1 Introduction

Drought is a normal recurrent feature of climate, and is considered by many to be the most complex but least understood of all natural hazards, affecting more people than any other hazard (Hagman 1984). The data from Emergency Events Database (www.em-dat.net) show that, throughout the world, droughts account for 5% of the natural disasters, but losses from droughts have caused up to 30% of all losses from disasters, ranking droughts first among all the natural hazards. Furthermore, with the escalation of global climate change, the frequency of droughts and the areas affected are expected to increase sharply (NCAR 2005; IPCC 2001). Worldwide losses from droughts have increased significantly in tandem with the increased number and severity of droughts (Wilhite 2000). The high economic cost and social vulnerability to droughts has led to increasing attention to the issue of drought vulnerability in recent years (Wilhite 1993, 2000; Keenan and Krannich 1997; Downing and Bakker 2000).

Vulnerability assessment provides a framework for identifying the social, economic and environmental causes of drought impacts. It is one of the main aspects of drought planning and mitigation (Wilhite 1993, 2000; Knutson et al. 1998), and bridges the gap between impact assessment and policy formulation by directing policy attention to underlying cause of vulnerability rather than the negative effects which follow triggering events such as droughts (Ribot et al. 1996). With a map of drought vulnerability, decision-makers can visualize the hazard risk and convey vulnerability information to other sectors to ensure that they will act in a timely and effectively way to tackle drought-related losses (Wilhelmi and Wilhite 2002).

Beginning in the 1980s, the study of social, economic and environmental vulnerability to disasters has become an important subject, and many scholars have carried out research on this (Kates 1985; Blaikie et al. 1994; Keenan and Krannich 1997; Downing and Bakker 2000). Downing and Bakker (2000) argue that vulnerability is a relative measure, therefore critical levels should be defined; however, because of the complexity of the issue of vulnerability, assessments can be complicated and are not always well understood, especially for vulnerability to drought, which still lacks a universal definition and onset criteria (Wilhite 1993, 2000). Vulnerability to drought is dynamic and is influenced by a multitude of factors, including increases and regional shifts in population, urbanization, technology, government policies, land use and other natural resource management practices, desertification processes, water-use trends, and increasing environmental awareness (Wilhite 2000). The Western Drought Coordination Council (WDCC) has defined drought vulnerability as follows: Characteristics of populations, activities, or the environment that make them susceptible to the effects of drought. The degree of vulnerability depends on the environmental and social characteristics of the region and is measured by the ability to anticipate, cope with, resist and recover from drought (Knutson et al. 1998). Additionally, WDCC developed a series of tree diagrams to demonstrate potential drought impacts, from which true drought vulnerabilities can be identified and subsequently addressed. In recent years, scholars have actively explored the methodology and techniques for quantitative agricultural drought vulnerability assessments (Wilhelmi and Wilhite 2002; Shahid and Behrawan 2007; Bella et al. 2005). The general method involves identifying vulnerability indicators to construct a drought vulnerability index, and using GIS techniques to develop thematic maps. The Wilhelmi and Wilhite (2002) study provides the starting point for the examination of drought vulnerability, this was done by using a weighted combination of data on probability of moisture deficiency, capacity of the soil to hold water, land use and irrigated cropland with the classes within each factor being assigned a vulnerability ranking. In this study, the drought vulnerability for a given area (Nebraska State) was estimated using GIS tools and RS data. To determine drought vulnerability, the most important and most difficult task is to select the factors and to determine the weighting of those factors, which are commonly subjective and may vary between regions. The main reason for this problem is that the mechanism for determining how agricultural drought vulnerability is produced is still unclear.

In China, drought is a persistent climatic problem. Droughts occur with high frequency, affecting a wide range of areas, causing huge losses in agriculture, and are currently the most serious type of meteorological disaster in China. The average area affected by drought is about 21.593 million hectares annually, accounting for 60% of the total area in China affected by all types of meteorological disasters. The annual grain losses due to drought are up to 10 billion kilograms. A number of studies have been carried out on the impact of droughts on agriculture (Fu 1991; Pan et al. 1996; Fang et al. 1997; Wang et al. 2002), and the temporal and spatial patterns of drought (Li et al. 1996), but little work has been done on the issue of agricultural drought vulnerability. The previous studies on agricultural drought vulnerability were carried out by constructing an evaluation index system and evaluation model, which were usually conducted on a relatively small scale, considering counties and provinces (Shang and Shi 1998; Shang 1999; Shang 2000a, b; Shang et al. 2006; Liu et al. 2002, 2005; Wang et al. 2005a, 2006; Su et al. 2005). These results met the needs of regional agricultural drought management, but due to the weak understanding of the potential cause of drought vulnerability, the indicators that were selected are subjective and too complicated to be applied in larger areas. There is an urgent need to construct a nationwide agricultural drought vulnerability map to enhance agricultural drought management.

The goal of this study is to determine key factors that contribute to agricultural drought vulnerability and to construct a grid-cell-based agricultural drought assessment model for major food crops (wheat, corn and corn) using a GIS environment. The advantage of this grid-cell analysis approach is that results are spatially continuous. The spatial variance of agricultural drought vulnerability in China will be analyzed using a 10-km grid-cell scale. Furthermore; we will explore an effective method to reduce the agricultural drought vulnerability, providing scientific guidance for the defense and mitigation of agricultural droughts, and protecting food production from drought.

2 Data and methodology

2.1 Methodology

2.1.1 Agricultural drought vulnerability indicator selection and calculations

A holistic drought vulnerability index should take into account the ecological, socio-economic and production conditions. However, indicators for all these conditions are not readily obtainable and/or quantifiable in China. After careful consideration of previous studies on drought vulnerability and obtainable and quantifiable socio-economic and physical indicators, three indicators were identified as important and pertinent to this study and were subsequently selected to represent the vulnerability to agricultural drought. These indicators are climate, soil and irrigation represented by seasonal crop water deficiency, soil water-holding capacity and irrigation availability, respectively. Based on the assumption that China’s agricultural drought vulnerability is affected by all three factors, we constructed the conceptual model for agricultural drought vulnerability assessment as follows:

$$ V = G(f(C),f(S),f(I)), $$
(1)

where V is the agricultural drought vulnerability, f(C) is the function of the climate factor, f(S) is the function of the soil factor; and f(I) is the function of the irrigation factor.

2.1.1.1 Climate

Agricultural drought becomes apparent when soil water decreases. The soil water condition was mainly determined by the balance between precipitation and evapotranspiration. In this study, we used the seasonal crop water deficiency (SCWD) to represent the climate indicator (Allen et al. 1998; Wilhelmi and Wilhite 2002). That is, without considering the irrigation condition, the difference between the crop water demand during the growing season and the precipitation during the crop growing season, as follows:

$$ {\text{SCWD}} = {\frac{{{\text{ET}} - P}}{\text{ET}}}, $$
(2)

where SCWD is the seasonal crop water deficiency, ET is the seasonal crop water use, and P is the precipitation during crop growing season.

Under well-watered conditions, ET was estimated using the mathematical model recommended by FAO (Allen et al. 1998), as follows:

$$ {\text{ET}} = {\text{ET}}_{0} \cdot {\text{Kc,}} $$
(3)

where ET0 is the potential evapotranspiration, which can be computed from meteorological data using the FAO Penman–Monteith method (Allen et al. 1998). This method overcomes shortcomings in the previous FAO Penman method and provides values more consistent with actual worldwide crop water-use data. It is popularly used and accurate for both arid and humid conditions. Studies carried out by Wu et al. (2005) and Yin et al. (2005) showed that it is also suitable for China. The model is as follows:

$$ {\text{ET}}_{0} = {\tfrac{{0.408\Updelta (R_{n} - G) + \gamma\, {\frac{900}{T + 273}}U_{2} (e_{s} - e_{a} )}}{{\Updelta + \gamma (1 + 0.34U_{2} )}}}, $$
(4)

where Δ represents the slope of the saturation vapor pressure–temperature relationship (kPa/°C−1), Rn is the net solar radiation (MJm−2d−1), G is the soil heat flux (MJm−2d−1), γ is the psychrometric constant (kPa/°C−1), T is mean daily air temperature at 2 m height (°C), e s is the saturation vapor pressure (kPa), e a is the actual vapor pressure deficit (kPa) and U 2 is the wind speed at 2 m height (m s−1).

In this model, the empirical parameters for solar radiation have been modified (Wang 1981; Wang et al. 1988; Zuo et al. 1993; Wu et al. 2005), making it more consistent with conditions in China. In this study, we used the modified model to calculate ET0.

Kc is a crop coefficient, determined from independent experiments by examining the ratio of ET–ET0 in various growth stages for a well-watered crop. Although the Kc values for main food crops (wheat, corn and rice) have been reported by FAO, additional research has been conducted on the Kc value of main food crops in China (Chen 1995; Liu et al. 1998; Lei et al. 1999; Liu and Pereira 2000; Sun et al. 2002; Zhang et al. 2002; Peng et al. 2007; Wang et al. 2005b; Liu and Kang 2006; Yang et al. 2006). Our study uses the modified Kc value for the three food crop (wheat, corn and rice) based on the above research.

At a regional scale, we use the following formula to calculate the area-weighted mean, ET, as follows:

$$ \overline{\text{ET}} = {\frac{{{\text{ET}}_{w} * \rho_{w} + {\text{ET}}_{c} * \rho_{c} + {\text{ET}}_{r} * \rho_{r} }}{{\rho_{w} + \rho_{c} + \rho_{r} }}}, $$
(5)

where ET w , ET c and ET r are the seasonal water uses of wheat, corn and rice, respectively; ρ w , ρ c and ρ r are the proportions of land under wheat, corn and rice compared with the total area of cropland in a single grid cell, the grid-cell size is 10 × 10 km.

Finally, the regional \( \overline{\text{SCWD}} \) is calculated using following model.

$$ \overline{\text{SCWD}} = {\frac{{\overline{\text{ET}} - P}}{{\overline{\text{ET}} }}}. $$
(6)
2.1.1.2 Soil

The available water-holding capacity (AWC) of soil is the volume of soil water that should be available to plants, which reflects the ability of different types of soils to buffer plants during periods of moisture deficiency. AWC is an important soil property and is commonly estimated as the difference in water content between field capacity and permanent wilting point with corrections for salinity, fragments and rooting depth (Scotter 1981). Studies show that AWC depends largely on soil porosity, which, in turn, depends on soil texture, structure, organic matter, bulk density and other parameters (Wilhelmi and Wilhite 2002; Marshall 1979).

The amount of available water that the soil can supply may be critical to sustain plants between rainfall events or periods of irrigation, as the soil effectively buffers the plant root environment against periods of water deficit (http://soils.usda.gov). Furthermore, the geographic pattern of soil water-holding capacity is also important for studying water stress in plants and is critical to water management planning for irrigation and dryland crops (Kern 1995; Klocke and Hergert 1990). For these reasons, we selected the AWC of soil as an important agricultural drought vulnerability indicator. Regions with low AWC values have a higher degree of agricultural drought vulnerability when compared with regions with high AWC values.

2.1.1.3 Irrigation

Irrigation is one important strategy to defend against and mitigate drought, and it is also important for improving crop yields, protecting water supplies, ensuring food security, increasing income and improving the ecological environment. Agriculture accounts for about 70% of all water use worldwide. Although the worldwide irrigated land accounts for only 18% of the total agricultural land, the grain output from irrigated land comprises 40% of the total grain output. In China, the spatial distribution of water resources does not match the distribution of cultivated land, water resources are distributed with 20% of the total in the north of China and 80% in the south, while the cultivated land is distributed with 62 and 38% in the north and south, respectively (Gao 2006). With this distribution pattern, irrigation is essential for agricultural production, and in this study, we assume that the access to irrigation effectively reduces drought impact and that irrigated lands have a lower drought vulnerability compared with rain-fed dry lands during short-term drought conditions.

2.1.2 Agricultural drought vulnerability assessment model

Based on the studies of scholars in China and internationally (Wilhelmi and Wilhite 2002; Shahid and Behrawan 2007; Bella et al. 2005; Liu et al. 2002; Shang 2000a, b), we set the weightings of the agricultural drought vulnerability factors using the following method (Table 1). Each of the selected indicators was divided into four classes, ranging from the lowest to the highest values. Then a weight was applied to each class using the value range from one to five, with level one relating to the lowest agricultural drought vulnerability, and level five the highest agricultural drought vulnerability. The choice of weights and the weight ranges was based on an informed assumption based on the relative contributions of each factor to the overall agricultural drought vulnerability. For example, compared with soil AWC, the climate condition may play a more important role in agricultural drought vulnerability, so the range of the climate condition weights are two to five, while the soil weights are one to four. Additionally, the availability of irrigation support is also given weights from one and four, which indicates the significant difference in the ability to withstand the lack of precipitation between dryland and irrigated land.

Table 1 Weighting scheme for assessing agricultural drought vulnerability in China
Full size table

Finally, we developed the agricultural drought assessment model as follows:

$$ {\text{DVI}} = W_{\text{awc}} + W_{\text{scwd}} + W_{\text{irr}} , $$
(7)

where DVI is the agricultural drought vulnerability index, W awc is the weighting of AWC, W scwd is the weighting of the seasonal crop water deficiency and W irr is the weighting of irrigation availability.

Based on GIS, a composite vulnerability map was generated by integrating the thematic maps of all indicators. The composite agricultural drought vulnerability index (DVI) for each single grid was calculated using Eq. 7. For example, for a single grid, if the W awc is 2, W scwd is 3 and W irr is 4, then the final DVI for this grid is 2 + 3 + 4 = 9. Finally, the value DVI was classified to four classes from high to low, identifying geographic areas with ‘low’, ‘moderate’, ‘high’ and ‘very high’ vulnerability. The natural break method was used to derive the classes. This method creates ranges according to an algorithm that uses the average of each range to distribute the data more evenly across the ranges. This method ensures that the ranges are well represented by their averages and that the data values within each range are fairly close together (Smith 1986).

2.2 Data and processes

2.2.1 Data

A large number of data were used in this study including meteorological data, crop phenological data, agricultural zoning data and crop distribution data, which were used to calculate the seasonal crop water deficiency. Nationally available soil AWC data and national irrigation data were also prepared for the computation of agricultural drought vulnerability. The sources of all data used in this study are listed in Table 2.

Table 2 Data Sources
Full size table

2.2.2 Data processing

To describe the spatial characteristics of agricultural drought vulnerability in China, using the Arc-GIS software package. Arc-GIS is an integrated collection of GIS software products that provides a standards-based platform for spatial analysis, data management and mapping. This study used a 10 km × 10 km grid cell as the computing units for the assessment of agricultural drought vulnerability in China. Consequently, all original data were processed into 10 km × 10 km raster data, using the following process: The daily meteorological (point) data were spatially interpolated using the ANUSPLIN interpolation method (Hutchinson 1991), to generate raster meteorological data with spatial resolution = 10 km. Using the same method as the analysis of climate, the ecological zoning of crops and the phenological data of three main food crops (wheat, corn, rice) were digitized in Arc-GIS to generate 10 km × 10 km raster data of crop phenology. The spatial distribution data for wheat, corn and rice were extracted from a 1:100 million vegetation map of China, and the area ratio of each crop in a 10 km grid cell was calculated using the method developed by Liu et al. (2003). The soil AWC data and the irrigation area data for China were extracted from the Global Profile Available Soil Water Capacity and Global Irrigation Area Map, respectively, and the extracted AWC data were reclassified into four classes using the natural breaks method.

3 Results

3.1 Spatial pattern of water deficiency ratio of principal crops in growing season

In this study, the seasonal crop water-deficiency ratio was used to track rainfall amounts related to crop phenological requirements and to provide a scientific basis for enhancing regional water use efficiency and developing water-saving agriculture. The 40 years average of SCWD for wheat, corn and rice during 1961–2000 was calculated using Eqs. 2–4, respectively. The digital maps at 10 km resolution for SCWD are shown below (Figs. 1, 2, 3). Crop distribution varies due to the climatic differences between northern and southern China. In the north, the weather is too dry and cold for rice cultivation, and the main food crop is wheat, while in the south it is warm and wet enough for the rice cultivation. Corn is widely distributed in China. According to the analysis results, the distributions SCWD for the main food crops all show a significant spatial variance. Generally, the SCWD values to the south of the line formed by the Qinling Mountains and Huai River and in the eastern part of northeast China are high, and SCWD values gradually increase from south to north and from east to west.

Fig. 1
figure 1

Spatial variation in seasonal moisture deficiency of wheat in China

Full size image
Fig. 2
figure 2

Spatial variation in seasonal moisture deficiency of rice in China

Full size image
Fig. 3
figure 3

Spatial variation in seasonal moisture deficiency of corn in China

Full size image

The areas with high SCWD for wheat are mainly distributed to the north of the line between the Huai River-Qinling Mountains, in northwest China and the Yunnan-Guizhou Plateau, which is located in southwest China. For the rice planted in the south of China, the rainfall during the growing season can typically meet the water requirements for rice, but in the north of China, rice has a relative high SCWD. The SCWD of corn is low in most area except for Xinjiang and the northeast of Inner Mongolia.

Using the 40 years average SCWD data for wheat, corn, rice and the plant area ratio for the three crops in each grid cell, the 40 years average crop SCWD was calculated using Eq. 6 and then classified into four levels from low to high (Fig. 4) at 10-km spatial resolution. As shown in Fig. 4, the SCWD shows an increasing trend from south to north and from east to west. The line along the Huai River-Qinling Mountains acts as a boundary, with the southern area having low SCWD values, as the rainfall can meet the crop water requirements in the growing season, while most of the northern area has high SCWD values, as the rainfall cannot satisfy the crop water needs during the growing season. The SCWD also shows significant variance from east to west, increasing gradually as a result of China’s terrain features that have three terrain steps. The areas with low SCWD values are mainly distributed to the south of Huai River-Qinling Mountains, the east of the Qinghai-Tibet Plateau and in the north of northeast China. The regions with high and very high SCWD values are mainly in the west of the Northeast Plain, and south of the Inner Mongolia Plateau, almost all of the Loess Plateau, Gan-Su Corridor, Xinjiang and Tibet. Finally, areas with moderate SCWD values are mainly located in the intermediate zones where between areas of low and high SCWD values.

Fig. 4
figure 4

Spatial variation in seasonal water deficiency of crops in China

Full size image

The spatial pattern of SCWD of major crops was consistent with the distribution of the aridity/humidity status of the land surface across China (Wu et al. 2005). In humid regions, precipitation can generally satisfy the crop water requirements in the growing season, while in arid and semiarid regions there is a large deficit between water requirement of crop and precipitation in growing season. The boundary between arid and humid areas often corresponds to the boundary between high and low SCWD areas, and this phenomenon is especially obvious in northeast China and the Qinling-Huaihe region.

3.2 Spatial pattern analysis of agricultural drought vulnerability

Using the agricultural drought vulnerability evaluation model established by this paper, the agricultural drought vulnerability was assessed at a 10-km grid scale. Figure 5 shows the spatial distribution of agricultural drought vulnerability in China, and there are apparent east–west and north–south differences. The low- and moderate-vulnerability areas are mainly distributed in the eastern and southern parts of China, while the areas with high and very high vulnerability to agricultural drought are mainly concentrated in the northern and western parts of China.

Fig. 5
figure 5

Spatial variation of agricultural drought vulnerability in China

Full size image

By integrating the drought vulnerability map with China’s agricultural zoning (edited by the committee of China’s agricultural encyclopedia, also shown on Fig. 5, we see that the low-vulnerability areas are mainly found in the center of the northwest region, the south of the Huang-Huai-Hai Region, the center of the southwest region, the north of the Middle-Low Reaches of Yangtze River region and the west of South China region, covering 25.2% of the total grain planting area. These areas are mainly described as humid and semi-humid areas, with abundant precipitation, low SCWD, high AWC, and have available irrigation from groundwater or surface water. Previous studies (Li et al. 1996; Fang et al. 1997; Pan et al. 1996; Wang et al. 2002) indicate that even if drought occurs frequently in these low-vulnerability areas, it does not typically result in serious disasters, which means that the reduced vulnerability in these areas effectively reduces the impacts of drought.

The moderate-vulnerability area covers 46.8% of the crop planting area and has the largest area of the four agricultural drought vulnerability levels. It occurs in the west of the northeast region, the north of the Huang-Huai-Hai Region, the north and south of the southwest region, the south of the middle-lower reaches of Yangtze River Region, and the east and west of the South China region. The moderate-vulnerability areas are more seriously affected by agricultural disasters. For example, the Huang-Huai-Hai Region is one of the most serious drought-affected areas in China (Li et al. 1996). In the moderate-vulnerability areas, conditions may be either humid to semi-humid or semi-arid. The humid and semi-humid areas are mainly located in the southwest, middle-lower reaches of Yangtze River, South China and northwest regions. These areas often have low SCWD and high AWC but irrigation is not easily available. For these areas, therefore, effectively reducing drought vulnerability should focus on the construction of water conservation measures to enlarge the irrigation area and enhance the emergency irrigation ability. The semi-arid conditions mainly occur in the Huang-Huai-Hai region, where high water demand crops of winter wheat and corn are common. This type of area is typified by high SCWD, but irrigation by ground water is available, and the AWC in this area is high. However, this area is experiencing a sharp decrease in ground water resources, and therefore reducing vulnerability should depend on not only water-saving irrigation and increasing input but also adjusting planting structures according to the capacity of the water resources under the present socioeconomic conditions and agricultural techniques.

The high vulnerability areas cover 17.3% of the total crop planting area and are mainly located in the Loess Plateau Region, the northwest and the southeast of the Inner Magnolia Region and small parts of the Gan-Xin and Qinghai-Tibet regions. These areas often are characterized by high or higher SCWD, low AWC and a lower irrigation ratio. The reduction in agricultural drought vulnerability in these areas should depend on both enhancing water conservation and water-saving techniques and also improving the utilization efficiency of water resources. The high vulnerability and frequent occurrences of drought in these areas mean that severe droughts may occur regularly.

The very high vulnerability area covers 10.7% of the crop planting area and mainly occurs in the Inner Mongolia Region, the northwest of the Loess Plateau Region, the southwest of the Qinghai-Tibet Region, and the majority of the Gan-Xin Region. The very high vulnerability of these areas is caused by many natural factors, including a xeric climate, frequent wind, high SCWD, poor soil quality, low AWC, and lack of large-scale irrigation due to water resources scarcity. Reducing vulnerability in this area should rely on inputs of knowledge, technology and financing, especially popularization of water-saving irrigation techniques. At the same time, this area is also a high ecological vulnerability area. To improve eco-environment areas those are not suitable for crop farming should be returned to livestock farming.

The risk of drought hazards mainly depends on variations in the climate conditions that will not change in the short term. Consequently, in high agricultural drought vulnerability areas, to effectively reduce the impacts of drought, the key factors should be identified to understand the mechanisms of vulnerability improve agricultural management and reduce agricultural drought vulnerability.

3.3 Analysis of agricultural drought vulnerability in different agriculture regions

As illustrated in Fig. 6 (the area ratio of different agricultural drought vulnerability levels in different agriculture regions), there are variations in the agricultural drought vulnerability between the different agricultural regions. The middle-lower reaches of the Yangtze River, South China, southwest, Huang-Huai-Hai and northeast regions are mainly low and moderate-vulnerability areas. A statistical analysis of the vulnerability factors in different agricultural districts found that the average SCWD in the South China, middle-lower reaches of the Yangtze River and southwest regions are −91.80, −69.10 and −48.80%, respectively. These areas have abundant rainfall, which can meet the needs of crop growth. The average SCWD in the northeast and Huang-Huai-Hai regions are 5.80 and 32.70%, respectively, but the irrigation rate is relatively high, which reduces the vulnerability of agricultural drought to some extent. The irrigation rate in the Huang-Huai-Hai Region is the highest at 88%. In addition, the AWC in all the above regions is relatively high.

Fig. 6
figure 6

Regional discrepancy of agricultural drought vulnerability in China

Full size image

Furthermore, within each region, variations in agricultural drought vulnerability are also apparent. For example, the vulnerability in the northern area of middle-lower reaches of the Yangtze River Region is lower than the southern part. Similarly, the eastern part of the northeast region has moderate vulnerability, while the central part has low vulnerability, and the northwest part has high vulnerability. The Huang-Huai-Hai Region shows significant north–south differences, as the southern area has low vulnerability, and the northern area is mainly moderate vulnerability. The low-vulnerability parts of the southwest and South China regions are mainly located in the central areas. The Loess Plateau, Inner Mongolia, Qinghai-Tibet and Gan-Xin regions are areas that have high and very high agricultural drought vulnerability in China. The high vulnerability area in the Gan-Xin region accounts for 75% of the region’s grain production area. The common feature of these areas is the high SCWD values, which in the Gan-Xin Region reaches 82.3%. The soil available water-holding capacity is typically low and is lowest in the Qinghai-Tibet area (166.20 mm). The irrigation rates are also low, at 1 and 8%, for Inner Mongolia and the Qinghai-Tibet Region, respectively. The Loess Plateau has the widest range of agricultural drought vulnerability. The vulnerabilities increase consistently from southeast to northwest (Table 3).

Table 3 Area ratio of different agricultural drought vulnerability levels in different agricultural regions and distribution of relative agricultural drought vulnerability factors
Full size table

4 Validations

Drought loss is a direct reflection of drought impact on agriculture. Fu Bojie studied the spatial characters of drought in China using agricultural loss data and figured out that the North Plain, the Loess Plateau, the northeast region and the Sichuan basin are the serious drought-affected areas (Fu 1991). Wang Jingai carried out similar study in China and point out that heavy drought area in northern China is relatively concentrated in the western part of Heilongjiang, the central of Inner Mongolia, the north of Hebei, the north of Shaanxi and Ningxia province; in the south are mainly distributed in the central five provinces (Anhui, Hubei, Hunan, Jiangxi and Henan) and the eastern part of Sichuan, Guizhou and Yunnan (Wang et al. 2002). Comparing these results to our study, we find that areas which were seriously affected by drought in the history are mostly located in the high and very high vulnerability areas.

5 Discussion and conclusion

In this study, we assessed overall agricultural drought vulnerability at a 10-km grid scale in China using a GIS-based agricultural drought vulnerability assessment model. The model was constructed by selecting three indicators (climate, soil and irrigation) which contribute to agricultural drought vulnerability. The detailed assessment included analysis of the spatial characteristics of SCWD and agricultural drought vulnerability in China.

The distribution of SCWD in China shows significant differences between the north and south of China, and between the east and west of China. The values for SCWD in the north and west are significantly higher than those in the east and south of China. The SCWD was highly consistent with the aridity/humidity status of the land surface in China.

Due to the variations in agricultural production conditions, obvious differences in agricultural drought vulnerability have been found both within and between different agricultural regions. The spatial distribution of agricultural drought vulnerability in China also shows apparent north–south and east–west differences, with the northern and western parts of China being far more vulnerable to drought than the southern and eastern parts. The areas of crop land with different levels of vulnerability, from low to very high, are 25.2, 46.8, 17.3 and 10.7%, respectively. The low- and moderate-vulnerability regions are mainly distributed in the east and south of China, while the high and very high vulnerability regions are mainly distributed in the north and west. The central part of southwest region, the area from the south of Huang-Huai-Hai region to the north of the middle-lower reaches of the Yangtze region, is the center of the low-vulnerability area in China. The high vulnerability areas are centered in the Loess Plateau, Inner Mongolia and Gan-Xin regions.

At a grid scale, this study constructed a simple model for agricultural drought vulnerability assessment, and using the simple model, we can study agricultural drought vulnerability moving from points to spaces and from small areas to large areas. The conclusions reached in this study could provide essential information to help address the issue of drought vulnerability in China and could also direct drought management strategies for mitigation purposes. Identifying regional vulnerabilities can lead to changing practices in water-dependent sectors and can help decision-makers to incorporate droughts into natural resource planning. For researchers, this study provides a scientific basis for further study of agricultural drought risk assessment.