Spatiotemporal Analysis of Soil Moisture Variability and Precipitation Response Across Soil Texture Classes in East Kazakhstan

Abstract

The study of the hydrological regimes of rivers in different regions of the globe has revealed the need to include the soil moisture content in flood prediction models. This paper investigates the nature of the dependence of soil moisture content on soil texture in the East Kazakhstan region. Data from ERA-5-land reanalysis, soil maps, hydrogeological maps, and the meteorological data of Kazhydromet were used. The years for analysis were selected due to their different moisture conditions. This study analyzed soil moisture within the root zone (0–28 cm depth). A JavaScript-based algorithm was developed in Google Earth Engine to analyze soil moisture and total precipitation across five Soil Texture Index categories during the growing seasons (April–September) of 2013, 2022, and 2023. Final cartographic processing and spatial distribution analysis were conducted using ESRI ArcGIS Pro 3.3. The study of soil moisture’s relationship with different soil textures in the East Kazakhstan region has revealed several key trends. The maximum values of soil moisture for each texture class change very slightly from year to year. The minimum soil moisture values fluctuate more strongly from year to year. The regression analysis demonstrates a statistically significant relationship between precipitation and soil moisture. The best performance is achieved when using a 1-day lag for 2013 and varying optimal lags for 2022 and 2023 (ranging from 1 to 3 days) during the high-precipitation period (months 6–9), with filtering applied to remove days with negligible rainfall.

1. Introduction

Soil moisture (SM) plays a crucial role in meteorology, hydrology, and ecology [1]. SM is in second place among the Essential Climate Variables established by the Global Climate Observing System (GCOS) [2,3,4]. SM is an important input variable in soil water management [5] and drought analysis [6]. In drought-prone areas, monitoring SM is very important as it can affect the water balance of agricultural crops [7].

SM is part of the hydrological cycle and the global water cycle [8]. Knowledge of SM is critical for developing an understanding of numerous hydrological processes. If we want to understand and predict extreme events, we should not focus specifically on rainfall trends, but rather on changes in soil water concentration. The regional mean state soil moisture changes are the primary drivers of future drought and flood risks [9]. Several authors have noted that it is the soil’s physical parameters that can largely control the magnitude of a flood event due to the non-linear nature of runoff response to precipitation and, therefore, it is very important to have a good record of previous soil moisture for flood forecasting purposes [10,11]. Orth and Destouni (2018) found that soil moisture deficit reduces runoff more strongly and faster than it reduces evapotranspiration. In the largest catchment of the Yangtze River in China, floods usually occur in summer when soil moisture saturation and extreme precipitation coincide [12,13].

SM monitoring over large areas is still an open research activity [14,15] and continuous observations of the soil moisture dynamics of different ecosystems still need to be improved [16]. Differently from precipitation, in situ monitoring networks for soil moisture are much less developed [4]. Ground-based SM observations are commonly considered to be highly accurate [8,17]. Ground-based soil moisture observations include (a) labor-intensive gravimetric measurements, (b) rapid measurements using hand-held devices, and (c) automated observation stations that are suitable for continuous observation at different depths [8,18,19,20]. Significant progress has been made since the early 2000s by Chinese researchers terms of in establishing ground-based measurement networks for extracting soil temperature and moisture at different depths in inaccessible areas [21]. The largest platform in the world for the collection and publication of SM measurements from more than 2800 sensors worldwide is the International Soil Moisture Network (ISMN) [22]. The ISMN platform also publishes SM data for the Kazakhstan area. These data were obtained from decade-long gravimetric measurements for agrometeorological purposes back in the Soviet period. Unfortunately, the number of stations where measurements were made is low (e.g., in East-Kazakhstan oblast–East-Kaza only (194a) with observations in 1987–1991).

Over recent decades, SM datasets have been used across a wide range of earth system applications. This is mainly attributed to the progress in remotely sensed soil moisture algorithms. The use of remote sensing data allows for continuous spatial coverage [17,23,24]. There are off-the-shelf RS-based products, such as European Space Agency (ESA): Soil Moisture and Ocean Salinity (SMOS) or National Aeronautics and Space Administration (NASA) Satellites: Aquarius, Soil Moisture Active Passive (SMAP) [17,19,25]. Meanwhile, individual authors have noted that it is SMAP that has the highest quality among all remotely sensed soil moisture products [26,27,28]. At the same time, Sentinel-1 and Sentinel-2 missions are more suitable for regional and localized studies where high spatial resolution is required [1,29,30,31,32]. Global models based on various data sources (remote sensing and in situ) have been developed as additional tools, such as the Global Land Data Assimilation System (GLDAS) developed by NASA [33], the Climate Change Initiative (CCI) developed by the European Space Agency (ESA) [34], the Global Land Evaporation Amsterdam Model (GLEAM) developed by European and US research staff [35], and the Remote Sensing-Based Global Surface Soil Moisture dataset (RSSSM) developed by the Research Center for Eco-Environmental Sciences, Chinese Academy of Sciences [36]. Recently, major advances in temporal coverage and spatial resolution have been made by the European Center for Medium-Range Weather Forecast (ECMWF) team, who have been developing the ERA-5 and ERA-5-Land products for a number of years [37,38]. ERA5 has been collecting hourly data on single levels from 1940 to present. The experience of applying the ERA-5 dataset to spring flood forecasting [39] is very interesting. Modelled soil moisture data is validated with observed soil moisture measurements from different places around the world, such as Russia, China, India, and the USA, and the validation results show that the modeled soil moisture simulates the seasonal to inter-annual variability of observed soil moisture very well in many locations [40,41].

Evaluation of the capacity of soil water reservoirs requires knowledge of some soil hydrological parameters. The amount of SM or the highest water content held in the soil after it has been thoroughly wetted and allowed to drain until drainage becomes practically negligible is termed field capacity, soil water-holding capacity (SWHC), or drained upper limit (DUL). Lower limit (LOL) or permanent wilting point (PWP) is the lowest field-measured water content of soil after plants have stopped extracting water and are at or near premature death, or dormant as a result of water stress. Potential extractable soil water (PLEXW) or total available water (TAW) is the difference in water content between DUL (SWHC) and LOL (PWP) [42,43,44,45].

The SWHC values can be obtained through field measurements, laboratory estimates, and, for large areas, from maps [46]: soil, geological, hydrogeological, or landscape. For example, some authors use proxies based on the landscape approach, which, under conditions of information scarcity, provides additional opportunities and allows them to find out to what extent regional soil–climatic conditions explain current runoff regimes and their sensitivity to climate change. Landscape hydrology deals with climate and land use composition and configuration, including soils, geology, and topography, and how their interactions affect the movement and storage of water in the landscape [47,48].

The main factors that affect SM are soil texture, soil depth, bulk density of soil, precipitation, and topography. It is well recognized that soil texture is the dominant determinant of soil water characteristics [46]. For example, the soil water content of the SWHC is highly correlated with soil texture. The soil texture is defined by the proportion of particle size fractions, called clay, silt, and sand [49,50]. The digital soil mapping models developed using environmental covariates as an input dataset showed that elevation and soil texture class were the most influential factors in predicting SWHC [51].

SM exhibits a tremendous heterogeneity across space and time [52]. The temporal variability of SM (natural SM content) may be influenced by snowmelt, temperature, rainfall events, and land cover [53,54]. SW behavior has been explored in the main agricultural regions in Australia [55] and in different soils of the Loess Plateau and the Maowusu Desert in China [56,57]. The results of studies in the Russian regions bordering Kazakhstan with similar soil types should be noted, such as the study on the different threshold values of the water-holding capacity of southern chernozems in Altai Krai [58,59], and the study on chestnut soils in the Lower Volga region [60].

The variability of SM content makes accurate characterization of this parameter difficult at a large scale [61]. For example, in some studies, the spatial variability of near-surface water content was found to be greater in drier soils [62,63], while other studies showed variability to be greater in wetter soils [64]. To understand the patterns of water behavior in the soil, it is advisable to conduct an analysis over large areas according to generalized classes, such as soil textures, different climates, and different land covers. Research examining the relationship between soil texture and soil moisture dynamics is currently scarce. So, Brocca [65,66] developed a «bottom-up» approach for directly estimating precipitation rates from soil moisture observations only. The precipitation products corrected through soil moisture data were used for improving flood prediction [67,68].

Hydrological projections are subject to considerable uncertainties. For example, modern studies have found that the Satellite rainfall products need a bias correction. Thieming [69] pointed out that in order to improve rainfall runoff modelling, bias correction and model recalibration are necessary. This issue has largely been overlooked in impact studies so far, and one has to be aware of this additional source of uncertainty [70].

The primary purpose of this study was to analyze the soil water behavior of the different textured soils in the East Kazakhstan region. This unique region is very interesting, as it is characterized by high landscape diversity.

The detailed objectives of this study were as follows:

(1)

To describe the seasonal SM fluctuations of different textured soils;

(2)

To describe the soil water behavior during years with different levels of atmospheric humidification;

(3)

To describe the relationship between the SM of different textured soils and the precipitation of the study area.

2. Materials and Methods

2.1. Study Area

Within its existing borders, the East Kazakhstan region (EKR) occupies an area of 97.8 thousand square kilometers (Figure 1). The Altai Mountains and their foothills occupy most of the EKR. The EKR borders on the east with the highest massifs of Altai: Belukha (4506 m a.s.l.), Tabyn-Bodgo-Ola (4356 m a.s.l.), and the Ukok plateau (2200–2600 m a.s.l.). Kazakh Altai is divided into several main mountain ranges: Korgonsky Koksuisky Ivanovsky, Ubinsky, Kholzun, Narymsky, Sarym-Sakty, and Southern Altai [71,72]. The Kalbinsky ridge is separated from Altai by the Irtysh River. The Saur and Tarbagatay ranges are located in the southern part of the region between Altai and the Tian-Shan Mountains [73]. The ranges are separated from Altai by the Zaisan Depression.

The formation of the EKR relief is associated with the Caledonian and Hercynian orogeny, Mesozoic continental leveling, and Quaternary neotectonics [74,75]. The glacial relief was formed by periodic Quaternary glaciations. The sand and loess cover was formed at the plain during the Quaternary [76].

Continental climatic conditions with significant seasonal weather fluctuations are prevailing [77]. According to the Köppen–Geiger climate classification [78], the mountainous areas of the EKR belong to the Dfa type (humid continental, no dry season, hot summer), and the plain part to the Dfb type (moderate continental climate) [79]. The mean annual air temperature in the eastern parts is −4/+1 °C; the mean January temperature is −13/−17 °C; and the mean July temperature is +15/+19 °C, with absolute minimums of −50 °C and maximums of +45 °C. The differentiation in the distribution of precipitation is due to the diversity of relief—the northern and north-eastern parts of the territory are exposed to wet north-western Atlantic cyclones (350–700 mm), while the seven arid western and southern areas are in the barrier shadows of mountain ranges (250–400 mm) [80]. The EKR is mainly represented by steppe (43%) and semidesert (14%) foothill landscapes, and forest steppe (15%), forest (13%), alpine, and nival (15%) mountain landscapes [81].

2.2. Data

Due to the lack of measured soil data, we extracted soil texture data from the Soil Map of the Kazakh SSR (1976), and a series of hydrogeological maps were used in this research. The hydrogeological maps, at a scale of 1:200,000, were compiled during geological surveys in 1970–1980. The hydrogeological maps were part of the geological surveys. They were made with high detail. Therefore, the maps were suitable for use in our study.

The soil texture classes followed the criteria of the USDA Soil Textural Classification Study Guide (1987) and Bormann [82]. Five general classes of soil texture were used in this study:

(1)

Very light soils: sand;

(2)

Light soils: loamy sand;

(3)

Medium soils: sandy loam;

(4)

Heavy soils: loam + silt loam + sandy clay loam.

Soil type was not used as an explanatory variable in this study.

The unification of all loamy textures into one class is due to the fact that they are very often combined in the area. In addition, all loamy textures are characterized by similar water–physical properties. Silt and sand are the main soil components in the study area [83]. There are no very heavy soils with high clay content (clay, clay loam, and silty clay) in the study area. None of the modern classifications represent a large number of clusters with high clay contents, probably because such soils are relatively rare in the real world [82]. On the other hand, taking into account the fact that soil texture classifications derived from soil hydrology should be based on similar hydrological fluxes or a similar soil water regime [82], we have identified one more soil texture class—(0). It includes soils formed in mountains on bedrock, where the thickness of loose deposits does not exceed several tens of centimeters. This class was added to reflect the low retention capacity and rapid saturation dynamics typical for mountain slopes with shallow soils, and it occupies the largest area in the EKR—58,288.89 km². The other classes are distributed approximately evenly (1—10,170.86 km², 2—8432.06 km², 3—3627.68 km², 4—10,941.45 km²). Each soil texture class is represented in the northern, central, and southern parts of the EKR (see Figure 2).

The years to be used for analysis were selected based on their different moisture conditions. The dataset was obtained from 15 EKR weather stations managed by the National Hydrometeorological Service of the Republican State Enterprise (RSE) “Kazhydromet”. Data were collected from 1961 to 2023. The Selyaninov Hydrothermal Coefficient was computed for each year and station (Equation (1)). Over the observation period, the range of Hydrothermal Coefficient (HTC) values at the weather stations in the EKR fluctuated within a wide range. During the last 25 years, the average maximum value for EKO was 1.07 (2013—extremely humid), the average minimum value was 0.43 (2022—extremely dry), and the mean value was 0.72 (2023). These years were selected to represent distinct hydrothermal conditions.

$$\mathrm{H}\mathrm{T}\mathrm{C}={\displaystyle \frac{\sum {\mathrm{R}}_{5–9}}{0.1\sum {\mathrm{T}}_{5–9}}}$$

(1)

where ∑R_5–9 is the sum of precipitation for May–September and ∑T_5–9 is the sum of daily air temperatures above 10 °C for May–September.

The climate and soil moisture data were obtained from ERA5-Land (Daily Aggregated), a global land surface reanalysis dataset with a spatial resolution of 0.1° × 0.1°. The dataset is distributed by the European Centre for Medium-Range Weather Forecasts (ECMWF) [84] and is widely recognized for its high spatial and temporal resolutions, frequent updates, and comprehensive parameter coverage. As a derivative of ERA5, ERA5-Land offers enhanced resolution and improved accuracy in representing surface variables, but has limitations in mountainous terrain. The Daily Aggregated version provides daily averaged values computed from the original hourly ERA5-Land data.

The ERA5-Land «volumet-ric_soil_water» variable provides daily averaged values for four depth layers; we focused on the upper 0–28 cm root zone using a weighted average to reflect active plant–soil interactions.

Data acquisition and preliminary analysis were performed using the Google Earth Engine (GEE) platform, a cloud-based geospatial processing service developed by Google. This platform provides access to satellite imagery archives, meteorological datasets, and computational resources for large-scale data analysis. A JavaScript-based algorithm was developed in GEE to analyze root zone soil moisture (0–28 cm depth) and total precipitation across five Soil Texture Index (STI) categories (0–4) during the growing seasons (April–September) of 2013, 2022, and 2023.

Final cartographic processing and spatial distribution analysis were conducted using ESRI ArcGIS Pro 3.3.

2.3. Methods

Several approaches exist for integrating soil moisture data from different layers, depending on the research objectives and the soil’s physical characteristics (e.g., direct summation and weighted summation with correction factors). We employed a thickness-weighted averaging method that accounts for differential layer thicknesses (7 cm for «volumetric_soil_water_layer_1» and 21 cm for «volumetric_soil_water_layer_2»), providing physically representative values for root zone soil moisture analysis. This is calculated using Formula (2):

(L₁ × 7 + L₂ × 21)/28,

(2)

where L₁—‘volumetric_soil_water_layer_1’ (m³/m³), L₂—‘volumetric_soil_water_layer_2’ (m³/m³).

It should be noted that such approaches invariably disregard the nonlinear moisture dynamics at layer boundaries and differential soil density across layers.

The study utilized precipitation data from the ERA5-Land dataset, represented by the variable «total_precipitation_sum», which quantifies the accumulated liquid and solid atmospheric precipitation (rain, snow, etc.) over a 24 h period. The daily time interval was selected to ensure consistency with soil moisture data.

In this study, we investigated the relationship between precipitation and soil moisture (measured as moisture_0_28cm) across different years (2013, 2022, and 2023) and soil texture classes (0–4). Our primary objective was to assess how different lag times (from 0 to 5 days) influence this relationship. Note that “lag 0” refers to the model without any shift (i.e., using the precipitation of the current day), whereas for lag values > 0, the precipitation variable is shifted so that precipitation on day N is used to predict the soil moisture on day N + lag, capturing the delayed response of soil moisture to rainfall.

We selected lags of 1–5 days to span the typical timescale of soil moisture’s response to rainfall. Empirical hydrological studies indicate that the e-folding time of root zone moisture decay (under no-rain conditions) spans 1–3 days, depending on soil hydraulic conductivity and storage capacity [35,40]. Testing up to 5 days, therefore, encompasses both the rapid initial infiltration (lag 1) and the slower drainage phases (lags 2–5).

The following steps were implemented:

• Data Preparation and Filtering:
  • Data for each year and soil texture class were loaded from CSV files;
  • Two filtering strategies were applied: no filtering (using the entire dataset) and seasonal filtering (limiting the analysis to high-precipitation months, 6–9, and excluding days with near-zero precipitation, ≤0.001).
• Regression Analysis:
  • Data for each year and soil texture class were loaded from CSV files;
  • Statsmodels was used for regression analysis;
  • SciPy was used for statistical testing (e.g., Shapiro–Wilk test);
  • Matplotlib 3.10.0 and Seaborn v0.1 were used for data visualization.
• Software Tools:
  The analysis was conducted using Python (version 3.13) with the following libraries:
  • Pandas for data manipulation;
  • Statsmodels for regression analysis;
  • SciPy for statistical testing (e.g., Shapiro–Wilk test);
  • Matplotlib and Seaborn for data visualization.
  • We used a simple linear regression model to relate root zone soil moisture to lagged daily precipitation: moisture_0_28cm = α + β × precipitation (lag n), where n = 0–5 days, depending on the lag configuration. We applied ordinary least squares (OLS) regression using statsmodels, assessing the significance (p < 0.05) of intercept (α) and slope (β), with residuals tested for normality using the Shapiro–Wilk test.

3. Results and Discussion

3.1. Seasonal and Interannual SM Fluctuations

The SM in the EKR varies between years and seasons. Data analysis shows that the maximum values of SM for each class change very slightly from year to year (

Table 1. The SM fluctuations during three contrasting years.

Soil Texture Class	SM Fluctuations
	2013—Extremely Humid			2022—Extremely Dry			2023—Mean
	Max	Min	Mean	Max	Min	Mean	Max	Min	Mean
0	0.40	0.28	0.34	0.39	0.20	0.31	0.39	0.20	0.30
1	0.36	0.19	0.24	0.35	0.16	0.23	0.34	0.16	0.20
2	0.39	0.21	0.28	0.38	0.17	0.26	0.39	0.17	0.23
3	0.35	0.17	0.23	0.35	0.15	0.22	0.34	0.15	0.19
4	0.40	0.24	0.31	0.40	0.19	0.29	0.39	0.19	0.26

; Figure 3, Figure 4 and Figure 5). The main reason is the stable snow cover that forms in winter throughout the entire territory of the EKR. Concentrated input during the melt of a seasonal snowpack saturates soils, causing saturation excess. Therefore, the maximum values of SM are observed in the spring in most cases. The exception is 2023, when in the mountains (soil texture class 0) the maximum value of SM was recorded in September, after prolonged precipitation. According to data from the weather stations located in the mountains, 2023 had a very rainy autumn.

The per-pixel distribution of the maximum values of SM shows that there is a relationship between soil texture class and SM (Figure 6, Figure 7 and Figure 8). For example, a vast area with high SM is represented in the north-west of the EKR. Heavy surficial sediments are formed here, which are represented by loess-like loams (class 4). Loess-like loams, due to the peculiarities of their granulometric composition, have relatively good water permeability and moisture capacity and are superior to original loam and clay [85]. On the other hand, the Altai and Saur mountains are also located in the zone of high SM, despite the fact that the amount of precipitation within the mountains varies greatly.

The minimum SM values fluctuate from year to year more strongly, as they are directly associated with precipitation during the warm period of the year. It is significant that the largest interannual amplitudes of the minimum SM are also characteristic of soil class 0. The average annual SM values in the wet, average, and dry years are also at their maximum in class 0. This fact is another confirmation that the mountains are the main sources of flood in the EKR. The seasonal amplitude of SM in all soil texture classes is observed to increase from a wet to a dry year.

It can also be noted that the values of SM content often exceed the SWHC for the corresponding soil texture. For example, the SWHC values range from a minimum of 11.8 ±4.9% for sand to a maximum of 35.0 ± 6.2% for silty clay [42]. The presence of organic matter in the near-surface soil layer, compaction, and porosity are the main reasons for this. Several studies have found that reanalysis products tend to overestimate soil moisture conditions [86,87]. Thus, it is noted that ERA5-Land regularly overestimates SM values in the spring compared to field measurement data. This may be due to an error in the snowmelt onset dates incorporated into the H-TESSEL hydrological model used to create ERA5-Land.

3.2. The Relationship Between Soil Moisture Variability and Precipitation

The analysis identifies a significant relationship between precipitation and SM, particularly during the summer period (high precipitation and freeing the soil from winter moisture). We calculated the Pearson correlation coefficient r and its 95% confidence interval, together with the associated p-value, for each soil texture class with a 1-day lag during the high-precipitation period. All five texture classes show p < 0.01 at their optimal lag, confirming a highly significant relationship. Key findings are summarized below.

Impact of Filtering and Lag. When using the entire dataset, the Pearson correlation between the considered parameters is moderate. However, when the analysis is restricted to the high-precipitation period (months 6–9) and days with negligible rainfall are removed, the correlation increases substantially (

Table 2. A summary of the optimal regression models by year and soil texture class.

Year	Texture Class	Optimal Lag	α	β	Pearson Corr.	n_Samples
2013	0	1	0.31	4.25	0.60	77
2013	1	1	0.20	6.02	0.62	57
2013	2	1	0.23	6.07	0.55	63
2013	3	1	0.19	6.05	0.56	50
2013	4	1	0.26	5.97	0.62	58
2022	0	3	0.26	3.47	0.49	51
2022	1	2	0.17	2.60	0.54	27
2022	2	3	0.19	2.91	0.48	41
2022	3	2	0.16	3.40	0.72	31
2022	4	3	0.22	3.39	0.55	35
2023	0	1	0.27	4.59	0.44	63
2023	1	3	0.18	8.01	0.62	34
2023	2	3	0.23	5.36	0.37	52
2023	3	1	0.16	10.82	0.73	37
2023	4	1	0.26	4.42	0.35	46

). We removed all days with total_precipitation_sum ≤ 0.001 mm to exclude instrument noise and drizzle events that do not measurably affect SM. This threshold matches the native precision of the ERA5-Land daily rainfall product. Quantitatively, filtering raises the mean r from ~0.38 to ~0.59 across all textures (an increase of ~0.21), underscoring the benefit of focusing on active rainfall events. For example, the highest correlation in the humid period of 2013 (0.55–0.62) was achieved with a 1-day lag in the filtered dataset for all soil textures. Patterns with other lag values (e.g., 0, 2–5 days) generally showed lower Pearson correlation coefficients. The multi-panel line graph shows the trend of Pearson correlation coefficients versus lag (including lag 0, labeled «0») for each combination of year and soil texture class (Figure 9).

Many land surface models calculate evaporation by parametrizing the relative humidity at the soil surface (α method) or the soil water diffusion resistance (β method). The regression model with a 1-day lag during the high-precipitation period is represented by the following equation:

moisture_0_28cm = α + β × precipitation (lag 1).

(3)

The formula (units m³/m³ per mm) quantifies the incremental increase in root zone moisture per additional millimeter of rain after 1 day. For example, implies that 1 mm of rainfall raises root zone moisture by 0.7% of the soil volume on average.

Both parameters (α and β) were statistically significant (p < 0.05). Confidence intervals for the coefficients were narrow, and the Shapiro–Wilk test confirmed that the residuals are normally distributed. We found = 0.97–0.99 and p > 0.05 in every case, confirming no significant departure from normality. Thus, the standard confidence intervals and p-values for and are valid.

In contrast, for 2022 and 2023 the optimal patterns show different coefficient values and lags (1–3 days). In the dry period of 2022 the highest correlation was achieved with a 2- or 3-day lag, while in 2023, for texture class 3 the correlation reached 0.73 with a 1-day lag. We observe that coarser soils (sand, loamy sand) attain their maximum correlation with a 1-day lag, reflecting their high hydraulic conductivity and rapid infiltration. Conversely, finer soils (loam, silt loam) peak at 2–3 days, consistent with their slower percolation and longer storage times. This pattern quantitatively aligns with the measured saturated hydraulic conductivities (Ks) that roughly decrease by an order of magnitude from sands (~10⁻³ m/s) to silty loams (~10⁻⁴ m/s), yielding proportionally longer moisture response times.

Thus, precipitation has a clear impact on SM patterns in different texture classes. Thanks to the strong relationship between the temporal evolution of rainfall and soil moisture, the latter can be employed to improve the quality of satellite precipitation products [88] and improve flood prediction [67,68].

It is known that discharge and other hydrological variables such as SM are often a result of accumulated precipitation. Extremes in discharge, i.e., both high and low flows, are in most cases caused by precipitation events spanning more than a day and are strongly influenced by antecedent wetness conditions [70]. Given the essential role that SM plays in runoff generation, our results can be useful for improving the use of the corrected rainfall data in streamflow prediction.

4. Conclusions

SM is one of the key factors in predicting droughts and floods. High soil moisture can increase the risk of floods, while low soil moisture is linked with drought conditions, each having substantial economic and social impacts.

The analysis of soil moisture across different soil texture classes in the East Kazakhstan Region (EKR) revealed several key trends. The maximum values of SM for each texture class change very slightly from year to year. An increase in SM always occurs in spring as snowmelt drains into the ground. In the mountains this can set the stage for potential flooding. When the thin soil layer above bedrock becomes fully saturated, subsequent rainfall may not infiltrate, increasing flood risk. The minimum SM values fluctuate from year to year more strongly, as they are directly associated with precipitation during the warm period of the year.

The per-pixel distribution of the maximum values of SM shows that the soil texture has a clear impact on SM patterns. The largest interannual amplitudes of the minimum SM are characteristic of soil class 0. The average annual SM values in years with different moisture availability are also at their maximum in class 0. The seasonal amplitudes of SM in all soil texture classes are observed to increase from a wet to a dry year.

The values of natural SM content often exceed the SWHC for the corresponding soil texture. The presence of organic matter in the near-surface soil layer and the reanalysis products’ overestimation are the main reasons for this. Another reason is mistakes in the satellite products’ data. This may lead to incorrect discharge simulations.

The regression analysis demonstrates a statistically significant relationship between precipitation and soil moisture in the EKR. The best performance is achieved when using a 1-day lag for 2013 and varying optimal lags for 2022 and 2023 (ranging from 1 to 3 days) during the high-precipitation period (months 6–9), with filtering applied to remove days with negligible rainfall. Patterns with no lag or with lags greater than the optimal value generally exhibit lower Pearson correlation coefficients.

This study has significant prospects, especially due to the need for the prediction of streamflow in ungauged catchments. Many of the ungauged catchments are located in mountainous regions around the world. Preventive runoff assessments are important due to the fact that many people live in the lower reaches of mountainous rivers. Such studies can complement rainfall runoff modelling. In addition, the study’s results may be useful for agricultural purposes, preventing landslide hazards, etc. At the same time, additional field studies are needed to analyze the estimation of moisture variability in different textured soils.