Dynamic Assessment of Drought Risk of Sugarcane in Guangxi, China Using Coupled Multi-Source Data

Abstract

Globally, drought is an increasing threat to agricultural ecosystems, resulting in impaired crop yields, high food prices, and low incomes for farmers. Fluctuations in crop production and prices can exert a negative transboundary effect on food exporting and importing countries across the world through international trade. Therefore, it is important to regionally assess agricultural drought risk to reduce crop yield reduction by adapting existing systems. In this study, from the perspective of Chinese sugar security, a comprehensive assessment index of drought risk of sugarcane was constructed by considering the atmosphere–soil–crop continuum. Based on disaster-causing factors (hazards) and exposure, vulnerability, and mitigation capabilities of disaster subjects (disaster bearers), a risk assessment model of drought disaster of sugarcane in the growing season was established. Results of this study were three-fold. First, the maximum entropy model accurately reflected the reliability and relative importance of the disaster-causing factors of vegetation condition index (VCI), soil moisture condition index (SMCI), and standardized precipitation evapotranspiration index (SPEI), with the area under the curve value of the comprehensive drought risk of sugarcane being greater than 0.75. Second, the drought frequency and impact range in four growth stages of sugarcane significantly declined with the increasing drought severity. Light drought was prevalent in each growth stage, and the occurrence frequency of severe drought was relatively low. The drought frequency was significantly higher in the seedling and maturity stages than in the tillering and stem elongation stages, and the drought distribution was mainly concentrated in the southwest and central regions. Finally, the spatial distribution characteristics of drought risk significantly differed among the four growth stages of sugarcane. The risk level in the seedling stage declined from the southwest to the northeast. The high risk in the tillering stage was mainly concentrated in the southwest and northeast of the study region. In the stem elongation stage, the southwest became a low- risk area. In the maturity stage, the risk level was higher in the southeast than in the other areas. As sugarcane is majorly planted on dry slopes with uneven rainfall, a lack of good infrastructure, and the further intensification of global warming, sugarcane areas that were highly exposed to drought stress were highly vulnerable to drought risk, which in turn weakens farmers’ willingness to plant, thus threatening the security of sugar and biofuel production.

1. Introduction

The sixth assessment report (AR6) of the Intergovernmental Panel on Climate Change (IPCC) stated that the average air temperature was 1.09 °C and 0.99 °C warmer during 2011–2020 and 2001–2020, respectively, than during 1850–1900 (pre-industrial) [1]. With the increasing global warming, extreme weather events show a frequent, widespread, strong, and concurrent trend. Additionally, the pressure of weather and climate extremes on food systems is also intensifying. Climate change is expected to increase rainfall and drought extremes, thus threatening food security in different regions of the world to varying degrees. However, higher global temperatures are directly proportional to the severity of the corresponding food crisis [2]. According to data from the Food and Agriculture Organization of the United Nations (FAO), agriculture is one of the industries that is the most sensitive to and severely affected by climate change, thus being at most risk. Though advancing rapidly, science and technology still cannot help agricultural production to completely adapt to uncontrollable weather and climate conditions. Agricultural production also cannot withstand devastating meteorological disasters, such as drought, flood, low-temperature damage, and hail. Among these, agricultural droughts threaten crops to varying degrees, reduce food production, and exert adverse economic impacts on human beings [3,4,5,6,7,8]. Between 1998 and 2017, the global economic loss due to drought was approximately 124 billion US dollars, and drought exposure has increased by 29% since 2000. Droughts have profoundly widespread and underestimated impacts on societies, agriculture, ecosystems, and economies, adversely affecting approximately 1.4 billion people and causing the loss of 1820 million Mg of grain (maize, rice, and wheat) between 2000 and 2019 [9,10]. From 1950 to 2016, the average annual agricultural losses in China exceeded 20,659,700 ha, of which drought-related losses were 16,302 million kg per year, accounting for more than 60% of the total losses from all types of natural disasters.

Drought-affected regions are changing subtly owing to global warming. The Amazon in South America, Great Plains in Central America, and southern China are affected by drought. In recent years, originally humid southern China has been suffering from drought, and frequent droughts have had an adverse impact on agricultural production in southern China [11]. The sugar industry is an important part of Chinese economy and is the pillar industry of Guangxi, with sugarcane as the main raw materials for the production of sugar and biofuels as energy substitutes. Globally, over 100 countries produce sugarcane, and Brazil, India, and China are the three largest sugarcane producers. Guangxi has the highest cultivation land and production of sugarcane in China, accounting for 60% of the total national area and 70% of the total national yield of sugarcane, which plays a decisive role in the development of Chinese sugar industry [12,13]. Compared with sugarcane croplands in other regions of the world, over 90% of the sugarcane fields in Guangxi are on dry slopes with thin soil and a lack of infrastructure, such as irrigation and drainage, often resulting in drought stress during the critical growth and development stages of sugarcane and reduced yields [14]. In particular, during 2009–2010, the sugarcane-growing region of Guangxi was affected by the worst drought event in history, which resulted in the lowest sugarcane production level in history. Previous studies have shown that the agricultural production risks due to climate change will become more severe in the future, especially in vulnerable areas. The area suffering from extreme drought in the 21st century is estimated to rise by 29% [15,16], with an occurrence probability of nearly 100% [17].

As a cash crop, sugarcane plays an important role in the agricultural economy by increasing the incomes of farmers and in turn improving their quality of life as well as those of residents. Ensuring the effective supply of important cash crops and promoting farmers’ income remain a major strategic issue for the national economic development and social stability of China. Drought disasters in production areas of major cash crops have become the main obstacle to stable and increased production of cash crops. Not only does a large gap exist between China and developed countries in the world in terms of research on disaster reduction and prevention of economic crops, but there is also a lack of technologically effective responses to disasters, such as risk assessment and management of meteorological disasters for major economic crops, resulting in heavy losses after disasters. Therefore, the development of a dynamic assessment of agro-meteorological disaster risk in the entire production process of major cash crops can minimize the risks and losses of meteorological disasters due to global climate change and ensure a high and stable yield of cash crops, which in turn provides a scientific basis for better understanding spatiotemporally dynamic drought risks and preventive and mitigative measures toward agricultural sustainability. In this context, accurately assessing the drought disaster risk of sugarcane in Guangxi is of important strategic significance for the sustainable and healthy development of the sugar industry in China, thus ensuring the effective supply of domestic materials and maintaining national economic security.

Risk can be defined as the probability of loss to humans or ecosystems [18]. Agro-meteorological disaster risk refers to the probability of agricultural production and crops being exposed to meteorological disasters of different intensities and being subjected to possible losses of varying degrees [19]. At present, theories on the formation mechanism of natural disaster risk mainly include the two-factor, three-factor, and four-factor methods. The two-factor theory states that disaster risk is the product of the comprehensive action of hazard (H) and vulnerability (V) in a certain region. Hazard refers to the possibility of the occurrence of disaster-causing factors, and vulnerability refers to the possibility of a given disaster subject (bearer) resisting damage and loss due to a given natural disaster [20]. The three-factor theory holds that disasters are related to not only disaster-causing factors and vulnerability, but also the extent of exposure (E) of a disaster bearer in a specific area to the risk factors. In other words, the more people and property that are exposed to the risk factors in the area, the greater the disaster risk is; therefore, more potential losses occur due to disasters [18,21]. According to the four-factor theory, in addition to the above three factors, emergency response and recovery capability (C) are the important factors that prevent and mitigate disaster risks. The higher the social awareness level of natural disaster prevention is, the more restricted the roles of other disaster-causing factors; thus, the risk factors of disasters weaken correspondingly [22,23]. From the above understanding of the risk mechanism and its application to the assessment of agro-meteorological disasters, we can infer that the drought disaster risk of sugarcane in Guangxi is the product of hazard, exposure, vulnerability, and mitigation capability.

Drought hazards and their risk formation processes depend on not only the causative factors of agro-meteorological hazard and the hazard formation process, but also on multiple drivers, such as the gestation environment, causative factors, disaster bearers, and mitigation capabilities. Therefore, among several assessment indices of agricultural drought, selecting a reliable drought index is a crucial prerequisite for the accurate identification of agricultural drought events. Because numerous factors and their multiple interactions affect regional drought, drought assessments using a single index only consider individual factors affecting drought and fail to achieve a comprehensive and objective assessment of drought risks [24,25]. To overcome these shortcomings and deficiencies, a multivariate drought index oriented toward the monitoring and assessment of drought disasters has become a research hotspot, and the comprehensive drought index (CDI) has distinct implications and is highly comparable [26,27,28,29]. Wang et al. [28] explored changes in maize tolerance to drought and flood events in Jilin Province based on the Palmer drought severity index (PDSI), normalized difference vegetation index (NDVI), and soil moisture index (SMI). Li et al. [30] simultaneously considered the vegetation–soil–atmosphere continuum and high-temperature conditions to construct a combined assessment model to analyze agricultural drought hazards in northeastern China. Bijaber et al. [31] built CDI according to standardized precipitation index (SPI), NDVI, land surface temperature (LST), and evapotranspiration (ET) to monitor drought events in Morocco.

Agricultural drought is a water shortage phenomenon for crops due to the water imbalance driven by the external environment, which hinders their normal growth and development [32]. Agricultural drought is a complex process involving multiple factors (e.g., atmosphere, crop, soil, hydrosphere, and anthroposphere) and their interactions [8]. Based on the four-factor theory of natural disaster risks, this study aimed to establish a method to identify and quantify the drivers of drought disaster of sugarcane by drawing on the multivariate assessment methods and composite indices currently used, which is derived from the soil–crop–atmosphere system, different growth stages of sugarcane, and multi-source data and technology, such as meteorological, remote sensing, soil, and other statistical data. The assessment model of drought risk of sugarcane was constructed to help sugarcane farmers to better use and distribute agricultural resources and timely avoid the threat of drought. The comprehensive assessment index of drought risk developed in this study overcomes the shortcomings of traditional drought indices. The introduction of the new concept of potential exposure has allowed for considering the potential planting range of sugarcane under climate change, providing a new basis for the exposure assessment of cash crops. Thus, this study provides a new perspective on and actionable insights into the assessment of meteorological disaster risks of major economic crops of China.

The specific objectives were as follows: (1) to determine the relative importance of the soil moisture condition index (SMCI), standardized precipitation evapotranspiration index (SPEI), and vegetation condition index (VCI) for the sugarcane-growing season and construct CDI for the risk factors of drought disaster for sugarcane in Guangxi; (2) to conduct a comprehensive exposure assessment of sugarcane considering its actual planting area and potential planting fields under climate change; (3) to integrate natural and socioeconomic conditions to quantify drought mitigation capability and indicators for sugarcane to provide a scientific basis for climate change mitigation; and (4) to quantify the four drought risk factors during the sugarcane-growing season and identify their spatial distribution characteristics.

2. Study Area and Data

2.1. Study Area

Sugarcane, also known as potato cane, sugar cane, and yellow-skinned fruit cane, is a raw material for the production of sugar as well as ethanol for use as an energy substitute [33,34,35]. The growth environment of sugarcane requires sufficient water and heat conditions, and the most suitable growth area is between 10° and 23° of the north-south latitude. Located in southern China (104°28′–112°04′E, 20°54′–26°23′N), with the Tropic of Cancer crossing the central part of the country (Figure 1), Guangxi Province shows a mean annual temperature of 17.5–23.5 °C and mean annual precipitation of 841.2–3387.5 mm, which comprise the suitable growth conditions for sugarcane [12,19,36]. However, under a typical subtropical monsoon climate region, Guangxi exhibits pronounced seasonal changes in precipitation owing to the alternating influence of winter and summer monsoons, with distinct dry and wet seasons. April to September is the rainy season, accounting for 70–85% of the annual precipitation, whereas October to March is the dry season, accounting for only 15–30% of the annual precipitation. Moreover, low mountains and hills are characteristic of the Guangxi terrain. Karst landforms are distributed in 83.9% of counties, accounting for approximately 51% of the total area of the autonomous region [37]. The sugarcane fields of the study region mainly occupy mountainous and semi-mountainous areas with low and slow hilly dryland slopes, shallow soils, low organic matter content, and nutrient deficiency. The cultivation layer is relatively shallow and typical of rain-fed dry farming [38].

2.2. Data and Processing

This study collected and collated meteorological, sugarcane planting area, soil, vegetation, economic, social, historical disaster, sugarcane yield, and other statistical data for Guangxi Province. Based on these data, the drought risks and disaster-causing factors of sugarcane were quantified to generate a risk map for Guangxi. All data were obtained from multiple sources, and the necessary characteristics are summarized in

Table 1. Information on the data types and data sources used in the study.

Type of Data	Variable of Data	Sources of Data (1990–2020)
Meteorological data	Daily precipitation, daily average temperature, daily maximum temperature, daily minimum temperature, relative humidity, water vapor pressure, 2 m high wind speed	National Meteorological Information Center (http://data.cma.cn/, accessed on 10 April 2022)
Remote sensing data	Normalized difference vegetation index (NDVI)	MOD13A3 NDVI dataset (https://ladsweb.modaps.eosdis.nasa.go11v/, accessed on 10 April 2022) GIMMS 3g verison1.0 NDVI dataset (https://www.noaa.gov, accessed on 10 April 2022)
	Soil moisture	Terra Climate dataset (4 km × 4 km) (http://www.climatologylab.org, accessed on 10 April 2022)
Agricultural production conditions, social and economic data	Per capita GDP, land use data	Resource and Environment Science and Data Center (https://www.resdc.cn/, accessed on 10 April 2022)
	Total power of agriculture machinery, Sugarcane-planting area	Guangxi Statistical Yearbook
	Distribution of machine-made sugar mills	Yi. M. [39]
Historical disaster data and yield data	Disaster and damage in Guangxi Province, sugarcane yield	China Meteorological Disaster Dictionary—Guangxi Volume China Meteorological Disaster Yearbook
Occurrence data	Sugarcane occurrence data	Global Biodiversity Information Facility (GBIF, http://www.gbif.org/, accessed on 23 April 2022), Chinese Virtual Herbarium (http://www.cvh.ac.cn/, accessed on 23 April 2022), Chinese Field Herbarium (http://www.cfh.ac.cn/, accessed on 23 April 2022), Flora China (http://www.iplant.cn/, accessed on 23 April 2022), and scientific literature from Ruan [40], Luo et al. [41]

. Finally, analyzed via ArcGIS 10.4, the raster data used had a spatial resolution of 1 km × 1 km.

2.3. Components of Risk

Based on the risk statements of the United Nations International Strategy for Disaster Reduction [21], the IPCC [22], and the risk formation mechanism of natural disasters, this study maintained that drought risk of sugarcane in Guangxi is a function of the interaction of the following four factors: hazard, exposure, vulnerability, and mitigation capability of hazard bearers as follows (Figure 2):

$$Risk=\left(Hazard\times Exposure\times Vulnerability\right)/Capability$$

(1)

2.3.1. Comprehensive Drought Index (CDI)

Agricultural drought occurs when limited water supply due to insufficient soil moisture cannot meet water demand of a given crop for its normal growth and development during its growing season. However, the flow of water from soil through plants to the atmosphere is a continuously changing and moving system of water flow. Water in soil flows to the root epidermis, which in turn absorbs and transports it to leaves through the xylem of roots and stems. The water vaporizes into water vapor via the pores between the leaf cells and diffuses through the leaf stomata to the air layer near the leaf surface, and finally, to the external atmosphere. Therefore, this study aimed to derive a comprehensive index of drought disaster risk of sugarcane from the atmosphere–soil–crop continuum. First, we identified and quantified the hazard factors of sugarcane by calculating the remote sensing indices of VCI and SMCI and the meteorological drought index of SPEI. Second, we used the maximum entropy (MaxEnt) model to determine the relative contributions of VCI, SMCI, and SPEI to the occurrence of drought disaster of sugarcane. We established the CDI for the four growth stages of sugarcane (i.e., seedling, tillering, stem elongation, and maturity) from 1990 to 2020 based on SPEI, VCI, and SMCI as follows:

$$CD{I}_{ij}=W_1\times SPE{I}_{ij}+W_2\times SMC{I}_{ij}+W_3\times VC{I}_{ij}$$

(2)

where $W_1$, $W_2$, and $W_3$ are the coefficients of the three hazard indicators of SPEI, SMCI, and VCI, respectively.

As a meteorological drought index, SPEI outperforms the other drought indices for drought effect detection on agricultural production [42,43,44]. The SPEI value is a standardized characteristic quantity representing drought, thereby reflecting the degree of drought that deviates from the values during a normal year. The smaller the value was, the stronger the drought degree and the higher the disaster probability of crop. Therefore, for the drought classifications of SPEI and CDI, we mainly referred to the previous classification standards and the actual situation of the study region (

Table 2. Drought classifications of comprehensive drought index (CDI) and standardized precipitation evapotranspiration index (SPEI).

Classification	SPEI	CDI
Near normal	$\left(-0.5,0.3\right]$	$\left(0,0.3\right]$
Slight drought	$\left(-1,-0.5\right]$	$\left(-0.3,0\right]$
Moderate drought	$\left(-1.5,-1\right]$	$\left(-0.6,-0.3\right]$
Serious drought	$\left(-2,-1.5\right]$	$\left(-0.9,-0.6\right]$
Extreme drought	$\left(-\infty ,-2\right]$	$\left(-\infty ,0.9\right]$

) as follows: (1) we followed the corresponding SPEI classifications of Wang et al. [29], Li et al. [31], and Riao et al. [45] and defined the CDI categories through the equal interval method; (2) according to the weights, spatial distribution characteristics, and fitting degree of SPEI, SMCI, and VCI, we further specified the categories of CDI; and (3) based on the statistical drought data for sugarcane over the years, we considered the relationship between CDI and drought disaster of sugarcane and improved the classification of CDI.

Calculation of SPEI

To integrate precipitation and evapotranspiration and reasonably assess drought on multiple timescales, we calculated a daily time series of drought events for sugarcane in Guangxi from 1990 to 2020 based on SPEI [42,43] and carried out drought detection for the growing season of sugarcane based on daily-scale SPEI. In this study, ${\mathrm{ET}}_0$ was estimated from the FAO Penman–Monteith equation proposed by Allan et al. [46].

Calculation of Soil Moisture Condition Index

As soil drying is the direct cause of agricultural drought, SMCI as an indicator of soil moisture deficit was used as an indicator of agricultural drought as follows:

$$SMC{I}^{\prime }=\frac{S{M}^{\prime }-S{M^{\prime }}_{min}}{S{M^{\prime }}_{max}-S{M^{\prime }}_{min}}$$

(3)

where $SM$ is the pixel value of soil moisture in the root zone; and $S{M^{\prime }}_{max}$ and $S{M^{\prime }}_{min}$ are the highest and lowest values of soil moisture, respectively, in each pixel of sugarcane during the different growth stages. The smaller the SMCI value was, the lower the water content.

Calculation of Vegetation Condition Index (VCI) and Normalized Difference Vegetation Index (NDVI)

As an important parameter, NDVI reflects crop growth and nutrient information and is widely used for monitoring and forecasting global drought and agricultural production. Therefore, VCI was chosen to identify and quantify the degree of environmental stress of sugarcane, and thus, the effect of drought on sugarcane as follows:

$$NDVI=\frac{NI{R}^{\prime }-R^{\prime }}{NI{R}^{\prime }+R^{\prime }}$$

(4)

$$VCI=\frac{NDVI-NDV{I}_{min}}{NDV{I}_{max}-NDV{I}_{min}}$$

(5)

where NIR′ and R′ are near-infrared and visible red bands, respectively; $NDVI$, $NDV{I}_{max}$, and $NDV{I}_{min}$ are the instantaneous, maximum, and minimum values of $NDVI$, respectively, of each pixel from 1990 to 2020. The period of VCI spanned 1990–2020. The smaller the VCI value was, the greater the drought impact.

2.3.2. Vulnerability

The risk level largely depends on vulnerability because the degree of drought impact is mediated by the vulnerability of the drought-affected agricultural ecosystem; that is, its sensitivity and adaptability to drought events affect the risk level [19]. Vulnerability spatially varies between and within regions and changes over time. Therefore, in this study, based on the growth characteristics and environmental conditions of sugarcane, a vulnerability assessment model of drought disaster of sugarcane was constructed by comprehensively considering the sensitivity and adaptability indices as follows:

$${V^{\prime }}_i={S^{\prime }}_i/{A^{\prime }}_i$$

(6)

where, ${V^{\prime }}_i$ is the drought vulnerability of sugarcane in growth stage $i$; ${S^{\prime }}_i$ is the drought sensitivity of sugarcane in growth stage $i$; and ${A^{\prime }}_i$ is the drought adaptability of sugarcane in growth stage $i$.

Sensitivity

Sensitivity refers to the degree of loss after disaster risk, which is determined by the physical characteristics of the disaster bearer itself and reflects the ability of the disaster bearer to resist the disaster-causing factors. In this study, sensitivity was mainly described as the drought impact on the sugarcane yield at the different growth stages as a function of drought intensity and yield loss rate. Drought intensity (DS′) and yield loss rate were calculated as follows:

$$D{S}^{\prime }={{\displaystyle \sum }}_{i=1}^{N}CD{I}_i$$

(7)

where $n$ is the number of years when drought occurs at a certain level; and $CD{I}_i$ is the CDI value of a certain level of drought during the growth season in the study period. The smaller the $D{S}^{\prime }$ value was, the stronger the drought intensity.

As crop yield is a combination of meteorological conditions, scientific and technological progress, management ability, and variety improvement, the actual crop yield was separated into meteorological yield due to short-term fluctuations in climatic conditions and trend yield due to improved technological development, which showed an upward trend, as follows:

$$Y_c=Y-Y_t+\epsilon$$

(8)

where $Y$ is the actual crop yield; $Y_t$ is the trend yield (calculated using the three-year moving average method); $Y_c$ refers to the meteorological yield; and $\epsilon$ is the random noise.

To ensure that the meteorological yield is unconstrained by time and geographical shadows with comparable results, the ratio of meteorological yield to trend yield was defined as the relative meteorological yield ($Y_a$) as follows:

$$Y_a=\frac{Y_c}{Y_t}\times 100\%$$

(9)

The negative $Y_a$ value indicates a year of yield reduction, and its absolute value is the yield reduction rate of the year. The average rate of yield reduction reflects the average yield reduction level of multi-year production. The greater the average rate of yield reduction was, the greater the impact of agro-meteorological hazards on the yield in the region, and the greater the risk of meteorological hazards.

Adaptability

Adaptation plays a key role in reducing vulnerability to climate change. Self-adjustment of adaptation in agricultural ecosystems occurs through evolutionary processes. In this study, adaptability was regarded as the effect of drought stress on transpirational water consumption of sugarcane as follows:

$${A^{\prime }}_i=\frac{E{T^{\prime }}_i}{E{T^{\prime }}_{0i}}$$

(10)

where $A{\prime }_i$ is the ratio of evapotranspiration to potential evapotranspiration of sugarcane during its growing season. The values of $E{T^{\prime }}_i$ and $E{T^{\prime }}_{0i}$ are the evapotranspiration and potential evapotranspiration of sugarcane in the first growing stage, respectively. The higher the ${A^{\prime }}_i$ value was, the stronger the sugarcane adaptability to interference.

2.3.3. Exposure

Exposure refers to all life and property that may be at risk of hazardous elements. For sugarcane in Guangxi, the larger the planting proportion was, the greater the potential yield loss was, and the greater the drought disaster risk in the region. The IPCC AR6 clearly stated that the global climate is warming, the temperature in high-latitude areas is rising, and the agro-climatic land suitability of crops is expanding. Sugarcane also has the potential to be planted at high latitudes [47]. Therefore, from the perspective of actual versus potential exposure, this study conducted a comprehensive exposure assessment of drought disaster of sugarcane given not only the current planting area of sugarcane, but also its potential planting range under climate change. The assessment index of actual exposure is the ratio of the actual planting area of sugarcane to the entire cultivated range, and potential exposure is characterized by the climatic suitability of sugarcane.

We established a comprehensive exposure index based on the actual and potential exposure as follows:

$$E_c={{\displaystyle \sum }}_{i=1}^n{W^{\prime }}_i{E^{\prime }}_i$$

(11)

where $E$ is the comprehensive sugarcane exposure; ${E^{\prime }}_i$ is the exposure index; and ${W^{\prime }}_i$ is the weight of the exposure index.

Actual Exposure

In this study, sugarcane was the main target of this disaster. The disaster-affected ecosystem was mainly the exposed sugarcane-growing area. The greater the planting range was, the greater the exposure risk.

$${R^{\prime }}_A=\frac{{A^{\prime }}_p}{{A^{\prime }}_c}$$

(12)

where ${R^{\prime }}_A$ is the ratio of sugarcane cultivation area; ${A^{\prime }}_p$, is the area of sugarcane planting; and ${A^{\prime }}_c$ is the cultivated land in all cities of Guangxi Province.

Potential Exposure

We considered the sugarcane growth conditions, geographical location, temperature, and precipitation and evaluated the agro-climate suitability of the sugarcane growth season from 1990 to 2020 based on the analytic hierarchy process. A total of five climate factors were identified in the study: days consistently above 25 °C; accumulation of daily average temperature at ≥20 °C; precipitation at ≥20 °C; extreme minimum temperature in winter, and average annual temperature [48]. To refine the climatic suitability zones, we unified the spatial resolution of the meteorological datasets at 1 km × 1 km based on interpolation via ANUSPLIN software developed by the Australian National University. The spatial interpolation of the meteorological factors was based on the theory of the thin-disk spline function and the introduction of multiple covariates, which significantly improved the interpolation accuracy.

2.3.4. Mitigation Capability

The mitigation capability was used to partly quantify the ability of the disaster-affected ecosystem to recover from it. The regional disaster mitigation capability is inversely proportional to the drought occurrence for sugarcane and closely related to both economic and social factors. This study considered the total power of agricultural machinery, per-capita GDP, distribution of sugar factories, and effective irrigation area. These elements were considered an important basis for reflecting the local mitigation capability. Drought mitigation capability was estimated as follows:

$$C_a={{\displaystyle \sum }}_{i=1}^n{W^{\prime }}_i{C^{\prime }}_i$$

(13)

where $C_a$ is drought mitigation capability; ${C^{\prime }}_i$ is the value of the evaluation index; and ${W^{\prime }}_i$ is the weight of the mitigation capability.

2.4. MaxEnt Model and Input

The MaxEnt model is a novel machine-learning algorithm for imputing habitat suitability of organisms, which exclusively uses the existing data of species distribution to simulate the interaction between species emergence and environmental variables [49]. Owing to its outstanding simulation ability, it has widely been used in the simulation and prediction of species distributions in recent years [50,51,52]. The model provides a self-testing function, which can automatically generate receiver operating characteristic (ROC) curves for model simulation and prediction, with high accuracy [53]. The MaxEnt model can gauge the relative contribution and necessity of influencing factors according to percentage contribution, importance score, and jackknife method. In addition, the model can quantify the comprehensive effect of each predictor by ascertaining the occurrence probability of the expected impacts at the prediction location.

In the MaxEnt model, the characteristic function of the environmental variables is linear, and we quantified the relative contribution percentage of the three hazard indicators (SPEI, SMCI, and VCI) to the comprehensive drought risk of sugarcane in Guangxi. We randomly selected 25% of the entire data of the disaster indicators of sugarcane in Guangxi as the testing (validation) dataset and the remaining 75% as the training dataset, and the accuracy of the simulation results was evaluated using AUC under the ROC curve. In general, when AUC > 0.75, the model simulation results are considered good [54].

2.5. Entropy Weight Method

The basic idea of the entropy weight method is to determine the objective weight according to the size of the index variability. Based on this principle, the smaller the degree of variation of an indicator was, the less the amount of information reflected and the lower its corresponding weight. The entropy weight method deeply reflects the distinguishing ability of indicators and leads to increased objectivity to determine a representative weight with a theoretical basis and high credibility. The algorithm is simple and practical and does not require other software.

2.6. Statistical Analysis

2.6.1. Mann–Kendal (MK) Test

The Mann–Kendal (MK) test is a non-parametric test for quantification, commonly used in the trend analysis of time-series data with non-normal distributions, and can analyze and detect the overall trend without computational delay. In the MK test, an upward or downward trend in the sequence is indicated if the values of the statistical forward ($U{F}_k$) and backward ($U{B}_k$) series are greater than or less than 0, respectively. When the series exceed the critical straight line, they indicate that the rising or falling trend is significant, and the range beyond the critical line is determined as the time zone of the sudden change. If the curves of $U{F}_k$ and $U{B}_k$ intersect between the critical lines, the corresponding time of the intersection point represents the beginning of mutation. This algorithm was implemented using Matlab software programming.

2.6.2. Pearson’s Correlation Analysis

In this study, Pearson’s correlation analysis was used to detect the linear relationships between SPEI, VCI, SMCI, CDI, and other indicators as well as between sugarcane yield volatility and drought risk. The coefficient of correlation (r) values between two variables vary between [−1, +1], and the greater the absolute value was, the higher the correlation.

3. Results

3.1. Model Validation and Important Relative Importance of Variables

3.1.1. Importance of Disaster Factors According to the MaxEnt Model

Given the simulation results of the comprehensive index of drought disaster risk of sugarcane, the AUC values of the model based on the training and testing datasets were higher than 0.75, indicating that the model predictions were reliable (Figure 3). Thus, we determined the weight values of the disaster-causing factors during the sugarcane growth stages. The effects of SPEI, SMCI, and VCI during the sugarcane growth stages on the drought occurrence for sugarcane varied significantly (

Table 3. Weight coefficients of the hazard indicators of the comprehensive drought index in the four growth stages of sugarcane.

	Hazard Index	SPEI	SMCI	VCI
Growing Stage
Seeding Stage		0.39	0.49	0.12
Tillering Stage		0.48	0.32	0.20
Stem elongation stage		0.51	0.36	0.13
maturity stage		0.41	0.49	0.10

). In the seedling and maturity stages, the highest contribution of the disaster factors to the comprehensive risk belonged to SMCI, followed by SPEI and VCI. In the tillering and stem elongation stages, the largest influence belonged to SPEI, followed by SMCI, and finally VCI.

3.1.2. Comparison of SPEI and SMCI with CDI

Having assessed the relative importance of the disaster factors based on the MaxEnt model, we compared several indices in pairs. Having assessed the relative importance of the disaster factors based on the MaxEnt model, we compared several indices in pairs. Figure 4 shows the time-series trend and annual mutation of SPEI, CDI, and SMCI of the growth stages of sugarcane in Guangxi. The overall trends of SMCI and CDI in the seedling and maturity stages were highly similar and showed a downward trend in the seedling stage but an upward trend in the maturity stage. In the tillering and stem elongation stages, the trends of SPEI and CDI were highly similar and showed a downward trend in the tillering stage but an upward trend in the stem elongation stage. In terms of SPEI, SMCI, or CDI, the $U{F}_k$ value was less than 0 during 2009–2011, indicating a significant decline. This result was because the once-in-a-century drought in southern China between 2010 and 2011 substantially reduced sugarcane production and increased sugar prices, thus resulting in a huge gap between sugar import and export in China. According to historical data, at the beginning of 2000, sugar imports in China remained at 1–1.5 million tons. The import volume sharply rose in 2010, reaching 2.92 million tons in 2011, a 65.3% increase relative to 2010, which broke the tariff quota for the first time. In 2012, sugar imports further surged to 3.75 million tons, 28.4% increase over the previous year, and 1.805 million tons were imported beyond the tariff quota. In the past decade under climate warming, the drought severity has risen, in particular in the seedling stage, with the $U{F}_k$ value was less than 0, indicating a significant decline. Since 2000, spring drought has occurred in Guangxi Province for the three consecutive years, resulting in a severe impact on sugarcane production, thus weakening farmers’ willingness to plant sugarcane and rapidly reducing the population of the young labor force.

3.2. Drought Hazard Assessment

3.2.1. Drought Frequency for Sugarcane According to CDI

The results of the drought classification of CDI are presented in Table 2. We calculated the number of repetitions of the drought classes at the grids for the four growth stages via ENVI 5.0, and the results are shown in Figure 5. The frequency of the light drought was 0.03–0.55 in the seedling stage and decreased from the southwest to the northeast. The high-frequency area was mainly concentrated in the southwest of the study region, including the Cities of Chongzuo and Nanning; in the west of the study region, including the City of White; and in the south of the study region, including the Cities of Beihai, Yulin, and Qinzhou. The low-frequency areas were mainly concentrated in Guilin, Hezhou, and Hechi. The frequency of the moderate drought was mainly concentrated in the west of the Chongzuo, and the east of the Fangchenggang and Beihai Cities. The moderate drought events in the other regions were relatively low. The high-incidence areas of the severe drought for sugarcane were mainly concentrated in Baise City, in western Guangxi.

Overall, the frequency of the light drought in southwest Guangxi rose in the seedling stage, whereas the western region was highly prone to severe drought. In the tillering stage, the west, south, and northeast of Guangxi Province exhibited a high incidence of light drought, whereas the drought frequency in the central region was relatively low. The high-incidence areas of moderate drought were mainly concentrated in the northeast and southwest, whereas the frequency of moderate drought in other areas was low. In the stem elongation stage, areas with a high incidence of light drought were mainly concentrated in the northeast. The overall frequencies of the moderate and severe droughts were low (<0.03). In the maturity stage, the high incidence of drought was extensive, and the high incidence of light drought covered the four major sugarcane-producing cities (Chongzuo, Nanning, Laibin, and Liuzhou). The areas with a high incidence of the moderate drought were mainly concentrated in the eastern areas of the study region, including the Cities of Beihai, Wuzhou, Yulin, and Hezhou. The overall frequency of the severe drought was relatively high in the north of the study region and low (<0.03) in other areas.

The drought frequency in the maturity stage was most significant, followed by those in the seedling, tillering, and stem elongation stages. Overall, drought occurred frequently in the major sugarcane-producing cities. Therefore, sugarcane farmers should synchronize the irrigation and key growth stages of sugarcane to avoid drought and its yield-reducing effect.

3.2.2. Drought Hazard of Sugarcane Based on the CDI

Figure 6 shows the spatial characteristics of the drought hazards in the growth stages of sugarcane. In the seedling stage, the drought hazard level decreased from the southwest to the northeast, and the high hazard areas were mainly concentrated in southwest Guangxi Province. The extremely severe hazard areas were mainly concentrated in the south of the Cities of Chongzuo, Nanning, Qinzhou, Beihai, and Yulin. The low hazard was mainly concentrated in the northeast, including the Cities of Liuzhou, Hechi, and Guilin, and the north of Hezhou City. In the tillering stage, the extremely high hazard occurred in the west, and the distribution of the extremely high hazards began to appear in the northeast. The overall hazard levels in the southwest and northeast were high. The distribution of the high hazard levels in Baise City was expanding. The light hazard areas were mainly concentrated in the central region. In the stem elongation stage, the risk of the entire province remained notably low, and most areas were below the moderate hazard level, with high hazard levels only in the northeast and northern marginal areas. The hazard level was significantly higher in the maturity stage than in the other growth stages which decreased from the southeast to the northwest of the study region. The extremely high hazards were mainly concentrated in the southeast, including the Cities of Beihai, Qinzhou, Laibin, Guigang, Yulin, Wuzhou, and Hezhou. The high hazard levels encompassed the three major sugarcane-producing cities (Chongzuo, Nanning, and Liuzhou) and extended to Guilin. The low and medium hazard levels were mainly concentrated in the northwest. Among the four growth stages, the seedling, tillering, and maturity stages in Guangxi Province were adversely affected by the drought disasters, with a higher risk. The hazard of the major sugarcane-producing cities remained relatively high, where special attention should be paid to changes in the water supply and sugarcane demand. Proper and reasonable management practices of irrigation should be adopted to ensure sustainable sugarcane yields in the study region.

3.3. Drought Vulnerability Assessment

3.3.1. Analysis of Sugarcane Sensitivity

Figure 7 illustrates the zoning map of the drought disaster sensitivity of sugarcane in the four growth stages in Guangxi. As shown in Figure 7, the sugarcane sensitivity showed a distinct spatial difference. In the seedling stage, the sugarcane sensitivity to the mild drought disaster in Guangxi was mainly concentrated in the marginal areas, and the sensitivity value of the Cities of Guigang and Qinzhou reached 0.19, indicating that the sugarcane yield was less adversely affected by the drought intensity in these areas than in the others. The sensitivity of the four major sugarcane-producing cities (Chongzuo, Nanning, Laibin, and Liuzhou) remained relatively high, with the highest value of 0.50, indicating that the potential risk of yield reduction was relatively high in the growth stage during drought. In the tillering stage, the low sensitivity was mainly concentrated in the main sugarcane-producing cities, whereas the high sensitivity was primarily distributed in the west and east. In the stem elongation stage, the sensitivity was distributed along one axis and two wings. Low sensitivity was concentrated in the area from the southwest to the northeast, whereas high sensitivity was scattered on both sides. In the maturity stage, low sensitivity was mainly concentrated in the east, including the Cities of Laibin, Wuzhou, Guigang, Wuzhou, and Hezhou and the southern part of the Liuzhou City. High sensitivity was mainly distributed in the west, including the Cities of Beihai, Fangchenggang, Qinzhou, Chongzuo, Nanning, Baise, and Hechi and the northern part of Liuzhou City.

3.3.2. Analysis of Sugarcane Adaptability

The sugarcane adaptability in Guangxi was highest in the stem elongation stage, followed by the tillering, mature, and seedling stages (Figure 8). In the seedling stage, only Baise City in northwest Guangxi showed higher adaptability than other regions. In the tillering stage, the high adaptation was mainly concentrated in the west and southwest, whereas the low adaptability was mainly distributed in the northeast. In the stem elongation stage, the area with high adaptability increased significantly, covering the southeast of the study region, whereas the area with low adaptability was concentrated in the northwest, whose distribution was relatively limited. In the maturity stage, the overall adaptability in the other regions was low, except for the southeast, including the Cities of Fangchenggang, Yulin, Qinzhou, and Beihai.

3.3.3. Analysis of Sugarcane Vulnerability

Figure 9 illustrates the zoning map of the drought disaster vulnerability of sugarcane in the growth stages. As shown in Figure 9, the distinct spatial differences appeared in the drought disaster vulnerability of sugarcane. In the seedling stage, the vulnerability was distributed along one axis and two wings. The extremely high vulnerability was mainly concentrated in the southwest northeast region, including the Cities of Nanning, Laibin, Liuzhou, and Hezhou. The high and moderate vulnerabilities were mainly distributed in the Cities of Fangchenggang, Chongzuo, Hechi, Beihai, Yulin, Wuzhou, and Guilin; the east of Baise City; and the north of Qinzhou City. The low vulnerability was mainly distributed in the west of Baise City, the south of the Qinzhou City, and most areas of the Guigang City. In the tillering stage, the extremely high vulnerability was sporadically distributed in the northeast of Guangxi, including the junctions of the Cities of Hechi and Liuzhou and of Liuzhou and Guilin, and the patchy distribution in the Cities of Hezhou and Wuzhou. The high vulnerability was mainly concentrated in the northeast, and the medium vulnerability was mainly concentrated in the middle of Guangxi. The vulnerability in the southwest was relatively low, including the Cities of Chongzuo, Nanning, Fangchenggang, and Beihai; the southeast of Baise City; and the southwest of Qinzhou City. In the stem elongation stage, the high vulnerability was mainly concentrated in the areas with high dimensions, including the north of Baise, Hechi, Liuzhou, and Guizhou Cities. The low vulnerability was mainly concentrated in the southwest and most areas of the Liuzhou City. In the maturity stage, the extremely high vulnerability was relatively widespread in the northeast, whereas the low vulnerability was mainly concentrated in the south and northwest of Guangxi. The vulnerability of the other regions remained mainly moderate and high. The overall sugarcane vulnerability in the mature stage showed a decreasing trend from the northeast to the southwest.

3.4. Drought Exposure Assessment

(1)

Actual exposure: Figure 10A shows the drought exposure distribution of sugarcane in Guangxi Province. The extremely high exposure was mainly concentrated in southwest Guangxi and Chongzuo City. The high exposure was primarily distributed in Nanning, Qinzhou, Beihai, Laibin, and Liuzhou. The moderate exposure was mainly distributed in the Cities of Fangchenggang and Guigang, south of Baise City, and west of Yulin City. The low exposure was mainly distributed in the eastern part of Guangxi, including the eastern parts of the Yulin and Guigang Cities; northern part of Hechi City; and the Cities of Hezhou, Wuzhou, and Guilin.

(2)

Potential exposure: Figure 10B shows the potential exposure distribution based on the climatic zone suitability of sugarcane. The potential exposure level presented a decreasing distribution pattern from low to high latitudes. The extremely high exposure was mainly concentrated in the south of Guangxi, and the high exposure was mainly concentrated in the center of Guangxi. Compared with the other classes, their distribution range was widest. The moderate exposure was mainly distributed in the middle of northern Guangxi. The low exposure was mainly concentrated in the northern edge of Guangxi because the heat in the northern part of Guangxi was low.

(3)

Comprehensive exposure: The weighted sum of the actual and potential exposure provided a comprehensive drought exposure of sugarcane in Guangxi (Figure 10C). The overall exposure trend was consistent with the actual exposure but not with the potential exposure due to climate change. The region covered by moderate exposure increased drastically, whereas the area of low exposure declined. In addition, with the continuous expansion of the sugarcane-planting area, the high-exposure area in the southwest expanded gradually.

3.5. Drought Mitigation Capability

The mitigation capability was assumed to reflect the ability of the different regions to prevent and buffer disaster damage when drought disasters of sugarcane occurred. The high-value area indicated that, when disasters occurred, the abilities not only to prevent and resist disasters, but also to recover from disasters, were relatively strong, whereas the low-value area indicated that both abilities were weak.

Figure 11 shows the spatial distribution of the mitigation capability in Guangxi. Owing to the differences in social and economic production conditions in various regions, disaster prevention and mitigation capabilities showed spatial differentiation. The higher the distribution density of mechanized sugar plants was, the higher the investment and technical support by local governments were. The mechanized sugar factories were concentrated in the main sugarcane-producing cities (Chongzuo, Nanning, Laibin, and Liuzhou). Per capita GDP showed a relatively scattered distribution and was relatively high in the southwest of the study region. The total power of agricultural machinery and the effective irrigation area were highly pronounced in the northeastern and central parts of the study region. The overall mitigation capability was strong from the southwest to the northeast and relatively poor in the east, where increased investment is required to enhance the drought resilience of the local sugarcane cultivation.

3.6. Drought Risk Assessment

Based on the analysis and zoning of the four risk indicators of the disaster-causing factors, vulnerability and exposure of the disaster bearer, and prevention and mitigation capability for drought disaster of sugarcane in Guangxi, the comprehensive index of drought disaster risk of sugarcane in Guangxi was calculated by coupling the risk mechanism model of natural disasters. The weights of the hazard, exposure, vulnerability, and mitigation capability estimated via the entropy weighting method were 0.3179, 0.209, 0.2245, and 0.2485, respectively.

The spatial distribution of the drought disaster risk of sugarcane in Guangxi is shown in Figure 12. In the seedling stage, the risk level decreased from the southwest to the north. The extremely high risks were mainly concentrated in Beihai City in the south of the study region and Chongzuo City in the southwest. The high risks were mainly concentrated in the Fangchenggang and Qinzhou Cities, south of the Nanning and Wuzhou Cities, west of Yulin City, east of Guigang City, and a small part of the Laibin and Liuzhou Cities. The moderate-risk area was mainly distributed in the middle of the study region, including the east of Baise City, the middle of Nanning and Laibin Cities, and at the junction of Wuzhou and Hezhou Cities. The study region was relatively weak in the high-latitude areas. In the tillering stage, the extremely high risks were mainly concentrated in the southwest and northeast of the study region, including the Cities of Chongzuo, Guilin, and Hezhou. The high-risk areas were mainly concentrated in the Qinzhou City, some areas of the Fangchenggang City, and southeast of Baise City. The risk in the central part of the study region was relatively weak. In the stem elongation stage, the extremely high risks were mainly distributed in the Cities of Qinzhou, Laibin, and Hezhou and the west of Hezhou City. Compared with the seedling and tillering stages, in the stem elongation stage, the southwest of the study region became a weak-risk area. In the maturity stage, the extremely high-risk level was mainly concentrated in Chongzuo, Laibin, Qinzhou, Guigang, Wuzhou, and Hezhou. The risks in Baise, Hechi, and Guilin Cities were relatively weak. Overall, Chongzuo, the largest city for sugarcane production, showed a relatively high risk in all the growth stages, except for the stem elongation stage, because the vulnerability of the exposed agricultural ecosystem was a component of this risk. This result showed that a drought disaster may lead to large economic losses.

3.7. Verification of Risk Assessment Model

Sugarcane growth mainly depends on natural precipitation, typical of rain-fed agriculture. Guangxi accounted for more than 60% of the sugarcane production and cultivation land in China. Therefore, this study analyzed the linear relationship between the yield reduction rate during the growing season and the drought disaster risk (Figure 13). The results showed that the increased drought risk reduced the yield anomalies. In general, the degree of drought risk was significantly negatively correlated with the yield anomaly (p < 0.001). The seedling emergence of sugarcane occurred in March and April when drought posed a relatively weak threat to sugarcane production. The impact of the drought disaster risk on the yield anomalies was significantly higher in the maturity stage than in the other growth stages. The maturity of sugarcane differed from that of other crops as sugarcane was divided into technological and physiological maturity, which refer to the concentration process of sucrose in cane so as to meet the technological requirements of sugar processing, and the sexual reproductive maturity process of booting, flowering, and seed setting, respectively. As an important parameter for sugarcane harvesting, when the sugar content of sugarcane reaches 13%, and the gravity purity of sugarcane juice exceeds 80%, sugarcane can be harvested. The physiological maturity process of sugarcane exerts a negative impact on its yield and sugar content. Drought stress is the main limiting factor of sugarcane production. In other words, with the increased climate warming and extreme weather events, the drought risk of sugarcane in Guangxi may further rise. According to statistics, sugar production in China in 2021 will be 10.7 million tons, an upward trend compared with 2020. In China, sugarcane import volume reached 985,800 tons in 2017 and 1.77 million tons in 2021, indicating an increasing market demand for sugarcane. Unfortunately, in recent years, due to major factors, such as natural disasters, sugarcane production in China has not been ideal. Assessing the risk of the drought disasters is necessary for not only emergency drought relief, but also for improving sugarcane production efficiency.

4. Discussion

As the drought disasters reduce crop yields and farmers’ income, which directly adversely affect social and economic development, preventing and mitigating drought loss play an important role in ensuring healthy social and economic development [55,56,57,58]. Driven by the atmosphere–soil–crop continuum, we built a drought hazard model and assessed the comprehensive drought risk of sugarcane by considering exposure, vulnerability, and mitigation capability. The results showed that, under the different drought levels, the frequency of the mild and moderate droughts was significantly higher than that of the severe drought, which was consistent with the results of Chen et al. [59] and Chen et al. [60]. Overall, the drought disaster risk remained high in the seedling and mature stages of sugarcane (the seedling stage was from March to April, whereas the maturity stage was in November, corresponding to spring and winter, respectively). This result was in close agreement with the previous results [61,62]. The relative contributions of the disaster-causing factors (VCI, SMCI, and SPEI) in the growth stages to the comprehensive drought disaster risk differed significantly. In the seedling and maturity stages, SMCI exhibited the largest impact, whereas SPEI showed the largest impact in the tillering and stem elongation stages, and VCI yielded the smallest contribution regardless of the growth stage. Summer drought may promote the crop growth owing to the increasing solar radiation and did not exert a significant negative impact even when the lag effect was considered [61,63]. This result can partly be explained by the low contribution of VCI to the comprehensive drought disaster risk of sugarcane. In addition, the relatively low influence of VCI reflected the effect of the “returning farmland to forests” program to some extent [64]. Reduced precipitation in winter and spring and significantly increased solar radiation accelerated the water loss from the surface soil and canopies, resulting in severe soil water shortages in the dry season in winter and spring. This pattern also explained why SMCI had high values for sugarcane in the seedling and maturity stages, and SPEI exhibited the highest values in the tillering and stem elongation stages. The shortage of precipitation during the growing season may be related to the karst landform in Guangxi, which causes the surface water to quickly drain into the ground and aggravates the water shortage of the surface and shallow soil [65,66].

The sugarcane exposure assessment was used to understand whether sugarcane was changing its geographical scope, shrinking, or expanding. As climate change alters the suitability of sugarcane varieties and their spatiotemporal distributions, it is important for decision makers in sugarcane-growing areas to assess the potential land suitability for sugarcane planting. The results emphasized that the exposure level of sugarcane decreased from the southwest to the northeast of the study region. The overall exposure in low-latitude areas was high, which was consistent with the results of Su et al. [48]. With further global warming, areas where sugarcane is highly exposed to the drought impact may face increased challenges from the drought events. In particular, in areas where sugarcane employment does not bring enough income, sugarcane farmers’ willingness to grow it may even weaken, and the risk of livelihood transfer may increase. In addition, the drought may expose outdoor sugarcane farmers to dry heat. Given the trend of sugarcane potential exposure in a changing climate, we divided the period of 1960–2019 into two parts, with 1989 as the time node, and analyzed and compared the possible impact of climate warming on the sugarcane planting range in the later period, as well as the possible impact of drought on the spatial displacement of the planting range (Figure 14 and Figure 15). We found that, in the past 30 years, the risk value increased for the regions with the high climatic suitability of planting range in the seedling, tillering, and stem elongation stage; with the moderate climatic suitability of planting range in the seedling stage and entire growing season; and with the low climatic suitability of planting range in all the growth stages. On the one hand, the increased planting range of sugarcane with the onset of climate warming increased the accumulated temperature in the growth stage and reduced the cold risk of sugarcane; on the other hand, the expansion of the potential exposure area increased the drought risk of sugarcane. Thus, sugarcane farmers should take measures to combat drought and avoid arbitrarily expanding sugarcane cultivation.

The spatial characteristics of the drought disaster vulnerability of sugarcane in the four growth stages in Guangxi differed significantly from one another. This result was due to the different adaptabilities and sensitivities of the growth stages. The latest IPCC evaluation report pointed out that climate change poses a serious risk to climate-sensitive areas, ecosystems and economic sectors, and high-exposure/vulnerability and low-adaptation areas. In this study, the sugarcane seedling stage was most prominent in terms of vulnerability given the weak adaptability and high sensitivity of the main sugarcane-producing cities. In other words, the high sugarcane vulnerability in the seedling stage was mainly concentrated in the sugarcane-producing cities. In general, the drought impact on the sugarcane fields was the product of the combined hazards of the drought-causing factors and agricultural drought vulnerability. Reducing the drought vulnerability of sugarcane remains an important method to reduce the negative effects of drought on sugarcane. If timely drought-resistance measures are not implemented, the yield loss rate of sugarcane will continue to rise, and the drought risk will increase due to further drought attacks and the lack of enhanced adaptive and mitigative capability.

Government investment and regional development will alleviate the threat of drought to the sugarcane cultivation. The high distribution density of mechanized sugar plants in Guangxi led to better development history of the sugarcane industry in this area. In general, considering the freight and freshness of sugar, it is reasonable that processing plants, raw materials, and markets remain relatively proximate to the fields, which also conforms to the idea of “Agricultural Location Theory”. Not only were mechanized sugar farms in Guangxi mainly concentrated in the major sugarcane-producing cities, including Chongzuo, Nanning, Laibin, Liuzhou, and some urban areas south of the study region, but the main production areas also exhibited enhanced mitigation capability. However, the effective irrigation level was not significant in the major production cities. As more than 80% of sugarcane in Guangxi is planted on slopes that rely on rainfall, so the drought frequency and limited access to irrigation water are the key limitations. Overall, the main sugarcane-producing cities bore a significantly higher drought risk regardless of the growth stage, except for in the stem elongation stage, compared with the other regions, as they received the largest share of the sugarcane fields. Therefore, local governments should clarify the irrigation needs of the sugarcane-planting areas and sustainably manage water resources to meet the irrigation demand of sugarcane. For example, water management strategies to combat drought include the use of all available conditions, such as irrigation via wells, rivers, reservoirs, artificial sprinklers, drench, no or reduced defoliation of sugarcane, and covering sugarcane rows to reduce water loss through evaporation.

For areas with the high drought risk of sugarcane, interventions, such as artificial rainfall, environmental protection, and the improvement of microclimate, can be adopted as the management, prevention, and mitigation technologies. For areas with high vulnerability, the adaptability of crops can be improved through scientific management methods, such as planting drought-resistant and high-temperature-resistant varieties, reasonable intercropping, adjusting sowing time, timely irrigation, timely fertilization, spraying chemical agents, and improving soil conditions and water use efficiency of crops. Crop planting areas with high disaster exposure can reasonably be arranged according to the principle of maximizing benefits and minimizing disasters. Planting can be conducted in areas with high climatic suitability and with few disaster events instead of high-risk areas. Furthermore, the planting time can be adjusted to avoid high-risk periods. To enhance capability for disaster prevention and mitigation, improvements should be made to investment in technology, infrastructure, scientific response measures, drought relief plans, and facilities for disaster prevention and mitigation. Improving the information-sharing channel, adjusting the industrial structure, strengthening the construction of service systems for drought relief, enriching the quality of meteorological services, cultivating public awareness of disaster prevention, and last but not least, establishing the sharing and transfer mechanism of disaster risk through insurance and social security are the most likely methods to realize the reduction, control, and comprehensive management of the drought stress and risk of sugarcane. In addition, the subsidy policy in different regions of Guangxi can be leveraged to improve the willingness and efficiency of planting sugarcane according to the risk level of drought.

This study has some limitations that should be addressed in the future studies. For example, the process-based crop or ecosystem model remains to be considered for assessing the drought disaster risk of sugarcane cultivation. Crop growth models can quantify the spatiotemporally dynamic processes of crop growth, development, grain formation, and yield as a function of meteorological, soil, and hydrological conditions as well as management practices. In addition, this study lacks simulations under the future scenarios of climate change. We will consider the ecological and physiological processes of sugarcane in the context of climate change and drought based on the process-based biogeochemical models in the future.

5. Conclusions

This study focused on detecting the spatiotemporal distribution characteristics of drought disasters in the major sugarcane-growing region of China according to the risk theory of natural disasters. Based on the four-factor method of the risk theory, a comprehensive risk model was built to assess drought disasters in the sugarcane-growing area of Guangxi. The main conclusions of this study are as follows:

(1)

Using the MaxEnt model, the relative contributions of VCI, SMCI, and SPEI to the comprehensive drought risk of sugarcane and the weight values of the three hazard indicators were determined. For the comprehensive drought risk of sugarcane, the AUC values of the training and testing datasets of the model were higher than 0.75, indicating the reliable model predictions.

(2)

The spatial distribution characteristics of the drought frequency in the growing season were analyzed to determine the drought classes of sugarcane. The drought frequency significantly declined with the increasing drought severity, with a high frequency of mild drought and a low frequency of severe drought. The drought frequency in the seedling and maturity stages was significant, and it was mainly distributed in the southwest and main urban areas. The high drought incidence area in the stem elongation stage was mainly concentrated in the northeast of the study region, with no significant impact on the main sugarcane-producing cities.

(3)

As a function of hazard, vulnerability, exposure, and mitigation capability, according to the four-factor theory of disaster risks, a drought index was established to conduct a comprehensive assessment and zoning of the drought risk of sugarcane in Guangxi Province. The risk level increased from the southwest to the north in the seedling stage. In the tillering stage, the extremely high risks were mainly distributed in the southwest and northeast. Compared with the seedling and tillering stages, the stem elongation stage was at low risk in the southwest of the study region. In the maturity stage, the risk level increased in the southeast region. Chongzuo, the largest city for sugarcane production, was at relatively high risk regardless of the growth stage, except for stem elongation. This finding is because the major sugarcane-growing areas at greater risk of drought events often undermine the security of sugar and biofuel production.