precipitation | 9.33 | 1. 36 | 3.77 | 3.87 | 1.36 | 5.01 | 4.85 |
Wind speed | 7.79 | 7.71 | 3.21 | 3.74 | 7.71 | −1.07 | 6.33 |
Maximum temperature | −1.66 | 0 | −0.68 | −0.57 | 0 | −1.33 | 0.83 |
(d) MIROC-ESM-CHEM model | |||||||
Effective precipitation | 9.70 | 4.16 | 3.27 | 5.67 | 4.16 | 6.82 | 6.2 |
Wind speed | 29.44 | 6.22 | 2.80 | 37.49 | 6.22 | 43.83 | 18.17 |
Maximum temperature | −1.05 | −0.56 | −0.40 | −0.56 | −0.56 | −6.90 | 1.21 |
(e) NorESM1-M model | |||||||
Effective precipitation | 9.03 | 2.18 | 3.27 | 4.60 | 2.58 | 5.15 | 5.06 |
Wind speed | 6.40 | 10.12 | 3.01 | 4.20 | 10.12 | −2.64 | 6.89 |
Maximum temperature | −1.35 | −0.83 | −0.97 | −1.00 | −0.83 | −1.40 | 1.08 |
Among the three meteorological factors, the simulation error of maximum temperature produced by the five climate models was the lowest and that of wind speed was the highest. The relative error of wind speed produced by the MIROC-ESM-CHEM model during the growth periods of winter wheat, cotton, and vegetables reached up to 29.44%, 37.49%, and 43.83%, respectively, indicating that the MIROC-ESM-CHEM model is not suitable for Tangshan climate projection. Effective precipitation was the most important influential factor and had a strong impact on the blue WF of winter wheat, summer maize, and rice; the combined blue WF of these crops accounted for 81% of the regional blue WF. The second most important influential factor was wind speed, followed by maximum temperature. Table 1 shows that the TE values for effective precipitation and wind speed in HadGEM2-ES were 3.57 and 5.95 respectively, which correspond to the minimum error for the five models. The minimum TE for maximum temperature was 0.8 in the IPSL-CM5A-LR model. Therefore, the HadGEM2-ES model is considered most suitable for climate simulations of blue WF in Tangshan.
CWR under different future climate change scenarios
Three representative concentration pathway (RCP) scenarios (RCP2.6, RCP4.5, and RCP8.5) were considered for projections of future CWR conditions in Tangshan. The projected average annual CWR for winter wheat under climate scenarios from 2017 to 2050 was 488.02 mm, 512.31 mm, and 510.01 mm, respectively, smaller than the average value of 520.92 mm during 1991–2016 (Figure 2). Data from these three climate scenarios showed a significant upward trend (99% confidence level), opposite to the trend during 1991–2016. The CWR for summer maize increased significantly (99% confidence level) under the three scenarios, with average values of 477.09 mm, 485.92 mm and 494.39 mm, respectively. Values in RCP4.5 and RCP8.5 were bigger than the average value of 479.14 mm. The respective average water requirements of rice under the three scenarios were 852.61 mm, 861.30 mm, and 876.98 mm, which vary largely from the average value of 749.36 mm during 1991–2016. In comparison with average data from 1991 to 2016, the projected CWR for cotton, oil crops, and vegetables were all larger in 2017–2050. Accordingly, climate change may cause crop requirements to rise in the future.
Crop yields under climate change
Assessment of PLSR predictive results
Yield data from 1991 to 2010 were decomposed into trend yield and meteorological yield using a linear moving average method with 5-year moving windows. Results of the path analysis indicate that temperature difference, sunshine hours, and precipitation sufficiently reflect the influence of climatic factors on crop yield. These three meteorological factors were selected as dependent variables for the partial least squares regression (PLSR) model, with meteorological yield as the independent variable. The PLSR model was established using meteorological crop yield and average meteorological data from different growth periods. The sum of trend yield and simulated meteorological yield were considered as the simulated annual yield. Actual yield data from 2011 to 2016 were used to assess the simulation accuracy of different types of crops. Error for the six types of crops was all below 10%; thus, the PLSR model is suitable for crop yield prediction.
Crop yields under climate change
The corrected projection data and PLSR were used to simulate crop yield during 2017–2050. With the exception of winter wheat in all scenarios and summer maize in RCP2.6, the yield of all crops decreased. For winter wheat, summer maize, rice, and cotton, both yield gains and losses were highest under RCP8.5 and lowest under RCP2.6. The largest decrease in oil crop output occurred under the RCP4.5 scenario. The yield of winter wheat was almost the same under all scenarios; rice yield in RCP2.6 increased by 900 kg/ha compared with that in RCP4.5 and RCP8.5. No obvious changes in the yield of other crops were evident, showing that the influence of climate change on yield varies among crops.
Agricultural WF in 2017–2050
Figure 3 shows the WF and blue WF of crops under RCP2.6, RCP 4.5, and RCP8.5 scenarios over the period of 2017–2050. The predicted average footprint during 2017–2050 under RCP2.6, RCP 4.5, and RCP8.5 was 3.85 billion m3, 3.97 billion m3, and 4.02 billion m3 and the blue WF was 1.85 billion m3, 1.81 billion m3, and 1.83 billion m3, respectively, almost equal to the average value during 1991–2016. The WF and blue WF varied slightly among the three emission scenarios because of different impacts of climate change on crop water requirement and crop yield.
Implications of water resources management
As discussed in this paper, reducing winter wheat and summer maize yield can have a large impact on regional blue WF. In contrast, significant increases in vegetables cause only small changes to the regional blue WF. This is because the blue WF per unit mass of vegetables is only 14.72 m3/ton, whereas that of winter wheat and summer maize is 846.4 and 407.8 m3/ton, respectively. Therefore, adjustments to planting structure can be implemented to mitigate water scarcity. Tangshan city should plant more vegetables and limit wheat planting. In addition, an interregional virtual water compensation scheme can provide a practical solution for water shortages.
CONCLUSIONS
In this paper, we presented a comprehensive analysis of the effects of climate change on regional agricultural blue WF based on meteorological and economic data, using the city of Tangshan as an example. Winter wheat, summer maize, rice, cotton, oil crops, and vegetables account for 94% of the total agricultural WF, and blue WF contributes 53.08% of the total agricultural WF. This demonstrates that these six crop varieties are major factors in the regional agricultural water resources system. Effective precipitation, wind speed, and maximum temperature are the main factors impacting the agricultural blue WF. The HadGEM2-ES model has the minimum total error for 2017–2050 simulations, and thus was selected as the most applicable climate change model. PLSR model simulations were used to calculate the blue WF of six kinds of crops under RCP2.6, RCP4.5, and RCP8.5 emission scenarios. The results showed that average agricultural blue WF during 2017–2050 varied slightly compared with that during 1991–2016 because of the inverse effect of climate change on crop water requirement and yield.
The methods used in this paper provide an example of agricultural blue WF simulations in a specific region and can be used to guide management of water resources. However, there are some limitations of the method used in this paper. Firstly we analyzed only five climate change models, and more models should be investigated in future research. In addition, the same climatic parameters were used in the analysis of the simulation accuracy of climate change models, but the factors influencing blue WF are different for the studied crops and should be explored individually. Further studies should be implemented to address the limitations listed above to achieve more accurate predictions of blue WF.
ACKNOWLEDGEMENTS
This study was supported by the National Key Research and Development Program of China (grant no. 2016YFC0401400) and the National Natural Science Foundation of China (grant nos 51709107; 51679089; 51609082).