North China Plain is one of the most important grain-producing areas in China. Because of unevenly distributed precipitation in this semi-arid area, crop production largely relies on underground blue water (UBW) to irrigate, overexploitation of which causes a lot of environmental problems. In this paper, we first defined the representative division of winter wheat irrigation water production efficiency with a support vector machine and genetic algorithm coupling algorithm, then established a model for evaluating the comparative advantage of UBW production efficiency by combining the effective precipitation with the UBW depth, calculated its value in each city, and further proposed the reasonable irrigation water requirement and its distribution under different grain reserves at the targeting year. The result showed that when the typical precipitation (P) is 25% in a targeting year, there is no need to irrigate with UBW; when P = 50%, supplying 5% of total winter wheat production (TWP) could save 15% of UBW; when P = 75%, supplying 5% of TWP could save 11% of UBW; and when P = 75%, supplying 10% of TWP could save up to 28% of UBW.
North China Plain, located in the arid and semi-arid area, is one of the most important crop-producing areas in China. People use 18.3% of the area, which produces 22% of the crops in China. The precipitation in North China Plain is unevenly distributed. The annual average precipitation is 470–912 mm, but about 65–80% of precipitation is from June to September (Gong, 2003). In addition, a lack of surface water makes it difficult to satisfy the requirement for irrigation water. To solve this problem, more groundwater could be exploited, but this would cause severe overexploitation and environmental problems (Liu, 2003).
The concept of green water and blue water was raised in 1995 and the major blue water in North China Plain is groundwater. Thus, to make full use of green water and improve the efficiency of underground blue water (UBW), efficient regulation and management are important to solve groundwater overexploitation. Some scholars have evaluated the UBW production efficiency in some dry areas (Ferrara, 2007; Machiwal et al., 2011), and proved that it could be used to guide the management of groundwater resources in practice. In their studies, some scholars have used features such as photosynthetic available radiation, average accumulated temperature, and soil organic matter in crop-water production functions of different crops in different areas (Kipkorir et al., 2002; Liu et al., 2002). Although some studies focused on North China Plain (Luo et al., 2006), blue water and green water were not separated in crop-water production functions when calculated. It is important to pay more attention to the irrigation efficiency of UBW in order to solve issues like groundwater over-mining and vulnerable ecological environment protection. To this end, we first collected agricultural meteorological data and UBW depth of 21 cities in North China Plain, then obtained the winter wheat irrigation water production efficiency in representative divisions using a support vector machine and genetic algorithm (SVM-GA) coupling algorithm, and finally established a model to evaluate the comparative advantage of UBW production efficiency of each city by three features: winter wheat irrigation water production efficiency, the effective precipitation, and UBW depth. Based on the results, we put forward the reasonable irrigation water requirement and its distribution under different grain supply rates and typical targeting years.
Materials and methods
Model for evaluating the comparative advantage of UBW production division
Through the equation, the comparative advantage for production efficiency of winter wheat irrigated with UBW could be obtained.
Scenario analysis considering food security and UBW use quantity
In the above equations, the weight of the comparative advantage of UBW irrigated in different places was considered equal. However, UBW distribution was different in every area, thus their weights were not the same. To solve the problem, the following optimization was carried out: (1) evaluating the comparative advantage of UBW production efficiency and classifying all areas into several irrigation groups; (2) arranging the higher utilization rate of irrigation groups to meet the demand of winter wheat; and (3) arranging the lower utilization rate of irrigation groups to meet the quantity demand for food security.
Method to determine the water production efficiency and its parameters’ thresholds in representative divisions
Method for water production efficiency of representative divisions
Light, heat, water, air, and nutrients are five basic elements of crop growth, and they have the same importance and non-substitutability (Duan et al., 2004). To obtain the regression coefficients of the crop-water production function in each region in North China Plain based on known observation sites, water production efficiency in a representative division was calculated based on the average photosynthetic available radiation, the average accumulated temperature and the soil organic matter using an SVM–GA coupling algorithm.
Different kernel functions have different effects on the generalization ability of SVM. Among the most common kernel functions, radial basis function (RBF), which has a good learning ability and fewer parameters (Lin & Lin, 2003), has been widely used. It was chosen to classify winter wheat water production efficiency in this paper. And the genetic algorithm (GA), which was borrowed from the biological genetics point of view to obtain the optimal values (Koza, 1992), was used to search for the optimal solutions of the parameters of RBF, while cross inspection was used to search for the optimal threshold of parameters.
There were 17 experimental stations of winter wheat irrigation in North China Plain. These stations were classified into three categories in advance by regression coefficients of winter wheat in each site, resulting in six of them belonging to the first category, eight to the second category, and three to the third category. The steps are as follows.
First, data were collected from different departments: background information (including geographical location, latitude, longitude) came from the national irrigation experimental station nets; regression coefficients of the crop-water production function came from Irrigation Water Quota of Major Crops in North China (Duan et al., 2004); the spatial distribution grid figure of the average photosynthetically active radiation per year and the spatial distribution grid figure of the average accumulated temperature (≥0 °C) per year came from the National Meteorological Information Center (NMIC); the spatial distribution grid figure of soil organic matter in North China Plain came from a 1:14000000 map of ‘The Soil Organic Matter Map of China’ by Data Center for Resources and Environmental Sciences of the Chinese Academy of Sciences.
Second, in the ArcGIS 10.0 interface, data of each irrigation observation site were extracted by using the spatial distribution grid figures collected and taken as training points. Appropriate kernel function and classifier were chosen to obtain the classification function f(x).
Finally, the data of each city were put into the model to obtain the appropriate water production efficiency of representative divisions.
Determining the parameters' threshold of water production efficiency
Results and analysis
Spatial distribution and representative division of water production efficiency in North China Plain
Comparative advantage and the spatial distribution of UBW irrigation production efficiency in North China Plain
Rational management of UBW use quantity at different effective precipitations in typical targeting years
To manage and save the UBW, we proposed that the government could store or input some winter wheat when more UBW needed to be exploited. The winter wheat production in 2005 was used to predict the quantity of UBW needed under the conditions of three typical years. Assuming that when the effective precipitation P = 25%, there is no need to supply UBW to the winter wheat reserve, because the precipitation is sufficient to ensure food security at this condition, the total quantity of irrigation water used in the year with P = 25% was considered as a threshold.
Figure 6(b) shows the results when supplying 5% of total winter wheat production (TWP). When P = 50% in the targeting year, the UBW quantity exploited is 0.98 times that of P = 25%. Therefore, the supply of 5% would be considered as the threshold in the year with P = 50%. When P = 75% in the targeting year, the UBW quantity exploited is 1.15 times that of P = 25%, which is lower than the threshold. Thus, 5% of TWP is not the targeting water-saving irrigation.
Figure 6(c) shows the results when there is 10% of TWP. When P = 75% in the targeting year, the UBW quantity exploited is 0.92 times that of P = 25%, which is lower than the threshold in the first-category to third-category areas. Therefore, 5–10% of TWP is considered as the proper value in the targeting year with P = 75%.
From the above results, we could reach three conclusions:
Considering the spatial variability of photosynthetically active radiation, soil organic matter and accumulated temperature, the SVM-GA coupling method could be used to effectively determine water production efficiency of the representative division at every irrigation site in North China Plain.
The model for evaluating the comparative advantage of UBW production efficiency was constructed considering the effective precipitation, irrigation efficiency and depth of UBW. The comparative advantage of UBW decreased from north and south to the center but slightly changed from east to west. It was lower in the center than in other areas, and higher in the north than in the south.
Combined with different supplies to ensure food security, the way to manage irrigation water in different typical targeting years was obtained. When P = 25% in the targeting year, there is no need to supply. When P = 50% in the targeting year, supplying 5% of TWP is needed and could save 15% of total UBW. When P = 75% in the targeting year, supplying 5–10% of TWP is needed and could save 11–28% of total UBW.