1. Introduction
Drought occurs frequently in China and there has been a long history of these occurrences. From 206 BC to 1949, 1056 droughts occurred in China [
1]. From 1971 to 2016, the average annual disaster rates of droughts in Heilongjiang, Jilin, Liaoning, and Inner Mongolia Autonomous Region were 19.4%, 23.6%, 25.4% and 29.8%, respectively. The average annual disaster rates of droughts in Anhui, Hebei, Henan, Jiangsu, and Shandong provinces were 11.5%, 20.4%, 16.2%, 8.9%, and 18.3%, respectively [
2]. The Ministry of Emergency Management of the People’s Republic of China has notified that from July to November of 2019, droughts had affected a total of 1174 thousand hectares of crops in Jiangxi and Anhui provinces, resulting in a direct economic loss of 8.8 billion yuan [
3]. From January to April of 2020, 2.433 million people had been affected in 81 counties of 16 cities (prefectures) in Yunnan Province. A total of 662 thousand people had requested for life assistance due to droughts, and 534 thousand hectares of crops were affected, leading to direct economic loss of 1.41 billion yuan [
4].
Drought is considered as a slow-moving natural disaster that causes severe damage to water resources and to agriculture [
5]. The characteristics of drought include, but are not limited to, high frequency, long duration, and large area being influenced [
6]. Agricultural drought is a crucial part of drought and it refers to the situation where agricultural production is sensitive and vulnerable to drought stress [
7]. Agriculture utilizes natural resources directly and it is also a national anchoring industry. Agriculture is less capable of resisting and dealing with disasters. The resistance and handling capacity of agriculture to disasters is low so the adverse impact on agricultural production is most severe when drought occurs. In the same way, droughts can be intensified by poor land management [
8]. Therefore, the situation of agriculture and the extent of drought affect each other. According to the Assessment Report of the AR5 Climate Change 2014: Impacts, Adaptation, and Vulnerability: Vulnerability encompasses a variety of concepts and elements including sensitivity or susceptibility to harm and lack of capacity to cope and adapt [
9]. Taking the initiative via human activity is an effective way to alleviate the loss caused by a drought disaster [
5]. So, measuring agricultural drought vulnerability is a prerequisite for targeting interventions to improve and sustain the agricultural performance of both irrigated and rain-fed agriculture [
10].
Climate change has an increasing impact on production and people’s lives. In recent years, the topic of vulnerability to agricultural drought has gradually become the focus and research hotspot of scholars around the world.
Yi (2010) evaluated the agricultural droughts in Dalian, China. Ten evaluation indexes such as irrigation index, population density and proportion of paddy areas were selected [
11]. Yuan (2016) proposed a comprehensive index of regional drought vulnerability that includes exposure, sensitivity, and adaptability [
12]. The establishment of evaluation indicators cannot be applied to all since it is highly subjective to regional characteristics. However, different indexing systems provide more research possibilities in the field of drought vulnerability.
Yan (2012), Pang et al. (2013), Farhangfar et al. (2015), Liu et al. (2015), and others conducted quantitative evaluation on drought vulnerability of maize and wheat and obtained the severity and spatial changes of crops at different growth stages [
13,
14,
15,
16]. Kim et al. (2018) used multivariate statistical analysis method to assess the agricultural vulnerability to droughts in South Korea and the results showed that the Chungchongnam-Do area was most vulnerable [
17]. Lestari et al. (2018) used Arc GIS spatial overlay analysis to evaluate the agricultural drought vulnerability of Semarang Port City in India. The results showed that high vulnerability in six villages, medium vulnerability in seven villages, and low vulnerability in three villages [
18]. Based on super sufficiency DEA, Huang et al. (2019) evaluated the agricultural drought vulnerability of Hetao Irrigation Area in Inner Mongolia and the results showed that the drought vulnerability in the eastern part of Hetao Irrigation Area was much higher than that in the western part [
19]. Frischen et al. (2019) combined the result from spatial analysis of expert consultation and determined the drought vulnerability of Zimbabwe’s agricultural system. The results showed that the country’s drought vulnerability and the degree of impact vary greatly. The northern and southern part of Matabeleland, a province in southwestern part, have higher vulnerability level [
20]. Das et al. (2019) used Savitzky and Golay filtering methods to study the agricultural drought situation and vulnerability in India from 1982 to 2015. Results showed that the vulnerability of drought will continue to decrease over time [
21]. On the basis of selecting the research areas and constructing the evaluation index system, scholars have adopted different methods to evaluate the agricultural vulnerability to droughts. For example: Data envelopment analysis [
22,
23], analytic hierarchy process [
24,
25,
26,
27,
28], principal component analysis [
29,
30], entropy weight method [
31,
32,
33], etc. STATA [
34,
35], ArcGIS [
36,
37,
38] and other software have also been used to construct an evaluation model for quantitative analysis.
Rojas et al. (2011) and Zhang et al. (2016) used remote sensing technology to monitor and predict agricultural drought [
39,
40]. Guo et al. (2016) proposed a new method (vulnerability surfaces) for assessing vulnerability quantitatively and continuously by including the environmental variable as an additional perspective on exposure and assessed global drought risk of maize based on these surfaces [
41]. Chen et al. (2017) and Zeng et al. (2019) conducted drought risk assessment on Yunnan Province and Gansu Province respectively [
42,
43]. All the above studies have provided scientific methods for drought risk assessment and they have since enriched the assessment system for agricultural drought vulnerability.
Basing on a wide range of research areas and research methods, there exists the differences in the natural geographical environment, economic and social conditions, which has led to different influencing factors and various degrees of agricultural drought vulnerability. For example: Zarafshani et al. (2012) argued that the vulnerability of wheat farmers in the western part of Iran is mainly affected by economical, socio-cultural, psychological, technological, and infrastructural factors [
44]. Wu et al. (2017) believed that the water shortage rate and irrigation level in the growing season were the main factors affecting the vulnerability level of regional agricultural drought [
45]. Kamali et al. (2019) believed that the fertilization level is an important factor affecting the vulnerability of crop to drought in sub-Saharan Africa. Generally, countries with a higher food production index and better infrastructure perform better in terms of withstanding drought [
46].
To sum up, there are two methods namely qualitative research and quantitative research on agricultural drought vulnerability. Existing research on agricultural drought vulnerability in China mainly focused on certain regions for quantitative research [
7,
14,
32,
37,
45,
47,
48,
49,
50,
51]. There were only a few studies on the overall assessment of agricultural drought vulnerability and among those the research objects, conclusions and countermeasures are limited.
Therefore, this paper focuses on the agricultural drought vulnerability in China. Based on literature review and relative theories, the paper first constructs the vulnerability evaluation index system of agricultural drought. Then the paper uses entropy weight method, weighted comprehensive scoring method as well as k-means clustering algorithm to evaluate and categorize the vulnerability of agricultural drought in China. Finally, using the contribution model to analyze the influencing factors and the degrees of agricultural drought vulnerability in China, this paper proposes countermeasures to reduce agricultural drought vulnerability in China. In one aspect, the paper carries theoretical value for enriching vulnerability research. It is also conducive to a better understanding of drought conditions and influencing factors in various regions of China. In another aspect, the empirical analysis provides the basis for the government to formulate corresponding policies, to reduce losses caused by disasters, and to promote the sustainable development of agriculture in China.
2. Materials and Methods
2.1. Research Area Overview
The People’s Republic of China is located in East Asia and to the west coast of the Pacific Ocean. Liberated on 1 October 1949, China’s capital city is Beijing and the provincial administrative divisions are divided into twenty-three provinces, five autonomous regions, four municipalities, and two special administrative regions. China’s land area is about 9.6 million square kilometers. China is the world’s second largest economy, the world’s largest industrial country, and the world’s largest agricultural country. At the end of 2019, the total population of mainland China was more than 1.4 billion.
The terrain is high in the West and low in the East. Mountains, plateaus, and hills account for estimated 67% of the land area, basins, and plains account for around 33% of the total land area. The climate condition is complex and diverse.
Looking at the situation and distribution of China’s agricultural natural resources as a whole, the light and heat conditions are superior. However, there is a great regional differences of dry and wet conditions. The total amount of river runoff is large; however, the coordination and distribution of soil and water is not even. The absolute amount of land resources is large; however, the land occupied per capita is small. Agriculture still serves as the basic industry of China’s national economy.
2.2. Establishment of Indicator System and Data Sources
The establishment of evaluation index system is the prerequisite for evaluating agricultural drought vulnerability. Vulnerability is the root cause of drought disasters, which results from the interaction of natural environment and social economy system as well as the interactions of sensitivity and resilience in a certain space. Therefore, following the principles of science, comprehensiveness, pertinence, quantification, and availability of data [
47], we select two first-level indicators, namely, sensitivity and resilience and 12 second-level indicators to conduct an evaluation on 31 provincial administrative units (except for Hong Kong, Macao, and Taiwan) in China to establish an indicator system (as shown in
Table 1). The larger the indicator, the larger the vulnerability of agricultural drought. Hence, it is a positive indicator. On the contrary, it would be a negative indicator.
Sensitivity is the sum of all kinds of natural and social factors that would cause or aggravate drought and its impact on agricultural drought vulnerability is negative. That means the higher the sensitivity, the greater the vulnerability of agricultural drought. It includes agriculture in GDP proportion, multiple-crop index, rural population proportion, annual average temperature, annual sunshine duration, and annual precipitation.
Higher proportion of agriculture in GDP means that farmers rely heavily on agricultural income which is highly dependent on natural conditions. So the vulnerability of agricultural drought will increase. The higher the multiple-crop index, the more water the crop would need to grow. As a result, drought vulnerability will increase. The most severely impacted population at the time of drought is the agricultural population. Therefore, when the proportion of rural population increases, the degree of vulnerability will also increase. Moreover, higher the temperature and longer sunshine hours will lead to the increase of evaporation, and hence the agricultural drought vulnerability will increase together. Precipitation is the main factor affecting the growth of crops. The precipitation index can reflect the meteorological conditions of crops in this region and the impact of precipitation on vulnerability is negative.
Resilience refers to the ability of human society to prepare for, to respond to, and to recover from, disasters. It has a positive impact on agricultural drought vulnerability. That means the stronger the resilience, the lower the drought vulnerability. It includes forest coverage rate, net income per capita of rural residents, food production per capita, real GDP per capita, effective irrigation rate, and agricultural fertilizer per unit area.
The forest coverage rate reflects a country’s (or region) actual level of forest resources and forestry possession. Net income per capita of rural residents reflects the group of people’s economical ability to withstand and to resist drought. The higher the net income per capita of rural residents, the weaker the threats of agricultural drought. Food production per capita reflects the level of agricultural productivity. Real GDP per capita reflects the level of social and economic development. When the index is bigger, it means that the social and economic development level and the ability to withstand disasters is high. The effective irrigation rate reflects the degree of water conservancy and irrigation capacity. The increase of the amount of agricultural fertilizer per unit area is beneficial to enhance soil fertility, to improve soil structure and to increase the efficiency of land usage. The above indicators constitute the resilience of the agricultural system.
The agriculture in GDP proportion, the rural population proportion, the net income per capita of rural residents, the food production per capita, and the real GDP per capita affect the agricultural drought vulnerability from the economic and social perspectives. The multiple-crop index, the effective irrigation rate and the agricultural fertilizer per unit area affect the vulnerability of agricultural drought from the perspective of agricultural technology. The forest coverage rate, annual average temperature, annual sunshine duration, and precipitation affect the vulnerability of agriculture to drought from the perspective of natural conditions.
The indicator data in this paper comes from the website of the National Bureau of Statistics [
52] and the China Meteorological Administration [
53]. The annual precipitation, annual sunshine duration and annual average temperature are obtained from annual observations from 613 weather stations nationwide from China Meteorological Administration data network. In addition to the forest coverage rate, net income per capita of rural residents and real GDP (Gross Domestic Product) per capita can be directly obtained, other indicators need to be calculated. The descriptive statistical results of the complete sample are shown in
Table 2.
Table 1.
Index system and source of China’s agricultural drought vulnerability assessment.
Table 1.
Index system and source of China’s agricultural drought vulnerability assessment.
Indicators and Units | Calculation Formula | Source |
---|
Agriculture in GDP proportion (%) | Agricultural output value/GDP | [51,54] |
Multiple-crop index (%) | Cultivated area of crops/Total cultivated area | [49] |
Rural population proportion (%) | Rural population/Total population | [51,54] |
Annual average temperature (°C) | Annual average value of each meteorological station | [32] |
Annual sunshine duration (h) | Annual average value of each meteorological station | [55] |
Annual precipitation (mm) | Annual average value of each meteorological station | [51,54,56] |
The forest coverage rate (%) | Available directly | [56,57] |
Net income per capita of rural residents (yuan/per) | Available directly | [22,58,59] |
Food production per capita (kg/per) | Food production/Total population | [49] |
Real GDP per capita (yuan/per) | Available directly | [32,51,59] |
The effective irrigation rate (%) | Effective irrigation area/Total cultivated area | [31,56] |
Agricultural fertilizer per unit area (ton/hm2) | Amount of fertilizer used/Total cultivated area | [32] |
2.3. Data Processing
From
Table 1, each indicator has different dimensions; hence, direct comparison is not possible. Therefore, it is necessary to carry out the dimensionless standardization of each indicator. The positive and negative indicators have different influence directions on agricultural drought vulnerability so the treatment methods should be different.
Suppose there are provinces, years and evaluation indicators; then represents the indicator value of province in year . The normalized value after treatment is expressed as (0 < < 1). is the minimum value of the j indicator and is the maximum value of the j indicator.
2.4. Improved Entropy Weight Method
There are two methods to determine the weight: subjective weight method and objective weight method. This paper chooses the entropy weighting method (one of the objective weighting methods) for indicator weighting, which overcomes the subjective arbitrariness of the subjective weighting method and makes the weighting more scientific. The improved entropy weighting method has the following methods and steps [
60,
61]:
Build the matrix
:
Calculate indicator information entropy
:
Find indicator difference coefficient (redundancy)
:
The weight of each indicator
:
2.5. Vulnerability Assessment Model
This paper chooses the weighted comprehensive scoring method and uses
to represent the degree of vulnerability. The improved vulnerability assessment model of agricultural drought in China is as follows:
2.6. K-Means Clustering Algorithm
According to the above steps, to calculate the degree of vulnerability of the target year of China’s agricultural drought in various regions and put them in ascending order. After that, to use k-means clustering algorithm in Stata to grade the vulnerability of China’s agricultural drought disaster [
48,
62].
Algorithms usually use Euclidean distance to calculate the distance between samples. The calculation formula is as follows:
Suppose the class center of the
category is
, then the formula of
is updated as follows:
The clustering algorithm requires continuous iteration to re-classify and update
value. Whenever the maximum number of iterations has been reached or the objective function is less than the threshold value, the iteration ends. The objective function is as follows:
2.7. Contribution Model
The main contributing factors of agricultural drought vulnerability in China are analyzed by contribution model.
is the weight of the
criterion level indicator;
is the contribution degree of the
indicator factor to the vulnerability of the
evaluation object;
represents the contribution of the first level indicators to vulnerability;
is the weight of single indicator to total target;
is the indicator membership degree (that is to say the proportion of Single factor indicator accounts for in vulnerability results. In the obstacle degree model, the indicator deviation degree is the difference between the individual index factor evaluation value and 100%. Therefore, the factor membership in the contribution degree model is the single indicator factor evaluation value ratio 100%) [
32].