Seasonal Sensitivity of Drought Indices in Northern Kazakhstan: A Comparative Evaluation and Selection of Optimal Indicators

Abstract

Drought is one of the main climate-induced risks threatening agricultural sustainability in semi-arid regions. Northern Kazakhstan, a key grain-producing region in Central Asia, exhibits increasing vulnerability to droughts due to climatic variability and reliance on rainfed agriculture. This study evaluates the informativeness of drought indices based on the response of agricultural vegetation to dry conditions using remote sensing-based vegetation indices across Northern Kazakhstan from 1990 to 2024. Ground-based meteorological indices—the Standardized Precipitation Index (SPI), the Standardized Precipitation Evapotranspiration Index (SPEI), the Hydrothermal Coefficient (HTC), and the Modified China-Z Index (MCZI)—and vegetation indices—the Vegetation Condition Index (VCI), the Temperature Condition Index (TCI), and the Vegetation Health Index (VHI)—were analyzed using data from 11 representative meteorological stations. For the first time in Kazakhstan, the MCZI was calculated, demonstrating high sensitivity to local climate variability and strong agreement with the VHI. The SPI, MCZI, and HTC showed strong seasonal correlations with vegetation indices, whereas the SPEI had a weak correlation, limiting its applicability. The highest correlations (r ≥ 0.82) between meteorological and vegetation indices were recorded in summer, while spring and autumn were influenced by phenological and temperature factors. Persistent drying trends in the southern and southwestern areas contrasted with moderate wetting in the north. The combined use of the SPI, MCZI, HTC, and VHI proved effective for monitoring droughts. The results provide a reproducible foundation for local drought assessment and early warning systems, supporting climate-resilient agricultural planning and sustainable land and water resource management. The results also offer actionable insights to enhance adaptation strategies and support long-term agricultural and environmental sustainability in Central Asia and similar continental agroecosystems.

1. Introduction

Climate change has significantly increased global temperatures, surpassing 1.5–2 °C, with projections indicating that even an additional 0.5 °C of warming could further intensify extreme climate events. Without substantial reductions in CO₂ and other greenhouse gas emissions, these changes may become irreversible [1,2]. Rising temperatures accelerate evapotranspiration, reduce soil moisture, and disrupt precipitation patterns, intensifying drought severity worldwide, particularly in arid and semi-arid regions [2]. These changes contribute to land degradation, declining agricultural productivity, and the expansion of desertification, posing significant threats to food security and sustainable development. The frequency and duration of droughts have increased by nearly 30% since 2000, with future projections indicating even greater risks [3].

Central Asia is highly vulnerable to drought due to its continental climate, dependence on seasonal precipitation, and extensive reliance on agriculture [4]. The region is characterized by low annual precipitation, large temperature fluctuations, and low humidity, making it highly sensitive to changes in climate patterns [5,6]. Similar vulnerabilities can be observed in other semi-arid regions, such as the Middle East and North Africa, where increasing temperatures and declining water resources pose serious threats to agricultural productivity and food security [7,8,9].

While drought has long been a natural phenomenon in Central Asia, its intensity and frequency have increased significantly over the past three decades. The region has become more susceptible to prolonged dry periods, driven by climate change and rising anthropogenic pressures. These factors directly threaten agriculture, ecosystems, and water security, further complicating socio-economic stability [10,11,12,13]. The agricultural sector, a vital component of the region’s economy, suffers from reduced crop yields, pasture degradation, and increased desertification risks [14].

Kazakhstan has shown significant growth in its wheat export capacity over time. Northern Kazakhstan is particularly affected by drought, as it plays a crucial role in the country’s agricultural production. Between 2001 and 2018, the country doubled its wheat exports from 3.0 to 6.2 million tons [15]. According to FAO estimates, Kazakhstan’s wheat exports were expected to reach 4–5 million tons for the 2019–2020 marketing year [16]. Some sources reported higher figures, with exports reaching 8.5 million tons in 2017 [17].

Droughts are leading to significant economic, agricultural, and human losses and have surged in recent decades [1,18,19]. Globally, over 75% of the world’s population is affected by droughts, and by 2050, an estimated 4.8 to 5.7 billion people will experience water scarcity for at least part of the year [3]. Between 1970 and 2019, droughts accounted for 15% of natural disasters but caused 34% of disaster-related deaths, approximately 650,000 fatalities, making them the deadliest weather-related hazard during this period [3,20].

Economic losses from droughts have exceeded USD 124 billion in recent decades, with a significant upward trend [3]. Agriculture is particularly at risk, with global crop production losses exceeding 10% between 1964 and 2007, impacting around 454 million hectares of farmland [1,21,22].

The consequences of droughts are particularly severe in Central Asia, especially in Kazakhstan, where climate change has made weather conditions increasingly erratic. Although earthquakes pose the greatest economic risk among natural disasters in the region, followed by floods, mudflows, and landslides, droughts have affected the largest number of people—up to 70% of those impacted by natural disasters [23]. In Kazakhstan, droughts occur recurrently, severely reducing agricultural productivity. The agricultural sector is a crucial component of the country’s economy and has suffered yield losses of up to 50% during severe drought years, leading to economic losses amounting to hundreds of millions of dollars [14,24]. The 2014 drought alone resulted in an estimated loss of USD 1.5 billion, highlighting the vulnerability of the sector to climatic extremes [23]. An analysis of adverse agrometeorological events in Kazakhstan that have inflicted significant or total destruction on crops indicates that atmospheric and soil droughts account for about 80% of such losses, while heavy rains and hail contribute 14%, frost 2%, excessive soil moisture 2%, and extreme cold and strong winds 1% each [25]. These factors underscore the critical need for improved climate resilience and drought mitigation strategies in the region. Drought monitoring is crucial for mitigating impacts, optimizing water resource management, and supporting agricultural adaptation through accurate assessment and forecasting systems [26,27,28].

Despite the growing severity of droughts, their assessment remains a challenge due to the lack of a universally accepted index. There have been more than 100 different drought indices developed to address specific meteorological drought impacts, including 23 recommended by the WMO [12,29,30,31]. The lack of a standardized definition for drought further complicates assessments, leading to contradictory conclusions regarding its severity and characteristics [12,29,32,33].

However, studies have shown that different drought indices provide different information depending on the specific climatic conditions of each region, including the Standardized Precipitation Index (SPI), the Standardized Precipitation Evapotranspiration Index (SPEI), and other indices such as the Modified China Z Index (MCZI) and the Hydrothermal Coefficient of Selyaninov (HTC) [34,35]. For example, the SPI is effective for assessing precipitation anomalies during the winter months, while the SPEI is more suitable in the warmer seasons. Additionally, the MCZI has demonstrated better performance in more humid regions [36,37,38].

Moreover, the changing climate creates additional challenges for drought assessment. With the increasing frequency of climate extremes, insufficient attention is given to the characteristics of droughts, including their severity and duration, especially at the subregional scale [39]. This complicates prediction efforts and the development of adaptation strategies, which play a crucial role in reducing vulnerability to droughts [40,41,42]. Furthermore, the correlation between drought indices and crop yields has become more complex due to the introduction of drought-resistant crop varieties, making agricultural impact assessments even more challenging.

Despite these uncertainties, extensive research has been conducted worldwide to analyze the spatial and temporal patterns of droughts [36,43,44,45]. In Central Asia, drought frequency between 1930 and 2014 was 42.87%, while in Kazakhstan, it reached 43,1% [33]. Droughts in Central Asia typically begin in winter, intensify in spring, and weaken during the summer months [46]. However, in northern Central Asia, CMIP6 models indicate an increase in spring precipitation and dry periods during the summer season [47].

Recent studies increasingly incorporate multipurpose drought indices such as the VHI, VCI, and EVI to enhance drought monitoring using meteorological indices and improve impact assessments of different crop varieties. While the SPI and SPEI remain key indices for global drought assessment, local adaptations and additional indices can provide more precise insights into drought dynamics [48,49].

For example, the VHI shows a strong correlation with the SPI at 3- and 6-month timescales, while high correlation coefficients between the SPI and VCI can be observed at 9- to 12-month scales [50]. Additionally, the VCI has demonstrated a strong relationship with crop yield [51]. The MCZI also demonstrates a high correlation with the SPI but tends to respond to drought onset with a delay [52,53]. By using the VCI, researchers can effectively detect drought conditions and assess vegetation health in areas such as Northern China, where agricultural productivity is at high risk due to climate variations [54].

In Northern Kazakhstan, the frequency of meteorological droughts increases from 20% in the north to 90% in the south. Mild droughts are recorded almost every year, moderate droughts approximately once every 4 years, and severe droughts once every 11 years [55]. In addition to the previously mentioned indices, the HTC has proven to be a sensitive drought indicator for Northern Kazakhstan, as its values align closely with grain yield fluctuations [56].

Despite extensive research on droughts in Central Asia, a limited number of studies focus on the effectiveness of various drought indices in Northern Kazakhstan. The objective of this research is to evaluate the effectiveness of different meteorological drought indices for monitoring and assessing drought conditions in Northern Kazakhstan, considering the complexity of the correlation between drought indices and crop yields, which has been influenced by the introduction of drought-resistant crop varieties. To address this gap, we conduct a comparative analysis of meteorological indices (SPI, SPEI, HTC, and MCZI) and vegetation indices (VHI, VCI, and TCI) to determine their accuracy in reflecting drought conditions in this area.

2. Materials and Methods

2.1. Study Region

Northern Kazakhstan was chosen as the study area because it is a leading exporter of crops. The region covers 565,000 km², about 20.7% of the Republic of Kazakhstan, a Central Asian country encompassing various landscapes, of which 44% are deserts, 14% are semi-deserts, 26% are steppes, and 15% are forests. Geographically, Northern Kazakhstan lies at the heart of Eurasia, between 51.0 and 55.5° N latitude and 61.0–78.0° E longitude (Figure 1). It consists of four administrative units: Kostanay, North Kazakhstan, Akmolinsk, and Pavlodar regions. The region has a moderate—continental climate characterized by long, cold, harsh winters and relatively warm summers [57,58]. The mean annual temperature is +2.4 °C, with January being the coldest month (−18 °C) and June the warmest (+21.0 °C). Absolute temperature extremes range from −52 °C to +45 °C. The annual precipitation varies between 216 and 364 mm with an uneven spatial and temporal distribution, making this area susceptible to drought conditions [59].

In total, 40% of the region’s territory is covered by fertile chernozem soils, which support high-quality crop production, making NK a significant agricultural region in the global market [60]. This region is particularly renowned for substantial agricultural production, a key contributor to the national economy. Notably, it is recognized for grain production, with wheat, barley, and oats as key crops. It serves as a major exporter of these crops to China, Turkey, Iran, and Russia [61].

2.2. Dataset

To evaluate drought conditions in Northern Kazakhstan, this study used a combination of ground-based meteorological observations and satellite-derived remote sensing data. This integrated approach provides a comprehensive basis for calculating both meteorological and remote sensing-based vegetation indices over an extended period.

Figure 2 presents a flowchart outlining the main steps involved in processing the dataset. Ground-based daily observations were used to calculate indices such as the SPI, SPEI, MCZI, and HTC in R-Studio (2024.12.0+467). In parallel, the MODIS and Landsat products were processed on the Google Earth Engine (GEE) platform (updates 14 April 2025) to derive vegetation indices, including the VCI, TCI, and VHI. The resulting outputs were used to generate spatial maps for further analysis in QGIS version 3.40.11.

A flowchart illustrates the necessary steps for processing ground-based and satellite data for drought assessment.

2.2.1. In Situ Data

Meteorological data from 1990 to 2024, including monthly total precipitation, daily mean air temperature, and daily maximum and minimum air temperatures, were obtained from the Republican State Enterprise “Kazhydromet” under the Ministry of Ecology and Natural Resources. Data were collected from 11 meteorological stations (MSs) listed in

Table 1. Meteorological stations.

Station	Latitude (°N)	Longitude (°E)	Elevation (m)	Established
Torgai	49.70	63.50	123	1874
Amangeldy	50.10	65.20	142	1935
Astana	51.08	71.40	350	1870
Ereimentau	51.60	73.20	397	1954
Pavlodar	52.10	77.10	125	1891
Kushmuryn	52.50	64.70	110	1940
Ruzaevka	52.80	67.00	227	1935
Aktogai	53.00	75.59	780	1960
Kostanay	53.20	63.60	156	1962
Yavlenka	54.30	68.50	115	1902
Petropavlovsk	54.80	69.20	142	1890

. The stations were selected based on representativeness, and the station data were assessed for homogeneity.

2.2.2. Satellite Data

Satellite-based remote sensing products were employed to assess vegetation and thermal conditions across Northern Kazakhstan. Two primary satellite data sources were used: the MODIS and Landsat missions. These provide consistent long-term observations suitable for monitoring drought dynamics and vegetation stress across large areas.

The MODIS data included two key products: MOD13A2 v6.1, which offers 16-day composites of vegetation indices, such as the NDVI, and was used to calculate the Vegetation Condition Index (VCI), and MOD11A2 v6.1, which provides 8-day composite land surface temperature (LST) data used to derive the Temperature Condition Index (TCI). In addition, Landsat 8 imagery (Product: LANDSAT/LC08/C02/T1_TOA) was utilized to extract vegetation and thermal parameters at a higher spatial resolution.

All satellite datasets were processed using the Google Earth Engine (GEE) cloud-based geospatial platform, which enabled efficient data access, filtering, and computation.

2.3. Methodology

2.3.1. Calculation of Meteorological and Remote Sensing-Based Vegetation Indices

The key meteorological and remote sensing-based vegetation indices used in this study included the Standardized Precipitation Index (SPI), the Standardized Precipitation Evapotranspiration Index (SPEI), the Hydrothermal Coefficient (HTC), the Modified China-Z Index (MCZI), the Vegetation Condition Index (VCI), the Temperature Condition Index (TCI), and the Vegetation Health Index (VHI). These are critical indices for assessing vegetation health and monitoring conditions. All calculation formulas, input data, and short descriptions are summarized in

Table 2. Summary of meteorological and remote sensing-based vegetation indices recommended by the WMO.

Index	Formula	Input Data	Description
SPI	Based on gamma distribution, transformed to standard normal variable	Precipitation	Quantifies precipitation anomalies [62,63].
SPEI	Similar to SPI but uses difference (precipitation–PET)	Precipitation, potential evapotranspiration	Captures drought by integrating precipitation and temperature [62,63,64].
HTC	$HTC={\displaystyle \frac{\sum R}{0.1\sum t}}$, (1)	Total precipitation (when t > +10 °C), air temperature (>+10 °C)	Moisture availability indicator during growing period [65].
MCZI	Based on skewness-corrected Z-score (Equations (5)–(7))	Precipitation	Incorporates the skewness coefficient and median precipitation, making it less sensitive to extreme outliers [66].
VCI	${VCI}_i={\displaystyle \frac{{NDVI}_i-{NDVI}_{min}}{{NDVI}_{max}-{NDVI}_{min}}}$ (2)	NDVI	Measures vegetation health relative to NDVI extremes [67].
TCI	${TCI}_I={\displaystyle \frac{{LST}_{max}-{LST}_i}{{LST}_{max}-{LST}_{min}}}$ (3)	Land surface temperature	Assesses thermal stress on vegetation; complements VCI [67].
VHI	$VHI=\alpha \times VCI+\left(1-\alpha \right)\times TCI$ (4)	VCI, TCI, a = 0.5 (contribution of VCI and TCI),	Combines moisture and temperature stress; widely used in arid and semi-arid monitoring [67,68].

. The VCI, TCI, and VHI included in the analysis are recommended by the World Meteorological Organization (WMO) for drought monitoring and assessment.

MCZI is an adaptation of the China-Z Index (CZI), developed to improve the accuracy of drought monitoring across various climatic zones. The calculation uses the skewness coefficient and the standard Z-score, facilitating the consideration of precipitation values. A key feature of the index is that MCZI calculations use the median precipitation value, making the index less sensitive to anomalously high or low values [66]. The specific calculation formula is as follows:

$${CZ}_{ij}= {\displaystyle \frac{6}{C_{si}}} \left({\left({\displaystyle \frac{C_{si}}{2}}{Z}_{ij}+1\right)}^{{\displaystyle \frac{1}{3}}}\right)- {\displaystyle \frac{6}{C_{si}}}+ {\displaystyle \frac{C_{si}}{6}} ,$$

(5)

$$C_{si}={\displaystyle \frac{{\sum }_{j=1}^n{\left(x_{ij}-med\left(x_i\right)\right)}^3}{{n\sigma }_i^3}},$$

(6)

$$Z_{ij}={\displaystyle \frac{x_{ij}-med\left(x_i\right)}{{\sigma }_i}},$$

(7)

where $C_s$ is the coefficient of skewness, $Z_{ij}$ is a standardized variable, and $x_{ij}$ is monthly precipitation.

The original daily precipitation and temperature data were aggregated into monthly precipitation totals and monthly mean temperatures for 11 stations over the 1990–2024 period. Subsequently, the meteorological drought indices were calculated from these monthly series, while the vegetation indices derived from satellite data were converted into monthly values. The SPI timescales of 3, 6, 9, and 12 months were constructed as moving sums over the respective number of months, allowing the analysis of drought dynamics across different temporal scales. The use of these timescales on a monthly basis will make it possible to trace how cumulative precipitation anomalies, particularly over the 9- and 12-month periods, may affect vegetation conditions at different stages of the growing cycle. This approach will help identify how prolonged precipitation deficits or surpluses in preceding months exert cumulative effects on vegetation status and the development of drought conditions. Vegetation index values were computed as arithmetic means over the same consecutive months to ensure temporal alignment with the meteorological series.

The drought categories for each index were classified according to the thresholds established by the authors of these indices and recommended for operational drought monitoring. Positive SPI, SPEI, and MCZI values indicate wetter conditions, whereas negative values represent varying categories of drought severity. For the HTC, higher values correspond to wetter conditions and lower values to increasing drought severity. Detailed classifications for drought categories based on the HTC and the SPI, SPEI, and MCZI are summarized in

Table 3. Drought index classification.

Drought Categories	HTC	Drought Categories	SPI/SPEI/MCZI
Extremely wet	>2.0	Extremely wet	>2.0
Moderate wet	>1.0	Very wet	1.5 to 1.99
Dry	<1.0	Moderate wet	1.0 to 1.49
Mild drought	1–0.8	Normal	−0.99 to 0.99
Drought	0.8–0.6	Moderate dry	−1.0 to −1.49
Moderate drought	0.6–0.5	Very dry	−1.5 to −1.99
Severe drought	0.5–0.4	Extremely dry	<−2.0
Very severe drought	<0.4

.

The vegetation index classifications are widely used to assess vegetation health and drought impacts on crops. VCI, TCI, and VHI values close to 100 indicate healthy vegetation with no signs of drought stress, while values approaching 0 reflect severe vegetation stress or dying vegetation. Detailed classifications of drought severity based on the vegetation indices are presented in

Table 4. Vegetation index classification.

Drought Categories	VHI/TCI/VCI
Extreme drought	<0.1
Severe drought	<0.2
Moderate drought	<0.3
Mild drought	<0.4
No drought	>0.4

[69].

2.3.2. Mann–Kendall Trend Test and Sen’s Slope

To identify significant monthly drought trends and assess their statistical significance levels, the Sen’s slope estimator and the Modified Mann–Kendall (MMK) test were applied. The modified version of the non-parametric Mann–Kendall test offers improved robustness and reliability in the presence of autocorrelation, making it a preferred tool for analyzing climatic time series. When a statistically significant trend is present, the corresponding test statistic is associated with a low p-value, indicating a non-zero trend in the parameter [70,71,72,73].

Sen’s slope provides an effective approach for estimating both the magnitude and direction of trends. It is resistant to outliers and does not assume any specific data distribution. The method is based on calculating the median of all possible slopes between pairs of points in the time series:

$$\beta =Median\left({\displaystyle \frac{x_j-x_i}{j-i}}\right),$$

(8)

where $\beta$ is the trend estimate, and $x_j$ and $x_i$ are the values of the variable at time points $i$ and $j$, respectively.

Sen’s slope can not only determine the direction of a trend but also provide a reliable quantitative estimate of its magnitude. A negative slope indicates a drying trend, while a positive slope reflects a tendency toward increased moisture in the analyzed variable.

In this study, drying and wetting trends were assessed using the Sen’s slope estimator in combination with the Modified Mann–Kendall test, with statistical significance evaluated at the 0.05 level.

2.3.3. Pearson Correlation Analysis

Pearson correlation analysis was applied to quantify the relationship between meteorological drought indices (SPI, SPEI, MCZI, and HTC) and vegetation indices (VHI, TCI, and VCI). Correlations were calculated at multiple timescales (1, 3, 6, 9, and 12 months) and separately for the main vegetation periods (spring, summer, and autumn). The correlation strength was classified as very high (|r| ≥ 0.80), high (0.60 ≤ |r| < 0.79), moderate (0.40 ≤ |r| < 0.59), or weak (|r| < 0.40). This approach can identify which meteorological indices most strongly anticipate vegetation stress under rainfed agriculture in the temperate continental climate of Northern Kazakhstan.

3. Results

3.1. Correlation Analysis Between Meteorological and Vegetation Drought Indices

This study analyzed the seasonal correlation between meteorological drought indices (SPI, SPEI, MCZI, and HTC) and remote sensing-based vegetation indices (TCI, VCI, and VHI), as presented in (Figure 3a–c). For interpretation, we classify absolute correlation as very high (|r| ≥ 0.80), high (0.60–0.79), moderate (0.40–0.59), or weak (<0.40). Across spring (MAM), summer (JJA), and autumn (SON), the SPI and MCZI exhibit very high agreement at the same timescales (typically r ≥ 0.96–0.98), indicating that these precipitation-based indices are effectively interchangeable for correlation-based diagnostics. The relationship between the HTC and SPI/MCZI is season-dependent. It is mostly moderate in MAM at 1–3 months, becomes high to very high in JJA at the 1-month timescale (up to r ≈ 0.9), and weakens at longer timescales and in SON.

In summer, the VHI shows high to very high correlations with the SPI and MCZI across timescales (up to r ≈ 0.82), with the maximum near 9 months. Correlations with the HTC are high (around r ≈ 0.7). The VCI is strongly related to the SPI and MCZI (about 0.72–0.74) and correlates appreciably with the HTC (about 0.66), whereas the TCI exhibits moderate-to-high correlations (roughly 0.54–0.65). In spring, the relationships weaken; the VHI maintains moderate correlations with the SPI and MCZI (up to r ≈ 0.5), while the TCI and VCI are strongly inversely related to each other (r ≈ −0.96). In autumn, the VHI correlates most clearly with the meteorological indices at the 3-month timescale, but the correlation is weak at 1 and 6–12 months; the TCI and VCI continue to be strongly inversely related (about −0.87) and show generally weak correlations with meteorological indices.

The most informative pairing was observed between the VHI and either the SPI or the MCZI, while the HTC provided short-term agronomic context at the 1-month timescale. Notably, correlation strength generally increased with longer accumulation periods, with the highest values typically observed at the 9-month timescale and the lowest at the 1-month scale. Due to the very high internal consistency between the SPI and MCZI, they can be substituted for one another. The SPEI assesses water deficit while accounting for temperature effects, and it is particularly informative at short timescales. However, in moderate continental climate, its link to vegetation response weakens at longer timescales and is more sensitive to the chosen accumulation period than precipitation-based indices, such as the SPI and MCZI.

3.2. Temporal Evolution and Trend Analysis of Drought

Our temporal analysis of drought indices enhanced our understanding of drought variability across Northern Kazakhstan (Figure 4, Figure 5 and Figure 6) and (Appendix A, Figure A1, Figure A2, Figure A3 and Figure A4). These figures illustrate the time series dynamics of the drought indices (SPI, SPEI, MCZI, HTC, VHI, TCI, and VCI) from 1990 to 2024 across 11 meteorological stations (Aktogai, Amangeldy, Astana, Ereimentau, Kostanay, Kushmuryn, Pavlodar, Petropavlovsk, Ruzaevka, Torgai, and Yavlenka). The red dashed line indicates the critical threshold demarcating the transition from normal to drought conditions: SPI, SPEI, and MCZI values below –1, HTC values below 1, and vegetation indices (VHI, TCI, and VCI) below 0.4. Figure 7 and Appendix B, Figure A5, Figure A6 and Figure A7 show the trend estimates using Sen’s slope with significance assessed by the MMK test at the 0.05 significance level. In these figures, blue triangles denote significant positive (increasing) trends, red inverted triangles denote significant negative (decreasing) trends, and gray triangles indicate non-significant results. Titles have been revised to specify the corresponding variables.

The analysis showed that during the period 1990–2024, the SPI, MCZI, and SPEI generally agree in regard to the main dry and wet phases in Northern Kazakhstan. Dry phases were observed in 1991–1992, 1995–1998, 2003–2005, and 2007–2011, whereas the wet period of 2011–2016 was followed by a drier phase lasting until 2022. Over the past two years (2023–2024), most stations (except Torgai and Amangeldy) recorded sharp positive drought index values.

However, the trend analysis revealed substantial discrepancies. The precipitation-based SPI and MCZI capture both positive and negative changes depending on the station. Statistically significant wetting can be observed in Astana (SPI1–SPI12, p < 0.0004), Kostanay (SPI12, p = 0.043), and Yavlenka and Ruzaevka (SPI3–SPI12, p ≤ 0.031; MCZI3–MCZI12, p ≤ 0.027), whereas persistent drying is recorded in Amangeldy (SPI1–SPI12, p < 0.001), Torgai (SPI1–SPI9, p < 0.01; MCZI1–MCZI9, p ≤ 0.018), Aktogai (SPI9–SPI12, p < 0.004; negative MCZI in medium and long windows), and Petropavlovsk (negative SPI and MCZI at 6–12 months). By contrast, the SPEI demonstrates a positive trend at all 11 stations (SPEI1–SPEI12), although the level of significance varies. In Petropavlovsk, the negative signal at SPEI1 shifts to a statistically significant positive one at the 3-month scale (p = 0.001); in Aktogai, a significant increase is recorded at SPEI1–SPEI6 (p ≤ 0.007), but it loses significance at SPEI9–SPEI12, while in Amangeldy and Yavlenka, positive trends are noted at all scales, although they are not statistically confirmed. After 2020, the SPEI sharply rises at all stations, indicating improved moisture availability, while the SPI and MCZI continue to show negative tendencies in arid regions. A similar temperature sensitivity can be observed for the HTC, which also shows positive trends in Torgai (p ≈ 0) despite its overall arid setting, whereas the SPI and MCZI indicate long-term drying there.

Analysis of the HTC also revealed notable differences. The linear regression results indicate that most stations are characterized by weakly negative slopes (–0.001 to +0.002) with very low explanatory power (R² = 0.00–0.07), suggesting no consistent long-term linear changes and a slight overall decline in hydrothermal conditions. The longest drought phases (HTC < 1) occurred between 2003 and 2011, while wet periods were observed in 1991–1997, 2011–2016, and in the last 2–3 years. Despite these fluctuations, on average, drought conditions prevailed across the study period. The MMK trend analysis (Figure 5) revealed positive tendencies at most stations, especially in Torgai (p ≈ 0) despite its arid setting, whereas a significant negative trend was observed in Kostanay (p = 0.001). High interannual variability in the HTC was recorded in Pavlodar, Petropavlovsk, Ruzaevka, and Yavlenka, where both extremely wet (HTC > 2.5) and extreme drought conditions (HTC < 0.4) were observed. These results highlight that while HTC is sensitive to temperature-driven fluctuations, its long-term dynamics remain weakly negative in most locations.

The vegetation indices (VHI, TCI, and VCI) confirmed the reflection of climatic anomalies under vegetation conditions. Despite the weak linear trends in the TCI and VCI, the VHI showed a statistically significant negative trend (p = 0.018), indicating the increased frequency of drought impacts (Appendix B,

Table A1. Drought trend based on Modified Mann–Kendall test for TCI, VCI, and VHI.

Region	Index	Z	p-Value	Sen’s Slope
North Kazakhstan	TCI	−0.476	0.634	−0.000067
	VCI	−0.549	0.583	−0.000042
	VHI	−2.363	0.018	−0.000078

). The year 2010 proved critical, as this was when VHI reached its minimum (0.29), and from 2017 to 2023, the negative consequences of droughts intensified again, especially in 2022, followed by partial recovery after 2023.

4. Discussion

The results provide a comprehensive view of spatiotemporal drought dynamics in Northern Kazakhstan using both meteorological and vegetation indices, revealing their informativeness in monitoring and analyzing climate-related risks. Analysis of time series data from 1990 to 2024 demonstrated pronounced heterogeneity in the spatial distribution of drought conditions. Persistent and statistically significant drying in the southwest (Amangeldy and Torgai) contrasted with wetting tendencies in the southern, central, and northwestern parts of the region (Astana, Ruzaevka, Kostanay, and Yavlenka).

Even in years with regionally above-normal precipitation, negative MCZI and HTC values were intermittently recorded in the southwest, underscoring subregional moisture deficits. Mixed statistically significant trends in the east and north (drying at Aktogai and Petropavlovsk versus wetting at Pavlodar and Yavlenka) reflect the interplay of local circulation patterns, orography, land-surface features, and the influence of Arctic air masses reaching the region during the formation of anticyclonic blocking systems, as noted in previous studies [74].

A key contribution of this study is the explicit comparison between meteorological indices (SPI, SPEI, MCZI, and HTC) and vegetation indices (VHI, TCI, and VCI) across seasons and timescales, allowing us to identify which meteorological indices best anticipate vegetation stress. During the active growing season in summer, the VHI shows the strongest correlations with the SPI and MCZI (r ≥ 0.82), with peak relationships at the 3–12-month timescales and a maximum around the 9-month timescale. This indicates that cumulative precipitation anomalies over several months, reflecting atmospheric moisture deficits, are the primary driver of crop conditions, whereas contemporaneous weather fluctuations still play a role but are secondary (up to r = 0.66). This is consistent with findings presented in previous studies [75,76]. However, in spring and autumn, the sensitivity of the indices decreased and exhibited different dynamics. In spring, VHI was influenced by both short-term (SPI1 and MCZI1) and long-term (SPI9 and MCZI9) timescales, consistent with the joint influence of current precipitation and over-winter/early-season reserves. Depending on vegetation type, the response to these timescales may vary [76]. The correlation with the HTC was moderate, and despite this relationship, precipitation remained the determining factor during spring temperature fluctuations. In autumn, the VHI was dependent on mid-term timescales (SPI3 and MCZI3) and the HTC, exhibiting closer ties to temperature, as captured by the TCI. This highlights the growing importance of thermal stress in late phenological stages, as noted in previous research [77,78]. Overall, combining the SPI and MCZI at the 3–9-month timescale with the VHI in summer provides the clearest and most robust signal of agriculturally relevant drought.

Analysis of the VHI revealed a sustained negative trend beginning in 2004, with particularly pronounced declines during the severe drought years of 2010 and 2022. These findings suggest increasing climate-induced stress on agricultural vegetation. Although slight improvements were observed after 2023, they may prove transient in light of growing interannual variability. The downward trajectory of VHI is consistent with the dynamics captured by the SPI and MCZI, which also identified periods of significant drought. This convergence of evidence points to a gradual degradation of vegetation conditions under intensifying climatic stressors. The most critical VHI value was recorded in 2010 (VHI = 0.29), marking the peak of drought severity, with a comparable decline observed again in 2022.

This study paid particular attention to the MCZI, applied for the first time under the conditions of Northern Kazakhstan. It demonstrated a strong correlation with the SPI across all timescales and showed consistent sensitivity to regional climate variability. This index is resistant to outliers, accounts for median precipitation values, and is well adapted to semi-arid zones, making it a valuable tool for subregional drought analysis that demonstrates strong summer agreement with the VHI. This supports using the SPI and MCZI as the primary meteorological indicators for diagnosing vegetation-impacting droughts in rainfed systems of temperate continental interiors.

As shown in Figure 5, the SPEI has exhibited consistent positive trends since 2016, similar to findings reported in previous studies [39,79,80]. The SPEI shows weaker and more scale-dependent links to vegetation, particularly at short timescales. The reasons for the weak correlations of the SPEI with vegetation are related to the fact that the index accounts for potential evapotranspiration (PET) rather than actual evapotranspiration. In the moderately continental climate of Northern Kazakhstan, PET can significantly exceed actual evapotranspiration, as the water regime is constrained by limited available moisture. Given that rainfed agriculture predominates in the region, atmospheric precipitation (including snowmelt and spring recharge) remains the decisive factor for crops under such conditions. This highlights the effectiveness of the SPEI in arid conditions, but not in the moderately continental climate of the northern areas of Kazakhstan, as supported by other research studies [81,82]. As a result, the SPEI is useful for diagnosing heat-amplified water-balance stress, yet it is less effective than the SPI, MCZI, and HTC in explaining vegetation response during peak growth, when antecedent precipitation and cumulative moisture deficits are the primary controls. This explains the systematically lower correlations between the SPEI and the VHI, despite positive SPEI trends across all timescales according to the MMK test.

The 1991–1992, 1995–1998, 2003–2005, and 2007–2011 periods were particularly dry according to the SPI and MCZI. Between 2011 and 2016, a temporary increase in wetness was recorded in the northern part of the study area, followed by deteriorating conditions up to 2022. The nature of such alternations can be partially explained by the influence of global atmospheric oscillations, such as ENSO, NAO, and the Arctic Oscillation. Their impact on drought cycles has been confirmed by other studies [39,46], and incorporating teleconnections into drought forecasting algorithms could significantly improve the accuracy of climate risk assessments [39,83,84].

Unlike the SPI and MCZI, the HTC index showed less stable trends and demonstrated a significant relationship with SPI1 and MCZI1, which can be explained by its high sensitivity to short-term fluctuations in temperature and precipitation. This makes the HTC less reliable for assessing long-term changes, but it is a valuable tool for analyzing droughts during the growing seasons. Despite the generally neutral direction of HTC trends, its spatial and interannual variability partially aligned with the SPI and MCZI results. Spatial comparisons also revealed that the HTC better captures short-term climate disturbances, especially in dry years, as confirmed by its extreme values recorded at most stations. These peaks corresponded to negative SPI and MCZI values, particularly at the 1-month timescale.

A consistent decline in VHI values since 2004, with pronounced minimums in 2010 and 2022 and only partial recovery in subsequent years, reflects increasing climate-induced stress on agricultural systems in Northern Kazakhstan. The analysis revealed that precipitation-based indices, when applied at appropriate seasonal timescales, exhibit the strongest correlation with vegetation dynamics. In particular, the study identified a clear seasonal sensitivity of drought indices under the region’s moderately continental climate, which is strongly influenced by antecedent precipitation. For spring, SPI and MCZI at 1- and 9-month timescales are recommended to capture both short-term variability and long-term moisture accumulation. In summer, the 3- and 9-month timescales for these indices better reflect delayed responses to precipitation deficits, while in autumn, the SPI and MCZI at a 3-month scale effectively capture transitional drought conditions. The MCZI further enhances assessment accuracy by highlighting subregional contrasts in semi-arid zones. The HTC demonstrated consistent responsiveness across all seasons, offering agronomically relevant context, while the SPEI serves as a temperature-sensitive indicator of water balance variability.

Research findings recommend using the HTC, MCZI, and SPI to assess moisture conditions and monitor droughts in Northern Kazakhstan [56]. Comparison with international global drought-monitoring systems, such as the Global Drought Analysis and Prediction Tools (IRI) and the South Asia Drought Monitoring System (SADMS), confirms the validity of this approach. The IRI system employs a range of global drought indicators, including the SPI, SPEI, and PDSI; however, it does not incorporate indices sensitive to regional specifics, such as the MCZI or HTC. Therefore, our results can be used to adapt and expand global drought assessment systems by introducing locally relevant indicators, which will improve monitoring and forecasting accuracy.

5. Conclusions

This study provides a 30-year assessment of drought in Northern Kazakhstan by combining meteorological (SPI, SPEI, MCZI, and HTC) and remote-sensing vegetation indices (VHI, TCI, and VCI). Drought patterns are spatially heterogeneous: persistent drying is evident in the southwest (Amangeldy and Torgai), whereas Astana, Ruzaevka, Kostanay, and Yavlenka tend to be wet. These contrasts reflect regional circulation, orography, and land-surface controls.

The strongest link to vegetation appears in summer between the VHI and the precipitation-based indices, especially the SPI and MCZI. Introduced here for the first time in this region, the MCZI shows strong agreement with the SPI, resists outliers, and enhances subregional diagnostics. The HTC adds agronomic context during the growing season. Although the SPEI reflects heat-amplified drought signals, its applicability is limited in the temperate-continental climate due to weaker links with vegetation dynamics.

Considering the characteristics of the moderate continental climate of Northern Kazakhstan and the phases of the vegetation period, a seasonally adapted application of the SPI and MCZI is recommended for practical drought monitoring. In spring, 1-month accumulation periods provide the most informative results, capturing the immediate sensitivity to precipitation during the sowing phase. In summer, when crops experience peak water demand, 3–9-month accumulation scales are appropriate for reflecting the delayed impact of precipitation deficits on the root zone. In autumn, a 3-month timescale is optimal for detecting transitional conditions and the residual effects of summer droughts on the final stages of plant development. The HTC index, which accounts for the thermal characteristics of the season, can be used across all seasons to assess meteorological drought. For a comprehensive understanding of drought impacts on vegetation, the VHI should be additionally employed as an indicator of agricultural crop health.

This study highlights the importance of subregional drought analysis, with results that can inform both early warning systems and farm-level adaptation strategies, such as optimizing sowing dates, selecting drought-resistant varieties, and improving water-use efficiency. The proposed approach can be adapted to other parts of Kazakhstan and Central Asia and can contribute to improving international drought monitoring systems and open data platforms. However, correlations between meteorological and vegetation indices may differ under other climatic or land-use conditions, so careful validation is required before transferring these findings elsewhere.