Analysis of Future Drought Risk and Wheat Meteorological Disaster in Ningxia (Northwest China) Based on CMIP6 and SPEI

Abstract

In arid areas, droughts caused by climate change seriously impact wheat production. Therefore, research on spatial and temporal variability of dry and hot wind events and drought risk under different development patterns of future climate can provide a reference for wheat cultivation planning in the study area. Based on meteorological data under three scenarios of the CMIP6 (Sixth International Coupled Model Comparison Program) shared socio-economic path (SSP), we introduced wheat dry hot wind discrimination criteria and calculated the Standardized Precipitation–Evapotranspiration Index (SPEI). Future temperature changes within the Ningxia Province were consistent, increasing at a rate of 0.037, 0.15 and 0.45 °C·(10 a⁻¹) under SSP126, 245 and 585 scenarios, respectively. Simultaneously, average annual precipitation would increase by 17.77, 38.73 and 32.12 mm, respectively. Dry hot wind frequency differed spatially, being higher in northern Ningxia and western Ningxia, and lower in southern Ningxia and eastern Ningxia. During the wheat growing period, there is an obvious increasing drought risk trend under the SSP585 model in May, and the possibility of drought risk in the middle period was highest under the SSP126 model. In June, SPEI was generally higher than in May, and the risk of alternating drought and flood was greater under the SSP585 model, while near-medium drought risk was lower under the SSP126 and SSP245 models. The influence of DHW (dry and hot wind) on wheat yield will increase with the increase of warming level. However, when DHW occurs, effective irrigation can mitigate the harm. Irrigation water can be sourced from various channels, including rainfall, diversion, and groundwater. These results provide scientific reference for sustainable agricultural production, drought risk and wheat meteorological disaster forecast in inland arid areas affected by climate change.

1. Introduction

Drought is a major cause of economic, environmental and agricultural losses [1]. Many studies suggest that the effects of future global warming will intensify. The risk of various disasters caused by continuous drought has become a major threat to human beings [2,3,4]. While any high temperature, low humidity or high wind speed extremes can negatively impact agricultural production, a combination of these can more severely impact crop yields [5]. It has been recorded that the frequency of dry hot wind (DHW) in China has been increasing in recent years due to climate warming [6,7]. Therefore, appropriate methodologies are required to accurately assess likely future drought trends under different development models [8]. Dry hot wind (DHW) events can directly lead to wheat water imbalance, flower abortion and inhibit grain carbohydrate synthesis [9]. Furthermore, increased DHW frequency [10] may lead to late wheat seeding and shorter growing season [11,12], thus reducing crop yield per unit area [13]. Wheat is an important grain crop in the Ningxia irrigation area, where its output reaches 346,000 tons, accounting for 9% of the total grain output. DHW events mainly occur during the wheat mature growth stage [14]. To ensure stable wheat harvests and food security, it is critical to systematically assess the future frequency and spatiotemporal changes in DHW events.

Chinese research on DHW events began in the late 1950s [15], and by the early 1980s, the cooperative research group on the impact of DHW on wheat in north China made remarkable achievements in the DHW meteorological index and its damage mechanism. Previous studies have shown that the frequency of wheat DHW events in Ningxia has been increasing in the last 40 years, and the Z value of the trend test was greater than 0 at all experimental stations [16]. Additionally, Liu et al. [12] introduced wheat dry hot wind indices in the Ningxia irrigation area of the Yellow River diversion, while others explored ways to quantify DHW effects on wheat growth and yield through modeling and empirical studies [17]. Zhao et al. [18] found that average maximum temperature had the greatest influence on the key wheat phenological periods, followed by relative humidity and solar radiation, and the least sensitive climatic factors were precipitation, wind speed and reference crop evapotranspiration. Li et al. [9] found that spraying phosphorus and a sugar preparation on leaves alleviated the dry hot wind effect on winter wheat. Current studies are still limited to exploring DHW formation and mechanisms [19] and analyzing DHW spatiotemporal characteristics and changes in historical periods [12]. However, it is inadequate to simply combine local DHW changes with those caused by future global warming. Therefore, we combined these changes with CMIP6 data to analyze future DHW temporal and spatial changes in Ningxia during wheat growing period under different development paths, to provide references for prediction and prevention of DHW meteorological disasters.

Additionally, it is insufficient to discuss only DHW changes for the analysis of drought caused by future complex climate change. Quantification and classification of drought disasters are also important for prospective planning (risk analysis) and retrospective analysis [20]. As the need to understand weather and climate grows, so does the need for consistent quantitative procedures for extreme drought disaster mitigation and monitoring. Therefore, it is necessary to re-classify single-value drought by incorporating indices such as precipitation, temperature and evapotranspiration [20]. The standardized precipitation–evapotranspiration coefficient (SPEI) is based on precipitation and evapotranspiration, and has been widely used in drought disaster grade assessment research [21]. The reference evapotranspiration is calculated according to the Penman–Monteith formula [22]. SPEI uses a more comprehensive measure of water availability and climate water balance than the Standardized Precipitation Index (SPI). Therefore, drought disaster assessment has great potential [23], and many studies have assessed drought risk based on SPEI changes. Tan et al. used CMIP5 data combined with SPEI to analyze drought disaster risk in Ningxia Hui Autonomous Region under an RCP scenario [24]. Chen et al. used the Thornthwaite method and Penman–Monteith equation to calculate reference evapotranspiration, and calculated and compared the influence of the two methods on SPEI [25]. Li et al. used SPEI to analyze drought spatiotemporal evolution characteristics at different time scales in Yunnan Province [26]. Existing studies have proved that SPEI is suitable for evaluating drought risk levels over long time scales. However, to cope with future drought disasters and reduce their impacts, it is crucial to quantitatively predict their future risk, so we used SPEI to quantitatively assess drought disaster risk in Ningxia Province. Based on DHW and SPEI, drought hazard risk temporal and spatial changes under different development paths were analyzed, and especially during future wheat growing periods. Time variation characteristics were tested by Mann–Kendall, and GIS visualization spatial change characteristics were determined by the Kriging interpolation method. Using Global Climate Model (GCM) climate data and three representative shared socio-economic approach (SSP) scenarios, we predicted future drought hazards and their risk levels in Ningxia. This study’s innovation lies in its novel approach of using historical meteorological data as a reference to process future CMIP6 meteorological data through machine learning. This method not only aligns more closely with the actual meteorological changes in the study area, but also preserves the essential characteristics of these changes, thereby enhancing the data’s credibility. By analyzing the frequency of dry hot wind (DHW) events and the variations in the Standardized Precipitation–Evapotranspiration Index (SPEI) during the future period, this study aims to explore the impacts of drought risk on wheat production. This research not only provides valuable data support for drought risk management in other regions, but also offers a methodology that can be extended and applied to other areas, contributing to scientific references for sustainable agricultural production and drought risk evaluate in inland arid regions affected by climate change.

2. Materials and Methods

2.1. Study Area

Ningxia is located in northwest China, in the upper reaches of the Yellow River (between 35°14′–39°23′ N and 104°17′–107°39′ E). It is located on the western edge of China’s monsoon region and has a typical continental climate. Because of the barrier of Helan Mountain in the west, it retains the precipitation brought by the monsoon, and the abundant water resources brought by the Yellow River, so Ningxia lives up to the reputation of being the Oasis Beyond the Great Wall, and is one of the main production areas of spring wheat in China. However, the drought disaster caused by the typical high temperature and strong wind in northwest China has a great impact on the actual wheat production, so Ningxia Province is selected as the research area in this study. The northern part of the study area belongs to the plain area, located in the middle and upper reaches of the Yellow River, with good irrigation conditions. The central part is more arid due to low precipitation and high evaporation, and is far away from the irrigation water source. The southern part is at a higher elevation, with sufficient rainfall but lower temperatures, and the soil is more infertile. According to landform, climate, vegetation and agricultural production types, Ningxia can be divided into three agricultural ecological types: the Yellow River diversion irrigation area, the central arid area and the southern mountainous area (Figure 1). We selected the longitude and latitude coordinates of 12 major agricultural experimental sites in the research area as research test locations (

Table 1. Information of latitude and longitude of test sites in the study area.

Agro-Ecological Region	Site Name	Latitude	Longitude	Elevation (m)
Ningxia Yellow River diversion irrigation area	huinong	39.22	106.77	1091
	shizuishan	38.98	106.39	1101.6
	yinchuan	38.5	106.24	1111.4
	zhongwei	37.5	105.2	1225.7
	litong	37.98	106.18	1127.7
Arid area of central Ningxia	zhongning	37.48	105.68	1183.3
	yanchi	37.8	107.38	1347.8
	tongxin	36.96	105.9	1343.9
	huianpu	37.47	106.68	1512.3
Southern mountainous area of Ningxia	xiji	35.97	105.72	1916.5
	guyuan	36.02	106.24	1753.2
	haiyuan	36.57	105.65	1853.7

).

2.2. Data Sources

We obtained historical daily meteorological data from 2001 to 2019 from the National Data Center for Meteorological Sciences (http://data.cma.cn, accessed on 18 December 2024). Wheat yield data by region were acquired from the National Statistics Center of China (https://www.stats.gov.cn/, accessed on 18 December 2024) and the China Statistical Yearbook. Future meteorological data were derived from global climate models (GCMs) provided by the World Climate Research Programme (WCRP), encompassing three future periods: short-term (ST, 2021–2045), middle-term (MT, 2046–2070) and long-term (LT, 2071–2098). GCMs were developed under the sixth phase of the Coupled Model Intercomparison Project (CMIP6; https://pcmdi.llnl.gov/CMIP6/, accessed on 18 December 2024) sixth assessment report (CMIP6-IPCC-AR6) [27]. The sixth Assessment report proposes five new climate change scenarios based on shared socio-economic pathways (SSPs), namely, SSP119, SSP126, SSP245, SSP370 and SSP585. We solely focused on investigating representative future climate scenario models, including the low (SSP126), medium (SSP245) and high warming level (SSP585) scenarios. SSP126 is an optimistic estimate predicated on effectively controlling global warming levels, SSP245 scenario represents future climate projections for countries predominantly pursuing sustainable development, and SSP585 scenario captures the ramifications of unconventional advancements. Meteorological variables selected for analysis included precipitation (pr, mm/s), surface down-welling shortwave radiation (rsds, MJ/m²·day), surface wind speed (sfcWind, m/s), daily minimum near-surface recorded air temperature (T_min, Kelvin) and daily maximum near-surface air temperature (T_max, Kelvin). Temperature, rainfall and radiation data were obtained according to 0 °C = 273.15 Kelvin; 86,400 mm/s = 1 mm/day; 1 W/m²·s = 11.574 MJ/m²·day to convert to daily meteorological data. In the conducted study, we trained the machine learning model with Yinchuan 2015~2018 measured meteorological data and CMIP6 data, and used 2019~2022 CMIP6 data as the input of the trained machine learning model, and compared the results of eight machine learning methods on different kinds of meteorological data using the measured meteorological data as a reference. According to the results of the RMSE, lasso regression, support vector machine, neural network and random forest methods were selected to process RSDS, T_min, T_max and sfcWind data, respectively, in CMIP6 from 2020 to 2100. For detailed insights into the process and outcomes of applying machine learning methodologies to process CMIP6 data, please refer to the provided references [28]. The data for the selected GCM variables were all derived under an initial field of r1i1p1f1 (an experimental field with 1 for the number of runs, initial conditions, physical scheme and forcing data within the model).

2.3. DHW Discriminant Indices

According to Ningxia climatic characteristics and hazard nature [12], we determined the DHW indices. A maximum temperature tasmax ≥ 32 °C, minimum relative humidity hurs ≤ 30% and wind speed sfcWind ≥ 2 m/s on the day from the wheat filling stage to the maturity stage was deemed a light DHW day. Under these conditions, transpiration of wheat is enhanced and plant water loss is accelerated, affecting photosynthesis and nutrient accumulation, which leads to insufficient irrigation of wheat kernels and lower thousand-grain weight, ultimately affecting yield and quality. If tasmax ≥ 34 °C, hurs ≤ 25% and sfcWind ≥ 3 m/s, it was deemed to be a heavy DHW day. Such conditions can cause rapid water loss in wheat plants, curling or even drying out of leaves, leading to premature cessation of growth and withering of seeds before they are fully ripe, which not only affects wheat plumping, but may also lead directly to early maturity and green blight, resulting in severe yield loss. Meteorological factors for the spring and winter wheat filling period in Ningxia were analyzed to determine the number of DHW meteorological days during the wheat growing period.

2.4. Standardized Precipitation–Evapotranspiration Index (SPEI)

SPEI is a new drought index proposed by Vicente-Serrano et al. [29], based on precipitation and temperature data with the advantage of combining multi-scalar features with the ability to include the effects of temperature changes on drought assessment. SPEI combines the trend and multi-temporal nature of the SPI with the sensitivity of the Palmer Drought Index (PDSI) to changes in evaporation due to temperature fluctuations. SPEI reflects the climate water balance on different time scales, with positive values indicating wetness and negative values indicating dryness. Under global warming, only PDSI and SPEI can show that increased drought severity is related to increased evapotranspiration. However, compared to PDSI, SPEI has the advantage of multiple scalars, which is crucial for drought analysis and monitoring. Since the FAO-56 Penman–Monteith (PM) equation accounts for not only temperature, but also the effects of wind speed and relative humidity in calculating Potential Evapotranspiration (PET), it is considered more physically accurate than the Thornthwaite equation, which solely considers temperature. Consequently, for the calculation of the Standardized Precipitation–Evapotranspiration Index (SPEI), the FAO-PM equation is used instead of the Thornthwaite equation. We used CMIP6 meteorological data to calculate monthly SPEI; soil water-holding capacity is not taken into account in crop drought impact assessments. Since wheat fertility in the study area is affected by dry weather concentrated in May, June and July, this study focused on the analysis of SPEI in May, June and July, when DHWs occur in wheat. The specific value can be calculated by the following formula:

$$D=P_i-PE{T}_i$$

(1)

$$D_n^k={\displaystyle \sum _{i=0}^{k-1}{P}_{n-i}-PE{T}_{n-i}}$$

(2)

where D is the difference between potential evapotranspiration PET and precipitation P, k is the time scale (month) and n is the month of calculation. D^k represents the aggregate value of D on a time scale k. According to the Log-logistic distribution, the probability distribution function of D is given by the following formula:

$$F\left(x\right)={\left[1+{\left({\displaystyle \frac{\alpha }{x-\gamma }}\right)}^{\beta }\right]}^{-1}$$

(3)

where $\alpha$, $\beta$ and $\gamma$ are the scale, shape and origin parameters of the D value in the range, respectively. The SPEI value is calculated by:

$$SPEI=W-{\displaystyle \frac{C_0+C_{1}W+C_2{W}^2}{1+d_{1}W+d_{2}W+d_3{W}^3}}$$

(4)

where C₀ = 2.515517, C₁ = 0.802853, C₂ = 0.010328, d₁ = 1.432788, d₂ = 0.189269 and d₃ = 0.001308. The mean SPEI is 0 and the standard deviation is 1. SPEI is a standardized variable, so it can be compared with other SPEI values over time and space.

The correlation test for the two sets of observations [30] is described below. The statistic S is calculated according to the formula.

$$S={\displaystyle {\sum }_{i=1}^{n-1}{\displaystyle {\sum }_{j=i+1}^n\mathrm{sgn}(x_j-x_i)}}$$

(5)

where x_i and x_j are the sequential sequencing values, and n is the number of data sets.

2.5. WOFOST Model Optimization Method

The WOFOST model was jointly developed by Wageningen University and the World Food Research Center (CWFS) in the Netherlands. It is applicable to most crops by utilizing different crop parameters. The model’s calculations are primarily performed through three modules: climate, crop and soil. Using daily meteorological data, the model incorporates dynamic explanatory variables for growth, including soil, management and crop parameters [31]. This allows the simulation of crop growth under three scenarios: optimal conditions, water-limited conditions and nutrient-limited conditions. The previous study calibrated the crop parameters of the WOFOST model based on changes in measured wheat yields and leaf area index (LAI), and validated the simulation of potential wheat yields (with adequate water and nutrient supply) in conjunction with the soil and meteorological conditions of the experimental sites. Crop parameters for the WOFOST model are selected as outlined in

Table 2. WOFOST model crop parameter values.

Crop Parameter	AMAXTB0 (kg·hm−2·h−1)	AMAXTB1 (kg·hm−2·h−1)	AMAXTB1.3 (kg·hm−2·h−1)	CVO (kg·kg−1)	CVL (kg·kg−1)
Significance	Maximum CO2 assimilation rate at growth stage 0	Maximum CO2 assimilation rate at growth stage 1	Maximum CO2 assimilation rate at growth stage 1.3	Efficiency of dry matter conversion to seed	Efficiency of dry matter conversion to leaves
Value	35.25283831	39.0031	38.17566	0.765686	0.667321425
Parameter	CVS (kg·kg−1)	CVR (kg·kg−1)	EFFTB0 (kg·hm−2·h−1·J−1·m2·s)	EFFTB40 (kg·hm−2·h−1·J−1·m2·s)	KDIFTB0
Significance	Efficiency of conversion of dry matter to stems	Efficiency of dry matter conversion to roots	Single-leaf light energy utilization at 0 °C	Single-leaf light energy utilization at 40 °C	Scattered light extinction coefficient at growth stage 0
Value	0.72464208	0.667882	0.487338	0.476887	0.573684886
Parameter	KDIFTB2	SPAN (d)	SLATB0 (hm2·kg−1)	SLATB0.5 (hm2·kg−1)	SLATB2 (hm2·kg−1)
Significance	Scattered light extinction coefficient at growth stage 2	Leaf life cycle at 35 °C	Specific leaf area at growth stage 0	Specific leaf area at growth stage 0.5	Specific leaf area at growth stage 2
Value	0.654334798	33.02043	0.002004	0.001944	0.002072466
Parameter	TMPFTB0	TMPFTB10	TMPFTB15	TMPFTB25	TMPFTB35
Significance	Maximum CO2 assimilation rate discount factor at mean temperature 0 °C	Maximum CO2 assimilation rate discount factor at mean temperature 10 °C	Maximum CO2 assimilation rate discount factor at mean temperature 15 °C	Maximum CO2 assimilation rate discount factor at mean temperature 25 °C	Maximum CO2 assimilation rate discount factor at mean temperature 35 °C
Value	0.099834809	0.56683	0.937174	0.956712	0.015038988
Parameter	TBASE	TSUM1 (°C·d)	TSUM2 (°C·d)	TMNFTB0	TMNFTB3
Significance	Minimum temperature for seedling emergence	Accumulated temperature from seedling emergence to flowering	Accumulated temperature from flowering to maturity	Total assimilation rate discount factor at 0 °C minimum temperature	Total assimilation rate discount factor at 3 °C minimum temperature
Value	−1.530764949	1269.892	1263.679	0.078391	0.913047808
Parameter	LAIEM (hm2·hm−2)	FOTB1 (kg·kg−1)	CFET	FLTB0 (kg·kg−1)	FLTB0.25 (kg·kg−1)
Significance	LAI at seedling emergence	Proportion of above-ground dry matter allocated to storage organs	Evapotranspiration rate correction factor	Proportion of above-ground dry matter allocated to leaves at growth stage 0	Proportion of above-ground dry matter allocated to leaves at growth stage 0.25
Value	0.129033698	0.956577	0.929755	0.601276	0.641848876
Parameter	FLTB0.5 (kg·kg−1)	FLTB0.646 (kg·kg−1)	DEPNR	TDWI (kg·hm−2)	Q10
Significance	Proportion of above-ground dry matter allocated to leaves at growth stage 0.5	Proportion of above-ground dry matter allocated to leaves at growth stage 0.646	Number of soil water depleting crop groups	gross initial dry weight	Rate of change in respiration at a temperature change of 10 °C
Value	0.450768093	0.296593	4.648691	182.6195	1.870266321
Parameter	RDI (cm)	RRI (cm·d−1)	RDMCR (cm)	AMAXTB2 (kg·hm−2·h−1)
Significance	Initial root depth	Maximum daily increase in root depth	Maximum root depth	Maximum CO2 assimilation rate at growth stage 2
Value	11.94677171	1.290248	122.5433	4.84661

. Details on the optimization method and verification process of these model parameters can be found in Ref. [32].

Based on the revised WOFOST model and CMIP6 data, this study simulated the future growth and development of spring wheat under the climate change scenario year by year. For the detailed process and results of WOFOST correction and verification, and the simulation of spring wheat growth process year by year using WOFOST and CMIP6 data, see Ref. [28]. In this study, TRANSP and TWSO in the simulation results were used as the total transpiration of crops during the growing period, and TWSO was used as the yield. In addition, the change of growth period was based on the emergence date (DM), flowering date (FLWR) and harvest date (HALT) in the simulation results.

3. Results

3.1. Meteorological Changes Under Future Climate Scenarios

Meteorological variables (e.g., temperature, precipitation and wind speed) are important factors in climate change’s impact on wheat production. Therefore, we used CMIP6 data to derive the average monthly precipitation, evaporation and temperature of each test site in the study area from 2015 to 2100 under three future climate change models (SSP126, SSP245 and SSP585) (Figure 2). The CMIP6 data spatial distribution was closely related to the latitude and longitude of the test site. Temperature change of all test sites was consistent, and the trend of the three models was high in the south and west, and low in the north and east. The average temperature in the SSP245 model was the highest, about 0.2 °C higher than in the other two models. In MT, SSP585 > SSP245 > SSP126, and the difference was about 0.5 and 0.6 °C. In LT, SSP585 > SSP245 > SSP126, and the difference was about 1.7 and 0.8 °C. Except for the three sites in the southern mountain area, the average monthly rainfall was SSP585 > SSP245 > SSP126, and the difference was about 1 and 0.5 mm. MT average monthly rainfall was SSP585 > SSP126 > SSP245, and the difference was about 0.7 mm. For LT, SSP585 > SSP245 > SSP126, and the difference was about 0.5 and 2 mm. The average monthly precipitation in the northern region was about 17 mm, for Zhongning, Zhongwei and Litong sites in the central region about 21 mm, Tongxin and Huianpu sites about 34 mm, and for Yanchi was about 40 mm. Average monthly precipitation in the southern region exceeded 40 mm. Due to the temperature and precipitation increase, average monthly evaporation significantly increased from ST to LT under each model. In addition to the three sites in the ST southern mountain area, the average monthly evaporation is SSP585 > SSP245 > SSP126, with a difference of about 1 mm. In MT, SSP585 > SSP126 > SSP245, and the difference was higher than ST. In LT, SSP585 > SSP245 > SSP126, differences were >1 mm and <1 mm, and the southern region was more different than the northern region. Specific results are set out in

Table A1. Changes of near- and long-term average monthly temperature, precipitation and evaporation of each test point in the study area under SSP126, 245 and 585 models.

Research Site	Period	SSP126 Precipitation Monthly (mm)	SSP126 Evapotranspiration Monthly (mm)	SSP126 TasMean Monthly (°C)	SSP245 Precipitation Monthly (mm)	SSP245 Evapotranspiration Monthly (mm)	SSP245 TasMean Monthly (°C)	SSP585 Precipitation Monthly (mm)	SSP585 Evapotranspiration Monthly (mm)	SSP585 TasMean Monthly (°C)
A Guyuan	ST	46.69	38.61	8.07	46.68	37.86	8.32	46.63	38.58	8.17
	MT	49.79	40.61	8.26	49.23	39.82	8.93	49.32	40.54	9.41
	LT	48.99	40.08	8.30	51.81	40.82	9.12	52.18	42.30	10.79
B Haiyuan	ST	35.41	30.74	7.79	35.42	30.30	8.01	35.95	31.10	7.89
	MT	37.89	32.59	8.00	37.40	31.80	8.62	38.06	32.82	9.13
	LT	37.05	31.93	8.02	39.62	33.21	8.82	40.03	34.30	10.52
C Huianpu	ST	29.89	28.33	7.97	30.42	27.83	8.17	31.86	29.08	8.01
	MT	32.75	30.21	8.10	31.53	29.20	8.75	33.42	30.72	9.27
	LT	31.34	29.02	8.20	34.02	31.17	8.98	34.41	31.64	10.73
D Huinong	ST	15.80	15.12	7.40	16.86	15.94	7.62	18.00	17.01	7.40
	MT	17.52	16.61	7.46	16.95	16.16	8.12	17.97	16.96	8.71
	LT	17.32	16.41	7.54	18.13	17.18	8.43	18.95	17.95	10.26
E Litong	ST	20.05	19.49	7.94	20.50	19.44	8.13	21.79	20.59	8.01
	MT	22.06	21.06	8.10	21.31	20.24	8.71	22.95	21.72	9.28
	LT	20.87	20.11	8.15	22.98	22.01	8.97	23.36	22.49	10.75
F Shizuishan	ST	15.21	14.62	7.58	15.85	15.00	7.81	17.07	16.14	7.62
	MT	16.82	15.99	7.69	16.12	15.35	8.32	17.28	16.35	8.92
	LT	16.29	15.52	7.75	17.30	16.45	8.63	18.14	17.26	10.45
G Tongxin	ST	29.44	26.47	8.01	29.68	26.20	8.21	30.51	27.15	8.11
	MT	31.81	28.25	8.22	31.27	27.44	8.83	32.52	28.83	9.36
	LT	30.68	27.38	8.25	33.36	29.20	9.04	33.48	29.87	10.77
H Xiji	ST	45.79	37.92	7.64	45.46	37.13	7.90	45.55	37.79	7.76
	MT	48.56	39.86	7.86	48.13	39.08	8.50	48.17	39.73	8.99
	LT	48.02	39.43	7.87	50.76	40.09	8.69	51.21	41.65	10.35
I Yanchi	ST	35.45	32.94	7.70	36.27	32.44	7.93	37.79	33.75	7.67
	MT	38.81	34.96	7.74	37.23	33.97	8.45	38.96	35.11	8.95
	LT	37.76	33.98	7.88	40.09	35.61	8.69	40.83	36.30	10.44
J Yinchuan	ST	17.31	16.74	7.77	17.79	16.85	7.98	19.05	18.01	7.82
	MT	19.08	18.17	7.90	18.33	17.42	8.53	19.73	18.67	9.11
	LT	18.20	17.45	7.96	19.73	18.82	8.81	20.37	19.50	10.62
K Zhongning	ST	20.23	20.10	8.09	20.80	20.19	8.26	21.98	21.23	8.18
	MT	22.35	21.82	8.29	21.76	21.01	8.88	23.45	22.64	9.44
	LT	21.05	20.80	8.33	23.42	23.03	9.11	23.63	23.35	10.87
L Zhongwei	ST	19.22	19.30	8.04	19.99	19.64	8.21	21.04	20.58	8.04
	MT	21.37	21.07	8.23	20.80	20.32	8.82	22.23	21.74	8.23
	LT	20.38	20.25	8.26	22.28	22.08	9.05	22.73	22.57	8.26

.

The Ningxia DHW discriminating method was used to screen and derive statistics on future DHW days over the wheat growing period of the test sites in the SSP126, 245 and 585 model study areas. According to the regional distribution of wheat planting in Ningxia, Figure 3 shows the change trend of DHW in five representative test sites of Huinong, Yinchuan, Litong, Zhongning and Zhongwei. The frequency and intensity of DHW in the test stations located in the northern region are significantly higher compared to those in the central arid region. Under both the SSP126 and SSP245 models, the variation trend of DHW in the central arid region remains stable, while in the northern region, the DHW trend shows a slight increase over the long term. Test sites with the highest number of light and heavy DHW days in the three models during the 81 years from 2020 to 2100 were Yinchaun, with 359, 360 and 582 days, respectively. Under the SSP126 model, average annual DHW days were 3.67 in the southern mountainous area, 10.31 in the central arid area and 11.96 in the northern irrigation area. In the SSP245 model, the number of days were 2.98, 9.78 and 12.18 d, respectively, and in the SSP585 model, 6.54, 13.97 and 16.37 d, respectively.

In the study area overall, DHW occurrence days were closely related to geographical location, with more in the north and less in the south. Average annual DHW days were LT > ST > MT in the SSP126 model, and the differences between near- and long-term were small, all within 1 d. In the SSP245 model, average annual DHW days differed little in the near-, medium- and long-term, and LT was generally small. In the SSP585 model, average annual DHW days trended LT > MT > ST, and the difference was >1 d. Overall in the study area, the number of DHW days was highest in the SSP585 model, followed by SSP126 and finally SSP245. The difference of DHW days in the wheat growth period between SSP245 and SSP126 was only 0.26 d per year. The near- to long-term average annual DHW days of each site are listed in

Table A2. The average DHW days during wheat growing period were measured under SSP126, 245 and 585 models.

Dry Hot Wind Days (d)	Southern Mountainous Area of Ningxia			Arid Area of Central Ningxia				Northern Region of Ningxia
	guyaun	haiyuan	xiji	huianpu	tongxin	yanchi	zhongning	huinong	litong	shizuishan	yinchuan	zhongwei
SSP126 ST	3.47	5.47	2.67	8.27	6.60	6.97	10.57	13.83	10.63	13.53	12.53	9.57
SSP126 MT	2.23	4.69	2.08	7.04	5.42	6.31	9.42	12.54	9.15	12.85	11.27	8.19
SSP126 LT	3.80	5.53	3.10	8.43	6.23	7.70	10.93	15.57	11.10	15.13	13.20	10.37
SSP245 ST	2.17	3.23	1.63	6.40	5.00	5.67	9.07	13.63	9.30	13.17	11.57	8.17
SSP245 MT	2.96	4.46	2.08	7.73	6.46	6.81	10.77	14.19	11.58	14.81	13.27	9.19
SSP245 LT	3.43	4.53	2.30	7.90	5.70	6.57	9.50	15.60	11.30	15.50	12.57	8.93
SSP585 ST	3.77	5.20	4.13	7.90	6.27	7.30	10.20	13.60	10.17	14.00	12.37	8.73
SSP585 MT	5.50	7.42	6.50	11.81	8.73	11.46	14.58	19.58	15.73	19.85	17.92	14.27
SSP585 LT	7.90	10.37	8.07	13.60	11.03	12.70	17.13	22.30	17.67	22.07	20.17	17.17

.

3.2. Future Wheat DHW Temporal and Spatial Changes in Ningxia

Mann–Kendall analysis and statistical tests were conducted on the total number of DHW days from 2015 to 2100 and during the wheat growth period at each test site (

Table 3. Mann–Kendall change trend test of total DHW days and wheat growing days at each test site under SSP126, 245 and 585 models.

Z Value	Southern Mountainous Area of Ningxia			Arid Area of Central Ningxia				Northern Region of Ningxia
	guyaun	haiyuan	xiji	huianpu	tongxin	yanchi	zhongning	huinong	litong	shizuishan	yinchuan	zhongwei
SSP126	−0.37	−0.38	−0.03	−1.06	−1.28	−0.51	−0.64	0.15	−0.57	−0.14	−0.45	−0.36
confidence interval	-	-	-	-	-	-	-	-	-	-	-	-
SSP245	1.76	1.12	0.85	1.44	1.23	1.17	1.28	0.85	1.27	1.11	1.19	1.23
confidence interval	95%	-	-	90%	-	-	90%	-	-	-	-	-
SSP585	2.65	3.61	2.61	3.44	3.08	3.37	3.86	4.59	4.38	4.41	4.46	5.20
confidence interval	99%	99%	99%	99%	99%	99%	99%	99%	99%	99%	99%	99%

Note: “-” indicates no significance.

). The total number of DHW days in the SSP126 model decreased in all sites except Huinong. Under the SSP245 model, there was an obvious DHW day rising trend in Guyuan, Huianpu and Zhongning sites. Among them, the Guyuan upward trend was statistically significant, but for the other sites the trend was non-significant. In the SSP585 model, all sites showed an obvious and significant DHW day upward trend. A Z test score of >4 in the five sites in the northern region indicated that during the wheat growth period, there was a significant upward DHW day trend at each site.

We conducted Mann–Kendall tests to determine significant changes in DHW days at each research site under the three future climate models during the wheat growth period (Figure 4). Under the SSP585 model, DHW days during the wheat growing period all began to trend insignificant upward before 2020 in Guyuan, Haiyuan and Xiji, while the DHW days in LT show a significant upward trend; this transformation appeared around 2068a. In the arid areas of central, Huianbao, Tongxin, Yanchi and Zhongning, DHW days significantly increased after 2050a, 2083a, 2049a and 2050a, respectively. There was a very significant increase in LT, and this change occurred in 2047a, 2059a, 2047a and 2048a, respectively. In Huinong, Litong, Shizuishan, Yinchuan and Zhongwei in the northern region, DHW days significantly increased after 2052a, 2050a, 2052a, 2051a and 2051a, respectively. There was a highly significant increase in LT, and mutations occurred in 2054a, 2050a, 2052a, 2051a and 2048a, respectively. There was no significant upward or downward trend in the SSP126 and SSP245 models, and no significant mutations were detected. Overall, under the SSP585 scenario, the risk of DHW increases across the entire study area. The rate of increase in DHW risk significantly accelerates in the northern and central arid regions after the 2050s, while in the southern mountainous region, this change occurs after the 2070s.

We conducted a spatial analysis on DHW during the wheat growth period (Figure 5). There were significant spatial differences in DHW frequency, which was higher in the northern and western regions and lower in the southern and eastern regions. Overall, DHW frequency in the SSP585 model was significantly higher than in the other two models. For annual DHW days, MT was significantly lower in the SSP126 model. There is no significant difference between the three periods in the SSP245 model. In the SSP585 model, the spatial distribution was LT > MT > ST.

3.3. Drought Index Spatiotemporal Variation During the Wheat Growth Period

The identification of drought events and their severity characteristics are crucial for assessing and lowering drought disaster risk [33]. SPEI is derived from precipitation, evaporation and temperature changes in the form of water balance, and is currently a popular quantitative index for assessing drought degree and drought risk [34]. We used the Penman–Monteith equation to calculate future reference evapotranspiration in the study area, to calculate monthly SPEI [25]. According to the SPEI index drought classification criteria, 81a moderate drought months were identified (

Table 4. Number of drought risk months under three future climate models.

Future Climate Model	Drought Risk Level	Southern Mountainous Area of Ningxia			Arid Area of Central Ningxia				Northern Region of Ningxia
		Guyaun	Haiyuan	Xiji	Huianpu	Tongxin	Yanchi	Zhongning	Huinong	Litong	Shizuishan	Yinchuan	Zhongwei
SSP126	Moderate drought	103	109	106	80	95	109	68	79	80	75	83	74
	Severe drought	31	28	30	16	17	12	14	16	12	18	14	14
	Extreme drought	1	1	1	2	1	2	0	5	2	4	4	1
SSP245	Moderate drought	110	94	114	85	82	87	70	87	84	85	86	73
	Severe drought	24	18	26	12	15	11	16	17	16	20	18	17
	Extreme drought	2	4	2	0	3	0	1	2	2	3	4	2
SSP585	Moderate drought	93	102	101	80	89	89	84	91	91	93	95	72
	Severe drought	20	20	18	16	17	15	15	18	18	18	17	9
	Extreme drought	4	4	4	1	4	1	2	3	2	6	6	1
SSP126 (Growth period)	Moderate drought	22	25	22	25	30	34	28	19	27	19	25	24
	Severe drought	12	9	11	7	7	4	7	7	5	10	8	6
	Extreme drought	0	0	0	2	0	2	0	2	1	1	1	0
SSP245 (Growth period)	Moderate drought	32	31	31	33	33	32	26	28	33	29	29	28
	Severe drought	10	8	10	5	6	4	5	5	6	6	7	7
	Extreme drought	0	0	0	0	0	0	0	0	1	1	2	0
SSP585 (Growth period)	Moderate drought	37	39	40	31	36	33	31	26	34	24	29	26
	Severe drought	4	4	2	4	3	3	2	3	4	5	3	1
	Extreme drought	0	0	0	0	0	0	0	1	0	1	2	0

), and based on these, the study area drought risk was assessed as follows: southern mountain area > central arid area > northern region (SSP126 model); and southern mountainous area > northern region > central arid area (SSP245 and 585 models). For the wheat growth period in the study area, we selected the SPEI index from May to July of 2015 to 2100 for analysis. Drought risk was as follows: central arid area > southern mountainous area > northern region (SSP126 model); southern mountainous area > northern region > central arid area (SSP245 model); and southern mountainous area > central arid area > northern region (SSP585 model). Severe drought frequency was highest in the southern mountainous area, and extreme drought frequency was highest in the northern region. These results were consistent between the three future models.

For the wheat growth period in the study area, the SPEI index for May, June and July from 2015 to 2100a was used to examine temporal changes (Figure 6, Figure 7 and Figure 8). The SPEI index change trend at all sites was similar, with low and high values alternating periodically. In May, under SSP585 mode, the frequency and extreme value of high SPEI values in LT of each site were greater than ST, and the frequency and amplitude of low LTSPEI values were also significantly higher than MT and ST. There was a clear trend of increasing drought risk. In the SSP245 model, the frequency and amplitude of high and low values from ST to LT increased significantly, indicating that drought risk was becoming unstable. The frequency and amplitude of MT low values were greatest in the SSP126 model, and the possibility of MT drought risk was the highest. In June, SPEI values generally increased compared to May, and the frequency and amplitude of high and low LT in the SSP585 model were greater than ST and MT. In the SSP126 and SSP245 models, the frequency of high ST and MT was higher, and the amplitude of low values was smaller. In July, the frequency and amplitude of high values in the three models were greater than in May and June. The frequency and amplitude of low LT increased compared to June, but was still smaller than in May. The SPEI index in MT fluctuated less than ST and LT.

The optimized crop parameters were observed to effectively enhance the simulation results of the WOFOST model and minimize RMSE between the simulated and measured values. Furthermore, recent wheat yield simulations outperformed those from the early 21st century, with all stations except for the yields measured in Zhongning.

The SPEI index annual means under the three future climate models were selected and their spatial distribution analyzed by the Kriegin difference method (Figure 9, Figure 10 and Figure 11). The SPEI index spatial distribution was less affected by latitude than by DHW days. The drought risk index in May, June and July was higher in MT, LT and ST, respectively, under the SSP126 model. In the SSP245 model, it was greater in MT, LT and MT, respectively, and in the SSP585 model, it was highest in LT. Its spatial distribution was less affected by latitude and longitude. Under the SSP585 model, spatial variation was significant, and the drought degree in the northern region was increasing monthly. There was no significant difference in the overall study area SPEI index under the SSP245 model. Overall, climate change was more consistent, while in the SSP126 model, the ML climate gradually became wetter.

3.4. Effects of DHW and SPEI Index Changes on Wheat Yield

According to a previous study [28], we analyze the correlation between the total transpiration (TRANSP), yield (TWSO), flowering time (FLWR) and maximum leaf area index (LAIM) from the WOFOST model’s simulation results for wheat from 2020 to 2098 and the annual number of DHW events, as well as the SPEI index for May, June and July (Figure 12). The most significant factor affecting TRANSP is LAIM, with correlation coefficients reaching 0.7 across all three scenarios. FLWR and TWSO also show significant effects: DHW has a notable impact under the SSP126 scenario, but not under the SSP245 and SSP585 scenarios. For TWSO, the impact of DHW increases with rising warming levels, with correlation coefficients of −0.142, −0.412 and −0.54 under the three scenarios, respectively. Additionally, the SPEI index for June shows a significant effect, being positively correlated in the SSP245 and SSP585 scenarios with coefficients of 0.195 and 0.328, respectively. In contrast, it is negatively correlated in the SSP126 scenario with a coefficient of −0.271, indicating that increased rainfall in SSP245 and SSP585 scenarios promotes wheat yield, while the SPEI index for May only has a significant effect in the SSP585 scenario. FLWR has a significant positive correlation with the SPEI indices for May and June under the SSP126 scenario, and with the SPEI indices for May and July under the SSP245 and SSP585 scenarios, respectively. DHW shows a negative correlation with the SPEI index across all three scenarios, with the strongest effect in the SSP585 scenario, with correlation coefficients of −0.512, −0.612 and −0.219 for May, June and July, respectively.

4. Discussion

4.1. Wheat DHW Temporal and Spatial Variation Trends in Ningxia

The precipitation and temperature anomaly greatly influences drought variability. A decrease in regional precipitation and an increase in evaporation caused by warming may greatly exacerbate drought [35]. We statistically analyzed temperature, precipitation and wind speed changes in Ningxia province under different future climate scenarios by collating CMIP6 data. According to meteorological changes, the spatiotemporal variation of dry hot wind in Huinong, Shizuishan and the other test sites was analyzed and temperature variation of each test site was consistent. Under SSP126, 245 and 585 conditions, temperature increased 0.037, 0.15 and 0.45 °C·(10 a⁻¹), respectively. Recent average annual temperatures in the northern region, central arid area and southern mountainous area were 7.75, 7.94 and 7.83 °C, respectively, which is consistent with Ma Yang et al. [36]. Under the SSP126, 245 and 585 scenarios, average annual precipitation increased by 17.77 mm, 38.73 mm and 32.12 mm from near- to long-term, respectively. Furthermore, annual evaporation increased by 11.96 mm, 30.85 mm and 28.18 mm, respectively, consistent with Zhang Shiyan et al. [37]. Thus, the future Ningxia Province climate will shift towards higher temperature and humidity, and high carbon emissions (SSP585) will accelerate the risk of warming and will not result in more precipitation than balanced carbon emissions (SSP245). This would increase extreme dry weather frequency. Dry hot wind weather statistics in each test site showed that Huinong and Shizuishan sites have the driest hot wind days. Under the SSP126 model, average annual DHW days were 5.71 in the southern mountainous area, 16.08 in the central arid area and 26.35 in the northern irrigation area. In the SSP245 model, there were 3.93, 12.58 and 23.29 d, respectively, and in the SSP585 model, 9.85, 20.85 and 32.35 d, respectively. Mann–Kendall trend test analysis showed that high warming levels will significantly affect the increasing hot wind days trend. Mutation was detected at all sites under the SSP585 model, indicating that the increased rate of dry hot wind days will be greatly improved, and drought risk is increasing rapidly in the long-term. DHW frequency was higher in the northern and western regions, and lower in the southern and eastern regions.

Li et al. [38] showed that the study area under the SSP585 model would be drier than in the other two scenarios, and expected future drought to be more severe in the long-term, than in the near- and medium-term. Feng et al. [39] investigated spatiotemporal variation of multi-variable complex hot, dry and strong wind extreme weather in northwest China, and found that both the upward trend in incidence and the increased likelihood of compound extreme weather posed a strong risk in the future. Tan et al. [24], based on CMIP5 data, used 20 global climate models and drought data to predict future climate change and related drought disasters in Ningxia, and found that the climate may become warmer and wetter in the 21st century, but drought disasters may also increase, which is a similar finding to our study. Compared with previous studies, we highlighted the impact of extreme weather on wheat by introducing the dry hot wind criterion, and analyzed specific test sites. We showed that future dry hot wind spatiotemporal variation in the study area may be caused by warming levels, which provides a key scientific basis for actual wheat drought risk prevention and mitigation in Ningxia.

4.2. Future Drought Temporal and Spatial Changes in Ningxia Based on the Drought Index

Drought status changes in the study area can be explained primarily by precipitation and evapotranspiration [40]. The increase of vegetation water demand under climate change is greater than the increase of rainfall. Therefore, even though precipitation may increase in some areas, more severe ecological droughts will still occur [41]. To determine the relative influences of precipitation, temperature and evapotranspiration, we introduced the SPEI index to quantitatively study drought/climate change effects in northwest China. In May, there was a clear trend of increased drought risk under the SSP585 model, but there was no obvious change in drought risk under the SSP245 model, and the probability of drought risk appearing in the middle period was the highest under the SSP126 model. In June, SPEI was generally higher than in May. The risk of alternating drought and flood was greater under the SSP585 model, while the risk of near-medium-term drought was lower under the SSP126 and SSP245 models. In July, the near-medium-term drought risk was greater than in May and June, and the long-term risk was higher than in June, but still lower than in May. Finally, we conducted Pearson correlation analysis on DHW, SPEI and the related indicators of the wheat simulation results. The study of DHW correlation indicates that DHW will have a significant impact on wheat yield under all scenarios, and this impact increases with rising warming levels. However, increased rainfall tends to suppress the growth of DHW, with this effect being more pronounced under high-emission scenarios. In the SSP126 scenario, cooling caused by rainfall or post-rain dry hot winds will have a substantial effect on wheat yield, while in the SSP585 scenario, the advancement of high-temperature periods will also significantly affect wheat yield. High temperatures combined with high rainfall will lead to varying degrees of advancement in the flowering period. The results of the SPEI correlation analysis show that moderate high rainfall will promote wheat yield. However, the correlation results for DHW indicate that extreme high-temperature weather accompanied by high rainfall will negatively impact wheat yield. This highlights the limitations of studying the effects of a single indicator on wheat yield. Specific changes in the SPEI index categories are illustrated in Figure 13.

Based on the CMIP6 model, Liu et al. [42] found by analyzing the water balance between increasing evapotranspiration caused by rising temperature and increasing precipitation, that not all arid and semi-arid regions in northwest China will be wet in the future. In some areas, the drought trend would still be strong, and more severe in the high- than the low-emission scenario. Based on CMIP5 data, according to Ma et al. [43], future changes will be characterized by significant contraction of the humid region and expansion of the arid/humid transition zone, results which are similar to our study. However, Zhang et al. [44] show that by the end of the 21st century, the area and severity of drought (based on scPDSI and SSIS) in the arid northwest will significantly trend downward. The reason for these differences may lie in the different index used to assess drought risk. The SPEI index predicts drought risk based on meteorological factors (precipitation, evapotranspiration), while Zhang et al. used an index based more on soil moisture analysis. Additionally, Tan et al. [24] showed that drought disaster risk in Ningxia may increase, reaching a maximum in the short-term, followed by the medium- and long-term. Conversely, we showed that the long-term risk increased significantly in the SSP585 model, while there was no significant change in the other two models. These differences reflect the different development models selected, and that the SPEI index was introduced to analyze the influence of evapotranspiration. However, Tan et al. [24] believe that compared with RCP2.6, drought disaster risk under the RCP4.5 scenario may further increase in the three periods, which concurs with our findings.

Referring to the future yield and fertility changes in the previous study and, in conjunction with the analysis of the future drought disaster risk for wheat in the study area, from the holistic perspective of the study area, the central arid zone faces the highest risk of drought disaster under SSP126 and SSP245 scenarios, with a greater distance to reach sufficient irrigation water. The southern mountainous area has limited access to light and heat resources, thus making the northern irrigation zone a suitable region for wheat cultivation. Specifically, for the experimental stations within the northern Yellow River Irrigation District, with the changing climatic conditions, advancing the sowing date appropriately can help achieve higher yields. Other than Yinchuan, Qingtongxia, where sowing can be advanced by an average of 1–3 days compared to the norm, under the SSP585 model, flowering can occur 4–8 days ahead of the average, which aids in making full use of rain and heat resources. Additionally, under the SSP585 scenario, the risk of drought will significantly increase across all stations in the study area in the future. However, the southern mountainous areas will not experience more severe drought due to yield reduction, and the rapid increase in temperature makes wheat cultivation more favorable. Therefore, if the rate of climate warming increases in the future, efforts should be made to develop and expand arable land in the southern mountainous areas to ensure the overall yield of wheat production.

Risk assessment is the core component of risk management, and its main purpose is to provide suggestions for risk-based decision-making and disaster risk management. Our results show that drought frequency and severity risk will increase, which will have a serious impact on agricultural production and economic development in Ningxia. Therefore, ways to reduce drought risk are the key issue. High-risk years and areas can also be avoided by adjusting planting areas and rationally planning grain reserves in advance [45], which requires the government to take corresponding measures. Additionally, for actual production practice, drought disaster damage can be alleviated by spring sowing in advance or artificial precipitation. Another approach is to improve crop adaptability and screen excellent drought-resistant varieties to reduce vulnerability [46]. Finally, governments could increase drought relief efforts such as efficient and water-saving agriculture development, reservoir and canal construction/renovation, planting structure adjustment, promotion of new drought resistant varieties, establishment of disaster preparedness and rescue systems and stockpiling a reserve of emergency materials [47].

5. Conclusions

Three CMIP6 development models were used to analyze future climate and dry hot wind meteorological changes in 12 test sites in Ningxia. The monthly SPEI index was calculated to analyze future spatiotemporal variation of drought conditions. In the study area overall, although temperature and precipitation would significantly increase, the risk of dry hot wind disaster and drought would also increase. Additionally, long-term drought risk would increase significantly under the high warming level model. During the wheat growing period, there was a clear trend of increasing drought risk in May under the SSP585 model. Under the SSP245 model, there was no change in drought risk, and under the SSP126 model, the probability of drought risk appeared in the middle period and was the highest. In June, the SPEI index was generally higher than in May, and the risk of alternating drought and flood was greater under the SSP585 model, but the risk of near-medium drought was lower under the SSP126 and SSP245 models. The influence of DHW on wheat yield will increase with the increase of warming level, but the increase of rainfall will effectively inhibit the growth and harm of DHW, so the appropriate high rainfall will promote wheat yield.